Advanced Bash-Scripting Guide

An in-depth exploration of the art of shell scripting

Mendel Cooper


03 October 2004

Revision History
Revision 2.718 April 2004Revised by: mc
'MULBERRY' release: Minor update.
Revision 2.811 July 2004Revised by: mc
'ELDERBERRY' release: Minor update.
Revision 3.003 Oct 2004Revised by: mc
'LOGANBERRY' release: Major update.

This tutorial assumes no previous knowledge of scripting or programming, but progresses rapidly toward an intermediate/advanced level of instruction . . . all the while sneaking in little snippets of UNIX® wisdom and lore. It serves as a textbook, a manual for self-study, and a reference and source of knowledge on shell scripting techniques. The exercises and heavily-commented examples invite active reader participation, under the premise that the only way to really learn scripting is to write scripts.

This book is suitable for classroom use as a general introduction to programming concepts.

The latest update of this document, as an archived, bzip2-ed "tarball" including both the SGML source and rendered HTML, may be downloaded from the author's home site. A pdf version is also available. See the change log for a revision history.


For Anita, the source of all the magic

Table of Contents
Part 1. Introduction
1. Why Shell Programming?
2. Starting Off With a Sha-Bang
Part 2. Basics
3. Special Characters
4. Introduction to Variables and Parameters
5. Quoting
6. Exit and Exit Status
7. Tests
8. Operations and Related Topics
Part 3. Beyond the Basics
9. Variables Revisited
10. Loops and Branches
11. Internal Commands and Builtins
12. External Filters, Programs and Commands
13. System and Administrative Commands
14. Command Substitution
15. Arithmetic Expansion
16. I/O Redirection
17. Here Documents
18. Recess Time
Part 4. Advanced Topics
19. Regular Expressions
20. Subshells
21. Restricted Shells
22. Process Substitution
23. Functions
24. Aliases
25. List Constructs
26. Arrays
27. Files
28. /dev and /proc
29. Of Zeros and Nulls
30. Debugging
31. Options
32. Gotchas
33. Scripting With Style
34. Miscellany
35. Bash, versions 2 and 3
36. Endnotes
36.1. Author's Note
36.2. About the Author
36.3. Where to Go For Help
36.4. Tools Used to Produce This Book
36.5. Credits
A. Contributed Scripts
B. Reference Cards
C. A Sed and Awk Micro-Primer
C.1. Sed
C.2. Awk
D. Exit Codes With Special Meanings
E. A Detailed Introduction to I/O and I/O Redirection
F. Standard Command-Line Options
G. Important System Directories
H. Localization
I. History Commands
J. A Sample .bashrc File
K. Converting DOS Batch Files to Shell Scripts
L. Exercises
L.1. Analyzing Scripts
L.2. Writing Scripts
M. Revision History
N. To Do List
O. Copyright
List of Tables
11-1. Job identifiers
31-1. Bash options
34-1. Numbers representing colors in Escape Sequences
B-1. Special Shell Variables
B-2. TEST Operators: Binary Comparison
B-3. TEST Operators: Files
B-4. Parameter Substitution and Expansion
B-5. String Operations
B-6. Miscellaneous Constructs
C-1. Basic sed operators
C-2. Examples of sed operators
D-1. "Reserved" Exit Codes
K-1. Batch file keywords / variables / operators, and their shell equivalents
K-2. DOS commands and their UNIX equivalents
M-1. Revision History
List of Examples
2-1. cleanup: A script to clean up the log files in /var/log
2-2. cleanup: An improved clean-up script
2-3. cleanup: An enhanced and generalized version of above scripts.
3-1. Code blocks and I/O redirection
3-2. Saving the results of a code block to a file
3-3. Running a loop in the background
3-4. Backup of all files changed in last day
4-1. Variable assignment and substitution
4-2. Plain Variable Assignment
4-3. Variable Assignment, plain and fancy
4-4. Integer or string?
4-5. Positional Parameters
4-6. wh, whois domain name lookup
4-7. Using shift
5-1. Echoing Weird Variables
5-2. Escaped Characters
6-1. exit / exit status
6-2. Negating a condition using !
7-1. What is truth?
7-2. Equivalence of test, /usr/bin/test, [ ], and /usr/bin/[
7-3. Arithmetic Tests using (( ))
7-4. Testing for broken links
7-5. Arithmetic and string comparisons
7-6. Testing whether a string is null
7-7. zmost
8-1. Greatest common divisor
8-2. Using Arithmetic Operations
8-3. Compound Condition Tests Using && and ||
8-4. Representation of numerical constants
9-1. $IFS and whitespace
9-2. Timed Input
9-3. Once more, timed input
9-4. Timed read
9-5. Am I root?
9-6. arglist: Listing arguments with $* and $@
9-7. Inconsistent $* and $@ behavior
9-8. $* and $@ when $IFS is empty
9-9. Underscore variable
9-10. Inserting a blank line between paragraphs in a text file
9-11. Converting graphic file formats, with filename change
9-12. Alternate ways of extracting substrings
9-13. Using parameter substitution and error messages
9-14. Parameter substitution and "usage" messages
9-15. Length of a variable
9-16. Pattern matching in parameter substitution
9-17. Renaming file extensions:
9-18. Using pattern matching to parse arbitrary strings
9-19. Matching patterns at prefix or suffix of string
9-20. Using declare to type variables
9-21. Indirect References
9-22. Passing an indirect reference to awk
9-23. Generating random numbers
9-24. Picking a random card from a deck
9-25. Random between values
9-26. Rolling a single die with RANDOM
9-27. Reseeding RANDOM
9-28. Pseudorandom numbers, using awk
9-29. C-type manipulation of variables
10-1. Simple for loops
10-2. for loop with two parameters in each [list] element
10-3. Fileinfo: operating on a file list contained in a variable
10-4. Operating on files with a for loop
10-5. Missing in [list] in a for loop
10-6. Generating the [list] in a for loop with command substitution
10-7. A grep replacement for binary files
10-8. Listing all users on the system
10-9. Checking all the binaries in a directory for authorship
10-10. Listing the symbolic links in a directory
10-11. Symbolic links in a directory, saved to a file
10-12. A C-like for loop
10-13. Using efax in batch mode
10-14. Simple while loop
10-15. Another while loop
10-16. while loop with multiple conditions
10-17. C-like syntax in a while loop
10-18. until loop
10-19. Nested Loop
10-20. Effects of break and continue in a loop
10-21. Breaking out of multiple loop levels
10-22. Continuing at a higher loop level
10-23. Using "continue N" in an actual task
10-24. Using case
10-25. Creating menus using case
10-26. Using command substitution to generate the case variable
10-27. Simple string matching
10-28. Checking for alphabetic input
10-29. Creating menus using select
10-30. Creating menus using select in a function
11-1. printf in action
11-2. Variable assignment, using read
11-3. What happens when read has no variable
11-4. Multi-line input to read
11-5. Detecting the arrow keys
11-6. Using read with file redirection
11-7. Problems reading from a pipe
11-8. Changing the current working directory
11-9. Letting "let" do arithmetic.
11-10. Showing the effect of eval
11-11. Forcing a log-off
11-12. A version of "rot13"
11-13. Using eval to force variable substitution in a Perl script
11-14. Using set with positional parameters
11-15. Reassigning the positional parameters
11-16. "Unsetting" a variable
11-17. Using export to pass a variable to an embedded awk script
11-18. Using getopts to read the options/arguments passed to a script
11-19. "Including" a data file
11-20. A (useless) script that sources itself
11-21. Effects of exec
11-22. A script that exec's itself
11-23. Waiting for a process to finish before proceeding
11-24. A script that kills itself
12-1. Using ls to create a table of contents for burning a CDR disk
12-2. Hello or Good-bye
12-3. Badname, eliminate file names in current directory containing bad characters and whitespace.
12-4. Deleting a file by its inode number
12-5. Logfile: Using xargs to monitor system log
12-6. Copying files in current directory to another
12-7. Killing processes by name
12-8. Word frequency analysis using xargs
12-9. Using expr
12-10. Using date
12-11. Word Frequency Analysis
12-12. Which files are scripts?
12-13. Generating 10-digit random numbers
12-14. Using tail to monitor the system log
12-15. Emulating "grep" in a script
12-16. Checking words in a list for validity
12-17. toupper: Transforms a file to all uppercase.
12-18. lowercase: Changes all filenames in working directory to lowercase.
12-19. Du: DOS to UNIX text file conversion.
12-20. rot13: rot13, ultra-weak encryption.
12-21. Generating "Crypto-Quote" Puzzles
12-22. Formatted file listing.
12-23. Using column to format a directory listing
12-24. nl: A self-numbering script.
12-25. Using cpio to move a directory tree
12-26. Unpacking an rpm archive
12-27. Stripping comments from C program files
12-28. Exploring /usr/X11R6/bin
12-29. An "improved" strings command
12-30. Using cmp to compare two files within a script.
12-31. basename and dirname
12-32. Checking file integrity
12-33. Uudecoding encoded files
12-34. Checking a spam domain
12-35. Getting a stock quote
12-36. Using ssh
12-37. A script that mails itself
12-38. Monthly Payment on a Mortgage
12-39. Base Conversion
12-40. Invoking bc using a "here document"
12-41. Calculating PI
12-42. Converting a decimal number to hexadecimal
12-43. Factoring
12-44. Calculating the hypotenuse of a triangle
12-45. Using seq to generate loop arguments
12-46. Using getopt to parse command-line options
12-47. A script that copies itself
12-48. Exercising dd
12-49. Capturing Keystrokes
12-50. Securely deleting a file
12-51. Filename generator
12-52. Converting meters to miles
12-53. Using m4
13-1. Setting a new password
13-2. Setting an erase character
13-3. secret password: Turning off terminal echoing
13-4. Keypress detection
13-5. pidof helps kill a process
13-6. Checking a CD image
13-7. Creating a filesystem in a file
13-8. Adding a new hard drive
13-9. killall, from /etc/rc.d/init.d
14-1. Stupid script tricks
14-2. Generating a variable from a loop
16-1. Redirecting stdin using exec
16-2. Redirecting stdout using exec
16-3. Redirecting both stdin and stdout in the same script with exec
16-4. Avoiding a subshell
16-5. Redirected while loop
16-6. Alternate form of redirected while loop
16-7. Redirected until loop
16-8. Redirected for loop
16-9. Redirected for loop (both stdin and stdout redirected)
16-10. Redirected if/then test
16-11. Data file "" for above examples
16-12. Logging events
17-1. dummyfile: Creates a 2-line dummy file
17-2. broadcast: Sends message to everyone logged in
17-3. Multi-line message using cat
17-4. Multi-line message, with tabs suppressed
17-5. Here document with parameter substitution
17-6. Upload a file pair to "Sunsite" incoming directory
17-7. Parameter substitution turned off
17-8. A script that generates another script
17-9. Here documents and functions
17-10. "Anonymous" Here Document
17-11. Commenting out a block of code
17-12. A self-documenting script
17-13. Prepending a line to a file
20-1. Variable scope in a subshell
20-2. List User Profiles
20-3. Running parallel processes in subshells
21-1. Running a script in restricted mode
23-1. Simple functions
23-2. Function Taking Parameters
23-3. Passing an indirect reference to a function
23-4. Dereferencing a parameter passed to a function
23-5. Again, dereferencing a parameter passed to a function
23-6. Maximum of two numbers
23-7. Converting numbers to Roman numerals
23-8. Testing large return values in a function
23-9. Comparing two large integers
23-10. Real name from username
23-11. Local variable visibility
23-12. Recursion, using a local variable
23-13. The Towers of Hanoi
24-1. Aliases within a script
24-2. unalias: Setting and unsetting an alias
25-1. Using an "and list" to test for command-line arguments
25-2. Another command-line arg test using an "and list"
25-3. Using "or lists" in combination with an "and list"
26-1. Simple array usage
26-2. Formatting a poem
26-3. Various array operations
26-4. String operations on arrays
26-5. Loading the contents of a script into an array
26-6. Some special properties of arrays
26-7. Of empty arrays and empty elements
26-8. Initializing arrays
26-9. Copying and concatenating arrays
26-10. More on concatenating arrays
26-11. An old friend: The Bubble Sort
26-12. Embedded arrays and indirect references
26-13. Complex array application: Sieve of Eratosthenes
26-14. Emulating a push-down stack
26-15. Complex array application: Exploring a weird mathematical series
26-16. Simulating a two-dimensional array, then tilting it
28-1. Using /dev/tcp for troubleshooting
28-2. Finding the process associated with a PID
28-3. On-line connect status
29-1. Hiding the cookie jar
29-2. Setting up a swapfile using /dev/zero
29-3. Creating a ramdisk
30-1. A buggy script
30-2. Missing keyword
30-3. test24, another buggy script
30-4. Testing a condition with an "assert"
30-5. Trapping at exit
30-6. Cleaning up after Control-C
30-7. Tracing a variable
30-8. Running multiple processes (on an SMP box)
32-1. Numerical and string comparison are not equivalent
32-2. Subshell Pitfalls
32-3. Piping the output of echo to a read
34-1. shell wrapper
34-2. A slightly more complex shell wrapper
34-3. A shell wrapper around an awk script
34-4. Perl embedded in a Bash script
34-5. Bash and Perl scripts combined
34-6. A (useless) script that recursively calls itself
34-7. A (useful) script that recursively calls itself
34-8. Another (useful) script that recursively calls itself
34-9. A "colorized" address database
34-10. Drawing a box
34-11. Echoing colored text
34-12. Return value trickery
34-13. Even more return value trickery
34-14. Passing and returning arrays
34-15. Fun with anagrams
34-16. Widgets invoked from a shell script
35-1. String expansion
35-2. Indirect variable references - the new way
35-3. Simple database application, using indirect variable referencing
35-4. Using arrays and other miscellaneous trickery to deal four random hands from a deck of cards
A-1. manview: Viewing formatted manpages
A-2. mailformat: Formatting an e-mail message
A-3. rn: A simple-minded file rename utility
A-4. blank-rename: renames filenames containing blanks
A-5. encryptedpw: Uploading to an ftp site, using a locally encrypted password
A-6. copy-cd: Copying a data CD
A-7. Collatz series
A-8. days-between: Calculate number of days between two dates
A-9. Make a "dictionary"
A-10. Soundex conversion
A-11. "Game of Life"
A-12. Data file for "Game of Life"
A-13. behead: Removing mail and news message headers
A-14. ftpget: Downloading files via ftp
A-15. password: Generating random 8-character passwords
A-16. fifo: Making daily backups, using named pipes
A-17. Generating prime numbers using the modulo operator
A-18. tree: Displaying a directory tree
A-19. string functions: C-like string functions
A-20. Directory information
A-21. Object-oriented database
A-22. Mounting USB keychain storage devices
A-23. Preserving weblogs
A-24. Protecting literal strings
A-25. Unprotecting literal strings
A-26. Spammer Identification
A-27. Basics Reviewed
C-1. Counting Letter Occurrences
J-1. Sample .bashrc file
K-2. Shell Script Conversion of VIEWDATA.BAT

Part 1. Introduction

The shell is a command interpreter. More than just the insulating layer between the operating system kernel and the user, it's also a fairly powerful programming language. A shell program, called a script, is an easy-to-use tool for building applications by "gluing" together system calls, tools, utilities, and compiled binaries. Virtually the entire repertoire of UNIX commands, utilities, and tools is available for invocation by a shell script. If that were not enough, internal shell commands, such as testing and loop constructs, give additional power and flexibility to scripts. Shell scripts lend themselves exceptionally well to administrative system tasks and other routine repetitive jobs not requiring the bells and whistles of a full-blown tightly structured programming language.

Chapter 1. Why Shell Programming?

A working knowledge of shell scripting is essential to anyone wishing to become reasonably proficient at system administration, even if they do not anticipate ever having to actually write a script. Consider that as a Linux machine boots up, it executes the shell scripts in /etc/rc.d to restore the system configuration and set up services. A detailed understanding of these startup scripts is important for analyzing the behavior of a system, and possibly modifying it.

Writing shell scripts is not hard to learn, since the scripts can be built in bite-sized sections and there is only a fairly small set of shell-specific operators and options [1] to learn. The syntax is simple and straightforward, similar to that of invoking and chaining together utilities at the command line, and there are only a few "rules" to learn. Most short scripts work right the first time, and debugging even the longer ones is straightforward.

A shell script is a "quick and dirty" method of prototyping a complex application. Getting even a limited subset of the functionality to work in a shell script is often a useful first stage in project development. This way, the structure of the application can be tested and played with, and the major pitfalls found before proceeding to the final coding in C, C++, Java, or Perl.

Shell scripting hearkens back to the classic UNIX philosophy of breaking complex projects into simpler subtasks, of chaining together components and utilities. Many consider this a better, or at least more esthetically pleasing approach to problem solving than using one of the new generation of high powered all-in-one languages, such as Perl, which attempt to be all things to all people, but at the cost of forcing you to alter your thinking processes to fit the tool.

When not to use shell scripts

  • Resource-intensive tasks, especially where speed is a factor (sorting, hashing, etc.)

  • Procedures involving heavy-duty math operations, especially floating point arithmetic, arbitrary precision calculations, or complex numbers (use C++ or FORTRAN instead)

  • Cross-platform portability required (use C instead)

  • Complex applications, where structured programming is a necessity (need type-checking of variables, function prototypes, etc.)

  • Mission-critical applications upon which you are betting the ranch, or the future of the company

  • Situations where security is important, where you need to guarantee the integrity of your system and protect against intrusion, cracking, and vandalism

  • Project consists of subcomponents with interlocking dependencies

  • Extensive file operations required (Bash is limited to serial file access, and that only in a particularly clumsy and inefficient line-by-line fashion)

  • Need multi-dimensional arrays

  • Need data structures, such as linked lists or trees

  • Need to generate or manipulate graphics or GUIs

  • Need direct access to system hardware

  • Need port or socket I/O

  • Need to use libraries or interface with legacy code

  • Proprietary, closed-source applications (shell scripts put the source code right out in the open for all the world to see)

If any of the above applies, consider a more powerful scripting language -- perhaps Perl, Tcl, Python, Ruby -- or possibly a high-level compiled language such as C, C++, or Java. Even then, prototyping the application as a shell script might still be a useful development step.

We will be using Bash, an acronym for "Bourne-Again shell" and a pun on Stephen Bourne's now classic Bourne shell. Bash has become a de facto standard for shell scripting on all flavors of UNIX. Most of the principles dealt with in this book apply equally well to scripting with other shells, such as the Korn Shell, from which Bash derives some of its features, [2] and the C Shell and its variants. (Note that C Shell programming is not recommended due to certain inherent problems, as pointed out in an October, 1993 Usenet post by Tom Christiansen.)

What follows is a tutorial on shell scripting. It relies heavily on examples to illustrate various features of the shell. The example scripts work -- they've been tested -- and some of them are even useful in real life. The reader can play with the actual working code of the examples in the source archive ( or scriptname.bash), [3] give them execute permission (chmod u+rx scriptname), then run them to see what happens. Should the source archive not be available, then cut-and-paste from the HTML, pdf, or text rendered versions. Be aware that some of the scripts below introduce features before they are explained, and this may require the reader to temporarily skip ahead for enlightenment.

Unless otherwise noted, the author of this book wrote the example scripts that follow.

Chapter 2. Starting Off With a Sha-Bang

In the simplest case, a script is nothing more than a list of system commands stored in a file. At the very least, this saves the effort of retyping that particular sequence of commands each time it is invoked.

Example 2-1. cleanup: A script to clean up the log files in /var/log

# Cleanup
# Run as root, of course.

cd /var/log
cat /dev/null > messages
cat /dev/null > wtmp
echo "Logs cleaned up."

There is nothing unusual here, just a set of commands that could just as easily be invoked one by one from the command line on the console or in an xterm. The advantages of placing the commands in a script go beyond not having to retype them time and again. The script can easily be modified, customized, or generalized for a particular application.

Example 2-2. cleanup: An improved clean-up script

# Proper header for a Bash script.

# Cleanup, version 2

# Run as root, of course.
# Insert code here to print error message and exit if not root.

# Variables are better than hard-coded values.

cat /dev/null > messages
cat /dev/null > wtmp

echo "Logs cleaned up."

exit # The right and proper method of "exiting" from a script.

Now that's beginning to look like a real script. But we can go even farther . . .

Example 2-3. cleanup: An enhanced and generalized version of above scripts.

# Cleanup, version 3

#  Warning:
#  -------
#  This script uses quite a number of features that will be explained
#+ later on.
#  By the time you've finished the first half of the book,
#+ there should be nothing mysterious about it.

ROOT_UID=0     # Only users with $UID 0 have root privileges.
LINES=50       # Default number of lines saved.
E_XCD=66       # Can't change directory?
E_NOTROOT=67   # Non-root exit error.

# Run as root, of course.
if [ "$UID" -ne "$ROOT_UID" ]
  echo "Must be root to run this script."
  exit $E_NOTROOT

if [ -n "$1" ]
# Test if command line argument present (non-empty).
  lines=$LINES # Default, if not specified on command line.

#  Stephane Chazelas suggests the following,
#+ as a better way of checking command line arguments,
#+ but this is still a bit advanced for this stage of the tutorial.
#    E_WRONGARGS=65  # Non-numerical argument (bad arg format)
#    case "$1" in
#    ""      ) lines=50;;
#    *[!0-9]*) echo "Usage: `basename $0` file-to-cleanup"; exit $E_WRONGARGS;;
#    *       ) lines=$1;;
#    esac
#* Skip ahead to "Loops" chapter to decipher all this.


if [ `pwd` != "$LOG_DIR" ]  # or   if [ "$PWD" != "$LOG_DIR" ]
                            # Not in /var/log?
  echo "Can't change to $LOG_DIR."
  exit $E_XCD
fi  # Doublecheck if in right directory, before messing with log file.

# far more efficient is:
# cd /var/log || {
#   echo "Cannot change to necessary directory." >&2
#   exit $E_XCD;
# }

tail -$lines messages > mesg.temp # Saves last section of message log file.
mv mesg.temp messages             # Becomes new log directory.

# cat /dev/null > messages
#* No longer needed, as the above method is safer.

cat /dev/null > wtmp  #  ': > wtmp' and '> wtmp'  have the same effect.
echo "Logs cleaned up."

exit 0
#  A zero return value from the script upon exit
#+ indicates success to the shell.

Since you may not wish to wipe out the entire system log, this version of the script keeps the last section of the message log intact. You will constantly discover ways of refining previously written scripts for increased effectiveness.

The sha-bang ( #!) at the head of a script tells your system that this file is a set of commands to be fed to the command interpreter indicated. The #! is actually a two-byte [4] "magic number", a special marker that designates a file type, or in this case an executable shell script (see man magic for more details on this fascinating topic). Immediately following the sha-bang is a path name. This is the path to the program that interprets the commands in the script, whether it be a shell, a programming language, or a utility. This command interpreter then executes the commands in the script, starting at the top (line 1 of the script), ignoring comments. [5]

#!/bin/sed -f
#!/usr/awk -f

Each of the above script header lines calls a different command interpreter, be it /bin/sh, the default shell (bash in a Linux system) or otherwise. [6] Using #!/bin/sh, the default Bourne shell in most commercial variants of UNIX, makes the script portable to non-Linux machines, though you sacrifice Bash-specific features. The script will, however, conform to the POSIX [7] sh standard.

Note that the path given at the "sha-bang" must be correct, otherwise an error message -- usually "Command not found" -- will be the only result of running the script.

#! can be omitted if the script consists only of a set of generic system commands, using no internal shell directives. The second example, above, requires the initial #!, since the variable assignment line, lines=50, uses a shell-specific construct. Note again that #!/bin/sh invokes the default shell interpreter, which defaults to /bin/bash on a Linux machine.


This tutorial encourages a modular approach to constructing a script. Make note of and collect "boilerplate" code snippets that might be useful in future scripts. Eventually you can build a quite extensive library of nifty routines. As an example, the following script prolog tests whether the script has been invoked with the correct number of parameters.

if [ $# -ne $Number_of_expected args ]
  echo "Usage: `basename $0` script_parameters"
  exit $E_WRONG_ARGS

2.1. Invoking the script

Having written the script, you can invoke it by sh scriptname, [8] or alternatively bash scriptname. (Not recommended is using sh <scriptname, since this effectively disables reading from stdin within the script.) Much more convenient is to make the script itself directly executable with a chmod.


chmod 555 scriptname (gives everyone read/execute permission) [9]


chmod +rx scriptname (gives everyone read/execute permission)

chmod u+rx scriptname (gives only the script owner read/execute permission)

Having made the script executable, you may now test it by ./scriptname. [10] If it begins with a "sha-bang" line, invoking the script calls the correct command interpreter to run it.

As a final step, after testing and debugging, you would likely want to move it to /usr/local/bin (as root, of course), to make the script available to yourself and all other users as a system-wide executable. The script could then be invoked by simply typing scriptname [ENTER] from the command line.

2.2. Preliminary Exercises

  1. System administrators often write scripts to automate common tasks. Give several instances where such scripts would be useful.

  2. Write a script that upon invocation shows the time and date, lists all logged-in users, and gives the system uptime. The script then saves this information to a logfile.

Chapter 3. Special Characters

Special Characters Found In Scripts and Elsewhere


Comments. Lines beginning with a # (with the exception of #!) are comments.

# This line is a comment.

Comments may also occur at the end of a command.

echo "A comment will follow." # Comment here.

Comments may also follow whitespace at the beginning of a line.

	# A tab precedes this comment.


A command may not follow a comment on the same line. There is no method of terminating the comment, in order for "live code" to begin on the same line. Use a new line for the next command.


Of course, an escaped # in an echo statement does not begin a comment. Likewise, a # appears in certain parameter substitution constructs and in numerical constant expressions.
echo "The # here does not begin a comment."
echo 'The # here does not begin a comment.'
echo The \# here does not begin a comment.
echo The # here begins a comment.

echo ${PATH#*:}       # Parameter substitution, not a comment.
echo $(( 2#101011 ))  # Base conversion, not a comment.

# Thanks, S.C.
The standard quoting and escape characters (" ' \) escape the #.

Certain pattern matching operations also use the #.


Command separator [semicolon]. Permits putting two or more commands on the same line.

echo hello; echo there

if [ -x "$filename" ]; then    # Note that "if" and "then" need separation.
                               # Why?
  echo "File $filename exists."; cp $filename $filename.bak
  echo "File $filename not found."; touch $filename
fi; echo "File test complete."

Note that the ";" sometimes needs to be escaped.


Terminator in a case option [double semicolon].

case "$variable" in
abc)  echo "\$variable = abc" ;;
xyz)  echo "\$variable = xyz" ;;


"dot" command [period]. Equivalent to source (see Example 11-19). This is a bash builtin.


"dot", as a component of a filename. When working with filenames, a dot is the prefix of a "hidden" file, a file that an ls will not normally show.
bash$ touch .hidden-file
bash$ ls -l	      
total 10
 -rw-r--r--    1 bozo      4034 Jul 18 22:04 data1.addressbook
 -rw-r--r--    1 bozo      4602 May 25 13:58 data1.addressbook.bak
 -rw-r--r--    1 bozo       877 Dec 17  2000 employment.addressbook

bash$ ls -al	      
total 14
 drwxrwxr-x    2 bozo  bozo      1024 Aug 29 20:54 ./
 drwx------   52 bozo  bozo      3072 Aug 29 20:51 ../
 -rw-r--r--    1 bozo  bozo      4034 Jul 18 22:04 data1.addressbook
 -rw-r--r--    1 bozo  bozo      4602 May 25 13:58 data1.addressbook.bak
 -rw-r--r--    1 bozo  bozo       877 Dec 17  2000 employment.addressbook
 -rw-rw-r--    1 bozo  bozo         0 Aug 29 20:54 .hidden-file

When considering directory names, a single dot represents the current working directory, and two dots denote the parent directory.

bash$ pwd

bash$ cd .
bash$ pwd

bash$ cd ..
bash$ pwd

The dot often appears as the destination (directory) of a file movement command.

bash$ cp /home/bozo/current_work/junk/* .


"dot" character match. When matching characters, as part of a regular expression, a "dot" matches a single character.


partial quoting [double quote]. "STRING" preserves (from interpretation) most of the special characters within STRING. See also Chapter 5.


full quoting [single quote]. 'STRING' preserves all special characters within STRING. This is a stronger form of quoting than using ". See also Chapter 5.


comma operator. The comma operator links together a series of arithmetic operations. All are evaluated, but only the last one is returned.
let "t2 = ((a = 9, 15 / 3))"  # Set "a" and calculate "t2".


escape [backslash]. A quoting mechanism for single characters.

\X "escapes" the character X. This has the effect of "quoting" X, equivalent to 'X'. The \ may be used to quote " and ', so they are expressed literally.

See Chapter 5 for an in-depth explanation of escaped characters.


Filename path separator [forward slash]. Separates the components of a filename (as in /home/bozo/projects/Makefile).

This is also the division arithmetic operator.


command substitution. The `command` construct makes available the output of command for setting a variable. This is also known as backticks or backquotes.


null command [colon]. This is the shell equivalent of a "NOP" (no op, a do-nothing operation). It may be considered a synonym for the shell builtin true. The ":" command is a itself a Bash builtin, and its exit status is "true" (0).

echo $?   # 0

Endless loop:

while :

# Same as:
#    while true
#    do
#      ...
#    done

Placeholder in if/then test:

if condition
then :   # Do nothing and branch ahead

Provide a placeholder where a binary operation is expected, see Example 8-2 and default parameters.

: ${username=`whoami`}
# ${username=`whoami`}   without the leading : gives an error
#                        unless "username" is a command or builtin...

Provide a placeholder where a command is expected in a here document. See Example 17-10.

Evaluate string of variables using parameter substitution (as in Example 9-13).
: ${HOSTNAME?} ${USER?} ${MAIL?}
#Prints error message if one or more of essential environmental variables not set.

Variable expansion / substring replacement.

In combination with the > redirection operator, truncates a file to zero length, without changing its permissions. If the file did not previously exist, creates it.
: >   # File "" now empty.	      

# Same effect as   cat /dev/null >
# However, this does not fork a new process, since ":" is a builtin.
See also Example 12-14.

In combination with the >> redirection operator, has no effect on a pre-existing target file (: >> target_file). If the file did not previously exist, creates it.


This applies to regular files, not pipes, symlinks, and certain special files.

May be used to begin a comment line, although this is not recommended. Using # for a comment turns off error checking for the remainder of that line, so almost anything may be appear in a comment. However, this is not the case with :.
: This is a comment that generates an error, ( if [ $x -eq 3] ).

The ":" also serves as a field separator, in /etc/passwd, and in the $PATH variable.
bash$ echo $PATH


reverse (or negate) the sense of a test or exit status. The ! operator inverts the exit status of the command to which it is applied (see Example 6-2). It also inverts the meaning of a test operator. This can, for example, change the sense of "equal" ( = ) to "not-equal" ( != ). The ! operator is a Bash keyword.

In a different context, the ! also appears in indirect variable references.

In yet another context, from the command line, the ! invokes the Bash history mechanism (see Appendix I). Note that within a script, the history mechanism is disabled.


wild card [asterisk]. The * character serves as a "wild card" for filename expansion in globbing. By itself, it matches every filename in a given directory.

bash$ echo *

The * also represents any number (or zero) characters in a regular expression.


arithmetic operator. In the context of arithmetic operations, the * denotes multiplication.

A double asterisk, **, is the exponentiation operator.


test operator. Within certain expressions, the ? indicates a test for a condition.

In a double parentheses construct, the ? serves as a C-style trinary operator. See Example 9-29.

In a parameter substitution expression, the ? tests whether a variable has been set.


wild card. The ? character serves as a single-character "wild card" for filename expansion in globbing, as well as representing one character in an extended regular expression.


Variable substitution.

echo $var1     # 5
echo $var2     # 23skidoo

A $ prefixing a variable name indicates the value the variable holds.


end-of-line. In a regular expression, a "$" addresses the end of a line of text.

$*, $@

exit status variable. The $? variable holds the exit status of a command, a function, or of the script itself.


process ID variable. The $$ variable holds the process ID of the script in which it appears.


command group.
(a=hello; echo $a)


A listing of commands within parentheses starts a subshell.

Variables inside parentheses, within the subshell, are not visible to the rest of the script. The parent process, the script, cannot read variables created in the child process, the subshell.
( a=321; )	      

echo "a = $a"   # a = 123
# "a" within parentheses acts like a local variable.

array initialization.
Array=(element1 element2 element3)


Brace expansion.
grep Linux file*.{txt,htm*}
# Finds all instances of the word "Linux"
# in the files "fileA.txt", "file2.txt", "fileR.html", "file-87.htm", etc.

A command may act upon a comma-separated list of file specs within braces. [11] Filename expansion (globbing) applies to the file specs between the braces.


No spaces allowed within the braces unless the spaces are quoted or escaped.

echo {file1,file2}\ :{\ A," B",' C'}

file1 : A file1 : B file1 : C file2 : A file2 : B file2 : C


Block of code [curly brackets]. Also referred to as an "inline group", this construct, in effect, creates an anonymous function. However, unlike a function, the variables in a code block remain visible to the remainder of the script.

bash$ { local a;
	      a=123; }
bash: local: can only be used in a

{ a=321; }
echo "a = $a"   # a = 321   (value inside code block)

# Thanks, S.C.

The code block enclosed in braces may have I/O redirected to and from it.

Example 3-1. Code blocks and I/O redirection

# Reading lines in /etc/fstab.


read line1
read line2
} < $File

echo "First line in $File is:"
echo "$line1"
echo "Second line in $File is:"
echo "$line2"

exit 0

Example 3-2. Saving the results of a code block to a file


# Queries an rpm file for description, listing, and whether it can be installed.
# Saves output to a file.
# This script illustrates using a code block.


if [ -z "$1" ]
  echo "Usage: `basename $0` rpm-file"
  exit $E_NOARGS

  echo "Archive Description:"
  rpm -qpi $1       # Query description.
  echo "Archive Listing:"
  rpm -qpl $1       # Query listing.
  rpm -i --test $1  # Query whether rpm file can be installed.
  if [ "$?" -eq $SUCCESS ]
    echo "$1 can be installed."
    echo "$1 cannot be installed."
} > "$1.test"       # Redirects output of everything in block to file.

echo "Results of rpm test in file $1.test"

# See rpm man page for explanation of options.

exit 0


Unlike a command group within (parentheses), as above, a code block enclosed by {braces} will not normally launch a subshell. [12]

{} \;

pathname. Mostly used in find constructs. This is not a shell builtin.


The ";" ends the -exec option of a find command sequence. It needs to be escaped to protect it from interpretation by the shell.

[ ]


Test expression between [ ]. Note that [ is part of the shell builtin test (and a synonym for it), not a link to the external command /usr/bin/test.

[[ ]]


Test expression between [[ ]] (shell keyword).

See the discussion on the [[ ... ]] construct.

[ ]

array element.

In the context of an array, brackets set off the numbering of each element of that array.
echo ${Array[1]}

[ ]

range of characters.

As part of a regular expression, brackets delineate a range of characters to match.

(( ))

integer expansion.

Expand and evaluate integer expression between (( )).

See the discussion on the (( ... )) construct.

> &> >& >> <

scriptname >filename redirects the output of scriptname to file filename. Overwrite filename if it already exists.

command &>filename redirects both the stdout and the stderr of command to filename.

command >&2 redirects stdout of command to stderr.

scriptname >>filename appends the output of scriptname to file filename. If filename does not already exist, it will be created.



In a different context, the "<" and ">" characters act as string comparison operators.

In yet another context, the "<" and ">" characters act as integer comparison operators. See also Example 12-9.


redirection used in a here document.


redirection used in a here string.

<, >

ASCII comparison.

if [[ "$veg1" < "$veg2" ]]
  echo "Although $veg1 precede $veg2 in the dictionary,"
  echo "this implies nothing about my culinary preferences."
  echo "What kind of dictionary are you using, anyhow?"

\<, \>

bash$ grep '\<the\>' textfile


pipe. Passes the output of previous command to the input of the next one, or to the shell. This is a method of chaining commands together.

echo ls -l | sh
#  Passes the output of "echo ls -l" to the shell,
#+ with the same result as a simple "ls -l".

cat *.lst | sort | uniq
# Merges and sorts all ".lst" files, then deletes duplicate lines.

The output of a command or commands may be piped to a script.
# : Changes input to uppercase.

tr 'a-z' 'A-Z'
#  Letter ranges must be quoted
#+ to prevent filename generation from single-letter filenames.

exit 0
Now, let us pipe the output of ls -l to this script.
bash$ ls -l | ./
-RW-RW-R--    1 BOZO  BOZO       109 APR  7 19:49 1.TXT
 -RW-RW-R--    1 BOZO  BOZO       109 APR 14 16:48 2.TXT
 -RW-R--R--    1 BOZO  BOZO       725 APR 20 20:56 DATA-FILE


The stdout of each process in a pipe must be read as the stdin of the next. If this is not the case, the data stream will block, and the pipe will not behave as expected.
cat file1 file2 | ls -l | sort
# The output from "cat file1 file2" disappears.

A pipe runs as a child process, and therefore cannot alter script variables.
echo "new_value" | read variable
echo "variable = $variable"     # variable = initial_value

If one of the commands in the pipe aborts, this prematurely terminates execution of the pipe. Called a broken pipe, this condition sends a SIGPIPE signal.


force redirection (even if the noclobber option is set). This will forcibly overwrite an existing file.


OR logical operator. In a test construct, the || operator causes a return of 0 (success) if either of the linked test conditions is true.


Run job in background. A command followed by an & will run in the background.

bash$ sleep 10 &
[1] 850
[1]+  Done                    sleep 10

Within a script, commands and even loops may run in the background.

Example 3-3. Running a loop in the background


for i in 1 2 3 4 5 6 7 8 9 10            # First loop.
  echo -n "$i "
done & # Run this loop in background.
       # Will sometimes execute after second loop.

echo   # This 'echo' sometimes will not display.

for i in 11 12 13 14 15 16 17 18 19 20   # Second loop.
  echo -n "$i "

echo   # This 'echo' sometimes will not display.

# ======================================================

# The expected output from the script:
# 1 2 3 4 5 6 7 8 9 10 
# 11 12 13 14 15 16 17 18 19 20 

# Sometimes, though, you get:
# 11 12 13 14 15 16 17 18 19 20 
# 1 2 3 4 5 6 7 8 9 10 bozo $
# (The second 'echo' doesn't execute. Why?)

# Occasionally also:
# 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
# (The first 'echo' doesn't execute. Why?)

# Very rarely something like:
# 11 12 13 1 2 3 4 5 6 7 8 9 10 14 15 16 17 18 19 20 
# The foreground loop preempts the background one.

exit 0


A command run in the background within a script may cause the script to hang, waiting for a keystroke. Fortunately, there is a remedy for this.


AND logical operator. In a test construct, the && operator causes a return of 0 (success) only if both the linked test conditions are true.


option, prefix. Option flag for a command or filter. Prefix for an operator.

COMMAND -[Option1][Option2][...]

ls -al

sort -dfu $filename

set -- $variable

if [ $file1 -ot $file2 ]
  echo "File $file1 is older than $file2."

if [ "$a" -eq "$b" ]
  echo "$a is equal to $b."

if [ "$c" -eq 24 -a "$d" -eq 47 ]
  echo "$c equals 24 and $d equals 47."


redirection from/to stdin or stdout [dash].

(cd /source/directory && tar cf - . ) | (cd /dest/directory && tar xpvf -)
# Move entire file tree from one directory to another
# [courtesy Alan Cox <>, with a minor change]

# 1) cd /source/directory    Source directory, where the files to be moved are.
# 2) &&                     "And-list": if the 'cd' operation successful, then execute the next command.
# 3) tar cf - .              The 'c' option 'tar' archiving command creates a new archive,
#                            the 'f' (file) option, followed by '-' designates the target file as stdout,
#                            and do it in current directory tree ('.').
# 4) |                       Piped to...
# 5) ( ... )                 a subshell
# 6) cd /dest/directory      Change to the destination directory.
# 7) &&                     "And-list", as above
# 8) tar xpvf -              Unarchive ('x'), preserve ownership and file permissions ('p'),
#                            and send verbose messages to stdout ('v'),
#                            reading data from stdin ('f' followed by '-').
#                            Note that 'x' is a command, and 'p', 'v', 'f' are options.
# Whew!

# More elegant than, but equivalent to:
#   cd source/directory
#   tar cf - . | (cd ../dest/directory; tar xpvf -)
# cp -a /source/directory /dest/directory     also has same effect.

bunzip2 linux-2.4.3.tar.bz2 | tar xvf -
# --uncompress tar file--    | --then pass it to "tar"--
# If "tar" has not been patched to handle "bunzip2",
# this needs to be done in two discrete steps, using a pipe.
# The purpose of the exercise is to unarchive "bzipped" kernel source.

Note that in this context the "-" is not itself a Bash operator, but rather an option recognized by certain UNIX utilities that write to stdout, such as tar, cat, etc.

bash$ echo "whatever" | cat -

Where a filename is expected, - redirects output to stdout (sometimes seen with tar cf), or accepts input from stdin, rather than from a file. This is a method of using a file-oriented utility as a filter in a pipe.

bash$ file
Usage: file [-bciknvzL] [-f namefile] [-m magicfiles] file...
By itself on the command line, file fails with an error message.

Add a "-" for a more useful result. This causes the shell to await user input.
bash$ file -
standard input:              ASCII text

bash$ file -
standard input:              Bourne-Again shell script text executable
Now the command accepts input from stdin and analyzes it.

The "-" can be used to pipe stdout to other commands. This permits such stunts as prepending lines to a file.

Using diff to compare a file with a section of another:

grep Linux file1 | diff file2 -

Finally, a real-world example using - with tar.

Example 3-4. Backup of all files changed in last day


#  Backs up all files in current directory modified within last 24 hours
#+ in a "tarball" (tarred and gzipped file).

BACKUPFILE=backup-$(date +%m-%d-%Y)
#                 Embeds date in backup filename.
#                 Thanks, Joshua Tschida, for the idea.
#  If no backup-archive filename specified on command line,
#+ it will default to "backup-MM-DD-YYYY.tar.gz."

tar cvf - `find . -mtime -1 -type f -print` > $archive.tar
gzip $archive.tar
echo "Directory $PWD backed up in archive file \"$archive.tar.gz\"."

#  Stephane Chazelas points out that the above code will fail
#+ if there are too many files found
#+ or if any filenames contain blank characters.

# He suggests the following alternatives:
# -------------------------------------------------------------------
#   find . -mtime -1 -type f -print0 | xargs -0 tar rvf "$archive.tar"
#      using the GNU version of "find".

#   find . -mtime -1 -type f -exec tar rvf "$archive.tar" '{}' \;
#         portable to other UNIX flavors, but much slower.
# -------------------------------------------------------------------

exit 0


Filenames beginning with "-" may cause problems when coupled with the "-" redirection operator. A script should check for this and add an appropriate prefix to such filenames, for example ./-FILENAME, $PWD/-FILENAME, or $PATHNAME/-FILENAME.

If the value of a variable begins with a -, this may likewise create problems.
echo $var		
# Has the effect of "echo -n", and outputs nothing.


previous working directory. A cd - command changes to the previous working directory. This uses the $OLDPWD environmental variable.


Do not confuse the "-" used in this sense with the "-" redirection operator just discussed. The interpretation of the "-" depends on the context in which it appears.


Minus. Minus sign in an arithmetic operation.


Equals. Assignment operator
echo $a   # 28

In a different context, the "=" is a string comparison operator.


Plus. Addition arithmetic operator.

In a different context, the + is a Regular Expression operator.


Option. Option flag for a command or filter.

Certain commands and builtins use the + to enable certain options and the - to disable them.


modulo. Modulo (remainder of a division) arithmetic operation.

In a different context, the % is a pattern matching operator.


home directory [tilde]. This corresponds to the $HOME internal variable. ~bozo is bozo's home directory, and ls ~bozo lists the contents of it. ~/ is the current user's home directory, and ls ~/ lists the contents of it.
bash$ echo ~bozo

bash$ echo ~

bash$ echo ~/

bash$ echo ~:

bash$ echo ~nonexistent-user


current working directory. This corresponds to the $PWD internal variable.


previous working directory. This corresponds to the $OLDPWD internal variable.


beginning-of-line. In a regular expression, a "^" addresses the beginning of a line of text.

Control Characters

change the behavior of the terminal or text display. A control character is a CONTROL + key combination.

Control characters are not normally useful inside a script.

  • Ctl-B

    Backspace (nondestructive).

  • Ctl-C

    Break. Terminate a foreground job.

  • Ctl-D

    Log out from a shell (similar to exit).

    "EOF" (end of file). This also terminates input from stdin.

    When typing text on the console or in an xterm window, Ctl-D erases the character under the cursor. When there are no characters present, Ctl-D logs out of the session, as expected.

  • Ctl-G

    "BEL" (beep).

  • Ctl-H

    "Rubout" (destructive backspace).

    # Embedding Ctl-H in a string.
    a="^H^H"                  # Two Ctl-H's (backspaces).
    echo "abcdef"             # abcdef
    echo -n "abcdef$a "       # abcd f
    #  Space at end  ^              ^ Backspaces twice.
    echo -n "abcdef$a"        # abcdef
    #  No space at end                Doesn't backspace (why?).
                              # Results may not be quite as expected.
    echo; echo

  • Ctl-I

    Horizontal tab.

  • Ctl-J

    Newline (line feed).

  • Ctl-K

    Vertical tab.

    When typing text on the console or in an xterm window, Ctl-K erases from the character under the cursor to end of line.

  • Ctl-L

    Formfeed (clear the terminal screen). This has the same effect as the clear command.

  • Ctl-M

    Carriage return.

    # Thank you, Lee Maschmeyer, for this example.
    read -n 1 -s -p $'Control-M leaves cursor at beginning of this line. Press Enter. \x0d'
                                      # Of course, '0d' is the hex equivalent of Control-M.
    echo >&2   #  The '-s' makes anything typed silent,
               #+ so it is necessary to go to new line explicitly.
    read -n 1 -s -p $'Control-J leaves cursor on next line. \x0a'
    echo >&2   #  Control-J is linefeed.
    read -n 1 -s -p $'And Control-K\x0bgoes straight down.'
    echo >&2   #  Control-K is vertical tab.
    # A better example of the effect of a vertical tab is:
    var=$'\x0aThis is the bottom line\x0bThis is the top line\x0a'
    echo "$var"
    #  This works the same way as the above example. However:
    echo "$var" | col
    #  This causes the right end of the line to be higher than the left end.
    #  It also explains why we started and ended with a line feed --
    #+ to avoid a garbled screen.
    # As Lee Maschmeyer explains:
    # --------------------------
    #  In the [first vertical tab example] . . . the vertical tab
    #+ makes the printing go straight down without a carriage return.
    #  This is true only on devices, such as the Linux console,
    #+ that can't go "backward."
    #  The real purpose of VT is to go straight UP, not down.
    #  It can be used to print superscripts on a printer.
    #  The col utility can be used to emulate the proper behavior of VT.
    exit 0

  • Ctl-Q

    Resume (XON).

    This resumes stdin in a terminal.

  • Ctl-S

    Suspend (XOFF).

    This freezes stdin in a terminal. (Use Ctl-Q to restore input.)

  • Ctl-U

    Erase a line of input, from the cursor backward to beginning of line. In some settings, Ctl-U erases the entire line of input, regardless of cursor position.

  • Ctl-V

    When inputting text, Ctl-V permits inserting control characters. For example, the following two are equivalent:
    echo -e '\x0a'
    echo <Ctl-V><Ctl-J>

    Ctl-V is primarily useful from within a text editor.

  • Ctl-W

    When typing text on the console or in an xterm window, Ctl-W erases from the character under the cursor backwards to the first instance of whitespace. In some settings, Ctl-W erases backwards to first non-alphanumeric character.

  • Ctl-Z

    Pause a foreground job.


functions as a separator, separating commands or variables. Whitespace consists of either spaces, tabs, blank lines, or any combination thereof. In some contexts, such as variable assignment, whitespace is not permitted, and results in a syntax error.

Blank lines have no effect on the action of a script, and are therefore useful for visually separating functional sections.

$IFS, the special variable separating fields of input to certain commands, defaults to whitespace.

Chapter 4. Introduction to Variables and Parameters

Variables are at the heart of every programming and scripting language. They appear in arithmetic operations and manipulation of quantities, string parsing, and are indispensable for working in the abstract with symbols - tokens that represent something else. A variable is nothing more than a location or set of locations in computer memory holding an item of data.

4.1. Variable Substitution

The name of a variable is a placeholder for its value, the data it holds. Referencing its value is called variable substitution.


Let us carefully distinguish between the name of a variable and its value. If variable1 is the name of a variable, then $variable1 is a reference to its value, the data item it contains. The only time a variable appears "naked", without the $ prefix, is when declared or assigned, when unset, when exported, or in the special case of a variable representing a signal (see Example 30-5). Assignment may be with an = (as in var1=27), in a read statement, and at the head of a loop (for var2 in 1 2 3).

Enclosing a referenced value in double quotes (" ") does not interfere with variable substitution. This is called partial quoting, sometimes referred to as "weak quoting". Using single quotes (' ') causes the variable name to be used literally, and no substitution will take place. This is full quoting, sometimes referred to as "strong quoting". See Chapter 5 for a detailed discussion.

Note that $variable is actually a simplified alternate form of ${variable}. In contexts where the $variable syntax causes an error, the longer form may work (see Section 9.3, below).

Example 4-1. Variable assignment and substitution


# Variables: assignment and substitution


# No space permitted on either side of = sign when initializing variables.

#  If "VARIABLE =value",
#+ script tries to run "VARIABLE" command with one argument, "=value".

#  If "VARIABLE= value",
#+ script tries to run "value" command with
#+ the environmental variable "VARIABLE" set to "".

echo hello    # Not a variable reference, just the string "hello".

echo $hello
echo ${hello} # Identical to above.

echo "$hello"
echo "${hello}"


hello="A B  C   D"
echo $hello   # A B C D
echo "$hello" # A B  C   D
# As you see, echo $hello   and   echo "$hello"   give different results.
# Quoting a variable preserves whitespace.


echo '$hello'  # $hello
#  Variable referencing disabled by single quotes,
#+ which causes the "$" to be interpreted literally.

# Notice the effect of different types of quoting.

hello=    # Setting it to a null value.
echo "\$hello (null value) = $hello"
#  Note that setting a variable to a null value is not the same as
#+ unsetting it, although the end result is the same (see below).

# --------------------------------------------------------------

#  It is permissible to set multiple variables on the same line,
#+ if separated by white space.
#  Caution, this may reduce legibility, and may not be portable.

var1=variable1  var2=variable2  var3=variable3
echo "var1=$var1   var2=$var2  var3=$var3"

# May cause problems with older versions of "sh".

# --------------------------------------------------------------

echo; echo

numbers="one two three"
other_numbers="1 2 3"
# If whitespace within a variable, then quotes necessary.
echo "numbers = $numbers"
echo "other_numbers = $other_numbers"   # other_numbers = 1 2 3

echo "uninitialized_variable = $uninitialized_variable"
# Uninitialized variable has null value (no value at all).
uninitialized_variable=   #  Declaring, but not initializing it
                          #+ (same as setting it to a null value, as above).
echo "uninitialized_variable = $uninitialized_variable"
                          # It still has a null value.

uninitialized_variable=23       # Set it.
unset uninitialized_variable    # Unset it.
echo "uninitialized_variable = $uninitialized_variable"
                                # It still has a null value.


exit 0


An uninitialized variable has a "null" value - no assigned value at all (not zero!). Using a variable before assigning a value to it will usually cause problems.

It is nevertheless possible to perform arithmetic operations on an uninitialized variable.
echo "$uninitialized"                                # (blank line)
let "uninitialized += 5"                             # Add 5 to it.
echo "$uninitialized"                                # 5

#  Conclusion:
#  An uninitialized variable has no value, however
#+ it acts as if it were 0 in an arithmetic operation.
#  This is undocumented (and probably non-portable) behavior.
See also Example 11-20.

4.2. Variable Assignment


the assignment operator (no space before & after)


Do not confuse this with = and -eq, which test, rather than assign!

Note that = can be either an assignment or a test operator, depending on context.

Example 4-2. Plain Variable Assignment

# Naked variables


# When is a variable "naked", i.e., lacking the '$' in front?
# When it is being assigned, rather than referenced.

# Assignment
echo "The value of \"a\" is $a."

# Assignment using 'let'
let a=16+5
echo "The value of \"a\" is now $a."


# In a 'for' loop (really, a type of disguised assignment)
echo -n "Values of \"a\" in the loop are: "
for a in 7 8 9 11
  echo -n "$a "


# In a 'read' statement (also a type of assignment)
echo -n "Enter \"a\" "
read a
echo "The value of \"a\" is now $a."


exit 0

Example 4-3. Variable Assignment, plain and fancy


a=23              # Simple case
echo $a
echo $b

# Now, getting a little bit fancier (command substitution).

a=`echo Hello!`   # Assigns result of 'echo' command to 'a'
echo $a
#  Note that using an exclamation mark (!) in command substitution
#+ will not work from the command line,
#+ since this triggers the Bash "history mechanism."
#  Within a script, however, the history functions are disabled.

a=`ls -l`         # Assigns result of 'ls -l' command to 'a'
echo $a           # Unquoted, however, removes tabs and newlines.
echo "$a"         # The quoted variable preserves whitespace.
                  # (See the chapter on "Quoting.")

exit 0

Variable assignment using the $(...) mechanism (a newer method than backquotes)

# From /etc/rc.d/rc.local
R=$(cat /etc/redhat-release)
arch=$(uname -m)

4.3. Bash Variables Are Untyped

Unlike many other programming languages, Bash does not segregate its variables by "type". Essentially, Bash variables are character strings, but, depending on context, Bash permits integer operations and comparisons on variables. The determining factor is whether the value of a variable contains only digits.

Example 4-4. Integer or string?

# Integer or string?

a=2334                   # Integer.
let "a += 1"
echo "a = $a "           # a = 2335
echo                     # Integer, still.

b=${a/23/BB}             # Substitute "BB" for "23".
                         # This transforms $b into a string.
echo "b = $b"            # b = BB35
declare -i b             # Declaring it an integer doesn't help.
echo "b = $b"            # b = BB35

let "b += 1"             # BB35 + 1 =
echo "b = $b"            # b = 1

echo "c = $c"            # c = BB34
d=${c/BB/23}             # Substitute "23" for "BB".
                         # This makes $d an integer.
echo "d = $d"            # d = 2334
let "d += 1"             # 2334 + 1 =
echo "d = $d"            # d = 2335

# What about null variables?
echo "e = $e"            # e =
let "e += 1"             # Arithmetic operations allowed on a null variable?
echo "e = $e"            # e = 1
echo                     # Null variable transformed into an integer.

# What about undeclared variables?
echo "f = $f"            # f =
let "f += 1"             # Arithmetic operations allowed?
echo "f = $f"            # f = 1
echo                     # Undeclared variable transformed into an integer.

# Variables in Bash are essentially untyped.

exit 0

Untyped variables are both a blessing and a curse. They permit more flexibility in scripting (enough rope to hang yourself!) and make it easier to grind out lines of code. However, they permit errors to creep in and encourage sloppy programming habits.

The burden is on the programmer to keep track of what type the script variables are. Bash will not do it for you.

4.4. Special Variable Types

local variables

variables visible only within a code block or function (see also local variables in functions)

environmental variables

variables that affect the behavior of the shell and user interface


In a more general context, each process has an "environment", that is, a group of variables that hold information that the process may reference. In this sense, the shell behaves like any other process.

Every time a shell starts, it creates shell variables that correspond to its own environmental variables. Updating or adding new environmental variables causes the shell to update its environment, and all the shell's child processes (the commands it executes) inherit this environment.


The space allotted to the environment is limited. Creating too many environmental variables or ones that use up excessive space may cause problems.

bash$ eval "`seq 10000 | sed -e 's/.*/export var&=ZZZZZZZZZZZZZZ/'`"

bash$ du
bash: /usr/bin/du: Argument list too long

(Thank you, S. C. for the clarification, and for providing the above example.)

If a script sets environmental variables, they need to be "exported", that is, reported to the environment local to the script. This is the function of the export command.


A script can export variables only to child processes, that is, only to commands or processes which that particular script initiates. A script invoked from the command line cannot export variables back to the command line environment. Child processes cannot export variables back to the parent processes that spawned them.


positional parameters

arguments passed to the script from the command line - $0, $1, $2, $3... $0 is the name of the script itself, $1 is the first argument, $2 the second, $3 the third, and so forth. [13] After $9, the arguments must be enclosed in brackets, for example, ${10}, ${11}, ${12}.

The special variables $* and $@ denote all the positional parameters.

Example 4-5. Positional Parameters


# Call this script with at least 10 parameters, for example
# ./scriptname 1 2 3 4 5 6 7 8 9 10


echo "The name of this script is \"$0\"."
# Adds ./ for current directory
echo "The name of this script is \"`basename $0`\"."
# Strips out path name info (see 'basename')


if [ -n "$1" ]              # Tested variable is quoted.
 echo "Parameter #1 is $1"  # Need quotes to escape #

if [ -n "$2" ]
 echo "Parameter #2 is $2"

if [ -n "$3" ]
 echo "Parameter #3 is $3"

# ...

if [ -n "${10}" ]  # Parameters > $9 must be enclosed in {brackets}.
 echo "Parameter #10 is ${10}"

echo "-----------------------------------"
echo "All the command-line parameters are: "$*""

if [ $# -lt "$MINPARAMS" ]
  echo "This script needs at least $MINPARAMS command-line arguments!"


exit 0

The bracket notation for positional parameters leads to a fairly simple way of referencing the last argument passed to a script on the command line. This also requires indirect referencing.

args=$#           # Number of args passed.
lastarg=${!args}  # Note that lastarg=${!$#} doesn't work.

Some scripts can perform different operations, depending on which name they are invoked with. For this to work, the script needs to check $0, the name it was invoked by. There must also exist symbolic links to all the alternate names of the script. See Example 12-2.


If a script expects a command line parameter but is invoked without one, this may cause a null variable assignment, generally an undesirable result. One way to prevent this is to append an extra character to both sides of the assignment statement using the expected positional parameter.

# This will prevent an error, even if positional parameter is absent.


# The extra character can be stripped off later, if desired, like so.
variable1=${variable1_/_/}   # Side effects only if $variable1_ begins with "_".
# This uses one of the parameter substitution templates discussed in Chapter 9.
# Leaving out the replacement pattern results in a deletion.

#  A more straightforward way of dealing with this is
#+ to simply test whether expected positional parameters have been passed.
if [ -z $1 ]


Example 4-6. wh, whois domain name lookup


# Does a 'whois domain-name' lookup on any of 3 alternate servers:
#          ,,

# Place this script, named 'wh' in /usr/local/bin

# Requires symbolic links:
# ln -s /usr/local/bin/wh /usr/local/bin/wh-ripe
# ln -s /usr/local/bin/wh /usr/local/bin/wh-cw
# ln -s /usr/local/bin/wh /usr/local/bin/wh-radb

if [ -z "$1" ]
  echo "Usage: `basename $0` [domain-name]"
  exit 65

case `basename $0` in
# Checks script name and calls proper server
    "wh"     ) whois $;;
    "wh-ripe") whois $;;
    "wh-radb") whois $;;
    "wh-cw"  ) whois $;;
    *        ) echo "Usage: `basename $0` [domain-name]";;

exit 0


The shift command reassigns the positional parameters, in effect shifting them to the left one notch.

$1 <--- $2, $2 <--- $3, $3 <--- $4, etc.

The old $1 disappears, but $0 (the script name) does not change. If you use a large number of positional parameters to a script, shift lets you access those past 10, although {bracket} notation also permits this.

Example 4-7. Using shift

# Using 'shift' to step through all the positional parameters.

#  Name this script something like shft,
#+ and invoke it with some parameters, for example
#          ./shft a b c def 23 skidoo

until [ -z "$1" ]  # Until all parameters used up...
  echo -n "$1 "

echo               # Extra line feed.

exit 0


The shift command also works on parameters passed to a function. See Example 34-12.

Chapter 5. Quoting

Quoting means just that, bracketing a string in quotes. This has the effect of protecting special characters in the string from reinterpretation or expansion by the shell or shell script. (A character is "special" if it has an interpretation other than its literal meaning, such as the wild card character, *.)

bash$ ls -l [Vv]*
-rw-rw-r--    1 bozo  bozo       324 Apr  2 15:05 VIEWDATA.BAT
 -rw-rw-r--    1 bozo  bozo       507 May  4 14:25
 -rw-rw-r--    1 bozo  bozo       539 Apr 14 17:11

bash$ ls -l '[Vv]*'
ls: [Vv]*: No such file or directory


Certain programs and utilities reinterpret or expand special characters in a quoted string. An important use of quoting is protecting a command-line parameter from the shell, but still letting the calling program expand it.

bash$ grep '[Ff]irst' *.txt
file1.txt:This is the first line of file1.txt.
 file2.txt:This is the First line of file2.txt.

Note that the unquoted grep [Ff]irst *.txt works under the Bash shell, but not under tcsh.

When referencing a variable, it is generally advisable to enclose it in double quotes (" "). This preserves all special characters within the variable name, except $, ` (backquote), and \ (escape). [14] Keeping $ as a special character within double quotes permits referencing a quoted variable ("$variable"), that is, replacing the variable with its value (see Example 4-1, above).

Use double quotes to prevent word splitting. [15] An argument enclosed in double quotes presents itself as a single word, even if it contains whitespace separators.
variable1="a variable containing five words"
COMMAND This is $variable1    # Executes COMMAND with 7 arguments:
# "This" "is" "a" "variable" "containing" "five" "words"

COMMAND "This is $variable1"  # Executes COMMAND with 1 argument:
# "This is a variable containing five words"

variable2=""    # Empty.

COMMAND $variable2 $variable2 $variable2        # Executes COMMAND with no arguments. 
COMMAND "$variable2" "$variable2" "$variable2"  # Executes COMMAND with 3 empty arguments. 
COMMAND "$variable2 $variable2 $variable2"      # Executes COMMAND with 1 argument (2 spaces). 

# Thanks, S.C.


Enclosing the arguments to an echo statement in double quotes is necessary only when word splitting is an issue.

Example 5-1. Echoing Weird Variables

# Echoing weird variables.

echo $var        # '(]\{}$"
echo "$var"      # '(]\{}$"     Doesn't make a difference.


echo $var        # '(] {}$"     \ converted to space.
echo "$var"      # '(]\{}$"

# Examples above supplied by S.C.

exit 0

Single quotes (' ') operate similarly to double quotes, but do not permit referencing variables, since the special meaning of $ is turned off. Within single quotes, every special character except ' gets interpreted literally. Consider single quotes ("full quoting") to be a stricter method of quoting than double quotes ("partial quoting").


Since even the escape character (\) gets a literal interpretation within single quotes, trying to enclose a single quote within single quotes will not yield the expected result.
echo "Why can't I write 's between single quotes"


# The roundabout method.
echo 'Why can'\''t I write '"'"'s between single quotes'
#    |-------|  |----------|   |-----------------------|
# Three single-quoted strings, with escaped and quoted single quotes between.

# This example courtesy of Stephane Chazelas.

Escaping is a method of quoting single characters. The escape (\) preceding a character tells the shell to interpret that character literally.


With certain commands and utilities, such as echo and sed, escaping a character may have the opposite effect - it can toggle on a special meaning for that character.

Special meanings of certain escaped characters

used with echo and sed


means newline


means return


means tab


means vertical tab


means backspace


means "alert" (beep or flash)


translates to the octal ASCII equivalent of 0xx

Example 5-2. Escaped Characters

# escaped characters

echo; echo

echo "\v\v\v\v"      # Prints \v\v\v\v literally.
# Use the -e option with 'echo' to print escaped characters.
echo "============="
echo -e "\v\v\v\v"   # Prints 4 vertical tabs.
echo "=============="

echo -e "\042"       # Prints " (quote, octal ASCII character 42).
echo "=============="

# The $'\X' construct makes the -e option unnecessary.
echo; echo "NEWLINE AND BEEP"
echo $'\n'           # Newline.
echo $'\a'           # Alert (beep).

echo "==============="
# Version 2 and later of Bash permits using the $'\nnn' construct.
# Note that in this case, '\nnn' is an octal value.
echo $'\t \042 \t'   # Quote (") framed by tabs.

# It also works with hexadecimal values, in an $'\xhhh' construct.
echo $'\t \x22 \t'  # Quote (") framed by tabs.
# Thank you, Greg Keraunen, for pointing this out.
# Earlier Bash versions allowed '\x022'.
echo "==============="

# Assigning ASCII characters to a variable.
# ----------------------------------------
quote=$'\042'        # " assigned to a variable.
echo "$quote This is a quoted string, $quote and this lies outside the quotes."


# Concatenating ASCII chars in a variable.
triple_underline=$'\137\137\137'  # 137 is octal ASCII code for '_'.
echo "$triple_underline UNDERLINE $triple_underline"


ABC=$'\101\102\103\010'           # 101, 102, 103 are octal A, B, C.
echo $ABC

echo; echo

escape=$'\033'                    # 033 is octal for escape.
echo "\"escape\" echoes as $escape"
#                                   no visible output.

echo; echo

exit 0

See Example 35-1 for another example of the $' ' string expansion construct.


gives the quote its literal meaning

echo "Hello"                  # Hello
echo "\"Hello\", he said."    # "Hello", he said.


gives the dollar sign its literal meaning (variable name following \$ will not be referenced)

echo "\$variable01"  # results in $variable01


gives the backslash its literal meaning

echo "\\"  # Results in \

# Whereas . . .

echo "\"   # Invokes secondary prompt from the command line.
           # In a script, gives an error message.


The behavior of \ depends on whether it is itself escaped, quoted, or appearing within command substitution or a here document.
                      #  Simple escaping and quoting
echo \z               #  z
echo \\z              # \z
echo '\z'             # \z
echo '\\z'            # \\z
echo "\z"             # \z
echo "\\z"            # \z

                      #  Command substitution
echo `echo \z`        #  z
echo `echo \\z`       #  z
echo `echo \\\z`      # \z
echo `echo \\\\z`     # \z
echo `echo \\\\\\z`   # \z
echo `echo \\\\\\\z`  # \\z
echo `echo "\z"`      # \z
echo `echo "\\z"`     # \z

                      # Here document
cat <<EOF              
EOF                   # \z

cat <<EOF              
EOF                   # \z

# These examples supplied by Stephane Chazelas.

Elements of a string assigned to a variable may be escaped, but the escape character alone may not be assigned to a variable.
echo "$variable"
# Will not work - gives an error message:
# : command not found
# A "naked" escape cannot safely be assigned to a variable.
#  What actually happens here is that the "\" escapes the newline and
#+ the effect is        variable=echo "$variable"
#+                      invalid variable assignment

echo "$variable"        #  23skidoo
                        #  This works, since the second line
                        #+ is a valid variable assignment.

#        \^    escape followed by space
echo "$variable"        # space

echo "$variable"        # \

echo "$variable"
# Will not work - gives an error message:
# \: command not found
#  First escape escapes second one, but the third one is left "naked",
#+ with same result as first instance, above.

echo "$variable"        # \\
                        # Second and fourth escapes escaped.
                        # This is o.k.

Escaping a space can prevent word splitting in a command's argument list.
file_list="/bin/cat /bin/gzip /bin/more /usr/bin/less /usr/bin/emacs-20.7"
# List of files as argument(s) to a command.

# Add two files to the list, and list all.
ls -l /usr/X11R6/bin/xsetroot /sbin/dump $file_list

echo "-------------------------------------------------------------------------"

# What happens if we escape a couple of spaces?
ls -l /usr/X11R6/bin/xsetroot\ /sbin/dump\ $file_list
# Error: the first three files concatenated into a single argument to 'ls -l'
#        because the two escaped spaces prevent argument (word) splitting.

The escape also provides a means of writing a multi-line command. Normally, each separate line constitutes a different command, but an escape at the end of a line escapes the newline character, and the command sequence continues on to the next line.

(cd /source/directory && tar cf - . ) | \
(cd /dest/directory && tar xpvf -)
# Repeating Alan Cox's directory tree copy command,
# but split into two lines for increased legibility.

# As an alternative:
tar cf - -C /source/directory . |
tar xpvf - -C /dest/directory
# See note below.
# (Thanks, Stephane Chazelas.)


If a script line ends with a |, a pipe character, then a \, an escape, is not strictly necessary. It is, however, good programming practice to always escape the end of a line of code that continues to the following line.

echo "foo


echo 'foo
bar'    # No difference yet.


echo foo\
bar     # Newline escaped.


echo "foo\
bar"     # Same here, as \ still interpreted as escape within weak quotes.


echo 'foo\
bar'     # Escape character \ taken literally because of strong quoting.

# Examples suggested by Stephane Chazelas.

Chapter 6. Exit and Exit Status


...there are dark corners in the Bourne shell, and people use all of them.

 Chet Ramey

The exit command may be used to terminate a script, just as in a C program. It can also return a value, which is available to the script's parent process.

Every command returns an exit status (sometimes referred to as a return status ). A successful command returns a 0, while an unsuccessful one returns a non-zero value that usually may be interpreted as an error code. Well-behaved UNIX commands, programs, and utilities return a 0 exit code upon successful completion, though there are some exceptions.

Likewise, functions within a script and the script itself return an exit status. The last command executed in the function or script determines the exit status. Within a script, an exit nnn command may be used to deliver an nnn exit status to the shell (nnn must be a decimal number in the 0 - 255 range).


When a script ends with an exit that has no parameter, the exit status of the script is the exit status of the last command executed in the script (not counting the exit).



. . .

# Will exit with status of last command.


The equivalent of a bare exit is exit $? or even just omitting the exit.



. . .

# Will exit with status of last command.

exit $?



. . . 

# Will exit with status of last command.

$? reads the exit status of the last command executed. After a function returns, $? gives the exit status of the last command executed in the function. This is Bash's way of giving functions a "return value". After a script terminates, a $? from the command line gives the exit status of the script, that is, the last command executed in the script, which is, by convention, 0 on success or an integer in the range 1 - 255 on error.

Example 6-1. exit / exit status


echo hello
echo $?    # Exit status 0 returned because command executed successfully.

lskdf      # Unrecognized command.
echo $?    # Non-zero exit status returned because command failed to execute.


exit 113   # Will return 113 to shell.
           # To verify this, type "echo $?" after script terminates.

#  By convention, an 'exit 0' indicates success,
#+ while a non-zero exit value means an error or anomalous condition.

$? is especially useful for testing the result of a command in a script (see Example 12-30 and Example 12-16).


The !, the logical "not" qualifier, reverses the outcome of a test or command, and this affects its exit status.

Example 6-2. Negating a condition using !

true  # the "true" builtin.
echo "exit status of \"true\" = $?"     # 0

! true
echo "exit status of \"! true\" = $?"   # 1
# Note that the "!" needs a space.
#    !true   leads to a "command not found" error
# The '!' operator prefixing a command invokes the Bash history mechanism.

# No error this time, but no negation either.
# It just repeats the previous command (true).

# Thanks, Stephane Chazelas and Kristopher Newsome.


Certain exit status codes have reserved meanings and should not be user-specified in a script.

Chapter 7. Tests

Every reasonably complete programming language can test for a condition, then act according to the result of the test. Bash has the test command, various bracket and parenthesis operators, and the if/then construct.

7.1. Test Constructs

  • An if/then construct tests whether the exit status of a list of commands is 0 (since 0 means "success" by UNIX convention), and if so, executes one or more commands.

  • There exists a dedicated command called [ (left bracket special character). It is a synonym for test, and a builtin for efficiency reasons. This command considers its arguments as comparison expressions or file tests and returns an exit status corresponding to the result of the comparison (0 for true, 1 for false).

  • With version 2.02, Bash introduced the [[ ... ]] extended test command, which performs comparisons in a manner more familiar to programmers from other languages. Note that [[ is a keyword, not a command.

    Bash sees [[ $a -lt $b ]] as a single element, which returns an exit status.

    The (( ... )) and let ... constructs also return an exit status of 0 if the arithmetic expressions they evaluate expand to a non-zero value. These arithmetic expansion constructs may therefore be used to perform arithmetic comparisons.
    let "1<2" returns 0 (as "1<2" expands to "1")
    (( 0 && 1 )) returns 1 (as "0 && 1" expands to "0")

  • An if can test any command, not just conditions enclosed within brackets.
    if cmp a b &> /dev/null  # Suppress output.
    then echo "Files a and b are identical."
    else echo "Files a and b differ."
    # The very useful "if-grep" construct:
    # ----------------------------------- 
    if grep -q Bash file
    then echo "File contains at least one occurrence of Bash."
    if echo "$word" | grep -q "$letter_sequence"
    # The "-q" option to grep suppresses output.
      echo "$letter_sequence found in $word"
      echo "$letter_sequence not found in $word"
    then echo "Command succeeded."
    else echo "Command failed."

  • An if/then construct can contain nested comparisons and tests.
    if echo "Next *if* is part of the comparison for the first *if*."
      if [[ $comparison = "integer" ]]
        then (( a < b ))
        [[ $a < $b ]]
      echo '$a is less than $b'

    This detailed "if-test" explanation courtesy of Stephane Chazelas.

Example 7-1. What is truth?



echo "Testing \"0\""
if [ 0 ]      # zero
  echo "0 is true."
  echo "0 is false."
fi            # 0 is true.


echo "Testing \"1\""
if [ 1 ]      # one
  echo "1 is true."
  echo "1 is false."
fi            # 1 is true.


echo "Testing \"-1\""
if [ -1 ]     # minus one
  echo "-1 is true."
  echo "-1 is false."
fi            # -1 is true.


echo "Testing \"NULL\""
if [ ]        # NULL (empty condition)
  echo "NULL is true."
  echo "NULL is false."
fi            # NULL is false.


echo "Testing \"xyz\""
if [ xyz ]    # string
  echo "Random string is true."
  echo "Random string is false."
fi            # Random string is true.


echo "Testing \"\$xyz\""
if [ $xyz ]   # Tests if $xyz is null, but...
              # it's only an uninitialized variable.
  echo "Uninitialized variable is true."
  echo "Uninitialized variable is false."
fi            # Uninitialized variable is false.


echo "Testing \"-n \$xyz\""
if [ -n "$xyz" ]            # More pedantically correct.
  echo "Uninitialized variable is true."
  echo "Uninitialized variable is false."
fi            # Uninitialized variable is false.


xyz=          # Initialized, but set to null value.

echo "Testing \"-n \$xyz\""
if [ -n "$xyz" ]
  echo "Null variable is true."
  echo "Null variable is false."
fi            # Null variable is false.


# When is "false" true?

echo "Testing \"false\""
if [ "false" ]              #  It seems that "false" is just a string.
  echo "\"false\" is true." #+ and it tests true.
  echo "\"false\" is false."
fi            # "false" is true.


echo "Testing \"\$false\""  # Again, uninitialized variable.
if [ "$false" ]
  echo "\"\$false\" is true."
  echo "\"\$false\" is false."
fi            # "$false" is false.
              # Now, we get the expected result.


exit 0

Exercise. Explain the behavior of Example 7-1, above.

if [ condition-true ]
   command 1
   command 2
   # Optional (may be left out if not needed).
   # Adds default code block executing if original condition tests false.
   command 3
   command 4


When if and then are on same line in a condition test, a semicolon must terminate the if statement. Both if and then are keywords. Keywords (or commands) begin statements, and before a new statement on the same line begins, the old one must terminate.

if [ -x "$filename" ]; then

Else if and elif


elif is a contraction for else if. The effect is to nest an inner if/then construct within an outer one.

if [ condition1 ]
elif [ condition2 ]
# Same as else if

The if test condition-true construct is the exact equivalent of if [ condition-true ]. As it happens, the left bracket, [ , is a token which invokes the test command. The closing right bracket, ] , in an if/test should not therefore be strictly necessary, however newer versions of Bash require it.


The test command is a Bash builtin which tests file types and compares strings. Therefore, in a Bash script, test does not call the external /usr/bin/test binary, which is part of the sh-utils package. Likewise, [ does not call /usr/bin/[, which is linked to /usr/bin/test.

bash$ type test
test is a shell builtin
bash$ type '['
[ is a shell builtin
bash$ type '[['
[[ is a shell keyword
bash$ type ']]'
]] is a shell keyword
bash$ type ']'
bash: type: ]: not found

Example 7-2. Equivalence of test, /usr/bin/test, [ ], and /usr/bin/[



if test -z "$1"
  echo "No command-line arguments."
  echo "First command-line argument is $1."


if /usr/bin/test -z "$1"      # Same result as "test" builtin".
  echo "No command-line arguments."
  echo "First command-line argument is $1."


if [ -z "$1" ]                # Functionally identical to above code blocks.
#   if [ -z "$1"                should work, but...
#+  Bash responds to a missing close-bracket with an error message.
  echo "No command-line arguments."
  echo "First command-line argument is $1."


if /usr/bin/[ -z "$1"         # Again, functionally identical to above.
# if /usr/bin/[ -z "$1" ]     # Works, but gives an error message.
  echo "No command-line arguments."
  echo "First command-line argument is $1."


exit 0

The [[ ]] construct is the more versatile Bash version of [ ]. This is the extended test command, adopted from ksh88.


No filename expansion or word splitting takes place between [[ and ]], but there is parameter expansion and command substitution.


if [[ -e $file ]]
  echo "Password file exists."


Using the [[ ... ]] test construct, rather than [ ... ] can prevent many logic errors in scripts. For example, the &&, ||, <, and > operators work within a [[ ]] test, despite giving an error within a [ ] construct.


Following an if, neither the test command nor the test brackets ( [ ] or [[ ]] ) are strictly necessary.

if cd "$dir" 2>/dev/null; then   # "2>/dev/null" hides error message.
  echo "Now in $dir."
  echo "Can't change to $dir."
The "if COMMAND" construct returns the exit status of COMMAND.

Similarly, a condition within test brackets may stand alone without an if, when used in combination with a list construct.
[ "$var1" -ne "$var2" ] && echo "$var1 is not equal to $var2"

[ -d "$home" ] || echo "$home directory does not exist."

The (( )) construct expands and evaluates an arithmetic expression. If the expression evaluates as zero, it returns an exit status of 1, or "false". A non-zero expression returns an exit status of 0, or "true". This is in marked contrast to using the test and [ ] constructs previously discussed.

Example 7-3. Arithmetic Tests using (( ))

# Arithmetic tests.

# The (( ... )) construct evaluates and tests numerical expressions.
# Exit status opposite from [ ... ] construct!

(( 0 ))
echo "Exit status of \"(( 0 ))\" is $?."         # 1

(( 1 ))
echo "Exit status of \"(( 1 ))\" is $?."         # 0

(( 5 > 4 ))                                      # true
echo "Exit status of \"(( 5 > 4 ))\" is $?."     # 0

(( 5 > 9 ))                                      # false
echo "Exit status of \"(( 5 > 9 ))\" is $?."     # 1

(( 5 - 5 ))                                      # 0
echo "Exit status of \"(( 5 - 5 ))\" is $?."     # 1

(( 5 / 4 ))                                      # Division o.k.
echo "Exit status of \"(( 5 / 4 ))\" is $?."     # 0

(( 1 / 2 ))                                      # Division result < 1.
echo "Exit status of \"(( 1 / 2 ))\" is $?."     # Rounded off to 0.
                                                 # 1

(( 1 / 0 )) 2>/dev/null                          # Illegal division by 0.
echo "Exit status of \"(( 1 / 0 ))\" is $?."     # 1

# What effect does the "2>/dev/null" have?
# What would happen if it were removed?
# Try removing it, then rerunning the script.

exit 0

7.2. File test operators

Returns true if...


file exists


file is a regular file (not a directory or device file)


file is not zero size


file is a directory


file is a block device (floppy, cdrom, etc.)


file is a character device (keyboard, modem, sound card, etc.)


file is a pipe


file is a symbolic link


file is a symbolic link


file is a socket


file (descriptor) is associated with a terminal device

This test option may be used to check whether the stdin ([ -t 0 ]) or stdout ([ -t 1 ]) in a given script is a terminal.


file has read permission (for the user running the test)


file has write permission (for the user running the test)


file has execute permission (for the user running the test)


set-group-id (sgid) flag set on file or directory

If a directory has the sgid flag set, then a file created within that directory belongs to the group that owns the directory, not necessarily to the group of the user who created the file. This may be useful for a directory shared by a workgroup.


set-user-id (suid) flag set on file

A binary owned by root with set-user-id flag set runs with root privileges, even when an ordinary user invokes it. [16] This is useful for executables (such as pppd and cdrecord) that need to access system hardware. Lacking the suid flag, these binaries could not be invoked by a non-root user.
	      -rwsr-xr-t    1 root       178236 Oct  2  2000 /usr/sbin/pppd
A file with the suid flag set shows an s in its permissions.


sticky bit set

Commonly known as the "sticky bit", the save-text-mode flag is a special type of file permission. If a file has this flag set, that file will be kept in cache memory, for quicker access. [17] If set on a directory, it restricts write permission. Setting the sticky bit adds a t to the permissions on the file or directory listing.
	      drwxrwxrwt    7 root         1024 May 19 21:26 tmp/
If a user does not own a directory that has the sticky bit set, but has write permission in that directory, he can only delete files in it that he owns. This keeps users from inadvertently overwriting or deleting each other's files in a publicly accessible directory, such as /tmp.


you are owner of file


group-id of file same as yours


file modified since it was last read

f1 -nt f2

file f1 is newer than f2

f1 -ot f2

file f1 is older than f2

f1 -ef f2

files f1 and f2 are hard links to the same file


"not" -- reverses the sense of the tests above (returns true if condition absent).

Example 7-4. Testing for broken links

# Written by Lee bigelow <>
# Used with permission.

#A pure shell script to find dead symlinks and output them quoted
#so they can be fed to xargs and dealt with :)
#eg. /somedir /someotherdir|xargs rm
#This, however, is a better method:
#find "somedir" -type l -print0|\
#xargs -r0 file|\
#grep "broken symbolic"|
#sed -e 's/^\|: *broken symbolic.*$/"/g'
#but that wouldn't be pure bash, now would it.
#Caution: beware the /proc file system and any circular links!

#If no args are passed to the script set directorys to search 
#to current directory.  Otherwise set the directorys to search 
#to the agrs passed.
[ $# -eq 0 ] && directorys=`pwd` || directorys=$@

#Setup the function linkchk to check the directory it is passed 
#for files that are links and don't exist, then print them quoted.
#If one of the elements in the directory is a subdirectory then 
#send that send that subdirectory to the linkcheck function.
linkchk () {
    for element in $1/*; do
    [ -h "$element" -a ! -e "$element" ] && echo \"$element\"
    [ -d "$element" ] && linkchk $element
    # Of course, '-h' tests for symbolic link, '-d' for directory.

#Send each arg that was passed to the script to the linkchk function
#if it is a valid directoy.  If not, then print the error message
#and usage info.
for directory in $directorys; do
    if [ -d $directory ]
	then linkchk $directory
	    echo "$directory is not a directory"
	    echo "Usage: $0 dir1 dir2 ..."

exit 0

Example 29-1, Example 10-7, Example 10-3, Example 29-3, and Example A-2 also illustrate uses of the file test operators.

7.3. Other Comparison Operators

A binary comparison operator compares two variables or quantities. Note the separation between integer and string comparison.

integer comparison


is equal to

if [ "$a" -eq "$b" ]


is not equal to

if [ "$a" -ne "$b" ]


is greater than

if [ "$a" -gt "$b" ]


is greater than or equal to

if [ "$a" -ge "$b" ]


is less than

if [ "$a" -lt "$b" ]


is less than or equal to

if [ "$a" -le "$b" ]


is less than (within double parentheses)

(("$a" < "$b"))


is less than or equal to (within double parentheses)

(("$a" <= "$b"))


is greater than (within double parentheses)

(("$a" > "$b"))


is greater than or equal to (within double parentheses)

(("$a" >= "$b"))

string comparison


is equal to

if [ "$a" = "$b" ]


is equal to

if [ "$a" == "$b" ]

This is a synonym for =.


The == comparison operator behaves differently within a double-brackets test than within single brackets.
[[ $a == z* ]]    # True if $a starts with an "z" (pattern matching).
[[ $a == "z*" ]]  # True if $a is equal to z* (literal matching).

[ $a == z* ]      # File globbing and word splitting take place.
[ "$a" == "z*" ]  # True if $a is equal to z* (literal matching).

# Thanks, Stephane Chazelas


is not equal to

if [ "$a" != "$b" ]

This operator uses pattern matching within a [[ ... ]] construct.


is less than, in ASCII alphabetical order

if [[ "$a" < "$b" ]]

if [ "$a" \< "$b" ]

Note that the "<" needs to be escaped within a [ ] construct.


is greater than, in ASCII alphabetical order

if [[ "$a" > "$b" ]]

if [ "$a" \> "$b" ]

Note that the ">" needs to be escaped within a [ ] construct.

See Example 26-11 for an application of this comparison operator.


string is "null", that is, has zero length


string is not "null".


The -n test absolutely requires that the string be quoted within the test brackets. Using an unquoted string with ! -z, or even just the unquoted string alone within test brackets (see Example 7-6) normally works, however, this is an unsafe practice. Always quote a tested string. [18]

Example 7-5. Arithmetic and string comparisons



#  Here "a" and "b" can be treated either as integers or strings.
#  There is some blurring between the arithmetic and string comparisons,
#+ since Bash variables are not strongly typed.

#  Bash permits integer operations and comparisons on variables
#+ whose value consists of all-integer characters.
#  Caution advised.


if [ "$a" -ne "$b" ]
  echo "$a is not equal to $b"
  echo "(arithmetic comparison)"


if [ "$a" != "$b" ]
  echo "$a is not equal to $b."
  echo "(string comparison)"
  #     "4"  != "5"
  # ASCII 52 != ASCII 53

# In this particular instance, both "-ne" and "!=" work.


exit 0

Example 7-6. Testing whether a string is null

# Testing null strings and unquoted strings,
#+ but not strings and sealing wax, not to mention cabbages and kings . . .

# Using   if [ ... ]

# If a string has not been initialized, it has no defined value.
# This state is called "null" (not the same as zero).

if [ -n $string1 ]    # $string1 has not been declared or initialized.
  echo "String \"string1\" is not null."
  echo "String \"string1\" is null."
# Wrong result.
# Shows $string1 as not null, although it was not initialized.


# Lets try it again.

if [ -n "$string1" ]  # This time, $string1 is quoted.
  echo "String \"string1\" is not null."
  echo "String \"string1\" is null."
fi                    # Quote strings within test brackets!


if [ $string1 ]       # This time, $string1 stands naked.
  echo "String \"string1\" is not null."
  echo "String \"string1\" is null."
# This works fine.
# The [ ] test operator alone detects whether the string is null.
# However it is good practice to quote it ("$string1").
# As Stephane Chazelas points out,
#    if [ $string1 ]    has one argument, "]"
#    if [ "$string1" ]  has two arguments, the empty "$string1" and "]" 



if [ $string1 ]       # Again, $string1 stands naked.
  echo "String \"string1\" is not null."
  echo "String \"string1\" is null."
# Again, gives correct result.
# Still, it is better to quote it ("$string1"), because . . .

string1="a = b"

if [ $string1 ]       # Again, $string1 stands naked.
  echo "String \"string1\" is not null."
  echo "String \"string1\" is null."
# Not quoting "$string1" now gives wrong result!

exit 0
# Thank you, also, Florian Wisser, for the "heads-up".

Example 7-7. zmost


#View gzipped files with 'most'


if [ $# -eq 0 ] # same effect as:  if [ -z "$1" ]
# $1 can exist, but be empty:  zmost "" arg2 arg3
  echo "Usage: `basename $0` filename" >&2
  # Error message to stderr.
  exit $NOARGS
  # Returns 65 as exit status of script (error code).


if [ ! -f "$filename" ]   # Quoting $filename allows for possible spaces.
  echo "File $filename not found!" >&2
  # Error message to stderr.
  exit $NOTFOUND

if [ ${filename##*.} != "gz" ]
# Using bracket in variable substitution.
  echo "File $1 is not a gzipped file!"
  exit $NOTGZIP

zcat $1 | most

# Uses the file viewer 'most' (similar to 'less').
# Later versions of 'most' have file decompression capabilities.
# May substitute 'more' or 'less', if desired.

exit $?   # Script returns exit status of pipe.
# Actually "exit $?" is unnecessary, as the script will, in any case,
# return the exit status of the last command executed.

compound comparison


logical and

exp1 -a exp2 returns true if both exp1 and exp2 are true.


logical or

exp1 -o exp2 returns true if either exp1 or exp2 are true.

These are similar to the Bash comparison operators && and ||, used within double brackets.
[[ condition1 && condition2 ]]
The -o and -a operators work with the test command or occur within single test brackets.
if [ "$exp1" -a "$exp2" ]

Refer to Example 8-3 and Example 26-16 to see compound comparison operators in action.

7.4. Nested if/then Condition Tests

Condition tests using the if/then construct may be nested. The net result is identical to using the && compound comparison operator above.

if [ condition1 ]
  if [ condition2 ]
    do-something  # But only if both "condition1" and "condition2" valid.

See Example 35-4 for an example of nested if/then condition tests.

7.5. Testing Your Knowledge of Tests

The systemwide xinitrc file can be used to launch the X server. This file contains quite a number of if/then tests, as the following excerpt shows.

if [ -f $HOME/.Xclients ]; then
  exec $HOME/.Xclients
elif [ -f /etc/X11/xinit/Xclients ]; then
  exec /etc/X11/xinit/Xclients
     # failsafe settings.  Although we should never get here
     # (we provide fallbacks in Xclients as well) it can't hurt.
     xclock -geometry 100x100-5+5 &
     xterm -geometry 80x50-50+150 &
     if [ -f /usr/bin/netscape -a -f /usr/share/doc/HTML/index.html ]; then
             netscape /usr/share/doc/HTML/index.html &

Explain the "test" constructs in the above excerpt, then examine the entire file, /etc/X11/xinit/xinitrc, and analyze the if/then test constructs there. You may need to refer ahead to the discussions of grep, sed, and regular expressions.

Chapter 8. Operations and Related Topics

8.1. Operators


variable assignment

Initializing or changing the value of a variable


All-purpose assignment operator, which works for both arithmetic and string assignments.

category=minerals  # No spaces allowed after the "=".


Do not confuse the "=" assignment operator with the = test operator.

#    = as a test operator

if [ "$string1" = "$string2" ]
# if [ "X$string1" = "X$string2" ] is safer,
# to prevent an error message should one of the variables be empty.
# (The prepended "X" characters cancel out.) 

arithmetic operators










# Bash, version 2.02, introduced the "**" exponentiation operator.

let "z=5**3"
echo "z = $z"   # z = 125


modulo, or mod (returns the remainder of an integer division operation)

bash$ echo `expr 5 % 3`

This operator finds use in, among other things, generating numbers within a specific range (see Example 9-23 and Example 9-26) and formatting program output (see Example 26-15 and Example A-7). It can even be used to generate prime numbers, (see Example A-17). Modulo turns up surprisingly often in various numerical recipes.

Example 8-1. Greatest common divisor

# greatest common divisor
#         Uses Euclid's algorithm

#  The "greatest common divisor" (gcd) of two integers
#+ is the largest integer that will divide both, leaving no remainder.

#  Euclid's algorithm uses successive division.
#  In each pass,
#+ dividend <---  divisor
#+ divisor  <---  remainder
#+ until remainder = 0.
#+ The gcd = dividend, on the final pass.
#  For an excellent discussion of Euclid's algorithm, see
#  Jim Loy's site,

# ------------------------------------------------------
# Argument check

if [ $# -ne "$ARGS" ]
  echo "Usage: `basename $0` first-number second-number"
  exit $E_BADARGS
# ------------------------------------------------------

gcd ()

                                 #  Arbitrary assignment.
  dividend=$1                    #  It does not matter
  divisor=$2                     #+ which of the two is larger.
                                 #  Why?

  remainder=1                    #  If uninitialized variable used in loop,
                                 #+ it results in an error message
				 #+ on first pass through loop.

  until [ "$remainder" -eq 0 ]
    let "remainder = $dividend % $divisor"
    dividend=$divisor            # Now repeat with 2 smallest numbers.
  done                           # Euclid's algorithm

}                                # Last $dividend is the gcd.

gcd $1 $2

echo; echo "GCD of $1 and $2 = $dividend"; echo

# Exercise :
# --------
#  Check command-line arguments to make sure they are integers,
#+ and exit the script with an appropriate error message if not.

exit 0

"plus-equal" (increment variable by a constant)

let "var += 5" results in var being incremented by 5.


"minus-equal" (decrement variable by a constant)


"times-equal" (multiply variable by a constant)

let "var *= 4" results in var being multiplied by 4.


"slash-equal" (divide variable by a constant)


"mod-equal" (remainder of dividing variable by a constant)

Arithmetic operators often occur in an expr or let expression.

Example 8-2. Using Arithmetic Operations

# Counting to 11 in 10 different ways.

n=1; echo -n "$n "

let "n = $n + 1"   # let "n = n + 1"  also works.
echo -n "$n "

: $((n = $n + 1))
#  ":" necessary because otherwise Bash attempts
#+ to interpret "$((n = $n + 1))" as a command.
echo -n "$n "

(( n = n + 1 ))
#  A simpler alternative to the method above.
#  Thanks, David Lombard, for pointing this out.
echo -n "$n "

n=$(($n + 1))
echo -n "$n "

: $[ n = $n + 1 ]
#  ":" necessary because otherwise Bash attempts
#+ to interpret "$[ n = $n + 1 ]" as a command.
# Works even if "n" was initialized as a string.
echo -n "$n "

n=$[ $n + 1 ]
#  Works even if "n" was initialized as a string.
#* Avoid this type of construct, since it is obsolete and nonportable.
# Thanks, Stephane Chazelas.
echo -n "$n "

# Now for C-style increment operators.
# Thanks, Frank Wang, for pointing this out.

let "n++"          # let "++n"  also works.
echo -n "$n "

(( n++ ))          # (( ++n )  also works.
echo -n "$n "

: $(( n++ ))       # : $(( ++n )) also works.
echo -n "$n "

: $[ n++ ]         # : $[ ++n ]] also works
echo -n "$n "


exit 0


Integer variables in Bash are actually signed long (32-bit) integers, in the range of -2147483648 to 2147483647. An operation that takes a variable outside these limits will give an erroneous result.
echo "a = $a"      # a = 2147483646
let "a+=1"         # Increment "a".
echo "a = $a"      # a = 2147483647
let "a+=1"         # increment "a" again, past the limit.
echo "a = $a"      # a = -2147483648
                   #      ERROR (out of range)

As of version 2.05b, Bash supports 64-bit integers.


Bash does not understand floating point arithmetic. It treats numbers containing a decimal point as strings.

let "b = $a + 1.3"  # Error.
# let: b = 1.5 + 1.3: syntax error in expression (error token is ".5 + 1.3")

echo "b = $b"       # b=1
Use bc in scripts that that need floating point calculations or math library functions.

bitwise operators. The bitwise operators seldom make an appearance in shell scripts. Their chief use seems to be manipulating and testing values read from ports or sockets. "Bit flipping" is more relevant to compiled languages, such as C and C++, which run fast enough to permit its use on the fly.

bitwise operators


bitwise left shift (multiplies by 2 for each shift position)



let "var <<= 2" results in var left-shifted 2 bits (multiplied by 4)


bitwise right shift (divides by 2 for each shift position)


"right-shift-equal" (inverse of <<=)


bitwise and


"bitwise and-equal"


bitwise OR


"bitwise OR-equal"


bitwise negate


bitwise NOT


bitwise XOR


"bitwise XOR-equal"

logical operators


and (logical)

if [ $condition1 ] && [ $condition2 ]
# Same as:  if [ $condition1 -a $condition2 ]
# Returns true if both condition1 and condition2 hold true...

if [[ $condition1 && $condition2 ]]    # Also works.
# Note that && operator not permitted within [ ... ] construct.


&& may also, depending on context, be used in an and list to concatenate commands.


or (logical)

if [ $condition1 ] || [ $condition2 ]
# Same as:  if [ $condition1 -o $condition2 ]
# Returns true if either condition1 or condition2 holds true...

if [[ $condition1 || $condition2 ]]    # Also works.
# Note that || operator not permitted within [ ... ] construct.


Bash tests the exit status of each statement linked with a logical operator.

Example 8-3. Compound Condition Tests Using && and ||



if [ "$a" -eq 24 ] && [ "$b" -eq 47 ]
  echo "Test #1 succeeds."
  echo "Test #1 fails."

# ERROR:   if [ "$a" -eq 24 && "$b" -eq 47 ]
#+         attempts to execute  ' [ "$a" -eq 24 '
#+         and fails to finding matching ']'.
#  Note:  if [[ $a -eq 24 && $b -eq 24 ]]  works.
#  The double-bracket if-test is more flexible
#+ than the single-bracket version.       
#    (The "&&" has a different meaning in line 17 than in line 6.)
#    Thanks, Stephane Chazelas, for pointing this out.

if [ "$a" -eq 98 ] || [ "$b" -eq 47 ]
  echo "Test #2 succeeds."
  echo "Test #2 fails."

#  The -a and -o options provide
#+ an alternative compound condition test.
#  Thanks to Patrick Callahan for pointing this out.

if [ "$a" -eq 24 -a "$b" -eq 47 ]
  echo "Test #3 succeeds."
  echo "Test #3 fails."

if [ "$a" -eq 98 -o "$b" -eq 47 ]
  echo "Test #4 succeeds."
  echo "Test #4 fails."

if [ "$a" = rhino ] && [ "$b" = crocodile ]
  echo "Test #5 succeeds."
  echo "Test #5 fails."

exit 0

The && and || operators also find use in an arithmetic context.

bash$ echo $(( 1 && 2 )) $((3 && 0)) $((4 || 0)) $((0 || 0))
1 0 1 0

miscellaneous operators


comma operator

The comma operator chains together two or more arithmetic operations. All the operations are evaluated (with possible side effects), but only the last operation is returned.

let "t1 = ((5 + 3, 7 - 1, 15 - 4))"
echo "t1 = $t1"               # t1 = 11

let "t2 = ((a = 9, 15 / 3))"  # Set "a" and calculate "t2".
echo "t2 = $t2    a = $a"     # t2 = 5    a = 9

The comma operator finds use mainly in for loops. See Example 10-12.

8.2. Numerical Constants

A shell script interprets a number as decimal (base 10), unless that number has a special prefix or notation. A number preceded by a 0 is octal (base 8). A number preceded by 0x is hexadecimal (base 16). A number with an embedded # evaluates as BASE#NUMBER (with range and notational restrictions).

Example 8-4. Representation of numerical constants

# Representation of numbers in different bases.

# Decimal: the default
let "dec = 32"
echo "decimal number = $dec"             # 32
# Nothing out of the ordinary here.

# Octal: numbers preceded by '0' (zero)
let "oct = 032"
echo "octal number = $oct"               # 26
# Expresses result in decimal.
# --------- ------ -- -------

# Hexadecimal: numbers preceded by '0x' or '0X'
let "hex = 0x32"
echo "hexadecimal number = $hex"         # 50
# Expresses result in decimal.

# Other bases: BASE#NUMBER
# BASE between 2 and 64.
# NUMBER must use symbols within the BASE range, see below.

let "bin = 2#111100111001101"
echo "binary number = $bin"              # 31181

let "b32 = 32#77"
echo "base-32 number = $b32"             # 231

let "b64 = 64#@_"
echo "base-64 number = $b64"             # 4031
# This notation only works for a limited range (2 - 64)
# 10 digits + 26 lowercase characters + 26 uppercase characters + @ + _


echo $((36#zz)) $((2#10101010)) $((16#AF16)) $((53#1aA))
                                         # 1295 170 44822 3375

#  Important note:
#  --------------
#  Using a digit out of range of the specified base notation
#+ will give an error message.

let "bad_oct = 081"
# Error message output:
# let: oct = 081: value too great for base (error token is "081")
#             Octal numbers use only digits in the range 0 - 7.

exit 0       # Thanks, Rich Bartell and Stephane Chazelas, for clarification.

Chapter 9. Variables Revisited

Used properly, variables can add power and flexibility to scripts. This requires learning their subtleties and nuances.

9.1. Internal Variables

Builtin variables

variables affecting bash script behavior


the path to the Bash binary itself
bash$ echo $BASH


an environmental variable pointing to a Bash startup file to be read when a script is invoked


a 6-element array containing version information about the installed release of Bash. This is similar to $BASH_VERSION, below, but a bit more detailed.

# Bash version info:

for n in 0 1 2 3 4 5
  echo "BASH_VERSINFO[$n] = ${BASH_VERSINFO[$n]}"

# BASH_VERSINFO[0] = 2                      # Major version no.
# BASH_VERSINFO[1] = 05                     # Minor version no.
# BASH_VERSINFO[2] = 8                      # Patch level.
# BASH_VERSINFO[3] = 1                      # Build version.
# BASH_VERSINFO[4] = release                # Release status.
# BASH_VERSINFO[5] = i386-redhat-linux-gnu  # Architecture
                                            # (same as $MACHTYPE).


the version of Bash installed on the system

bash$ echo $BASH_VERSION

tcsh% echo $BASH_VERSION
BASH_VERSION: Undefined variable.

Checking $BASH_VERSION is a good method of determining which shell is running. $SHELL does not necessarily give the correct answer.


the top value in the directory stack (affected by pushd and popd)

This builtin variable corresponds to the dirs command, however dirs shows the entire contents of the directory stack.


the default editor invoked by a script, usually vi or emacs.


"effective" user ID number

Identification number of whatever identity the current user has assumed, perhaps by means of su.


The $EUID is not necessarily the same as the $UID.


name of the current function

xyz23 ()
  echo "$FUNCNAME now executing."  # xyz23 now executing.


                                   # Null value outside a function.


A list of filename patterns to be excluded from matching in globbing.


groups current user belongs to

This is a listing (array) of the group id numbers for current user, as recorded in /etc/passwd.

root# echo $GROUPS

root# echo ${GROUPS[1]}

root# echo ${GROUPS[5]}


home directory of the user, usually /home/username (see Example 9-13)


The hostname command assigns the system name at bootup in an init script. However, the gethostname() function sets the Bash internal variable $HOSTNAME. See also Example 9-13.


host type

Like $MACHTYPE, identifies the system hardware.

bash$ echo $HOSTTYPE

internal field separator

This variable determines how Bash recognizes fields, or word boundaries when it interprets character strings.

$IFS defaults to whitespace (space, tab, and newline), but may be changed, for example, to parse a comma-separated data file. Note that $* uses the first character held in $IFS. See Example 5-1.

bash$ echo $IFS | cat -vte

bash$ bash -c 'set w x y z; IFS=":-;"; echo "$*"'


$IFS does not handle whitespace the same as it does other characters.

Example 9-1. $IFS and whitespace

# $IFS treats whitespace differently than other characters.

  for arg
  do echo "[$arg]"

echo; echo "IFS=\" \""
echo "-------"

IFS=" "
var=" a  b c   "
output_args_one_per_line $var  # output_args_one_per_line `echo " a  b c   "`
# [a]
# [b]
# [c]

echo; echo "IFS=:"
echo "-----"

var=":a::b:c:::"               # Same as above, but substitute ":" for " ".
output_args_one_per_line $var
# []
# [a]
# []
# [b]
# [c]
# []
# []
# []

# The same thing happens with the "FS" field separator in awk.

# Thank you, Stephane Chazelas.


exit 0

(Thanks, S. C., for clarification and examples.)

See also Example 12-34 for an instructive example of using $IFS.


ignore EOF: how many end-of-files (control-D) the shell will ignore before logging out.


Often set in the .bashrc or /etc/profile files, this variable controls collation order in filename expansion and pattern matching. If mishandled, LC_COLLATE can cause unexpected results in filename globbing.


As of version 2.05 of Bash, filename globbing no longer distinguishes between lowercase and uppercase letters in a character range between brackets. For example, ls [A-M]* would match both File1.txt and file1.txt. To revert to the customary behavior of bracket matching, set LC_COLLATE to C by an export LC_COLLATE=C in /etc/profile and/or ~/.bashrc.


This internal variable controls character interpretation in globbing and pattern matching.


This variable is the line number of the shell script in which this variable appears. It has significance only within the script in which it appears, and is chiefly useful for debugging purposes.

last_cmd_arg=$_  # Save it.

echo "At line number $LINENO, variable \"v1\" = $v1"
echo "Last command argument processed = $last_cmd_arg"


machine type

Identifies the system hardware.

bash$ echo $MACHTYPE

old working directory ("OLD-print-working-directory", previous directory you were in)


operating system type

bash$ echo $OSTYPE

path to binaries, usually /usr/bin/, /usr/X11R6/bin/, /usr/local/bin, etc.

When given a command, the shell automatically does a hash table search on the directories listed in the path for the executable. The path is stored in the environmental variable, $PATH, a list of directories, separated by colons. Normally, the system stores the $PATH definition in /etc/profile and/or ~/.bashrc (see Chapter 27).

bash$ echo $PATH

PATH=${PATH}:/opt/bin appends the /opt/bin directory to the current path. In a script, it may be expedient to temporarily add a directory to the path in this way. When the script exits, this restores the original $PATH (a child process, such as a script, may not change the environment of the parent process, the shell).


The current "working directory", ./, is usually omitted from the $PATH as a security measure.


Array variable holding exit status(es) of last executed foreground pipe. Interestingly enough, this does not necessarily give the same result as the exit status of the last executed command.

bash$ echo $PIPESTATUS

bash$ ls -al | bogus_command
bash: bogus_command: command not found
bash$ echo $PIPESTATUS

bash$ ls -al | bogus_command
bash: bogus_command: command not found
bash$ echo $?

The members of the $PIPESTATUS array hold the exit status of each respective command executed in a pipe. $PIPESTATUS[0] holds the exit status of the first command in the pipe, $PIPESTATUS[1] the exit status of the second command, and so on.


The $PIPESTATUS variable may contain an erroneous 0 value in a login shell (in releases prior to 3.0 of Bash).

tcsh% bash

bash$ who | grep nobody | sort
bash$ echo ${PIPESTATUS[*]}

The above lines contained in a script would produce the expected 0 1 0 output.

Thank you, Wayne Pollock for pointing this out and supplying the above example.


In the version 2.05b.0(1)-release of Bash, and possibly earlier versions as well, the $PIPESTATUS variable appears to be broken. This is a bug that has been (mostly) fixed in Bash, version 3.0 and later.

bash$ echo $BASH_VERSION

bash$ $ ls | bogus_command | wc
bash: bogus_command: command not found
 0       0       0

bash$ echo ${PIPESTATUS[@]}
141 127 0

bash$ echo $BASH_VERSION

bash$ $ ls | bogus_command | wc
bash: bogus_command: command not found
 0       0       0

bash$ echo ${PIPESTATUS[@]}
0 127 0


$PIPESTATUS is a "volatile" variable. It needs to be captured immediately after the pipe in question, before any other command intervenes.

bash$ $ ls | bogus_command | wc
bash: bogus_command: command not found
 0       0       0

bash$ echo ${PIPESTATUS[@]}
0 127 0

bash$ echo ${PIPESTATUS[@]}


The $PPID of a process is the process ID (pid) of its parent process. [19]

Compare this with the pidof command.


A variable holding a command to be executed just before the primary prompt, $PS1 is to be displayed.


This is the main prompt, seen at the command line.


The secondary prompt, seen when additional input is expected. It displays as ">".


The tertiary prompt, displayed in a select loop (see Example 10-29).


The quartenary prompt, shown at the beginning of each line of output when invoking a script with the -x option. It displays as "+".


working directory (directory you are in at the time)

This is the analog to the pwd builtin command.



clear # Clear screen.


cd $TargetDirectory
echo "Deleting stale files in $TargetDirectory."

if [ "$PWD" != "$TargetDirectory" ]
then    # Keep from wiping out wrong directory by accident.
  echo "Wrong directory!"
  echo "In $PWD, rather than $TargetDirectory!"
  echo "Bailing out!"

rm -rf *
rm .[A-Za-z0-9]*    # Delete dotfiles.
# rm -f .[^.]* ..?*   to remove filenames beginning with multiple dots.
# (shopt -s dotglob; rm -f *)   will also work.
# Thanks, S.C. for pointing this out.

# Filenames may contain all characters in the 0 - 255 range, except "/".
# Deleting files beginning with weird characters is left as an exercise.

# Various other operations here, as necessary.

echo "Done."
echo "Old files deleted in $TargetDirectory."

exit 0


The default value when a variable is not supplied to read. Also applicable to select menus, but only supplies the item number of the variable chosen, not the value of the variable itself.


# REPLY is the default value for a 'read' command.

echo -n "What is your favorite vegetable? "

echo "Your favorite vegetable is $REPLY."
#  REPLY holds the value of last "read" if and only if
#+ no variable supplied.

echo -n "What is your favorite fruit? "
read fruit
echo "Your favorite fruit is $fruit."
echo "but..."
echo "Value of \$REPLY is still $REPLY."
#  $REPLY is still set to its previous value because
#+ the variable $fruit absorbed the new "read" value.


exit 0


The number of seconds the script has been running.



echo "Hit Control-C to exit before $TIME_LIMIT seconds."

while [ "$SECONDS" -le "$TIME_LIMIT" ]
  if [ "$SECONDS" -eq 1 ]

  echo "This script has been running $SECONDS $units."
  #  On a slow or overburdened machine, the script may skip a count
  #+ every once in a while.
  sleep $INTERVAL

echo -e "\a"  # Beep!

exit 0


the list of enabled shell options, a readonly variable
bash$ echo $SHELLOPTS


Shell level, how deeply Bash is nested. If, at the command line, $SHLVL is 1, then in a script it will increment to 2.


If the $TMOUT environmental variable is set to a non-zero value time, then the shell prompt will time out after time seconds. This will cause a logout.

As of version 2.05b of Bash, it is now possible to use $TMOUT in a script in combination with read.

# Works in scripts for Bash, versions 2.05b and later.

TMOUT=3    # Prompt times out at three seconds.

echo "What is your favorite song?"
echo "Quickly now, you only have $TMOUT seconds to answer!"
read song

if [ -z "$song" ]
  song="(no answer)"
  # Default response.

echo "Your favorite song is $song."

There are other, more complex, ways of implementing timed input in a script. One alternative is to set up a timing loop to signal the script when it times out. This also requires a signal handling routine to trap (see Example 30-5) the interrupt generated by the timing loop (whew!).

Example 9-2. Timed Input


# TMOUT=3            useless in a script

TIMELIMIT=3  # Three seconds in this instance, may be set to different value.

  if [ "$answer" = TIMEOUT ]
    echo $answer
  else       # Don't want to mix up the two instances. 
    echo "Your favorite veggie is $answer"
    kill $!  # Kills no longer needed TimerOn function running in background.
             # $! is PID of last job running in background.


  sleep $TIMELIMIT && kill -s 14 $$ &
  # Waits 3 seconds, then sends sigalarm to script.

  exit 14

trap Int14Vector 14   # Timer interrupt (14) subverted for our purposes.

echo "What is your favorite vegetable "
read answer

#  Admittedly, this is a kludgy implementation of timed input,
#+ however the "-t" option to "read" simplifies this task.
#  See "", below.

#  If you need something really elegant...
#+ consider writing the application in C or C++,
#+ using appropriate library functions, such as 'alarm' and 'setitimer'.

exit 0

An alternative is using stty.

Example 9-3. Once more, timed input


# Written by Stephane Chazelas,
# and modified by the document author.

INTERVAL=5                # timeout interval

timedout_read() {
  old_tty_settings=`stty -g`
  stty -icanon min 0 time ${timeout}0
  eval read $varname      # or just    read $varname
  stty "$old_tty_settings"
  # See man page for "stty".

echo; echo -n "What's your name? Quick! "
timedout_read $INTERVAL your_name

# This may not work on every terminal type.
# The maximum timeout depends on the terminal.
# (it is often 25.5 seconds).


if [ ! -z "$your_name" ]  # If name input before timeout...
  echo "Your name is $your_name."
  echo "Timed out."


# The behavior of this script differs somewhat from "".
# At each keystroke, the counter resets.

exit 0

Perhaps the simplest method is using the -t option to read.

Example 9-4. Timed read

# Inspired by a suggestion from "syngin seven" (thanks).

TIMELIMIT=4         # 4 seconds

read -t $TIMELIMIT variable <&1


if [ -z "$variable" ]
  echo "Timed out, variable still unset."
  echo "variable = $variable"

exit 0

# Exercise for the reader:
# -----------------------
# Why is the redirection (<&1) necessary in line 8?
# What happens if it is omitted?

user ID number

current user's user identification number, as recorded in /etc/passwd

This is the current user's real id, even if she has temporarily assumed another identity through su. $UID is a readonly variable, not subject to change from the command line or within a script, and is the counterpart to the id builtin.

Example 9-5. Am I root?

#   Am I root or not?

ROOT_UID=0   # Root has $UID 0.

if [ "$UID" -eq "$ROOT_UID" ]  # Will the real "root" please stand up?
  echo "You are root."
  echo "You are just an ordinary user (but mom loves you just the same)."

exit 0

# ============================================================= #
# Code below will not execute, because the script already exited.

# An alternate method of getting to the root of matters:


username=`id -nu`              # Or...   username=`whoami`
if [ "$username" = "$ROOTUSER_NAME" ]
  echo "Rooty, toot, toot. You are root."
  echo "You are just a regular fella."

See also Example 2-3.


The variables $ENV, $LOGNAME, $MAIL, $TERM, $USER, and $USERNAME are not Bash builtins. These are, however, often set as environmental variables in one of the Bash startup files. $SHELL, the name of the user's login shell, may be set from /etc/passwd or in an "init" script, and it is likewise not a Bash builtin.

tcsh% echo $LOGNAME
tcsh% echo $SHELL
tcsh% echo $TERM

bash$ echo $LOGNAME
bash$ echo $SHELL
bash$ echo $TERM

Positional Parameters

$0, $1, $2, etc.

positional parameters, passed from command line to script, passed to a function, or set to a variable (see Example 4-5 and Example 11-14)


number of command line arguments [20] or positional parameters (see Example 34-2)


All of the positional parameters, seen as a single word


"$*" must be quoted.


Same as $*, but each parameter is a quoted string, that is, the parameters are passed on intact, without interpretation or expansion. This means, among other things, that each parameter in the argument list is seen as a separate word.


Of course, "$@" should be quoted.

Example 9-6. arglist: Listing arguments with $* and $@

# Invoke this script with several arguments, such as "one two three".


if [ ! -n "$1" ]
  echo "Usage: `basename $0` argument1 argument2 etc."
  exit $E_BADARGS


index=1          # Initialize count.

echo "Listing args with \"\$*\":"
for arg in "$*"  # Doesn't work properly if "$*" isn't quoted.
  echo "Arg #$index = $arg"
  let "index+=1"
done             # $* sees all arguments as single word. 
echo "Entire arg list seen as single word."


index=1          # Reset count.
                 # What happens if you forget to do this?

echo "Listing args with \"\$@\":"
for arg in "$@"
  echo "Arg #$index = $arg"
  let "index+=1"
done             # $@ sees arguments as separate words. 
echo "Arg list seen as separate words."


index=1          # Reset count.

echo "Listing args with \$* (unquoted):"
for arg in $*
  echo "Arg #$index = $arg"
  let "index+=1"
done             # Unquoted $* sees arguments as separate words. 
echo "Arg list seen as separate words."

exit 0

Following a shift, the $@ holds the remaining command-line parameters, lacking the previous $1, which was lost.
# Invoke with ./scriptname 1 2 3 4 5

echo "$@"    # 1 2 3 4 5
echo "$@"    # 2 3 4 5
echo "$@"    # 3 4 5

# Each "shift" loses parameter $1.
# "$@" then contains the remaining parameters.

The $@ special parameter finds use as a tool for filtering input into shell scripts. The cat "$@" construction accepts input to a script either from stdin or from files given as parameters to the script. See Example 12-20 and Example 12-21.


The $* and $@ parameters sometimes display inconsistent and puzzling behavior, depending on the setting of $IFS.

Example 9-7. Inconsistent $* and $@ behavior


#  Erratic behavior of the "$*" and "$@" internal Bash variables,
#+ depending on whether they are quoted or not.
#  Inconsistent handling of word splitting and linefeeds.

set -- "First one" "second" "third:one" "" "Fifth: :one"
# Setting the script arguments, $1, $2, etc.


echo 'IFS unchanged, using "$*"'
for i in "$*"               # quoted
do echo "$((c+=1)): [$i]"   # This line remains the same in every instance.
                            # Echo args.
echo ---

echo 'IFS unchanged, using $*'
for i in $*                 # unquoted
do echo "$((c+=1)): [$i]"
echo ---

echo 'IFS unchanged, using "$@"'
for i in "$@"
do echo "$((c+=1)): [$i]"
echo ---

echo 'IFS unchanged, using $@'
for i in $@
do echo "$((c+=1)): [$i]"
echo ---

echo 'IFS=":", using "$*"'
for i in "$*"
do echo "$((c+=1)): [$i]"
echo ---

echo 'IFS=":", using $*'
for i in $*
do echo "$((c+=1)): [$i]"
echo ---

echo 'IFS=":", using "$var" (var=$*)'
for i in "$var"
do echo "$((c+=1)): [$i]"
echo ---

echo 'IFS=":", using $var (var=$*)'
for i in $var
do echo "$((c+=1)): [$i]"
echo ---

echo 'IFS=":", using $var (var="$*")'
for i in $var
do echo "$((c+=1)): [$i]"
echo ---

echo 'IFS=":", using "$var" (var="$*")'
for i in "$var"
do echo "$((c+=1)): [$i]"
echo ---

echo 'IFS=":", using "$@"'
for i in "$@"
do echo "$((c+=1)): [$i]"
echo ---

echo 'IFS=":", using $@'
for i in $@
do echo "$((c+=1)): [$i]"
echo ---

echo 'IFS=":", using $var (var=$@)'
for i in $var
do echo "$((c+=1)): [$i]"
echo ---

echo 'IFS=":", using "$var" (var=$@)'
for i in "$var"
do echo "$((c+=1)): [$i]"
echo ---

echo 'IFS=":", using "$var" (var="$@")'
for i in "$var"
do echo "$((c+=1)): [$i]"
echo ---

echo 'IFS=":", using $var (var="$@")'
for i in $var
do echo "$((c+=1)): [$i]"


# Try this script with ksh or zsh -y.

exit 0

# This example script by Stephane Chazelas,
# and slightly modified by the document author.


The $@ and $* parameters differ only when between double quotes.

Example 9-8. $* and $@ when $IFS is empty


# If $IFS set, but empty,
# then "$*" and "$@" do not echo positional params as expected.

mecho ()       # Echo positional parameters.
echo "$1,$2,$3";

IFS=""         # Set, but empty.
set a b c      # Positional parameters.

mecho "$*"     # abc,,
mecho $*       # a,b,c

mecho $@       # a,b,c
mecho "$@"     # a,b,c

# The behavior of $* and $@ when $IFS is empty depends
# on whatever Bash or sh version being run.
# It is therefore inadvisable to depend on this "feature" in a script.

# Thanks, S.C.

exit 0

Other Special Parameters


Flags passed to script (using set). See Example 11-14.


This was originally a ksh construct adopted into Bash, and unfortunately it does not seem to work reliably in Bash scripts. One possible use for it is to have a script self-test whether it is interactive.


PID (process ID) of last job run in background


COMMAND1="sleep 100"

echo "Logging PIDs background commands for script: $0" >> "$LOG"
# So they can be monitored, and killed as necessary.
echo >> "$LOG"

# Logging commands.

echo -n "PID of \"$COMMAND1\":  " >> "$LOG"
echo $! >> "$LOG"
# PID of "sleep 100":  1506

# Thank you, Jacques Lederer, for suggesting this.

possibly_hanging_job & { sleep ${TIMEOUT}; eval 'kill -9 $!' &> /dev/null; }
# Forces completion of an ill-behaved program.
# Useful, for example, in init scripts.

# Thank you, Sylvain Fourmanoit, for this creative use of the "!" variable.


Special variable set to last argument of previous command executed.

Example 9-9. Underscore variable


echo $_              # /bin/bash
                     # Just called /bin/bash to run the script.

du >/dev/null        # So no output from command.
echo $_              # du

ls -al >/dev/null    # So no output from command.
echo $_              # -al  (last argument)

echo $_              # :

Exit status of a command, function, or the script itself (see Example 23-6)


Process ID of the script itself. The $$ variable often finds use in scripts to construct "unique" temp file names (see Example A-14, Example 30-6, Example 12-26, and Example 11-24). This is usually simpler than invoking mktemp.

9.2. Manipulating Strings

Bash supports a surprising number of string manipulation operations. Unfortunately, these tools lack a unified focus. Some are a subset of parameter substitution, and others fall under the functionality of the UNIX expr command. This results in inconsistent command syntax and overlap of functionality, not to mention confusion.

String Length


expr length $string

expr "$string" : '.*'


echo ${#stringZ}                 # 15
echo `expr length $stringZ`      # 15
echo `expr "$stringZ" : '.*'`    # 15

Example 9-10. Inserting a blank line between paragraphs in a text file


# Inserts a blank line between paragraphs of a single-spaced text file.
# Usage: $0 <FILENAME

MINLEN=45        # May need to change this value.
#  Assume lines shorter than $MINLEN characters
#+ terminate a paragraph.

while read line  # For as many lines as the input file has...
  echo "$line"   # Output the line itself.

  if [ "$len" -lt "$MINLEN" ]
    then echo    # Add a blank line after short line.

exit 0

Length of Matching Substring at Beginning of String

expr match "$string" '$substring'

$substring is a regular expression.

expr "$string" : '$substring'

$substring is a regular expression.

#       |------|

echo `expr match "$stringZ" 'abc[A-Z]*.2'`   # 8
echo `expr "$stringZ" : 'abc[A-Z]*.2'`       # 8


expr index $string $substring

Numerical position in $string of first character in $substring that matches.

echo `expr index "$stringZ" C12`             # 6
                                             # C position.

echo `expr index "$stringZ" 1c`              # 3
# 'c' (in #3 position) matches before '1'.

This is the near equivalent of strchr() in C.

Substring Extraction


Extracts substring from $string at $position.

If the $string parameter is "*" or "@", then this extracts the positional parameters, [21] starting at $position.


Extracts $length characters of substring from $string at $position.

#       0123456789.....
#       0-based indexing.

echo ${stringZ:0}                            # abcABC123ABCabc
echo ${stringZ:1}                            # bcABC123ABCabc
echo ${stringZ:7}                            # 23ABCabc

echo ${stringZ:7:3}                          # 23A
                                             # Three characters of substring.

# Is it possible to index from the right end of the string?
echo ${stringZ:-4}                           # abcABC123ABCabc
# Defaults to full string, as in ${parameter:-default}.
# However . . .

echo ${stringZ:(-4)}                         # Cabc 
echo ${stringZ: -4}                          # Cabc
# Now, it works.
# Parentheses or added space "escape" the position parameter.

# Thank you, Dan Jacobson, for pointing this out.

If the $string parameter is "*" or "@", then this extracts a maximum of $length positional parameters, starting at $position.

echo ${*:2}          # Echoes second and following positional parameters.
echo ${@:2}          # Same as above.

echo ${*:2:3}        # Echoes three positional parameters, starting at second.

expr substr $string $position $length

Extracts $length characters from $string starting at $position.

#       123456789......
#       1-based indexing.

echo `expr substr $stringZ 1 2`              # ab
echo `expr substr $stringZ 4 3`              # ABC

expr match "$string" '\($substring\)'

Extracts $substring at beginning of $string, where $substring is a regular expression.

expr "$string" : '\($substring\)'

Extracts $substring at beginning of $string, where $substring is a regular expression.

#       =======	    

echo `expr match "$stringZ" '\(.[b-c]*[A-Z]..[0-9]\)'`   # abcABC1
echo `expr "$stringZ" : '\(.[b-c]*[A-Z]..[0-9]\)'`       # abcABC1
echo `expr "$stringZ" : '\(.......\)'`                   # abcABC1
# All of the above forms give an identical result.

expr match "$string" '.*\($substring\)'

Extracts $substring at end of $string, where $substring is a regular expression.

expr "$string" : '.*\($substring\)'

Extracts $substring at end of $string, where $substring is a regular expression.

#                ======

echo `expr match "$stringZ" '.*\([A-C][A-C][A-C][a-c]*\)'`    # ABCabc
echo `expr "$stringZ" : '.*\(......\)'`                       # ABCabc

Substring Removal


Strips shortest match of $substring from front of $string.


Strips longest match of $substring from front of $string.

#       |----|
#       |----------|

echo ${stringZ#a*C}      # 123ABCabc
# Strip out shortest match between 'a' and 'C'.

echo ${stringZ##a*C}     # abc
# Strip out longest match between 'a' and 'C'.


Strips shortest match of $substring from back of $string.


Strips longest match of $substring from back of $string.

#                    ||
#        |------------|

echo ${stringZ%b*c}      # abcABC123ABCa
# Strip out shortest match between 'b' and 'c', from back of $stringZ.

echo ${stringZ%%b*c}     # a
# Strip out longest match between 'b' and 'c', from back of $stringZ.

Example 9-11. Converting graphic file formats, with filename change

#  Converts all the MacPaint image files in a directory to "pbm" format.

#  Uses the "macptopbm" binary from the "netpbm" package,
#+ which is maintained by Brian Henderson (
#  Netpbm is a standard part of most Linux distros.

SUFFIX=pbm          # New filename suffix. 

if [ -n "$1" ]
  directory=$1      # If directory name given as a script argument...
  directory=$PWD    # Otherwise use current working directory.
#  Assumes all files in the target directory are MacPaint image files,
# + with a ".mac" suffix.

for file in $directory/*    # Filename globbing.
  filename=${file%.*c}      #  Strip ".mac" suffix off filename
                            #+ ('.*c' matches everything
			    #+ between '.' and 'c', inclusive).
  $OPERATION $file > "$filename.$SUFFIX"
                            # Redirect conversion to new filename.
  rm -f $file               # Delete original files after converting.   
  echo "$filename.$SUFFIX"  # Log what is happening to stdout.

exit 0

# Exercise:
# --------
#  As it stands, this script converts *all* the files in the current
#+ working directory.
#  Modify it to work *only* on files with a ".mac" suffix.

Substring Replacement


Replace first match of $substring with $replacement.


Replace all matches of $substring with $replacement.


echo ${stringZ/abc/xyz}           # xyzABC123ABCabc
                                  # Replaces first match of 'abc' with 'xyz'.

echo ${stringZ//abc/xyz}          # xyzABC123ABCxyz
                                  # Replaces all matches of 'abc' with # 'xyz'.


If $substring matches front end of $string, substitute $replacement for $substring.


If $substring matches back end of $string, substitute $replacement for $substring.


echo ${stringZ/#abc/XYZ}          # XYZABC123ABCabc
                                  # Replaces front-end match of 'abc' with 'XYZ'.

echo ${stringZ/%abc/XYZ}          # abcABC123ABCXYZ
                                  # Replaces back-end match of 'abc' with 'XYZ'.

9.2.1. Manipulating strings using awk

A Bash script may invoke the string manipulation facilities of awk as an alternative to using its built-in operations.

Example 9-12. Alternate ways of extracting substrings


#      012345678    Bash
#      123456789    awk
# Note different string indexing system:
# Bash numbers first character of string as '0'.
# Awk  numbers first character of string as '1'.

echo ${String:2:4} # position 3 (0-1-2), 4 characters long
                                         # skid

# The awk equivalent of ${string:pos:length} is substr(string,pos,length).
echo | awk '
{ print substr("'"${String}"'",3,4)      # skid
#  Piping an empty "echo" to awk gives it dummy input,
#+ and thus makes it unnecessary to supply a filename.

exit 0

9.2.2. Further Discussion

For more on string manipulation in scripts, refer to Section 9.3 and the relevant section of the expr command listing. For script examples, see:

  1. Example 12-9

  2. Example 9-15

  3. Example 9-16

  4. Example 9-17

  5. Example 9-19

9.3. Parameter Substitution

Manipulating and/or expanding variables


Same as $parameter, i.e., value of the variable parameter. In certain contexts, only the less ambiguous ${parameter} form works.

May be used for concatenating variables with strings.

echo "$your_id"
echo "Old \$PATH = $PATH"
PATH=${PATH}:/opt/bin  #Add /opt/bin to $PATH for duration of script.
echo "New \$PATH = $PATH"

${parameter-default}, ${parameter:-default}

If parameter not set, use default.

echo ${username-`whoami`}
# Echoes the result of `whoami`, if variable $username is still unset.


${parameter-default} and ${parameter:-default} are almost equivalent. The extra : makes a difference only when parameter has been declared, but is null.


#  Whether a variable has been declared
#+ affects triggering of the default option
#+ even if the variable is null.

# username0 has been declared, but is set to null.
echo "username0 = ${username0-`whoami`}"
# Will not echo.

echo "username1 = ${username1-`whoami`}"
# username1 has not been declared.
# Will echo.

# username2 has been declared, but is set to null.
echo "username2 = ${username2:-`whoami`}"
# Will echo because of :- rather than just - in condition test.
# Compare to first instance, above.

exit 0

The default parameter construct finds use in providing "missing" command-line arguments in scripts.
#  If not otherwise specified, the following command block operates
#+ on the file "".
#  Commands follow.

See also Example 3-4, Example 29-2, and Example A-7.

Compare this method with using an and list to supply a default command-line argument.

${parameter=default}, ${parameter:=default}

If parameter not set, set it to default.

Both forms nearly equivalent. The : makes a difference only when $parameter has been declared and is null, [22] as above.

echo ${username=`whoami`}
# Variable "username" is now set to `whoami`.

${parameter+alt_value}, ${parameter:+alt_value}

If parameter set, use alt_value, else use null string.

Both forms nearly equivalent. The : makes a difference only when parameter has been declared and is null, see below.

echo "###### \${parameter+alt_value} ########"

echo "a = $a"      # a =

echo "a = $a"      # a = xyz

echo "a = $a"      # a = xyz

echo "###### \${parameter:+alt_value} ########"

echo "a = $a"      # a =

echo "a = $a"      # a =
# Different result from   a=${param5+xyz}

echo "a = $a"      # a = xyz

${parameter?err_msg}, ${parameter:?err_msg}

If parameter set, use it, else print err_msg.

Both forms nearly equivalent. The : makes a difference only when parameter has been declared and is null, as above.

Example 9-13. Using parameter substitution and error messages


#  Check some of the system's environmental variables.
#  This is good preventative maintenance.
#  If, for example, $USER, the name of the person at the console, is not set,
#+ the machine will not recognize you.

: ${HOSTNAME?} ${USER?} ${HOME?} ${MAIL?}
  echo "Name of the machine is $HOSTNAME."
  echo "You are $USER."
  echo "Your home directory is $HOME."
  echo "Your mail INBOX is located in $MAIL."
  echo "If you are reading this message,"
  echo "critical environmental variables have been set."

# ------------------------------------------------------

#  The ${variablename?} construction can also check
#+ for variables set within the script.

#  Note, by the way, that string variables may be set
#+ to characters disallowed in their names.
: ${ThisVariable?}
echo "Value of ThisVariable is $ThisVariable".

: ${ZZXy23AB?"ZZXy23AB has not been set."}
#  If ZZXy23AB has not been set,
#+ then the script terminates with an error message.

# You can specify the error message.
# : ${ZZXy23AB?"ZZXy23AB has not been set."}

# Same result with:    dummy_variable=${ZZXy23AB?}
#                      dummy_variable=${ZZXy23AB?"ZXy23AB has not been set."}
#                      echo ${ZZXy23AB?} >/dev/null

#  Compare these methods of checking whether a variable has been set
#+ with "set -u" . . .

echo "You will not see this message, because script already terminated."

exit $HERE   # Will NOT exit here.

Example 9-14. Parameter substitution and "usage" messages


: ${1?"Usage: $0 ARGUMENT"}
#  Script exits here if command-line parameter absent,
#+ with following error message.
# 1: Usage: ARGUMENT

echo "These two lines echo only if command-line parameter given."
echo "command line parameter = \"$1\""

exit 0  # Will exit here only if command-line parameter present.

# Check the exit status, both with and without command-line parameter.
# If command-line parameter present, then "$?" is 0.
# If not, then "$?" is 1.

Parameter substitution and/or expansion. The following expressions are the complement to the match in expr string operations (see Example 12-9). These particular ones are used mostly in parsing file path names.

Variable length / Substring removal


String length (number of characters in $var). For an array, ${#array} is the length of the first element in the array.



  • ${#*} and ${#@} give the number of positional parameters.

  • For an array, ${#array[*]} and ${#array[@]} give the number of elements in the array.

Example 9-15. Length of a variable



if [ $# -eq 0 ]  # Must have command-line args to demo script.
  echo "Invoke this script with one or more command-line arguments."
  exit $E_NO_ARGS


echo "var01 = ${var01}"
echo "Length of var01 = ${#var01}"

echo "Number of command-line arguments passed to script = ${#@}"
echo "Number of command-line arguments passed to script = ${#*}"

exit 0
${var#Pattern}, ${var##Pattern}

Remove from $var the shortest/longest part of $Pattern that matches the front end of $var.

A usage illustration from Example A-8:
# Function from "" example.
# Strips leading zero(s) from argument passed.

strip_leading_zero () #  Strip possible leading zero(s)
{                     #+ from argument passed.
  return=${1#0}       #  The "1" refers to "$1" -- passed arg.
}                     #  The "0" is what to remove from "$1" -- strips zeros.

Manfred Schwarb's more elaborate variation of the above:
strip_leading_zero2 () # Strip possible leading zero(s), since otherwise
{                      # Bash will interpret such numbers as octal values.
  shopt -s extglob     # Turn on extended globbing.
  local val=${1##+(0)} # Use local variable, longest matching series of 0's.
  shopt -u extglob     # Turn off extended globbing.
                       # If input was 0, return 0 instead of "".

Another usage illustration:
echo `basename $PWD`        # Basename of current working directory.
echo "${PWD##*/}"           # Basename of current working directory.
echo `basename $0`          # Name of script.
echo $0                     # Name of script.
echo "${0##*/}"             # Name of script.
echo "${filename##*.}"      # data
                            # Extension of filename.

${var%Pattern}, ${var%%Pattern}

Remove from $var the shortest/longest part of $Pattern that matches the back end of $var.

Version 2 of Bash added additional options.

Example 9-16. Pattern matching in parameter substitution


# Pattern matching  using the # ## % %% parameter substitution operators.

pattern1=a*c  # * (wild card) matches everything between a - c.

echo "var1 = $var1"           # abcd12345abc6789
echo "var1 = ${var1}"         # abcd12345abc6789
                              # (alternate form)
echo "Number of characters in ${var1} = ${#var1}"

echo "pattern1 = $pattern1"   # a*c  (everything between 'a' and 'c')
echo "--------------"
echo '${var1#$pattern1}  =' "${var1#$pattern1}"    #         d12345abc6789
# Shortest possible match, strips out first 3 characters  abcd12345abc6789
#                                     ^^^^^               |-|
echo '${var1##$pattern1} =' "${var1##$pattern1}"   #                  6789      
# Longest possible match, strips out first 12 characters  abcd12345abc6789
#                                    ^^^^^                |----------|

echo; echo; echo

pattern2=b*9            # everything between 'b' and '9'
echo "var1 = $var1"     # Still  abcd12345abc6789
echo "pattern2 = $pattern2"
echo "--------------"
echo '${var1%pattern2}  =' "${var1%$pattern2}"     #     abcd12345a
# Shortest possible match, strips out last 6 characters  abcd12345abc6789
#                                     ^^^^                         |----|
echo '${var1%%pattern2} =' "${var1%%$pattern2}"    #     a
# Longest possible match, strips out last 12 characters  abcd12345abc6789
#                                    ^^^^                 |-------------|

# Remember, # and ## work from the left end (beginning) of string,
#           % and %% work from the right end.


exit 0

Example 9-17. Renaming file extensions:


#                 rfe
#                 ---

# Renaming file extensions.
#         rfe old_extension new_extension
# Example:
# To rename all *.gif files in working directory to *.jpg,
#          rfe gif jpg


if [ $# -ne "$ARGS" ]
  echo "Usage: `basename $0` old_file_suffix new_file_suffix"
  exit $E_BADARGS

for filename in *.$1
# Traverse list of files ending with 1st argument.
  mv $filename ${filename%$1}$2
  #  Strip off part of filename matching 1st argument,
  #+ then append 2nd argument.

exit 0

Variable expansion / Substring replacement

These constructs have been adopted from ksh.


Variable var expanded, starting from offset pos.


Expansion to a max of len characters of variable var, from offset pos. See Example A-15 for an example of the creative use of this operator.


First match of Pattern, within var replaced with Replacement.

If Replacement is omitted, then the first match of Pattern is replaced by nothing, that is, deleted.


Global replacement. All matches of Pattern, within var replaced with Replacement.

As above, if Replacement is omitted, then all occurrences of Pattern are replaced by nothing, that is, deleted.

Example 9-18. Using pattern matching to parse arbitrary strings


echo "var1 = $var1"

echo "var1 (with everything, up to and including first - stripped out) = $t"
#  t=${var1#*-}  works just the same,
#+ since # matches the shortest string,
#+ and * matches everything preceding, including an empty string.
# (Thanks, S. C. for pointing this out.)

echo "If var1 contains a \"-\", returns empty string...   var1 = $t"

echo "var1 (with everything from the last - on stripped out) = $t"


# -------------------------------------------
# -------------------------------------------
echo "path_name = $path_name"
echo "path_name, stripped of prefixes = $t"
# Same effect as   t=`basename $path_name` in this particular case.
#  t=${path_name%/}; t=${t##*/}   is a more general solution,
#+ but still fails sometimes.
#  If $path_name ends with a newline, then `basename $path_name` will not work,
#+ but the above expression will.
# (Thanks, S.C.)

# Same effect as   t=`dirname $path_name`
echo "path_name, stripped of suffixes = $t"
# These will fail in some cases, such as "../", "/foo////", # "foo/", "/".
#  Removing suffixes, especially when the basename has no suffix,
#+ but the dirname does, also complicates matters.
# (Thanks, S.C.)


echo "$path_name, with first 11 chars stripped off = $t"
echo "$path_name, with first 11 chars stripped off, length 5 = $t"


echo "$path_name with \"bozo\" replaced  by \"clown\" = $t"
echo "$path_name with \"today\" deleted = $t"
echo "$path_name with all o's capitalized = $t"
echo "$path_name with all o's deleted = $t"

exit 0

If prefix of var matches Pattern, then substitute Replacement for Pattern.


If suffix of var matches Pattern, then substitute Replacement for Pattern.

Example 9-19. Matching patterns at prefix or suffix of string

# Pattern replacement at prefix / suffix of string.

v0=abc1234zip1234abc    # Original variable.
echo "v0 = $v0"         # abc1234zip1234abc

# Match at prefix (beginning) of string.
v1=${v0/#abc/ABCDEF}    # abc1234zip1234abc
                        # |-|
echo "v1 = $v1"         # ABCDEF1234zip1234abc
                        # |----|

# Match at suffix (end) of string.
v2=${v0/%abc/ABCDEF}    # abc1234zip123abc
                        #              |-|
echo "v2 = $v2"         # abc1234zip1234ABCDEF
                        #               |----|


#  ----------------------------------------------------
#  Must match at beginning / end of string,
#+ otherwise no replacement results.
#  ----------------------------------------------------
v3=${v0/#123/000}       # Matches, but not at beginning.
echo "v3 = $v3"         # abc1234zip1234abc
                        # NO REPLACEMENT.
v4=${v0/%123/000}       # Matches, but not at end.
echo "v4 = $v4"         # abc1234zip1234abc
                        # NO REPLACEMENT.

exit 0			
${!varprefix*}, ${!varprefix@}

Matches all previously declared variables beginning with varprefix.

a=${!xyz*}      # Expands to names of declared variables beginning with "xyz".
echo "a = $a"   # a = xyz23 xyz24
a=${!xyz@}      # Same as above.
echo "a = $a"   # a = xyz23 xyz24

# Bash, version 2.04, adds this feature.

9.4. Typing variables: declare or typeset

The declare or typeset builtins (they are exact synonyms) permit restricting the properties of variables. This is a very weak form of the typing available in certain programming languages. The declare command is specific to version 2 or later of Bash. The typeset command also works in ksh scripts.

declare/typeset options

-r readonly

declare -r var1

(declare -r var1 works the same as readonly var1)

This is the rough equivalent of the C const type qualifier. An attempt to change the value of a readonly variable fails with an error message.

-i integer

declare -i number
# The script will treat subsequent occurrences of "number" as an integer.		

echo "Number = $number"     # Number = 3

echo "Number = $number"     # Number = 0
# Tries to evaluate the string "three" as an integer.

Certain arithmetic operations are permitted for declared integer variables without the need for expr or let.

echo "n = $n"       # n = 6/3

declare -i n
echo "n = $n"       # n = 2

-a array

declare -a indices

The variable indices will be treated as an array.

-f functions

declare -f

A declare -f line with no arguments in a script causes a listing of all the functions previously defined in that script.

declare -f function_name

A declare -f function_name in a script lists just the function named.

-x export

declare -x var3

This declares a variable as available for exporting outside the environment of the script itself.

-x var=$value

declare -x var3=373

The declare command permits assigning a value to a variable in the same statement as setting its properties.

Example 9-20. Using declare to type variables


func1 ()
echo This is a function.

declare -f        # Lists the function above.


declare -i var1   # var1 is an integer.
echo "var1 declared as $var1"
var1=var1+1       # Integer declaration eliminates the need for 'let'.
echo "var1 incremented by 1 is $var1."
# Attempt to change variable declared as integer
echo "Attempting to change var1 to floating point value, 2367.1."
var1=2367.1       # Results in error message, with no change to variable.
echo "var1 is still $var1"


declare -r var2=13.36         # 'declare' permits setting a variable property
                              #+ and simultaneously assigning it a value.
echo "var2 declared as $var2" # Attempt to change readonly variable.
var2=13.37                    # Generates error message, and exit from script.

echo "var2 is still $var2"    # This line will not execute.

exit 0                        # Script will not exit here.

9.5. Indirect References to Variables

Assume that the value of a variable is the name of a second variable. Is it somehow possible to retrieve the value of this second variable from the first one? For example, if a=letter_of_alphabet and letter_of_alphabet=z, can a reference to a return z? This can indeed be done, and it is called an indirect reference. It uses the unusual eval var1=\$$var2 notation.

Example 9-21. Indirect References

# Indirect variable referencing.



# Direct reference.
echo "a = $a"

# Indirect reference.
eval a=\$$a
echo "Now a = $a"


# Now, let's try changing the second order reference.

echo "\"table_cell_3\" = $table_cell_3"
echo -n "dereferenced \"t\" = "; eval echo \$$t
# In this simple case,
#   eval t=\$$t; echo "\"t\" = $t"
# also works (why?).


echo "Changing value of \"table_cell_3\" to $NEW_VAL."
echo "\"table_cell_3\" now $table_cell_3"
echo -n "dereferenced \"t\" now "; eval echo \$$t
# "eval" takes the two arguments "echo" and "\$$t" (set equal to $table_cell_3)

# (Thanks, S.C., for clearing up the above behavior.)

# Another method is the ${!t} notation, discussed in "Bash, version 2" section.
# See also example "".

exit 0

Example 9-22. Passing an indirect reference to awk


#  Another version of the "column totaler" script
#+ that adds up a specified column (of numbers) in the target file.
#  This uses indirect references.


if [ $# -ne "$ARGS" ] # Check for proper no. of command line args.
   echo "Usage: `basename $0` filename column-number"
   exit $E_WRONGARGS


#===== Same as original script, up to this point =====#

# A multi-line awk script is invoked by   awk ' ..... '

# Begin awk script.
# ------------------------------------------------
awk "

{ total += \$${column_number} # indirect reference
     print total

     " "$filename"
# ------------------------------------------------
# End awk script.

#  Indirect variable reference avoids the hassles
#+  of referencing a shell variable within the embedded awk script.
#  Thanks, Stephane Chazelas.

exit 0


This method of indirect referencing is a bit tricky. If the second order variable changes its value, then the first order variable must be properly dereferenced (as in the above example). Fortunately, the ${!variable} notation introduced with version 2 of Bash (see Example 35-2) makes indirect referencing more intuitive.

9.6. $RANDOM: generate random integer

$RANDOM is an internal Bash function (not a constant) that returns a pseudorandom integer in the range 0 - 32767. $RANDOM should not be used to generate an encryption key.

Example 9-23. Generating random numbers


# $RANDOM returns a different random integer at each invocation.
# Nominal range: 0 - 32767 (signed 16-bit integer).


echo "$MAXCOUNT random numbers:"
echo "-----------------"
while [ "$count" -le $MAXCOUNT ]      # Generate 10 ($MAXCOUNT) random integers.
  echo $number
  let "count += 1"  # Increment count.
echo "-----------------"

# If you need a random int within a certain range, use the 'modulo' operator.
# This returns the remainder of a division operation.



let "number %= $RANGE"
echo "Random number less than $RANGE  ---  $number"


# If you need a random int greater than a lower bound,
# then set up a test to discard all numbers below that.


number=0   #initialize
while [ "$number" -le $FLOOR ]
echo "Random number greater than $FLOOR ---  $number"

# May combine above two techniques to retrieve random number between two limits.
number=0   #initialize
while [ "$number" -le $FLOOR ]
  let "number %= $RANGE"  # Scales $number down within $RANGE.
echo "Random number between $FLOOR and $RANGE ---  $number"

# Generate binary choice, that is, "true" or "false" value.

let "number %= $BINARY"
#  Note that    let "number >>= 14"    gives a better random distribution
#+ (right shifts out everything except last binary digit).
if [ "$number" -eq $T ]
  echo "TRUE"
  echo "FALSE"


# Generate toss of the dice.
SPOTS=6   # Modulo 6 gives range 0 - 5.
          # Incrementing by 1 gives desired range of 1 - 6.
          # Thanks, Paulo Marcel Coelho Aragao, for the simplification.

# Tosses each die separately, and so gives correct odds.

    let "die1 = $RANDOM % $SPOTS +1" # Roll first one.
    let "die2 = $RANDOM % $SPOTS +1" # Roll second one.

let "throw = $die1 + $die2"
echo "Throw of the dice = $throw"

exit 0

Example 9-24. Picking a random card from a deck


# This is an example of choosing a random element of an array.

# Pick a card, any card.



suite=($Suites)                # Read into array variable.

num_suites=${#suite[*]}        # Count how many elements.

echo -n "${denomination[$((RANDOM%num_denominations))]} of "
echo ${suite[$((RANDOM%num_suites))]}

# $bozo sh
# Jack of Clubs

# Thank you, "jipe," for pointing out this use of $RANDOM.
exit 0

Jipe points out a set of techniques for generating random numbers within a range.
#  Generate random number between 6 and 30.

#  Generate random number in the same 6 - 30 range,
#+ but the number must be evenly divisible by 3.

# Note that this will not work all the time.
# It fails if $RANDOM returns 0.

#  Exercise: Try to figure out the pattern here.

Bill Gradwohl came up with an improved formula that works for positive numbers.

Here Bill presents a versatile function that returns a random number between two specified values.

Example 9-25. Random between values

# Random number between two specified values. 
# Script by Bill Gradwohl, with minor modifications by the document author.
# Used with permission.

randomBetween() {
   #  Generates a positive or negative random number
   #+ between $min and $max
   #+ and divisible by $divisibleBy.
   #  Gives a "reasonably random" distribution of return values.
   #  Bill Gradwohl - Oct 1, 2003

   syntax() {
   # Function embedded within function.
      echo    "Syntax: randomBetween [min] [max] [multiple]"
      echo    "Expects up to 3 passed parameters, but all are completely optional."
      echo    "min is the minimum value"
      echo    "max is the maximum value"
      echo    "multiple specifies that the answer must be a multiple of this value."
      echo    "    i.e. answer must be evenly divisible by this number."
      echo    "If any value is missing, defaults area supplied as: 0 32767 1"
      echo    "Successful completion returns 0, unsuccessful completion returns"
      echo    "function syntax and 1."
      echo    "The answer is returned in the global variable randomBetweenAnswer"
      echo    "Negative values for any passed parameter are handled correctly."

   local min=${1:-0}
   local max=${2:-32767}
   local divisibleBy=${3:-1}
   # Default values assigned, in case parameters not passed to function.

   local x
   local spread

   # Let's make sure the divisibleBy value is positive.
   [ ${divisibleBy} -lt 0 ] && divisibleBy=$((0-divisibleBy))

   # Sanity check.
   if [ $# -gt 3 -o ${divisibleBy} -eq 0 -o  ${min} -eq ${max} ]; then 
      return 1

   # See if the min and max are reversed.
   if [ ${min} -gt ${max} ]; then
      # Swap them.

   #  If min is itself not evenly divisible by $divisibleBy,
   #+ then fix the min to be within range.
   if [ $((min/divisibleBy*divisibleBy)) -ne ${min} ]; then 
      if [ ${min} -lt 0 ]; then

   #  If max is itself not evenly divisible by $divisibleBy,
   #+ then fix the max to be within range.
   if [ $((max/divisibleBy*divisibleBy)) -ne ${max} ]; then 
      if [ ${max} -lt 0 ]; then

   #  ---------------------------------------------------------------------
   #  Now do the real work.

   #  Note that to get a proper distribution for the end points, the
   #+ range of random values has to be allowed to go between 0 and
   #+ abs(max-min)+divisibleBy, not just abs(max-min)+1.

   #  The slight increase will produce the proper distribution for the
   #+ end points.

   #  Changing the formula to use abs(max-min)+1 will still produce
   #+ correct answers, but the randomness of those answers is faulty in
   #+ that the number of times the end points ($min and $max) are returned
   #+ is considerably lower than when the correct formula is used.
   #  ---------------------------------------------------------------------

   [ ${spread} -lt 0 ] && spread=$((0-spread))
   let spread+=divisibleBy

   return 0

   #  However, Paulo Marcel Coelho Aragao points out that
   #+ when $max and $min are not divisible by $divisibleBy,
   #+ the formula fails.
   #  He suggests instead the following formula:
   #    rnumber = $(((RANDOM%(max-min+1)+min)/divisibleBy*divisibleBy))


# Let's test the function.

#  Generate an array of expected answers and check to make sure we get
#+ at least one of each answer if we loop long enough.

declare -a answer
   if [ $((minimum/divisibleBy*divisibleBy)) -ne ${minimum} ]; then 
      if [ ${minimum} -lt 0 ]; then

   #  If max is itself not evenly divisible by $divisibleBy,
   #+ then fix the max to be within range.

   if [ $((maximum/divisibleBy*divisibleBy)) -ne ${maximum} ]; then 
      if [ ${maximum} -lt 0 ]; then

#  We need to generate only positive array subscripts,
#+ so we need a displacement that that will guarantee
#+ positive results.

for ((i=${minimum}; i<=${maximum}; i+=divisibleBy)); do

# Now loop a large number of times to see what we get.
loopIt=1000   #  The script author suggests 100000,
              #+ but that takes a good long while.

for ((i=0; i<${loopIt}; ++i)); do

   #  Note that we are specifying min and max in reversed order here to
   #+ make the function correct for this case.

   randomBetween ${max} ${min} ${divisibleBy}

   # Report an error if an answer is unexpected.
   [ ${randomBetweenAnswer} -lt ${min} -o ${randomBetweenAnswer} -gt ${max} ] && echo MIN or MAX error - ${randomBetweenAnswer}!
   [ $((randomBetweenAnswer%${divisibleBy})) -ne 0 ] && echo DIVISIBLE BY error - ${randomBetweenAnswer}!

   # Store the answer away statistically.

# Let's check the results

for ((i=${minimum}; i<=${maximum}; i+=divisibleBy)); do
   [ ${answer[i+displacement]} -eq 0 ] && echo "We never got an answer of $i." || echo "${i} occurred ${answer[i+displacement]} times."

exit 0

Just how random is $RANDOM? The best way to test this is to write a script that tracks the distribution of "random" numbers generated by $RANDOM. Let's roll a $RANDOM die a few times...

Example 9-26. Rolling a single die with RANDOM

# How random is RANDOM?

RANDOM=$$       # Reseed the random number generator using script process ID.

PIPS=6          # A die has 6 pips.
MAXTHROWS=600   # Increase this, if you have nothing better to do with your time.
throw=0         # Throw count.

ones=0          #  Must initialize counts to zero,
twos=0          #+ since an uninitialized variable is null, not zero.

print_result ()
echo "ones =   $ones"
echo "twos =   $twos"
echo "threes = $threes"
echo "fours =  $fours"
echo "fives =  $fives"
echo "sixes =  $sixes"

case "$1" in
  0) let "ones += 1";;   # Since die has no "zero", this corresponds to 1.
  1) let "twos += 1";;   # And this to 2, etc.
  2) let "threes += 1";;
  3) let "fours += 1";;
  4) let "fives += 1";;
  5) let "sixes += 1";;


while [ "$throw" -lt "$MAXTHROWS" ]
  let "die1 = RANDOM % $PIPS"
  update_count $die1
  let "throw += 1"


#  The scores should distribute fairly evenly, assuming RANDOM is fairly random.
#  With $MAXTHROWS at 600, all should cluster around 100, plus-or-minus 20 or so.
#  Keep in mind that RANDOM is a pseudorandom generator,
#+ and not a spectacularly good one at that.

#  Randomness is a deep and complex subject.
#  Sufficiently long "random" sequences may exhibit
#+ chaotic and other "non-random" behavior.

# Exercise (easy):
# ---------------
# Rewrite this script to flip a coin 1000 times.
# Choices are "HEADS" and "TAILS".

exit 0

As we have seen in the last example, it is best to "reseed" the RANDOM generator each time it is invoked. Using the same seed for RANDOM repeats the same series of numbers. (This mirrors the behavior of the random() function in C.)

Example 9-27. Reseeding RANDOM

# Seeding the RANDOM variable.

MAXCOUNT=25       # How many numbers to generate.

random_numbers ()
while [ "$count" -lt "$MAXCOUNT" ]
  echo -n "$number "
  let "count += 1"

echo; echo

RANDOM=1          # Setting RANDOM seeds the random number generator.

echo; echo

RANDOM=1          # Same seed for RANDOM...
random_numbers    # ...reproduces the exact same number series.
                  # When is it useful to duplicate a "random" number series?

echo; echo

RANDOM=2          # Trying again, but with a different seed...
random_numbers    # gives a different number series.

echo; echo

# RANDOM=$$  seeds RANDOM from process id of script.
# It is also possible to seed RANDOM from 'time' or 'date' commands.

# Getting fancy...
SEED=$(head -1 /dev/urandom | od -N 1 | awk '{ print $2 }')
#  Pseudo-random output fetched
#+ from /dev/urandom (system pseudo-random device-file),
#+ then converted to line of printable (octal) numbers by "od",
#+ finally "awk" retrieves just one number for SEED.

echo; echo

exit 0


The /dev/urandom device-file provides a means of generating much more "random" pseudorandom numbers than the $RANDOM variable. dd if=/dev/urandom of=targetfile bs=1 count=XX creates a file of well-scattered pseudorandom numbers. However, assigning these numbers to a variable in a script requires a workaround, such as filtering through od (as in above example) or using dd (see Example 12-50).

There are also other means of generating pseudorandom numbers in a script. Awk provides a convenient means of doing this.

Example 9-28. Pseudorandom numbers, using awk

# Returns a pseudorandom number in the range 0 - 1.
# Uses the awk rand() function.

AWKSCRIPT=' { srand(); print rand() } '
#            Command(s) / parameters passed to awk
# Note that srand() reseeds awk's random number generator.

echo -n "Random number between 0 and 1 = "

echo | awk "$AWKSCRIPT"
# What happens if you leave out the 'echo'?

exit 0

# Exercises:
# ---------

# 1) Using a loop construct, print out 10 different random numbers.
#      (Hint: you must reseed the "srand()" function with a different seed
#+     in each pass through the loop. What happens if you fail to do this?)

# 2) Using an integer multiplier as a scaling factor, generate random numbers 
#+   in the range between 10 and 100.

# 3) Same as exercise #2, above, but generate random integers this time.

The date command also lends itself to generating pseudorandom integer sequences.

9.7. The Double Parentheses Construct

Similar to the let command, the ((...)) construct permits arithmetic expansion and evaluation. In its simplest form, a=$(( 5 + 3 )) would set "a" to "5 + 3", or 8. However, this double parentheses construct is also a mechanism for allowing C-type manipulation of variables in Bash.

Example 9-29. C-type manipulation of variables

# Manipulating a variable, C-style, using the ((...)) construct.


(( a = 23 ))  # Setting a value, C-style, with spaces on both sides of the "=".
echo "a (initial value) = $a"

(( a++ ))     # Post-increment 'a', C-style.
echo "a (after a++) = $a"

(( a-- ))     # Post-decrement 'a', C-style.
echo "a (after a--) = $a"

(( ++a ))     # Pre-increment 'a', C-style.
echo "a (after ++a) = $a"

(( --a ))     # Pre-decrement 'a', C-style.
echo "a (after --a) = $a"


(( t = a<45?7:11 ))   # C-style trinary operator.
echo "If a < 45, then t = 7, else t = 11."
echo "t = $t "        # Yes!


# -----------------
# Easter Egg alert!
# -----------------
#  Chet Ramey apparently snuck a bunch of undocumented C-style constructs
#+ into Bash (actually adapted from ksh, pretty much).
#  In the Bash docs, Ramey calls ((...)) shell arithmetic,
#+ but it goes far beyond that.
#  Sorry, Chet, the secret is now out.

# See also "for" and "while" loops using the ((...)) construct.

# These work only with Bash, version 2.04 or later.

exit 0

See also Example 10-12.

Chapter 10. Loops and Branches

Operations on code blocks are the key to structured, organized shell scripts. Looping and branching constructs provide the tools for accomplishing this.

10.1. Loops

A loop is a block of code that iterates (repeats) a list of commands as long as the loop control condition is true.

for loops

for (in)

This is the basic looping construct. It differs significantly from its C counterpart.

for arg in [list]


During each pass through the loop, arg takes on the value of each successive variable in the list.

for arg in "$var1" "$var2" "$var3" ... "$varN"  
# In pass 1 of the loop, $arg = $var1	    
# In pass 2 of the loop, $arg = $var2	    
# In pass 3 of the loop, $arg = $var3	    
# ...
# In pass N of the loop, $arg = $varN

# Arguments in [list] quoted to prevent possible word splitting.

The argument list may contain wild cards.

If do is on same line as for, there needs to be a semicolon after list.

for arg in [list] ; do

Example 10-1. Simple for loops

# Listing the planets.

for planet in Mercury Venus Earth Mars Jupiter Saturn Uranus Neptune Pluto
  echo $planet  # Each planet on a separate line.


for planet in "Mercury Venus Earth Mars Jupiter Saturn Uranus Neptune Pluto"
# All planets on same line.
# Entire 'list' enclosed in quotes creates a single variable.
  echo $planet

exit 0


Each [list] element may contain multiple parameters. This is useful when processing parameters in groups. In such cases, use the set command (see Example 11-14) to force parsing of each [list] element and assignment of each component to the positional parameters.

Example 10-2. for loop with two parameters in each [list] element

# Planets revisited.

# Associate the name of each planet with its distance from the sun.

for planet in "Mercury 36" "Venus 67" "Earth 93"  "Mars 142" "Jupiter 483"
  set -- $planet  # Parses variable "planet" and sets positional parameters.
  # the "--" prevents nasty surprises if $planet is null or begins with a dash.

  # May need to save original positional parameters, since they get overwritten.
  # One way of doing this is to use an array,
  #        original_params=("$@")

  echo "$1		$2,000,000 miles from the sun"
  #-------two  tabs---concatenate zeroes onto parameter $2

# (Thanks, S.C., for additional clarification.)

exit 0

A variable may supply the [list] in a for loop.

Example 10-3. Fileinfo: operating on a file list contained in a variable


/sbin/ypbind"     # List of files you are curious about.
                  # Threw in a dummy file, /usr/bin/fakefile.


for file in $FILES

  if [ ! -e "$file" ]       # Check if file exists.
    echo "$file does not exist."; echo
    continue                # On to next.

  ls -l $file | awk '{ print $9 "         file size: " $5 }'  # Print 2 fields.
  whatis `basename $file`   # File info.

exit 0

The [list] in a for loop may contain filename globbing, that is, using wildcards for filename expansion.

Example 10-4. Operating on files with a for loop

# Generating [list] in a for-loop using "globbing".


for file in *
  ls -l "$file"  # Lists all files in $PWD (current directory).
  # Recall that the wild card character "*" matches every filename,
  # however, in "globbing", it doesn't match dot-files.

  # If the pattern matches no file, it is expanded to itself.
  # To prevent this, set the nullglob option
  # (shopt -s nullglob).
  # Thanks, S.C.

echo; echo

for file in [jx]*
  rm -f $file    # Removes only files beginning with "j" or "x" in $PWD.
  echo "Removed file \"$file\"".


exit 0

Omitting the in [list] part of a for loop causes the loop to operate on $@, the list of arguments given on the command line to the script. A particularly clever illustration of this is Example A-17.

Example 10-5. Missing in [list] in a for loop


#  Invoke this script both with and without arguments,
#+ and see what happens.

for a
 echo -n "$a "

#  The 'in list' missing, therefore the loop operates on '$@'
#+ (command-line argument list, including whitespace).


exit 0

It is possible to use command substitution to generate the [list] in a for loop. See also Example 12-45, Example 10-10 and Example 12-39.

Example 10-6. Generating the [list] in a for loop with command substitution

# for-loop with [list]
#+ generated by command substitution.

NUMBERS="9 7 3 8 37.53"

for number in `echo $NUMBERS`  # for number in 9 7 3 8 37.53
  echo -n "$number "

exit 0

This is a somewhat more complex example of using command substitution to create the [list].

Example 10-7. A grep replacement for binary files

# Locates matching strings in a binary file.

# A "grep" replacement for binary files.
# Similar effect to "grep -a"


if [ $# -ne 2 ]
  echo "Usage: `basename $0` search_string filename"
  exit $E_BADARGS

if [ ! -f "$2" ]
  echo "File \"$2\" does not exist."
  exit $E_NOFILE

IFS="\n"         # Per suggestion of Paulo Marcel Coelho Aragao.
for word in $( strings "$2" | grep "$1" )
# The "strings" command lists strings in binary files.
# Output then piped to "grep", which tests for desired string.
  echo $word

# As S.C. points out, lines 23 - 29 could be replaced with the simpler
#    strings "$2" | grep "$1" | tr -s "$IFS" '[\n*]'

# Try something like  "./ mem /bin/ls"  to exercise this script.

exit 0

More of the same.

Example 10-8. Listing all users on the system


n=1           # User number

for name in $(awk 'BEGIN{FS=":"}{print $1}' < "$PASSWORD_FILE" )
# Field separator = :    ^^^^^^
# Print first field              ^^^^^^^^
# Get input from password file               ^^^^^^^^^^^^^^^^^
  echo "USER #$n = $name"
  let "n += 1"

# USER #1 = root
# USER #2 = bin
# USER #3 = daemon
# ...
# USER #30 = bozo

exit 0

A final example of the [list] resulting from command substitution.

Example 10-9. Checking all the binaries in a directory for authorship

# Find a particular string in binaries in a specified directory.

fstring="Free Software Foundation"  # See which files come from the FSF.

for file in $( find $directory -type f -name '*' | sort )
  strings -f $file | grep "$fstring" | sed -e "s%$directory%%"
  #  In the "sed" expression,
  #+ it is necessary to substitute for the normal "/" delimiter
  #+ because "/" happens to be one of the characters filtered out.
  #  Failure to do so gives an error message (try it).

exit 0

#  Exercise (easy):
#  ---------------
#  Convert this script to taking command-line parameters
#+ for $directory and $fstring.

The output of a for loop may be piped to a command or commands.

Example 10-10. Listing the symbolic links in a directory

# Lists symbolic links in a directory.

#  Defaults to current working directory,
#+ if not otherwise specified.
#  Equivalent to code block below.
# ----------------------------------------------------------
# ARGS=1                 # Expect one command-line argument.
# if [ $# -ne "$ARGS" ]  # If not 1 arg...
# then
#   directory=`pwd`      # current working directory
# else
#   directory=$1
# fi
# ----------------------------------------------------------

echo "symbolic links in directory \"$directory\""

for file in "$( find $directory -type l )"   # -type l = symbolic links
  echo "$file"
done | sort                                  # Otherwise file list is unsorted.
#  Strictly speaking, a loop isn't really necessary here,
#+ since the output of the "find" command is expanded into a single word.
#  However, it's easy to understand and illustrative this way.

#  As Dominik 'Aeneas' Schnitzer points out,
#+ failing to quote  $( find $directory -type l )
#+ will choke on filenames with embedded whitespace.
#  Even this will only pick up the first field of each argument.

exit 0

# Jean Helou proposes the following alternative:

echo "symbolic links in directory \"$directory\""
# Backup of the current IFS. One can never be too cautious.

for file in $(find $directory -type l -printf "%p$IFS")
do     #                              ^^^^^^^^^^^^^^^^
       echo "$file"

The stdout of a loop may be redirected to a file, as this slight modification to the previous example shows.

Example 10-11. Symbolic links in a directory, saved to a file

# Lists symbolic links in a directory.

OUTFILE=symlinks.list                         # save file

#  Defaults to current working directory,
#+ if not otherwise specified.

echo "symbolic links in directory \"$directory\"" > "$OUTFILE"
echo "---------------------------" >> "$OUTFILE"

for file in "$( find $directory -type l )"    # -type l = symbolic links
  echo "$file"
done | sort >> "$OUTFILE"                     # stdout of loop
#           ^^^^^^^^^^^^^                       redirected to save file.

exit 0

There is an alternative syntax to a for loop that will look very familiar to C programmers. This requires double parentheses.

Example 10-12. A C-like for loop

# Two ways to count up to 10.


# Standard syntax.
for a in 1 2 3 4 5 6 7 8 9 10
  echo -n "$a "

echo; echo

# +==========================================+

# Now, let's do the same, using C-like syntax.


for ((a=1; a <= LIMIT ; a++))  # Double parentheses, and "LIMIT" with no "$".
  echo -n "$a "
done                           # A construct borrowed from 'ksh93'.

echo; echo

# +=========================================================================+

# Let's use the C "comma operator" to increment two variables simultaneously.

for ((a=1, b=1; a <= LIMIT ; a++, b++))  # The comma chains together operations.
  echo -n "$a-$b "

echo; echo

exit 0

See also Example 26-15, Example 26-16, and Example A-7.


Now, a for-loop used in a "real-life" context.

Example 10-13. Using efax in batch mode



if [ $# -ne $EXPECTED_ARGS ]
# Check for proper no. of command line args.
   echo "Usage: `basename $0` phone# text-file"
   exit $E_BADARGS

if [ ! -f "$2" ]
  echo "File $2 is not a text file"
  exit $E_BADARGS

fax make $2              # Create fax formatted files from text files.

for file in $(ls $2.0*)  # Concatenate the converted files.
                         # Uses wild card in variable list.
  fil="$fil $file"

efax -d /dev/ttyS3 -o1 -t "T$1" $fil   # Do the work.

# As S.C. points out, the for-loop can be eliminated with
#    efax -d /dev/ttyS3 -o1 -t "T$1" $2.0*
# but it's not quite as instructive [grin].

exit 0

This construct tests for a condition at the top of a loop, and keeps looping as long as that condition is true (returns a 0 exit status). In contrast to a for loop, a while loop finds use in situations where the number of loop repetitions is not known beforehand.

while [condition]

As is the case with for/in loops, placing the do on the same line as the condition test requires a semicolon.

while [condition] ; do

Note that certain specialized while loops, as, for example, a getopts construct, deviate somewhat from the standard template given here.

Example 10-14. Simple while loop



while [ "$var0" -lt "$LIMIT" ]
  echo -n "$var0 "        # -n suppresses newline.
  #             ^           Space, to separate printed out numbers.

  var0=`expr $var0 + 1`   # var0=$(($var0+1))  also works.
                          # var0=$((var0 + 1)) also works.
                          # let "var0 += 1"    also works.
done                      # Various other methods also work.


exit 0

Example 10-15. Another while loop



while [ "$var1" != "end" ]     # while test "$var1" != "end"
do                             # also works.
  echo "Input variable #1 (end to exit) "
  read var1                    # Not 'read $var1' (why?).
  echo "variable #1 = $var1"   # Need quotes because of "#".
  # If input is 'end', echoes it here.
  # Does not test for termination condition until top of loop.

exit 0

A while loop may have multiple conditions. Only the final condition determines when the loop terminates. This necessitates a slightly different loop syntax, however.

Example 10-16. while loop with multiple conditions



while echo "previous-variable = $previous"
      [ "$var1" != end ] # Keeps track of what $var1 was previously.
      # Four conditions on "while", but only last one controls loop.
      # The *last* exit status is the one that counts.
echo "Input variable #1 (end to exit) "
  read var1
  echo "variable #1 = $var1"

# Try to figure out how this all works.
# It's a wee bit tricky.

exit 0

As with a for loop, a while loop may employ C-like syntax by using the double parentheses construct (see also Example 9-29).

Example 10-17. C-like syntax in a while loop

# Count to 10 in a "while" loop.


while [ "$a" -le $LIMIT ]
  echo -n "$a "
  let "a+=1"
done           # No surprises, so far.

echo; echo

# +=================================================================+

# Now, repeat with C-like syntax.

((a = 1))      # a=1
# Double parentheses permit space when setting a variable, as in C.

while (( a <= LIMIT ))   # Double parentheses, and no "$" preceding variables.
  echo -n "$a "
  ((a += 1))   # let "a+=1"
  # Yes, indeed.
  # Double parentheses permit incrementing a variable with C-like syntax.


# Now, C programmers can feel right at home in Bash.

exit 0


A while loop may have its stdin redirected to a file by a < at its end.

A while loop may have its stdin supplied by a pipe.


This construct tests for a condition at the top of a loop, and keeps looping as long as that condition is false (opposite of while loop).

until [condition-is-true]

Note that an until loop tests for the terminating condition at the top of the loop, differing from a similar construct in some programming languages.

As is the case with for/in loops, placing the do on the same line as the condition test requires a semicolon.

until [condition-is-true] ; do

Example 10-18. until loop



until [ "$var1" = "$END_CONDITION" ]
# Tests condition here, at top of loop.
  echo "Input variable #1 "
  echo "($END_CONDITION to exit)"
  read var1
  echo "variable #1 = $var1"

exit 0

10.2. Nested Loops

A nested loop is a loop within a loop, an inner loop within the body of an outer one. What happens is that the first pass of the outer loop triggers the inner loop, which executes to completion. Then the second pass of the outer loop triggers the inner loop again. This repeats until the outer loop finishes. Of course, a break within either the inner or outer loop may interrupt this process.

Example 10-19. Nested Loop

# Nested "for" loops.

outer=1             # Set outer loop counter.

# Beginning of outer loop.
for a in 1 2 3 4 5
  echo "Pass $outer in outer loop."
  echo "---------------------"
  inner=1           # Reset inner loop counter.

  # Beginning of inner loop.
  for b in 1 2 3 4 5
    echo "Pass $inner in inner loop."
    let "inner+=1"  # Increment inner loop counter.
  # End of inner loop.

  let "outer+=1"    # Increment outer loop counter. 
  echo              # Space between output in pass of outer loop.
# End of outer loop.

exit 0

See Example 26-11 for an illustration of nested "while" loops, and Example 26-13 to see a "while" loop nested inside an "until" loop.

10.3. Loop Control

Commands Affecting Loop Behavior

break, continue

The break and continue loop control commands [23] correspond exactly to their counterparts in other programming languages. The break command terminates the loop (breaks out of it), while continue causes a jump to the next iteration of the loop, skipping all the remaining commands in that particular loop cycle.

Example 10-20. Effects of break and continue in a loop


LIMIT=19  # Upper limit

echo "Printing Numbers 1 through 20 (but not 3 and 11)."


while [ $a -le "$LIMIT" ]

 if [ "$a" -eq 3 ] || [ "$a" -eq 11 ]  # Excludes 3 and 11
   continue  # Skip rest of this particular loop iteration.

 echo -n "$a "

# Exercise:
# Why does loop print up to 20?

echo; echo

echo Printing Numbers 1 through 20, but something happens after 2.


# Same loop, but substituting 'break' for 'continue'.


while [ "$a" -le "$LIMIT" ]

 if [ "$a" -gt 2 ]
   break  # Skip entire rest of loop.

 echo -n "$a "

echo; echo; echo

exit 0

The break command may optionally take a parameter. A plain break terminates only the innermost loop in which it is embedded, but a break N breaks out of N levels of loop.

Example 10-21. Breaking out of multiple loop levels

# Breaking out of loops.

# "break N" breaks out of N level loops.

for outerloop in 1 2 3 4 5
  echo -n "Group $outerloop:   "

  for innerloop in 1 2 3 4 5
    echo -n "$innerloop "

    if [ "$innerloop" -eq 3 ]
      break  # Try   break 2   to see what happens.
             # ("Breaks" out of both inner and outer loops.)



exit 0

The continue command, similar to break, optionally takes a parameter. A plain continue cuts short the current iteration within its loop and begins the next. A continue N terminates all remaining iterations at its loop level and continues with the next iteration at the loop N levels above.

Example 10-22. Continuing at a higher loop level

# The "continue N" command, continuing at the Nth level loop.

for outer in I II III IV V           # outer loop
  echo; echo -n "Group $outer: "

  for inner in 1 2 3 4 5 6 7 8 9 10  # inner loop

    if [ "$inner" -eq 7 ]
      continue 2  # Continue at loop on 2nd level, that is "outer loop".
                  # Replace above line with a simple "continue"
                  # to see normal loop behavior.

    echo -n "$inner "  # 7 8 9 10 will never echo.


echo; echo

# Exercise:
# Come up with a meaningful use for "continue N" in a script.

exit 0

Example 10-23. Using "continue N" in an actual task

# Albert Reiner gives an example of how to use "continue N":
# ---------------------------------------------------------

#  Suppose I have a large number of jobs that need to be run, with
#+ any data that is to be treated in files of a given name pattern in a
#+ directory. There are several machines that access this directory, and
#+ I want to distribute the work over these different boxen. Then I
#+ usually nohup something like the following on every box:

while true
  for n in .iso.*
    [ "$n" = ".iso.opts" ] && continue
    [ -r .Iso.$beta ] && continue
    [ -r .lock.$beta ] && sleep 10 && continue
    lockfile -r0 .lock.$beta || continue
    echo -n "$beta: " `date`
    run-isotherm $beta
    ls -alF .Iso.$beta
    [ -r .Iso.$beta ] && rm -f .lock.$beta
    continue 2

#  The details, in particular the sleep N, are particular to my
#+ application, but the general pattern is:

while true
  for job in {pattern}
    {job already done or running} && continue
    {mark job as running, do job, mark job as done}
    continue 2
  break        # Or something like `sleep 600' to avoid termination.

#  This way the script will stop only when there are no more jobs to do
#+ (including jobs that were added during runtime). Through the use
#+ of appropriate lockfiles it can be run on several machines
#+ concurrently without duplication of calculations [which run a couple
#+ of hours in my case, so I really want to avoid this]. Also, as search
#+ always starts again from the beginning, one can encode priorities in
#+ the file names. Of course, one could also do this without `continue 2',
#+ but then one would have to actually check whether or not some job
#+ was done (so that we should immediately look for the next job) or not
#+ (in which case we terminate or sleep for a long time before checking
#+ for a new job).


The continue N construct is difficult to understand and tricky to use in any meaningful context. It is probably best avoided.

10.4. Testing and Branching

The case and select constructs are technically not loops, since they do not iterate the execution of a code block. Like loops, however, they direct program flow according to conditions at the top or bottom of the block.

Controlling program flow in a code block

case (in) / esac

The case construct is the shell equivalent of switch in C/C++. It permits branching to one of a number of code blocks, depending on condition tests. It serves as a kind of shorthand for multiple if/then/else statements and is an appropriate tool for creating menus.

case "$variable" in

 "$condition1" )

 "$condition2" )



  • Quoting the variables is not mandatory, since word splitting does not take place.

  • Each test line ends with a right paren ).

  • Each condition block ends with a double semicolon ;;.

  • The entire case block terminates with an esac (case spelled backwards).

Example 10-24. Using case

# Testing ranges of characters.

echo; echo "Hit a key, then hit return."
read Keypress

case "$Keypress" in
  [[:lower:]]   ) echo "Lowercase letter";;
  [[:upper:]]   ) echo "Uppercase letter";;
  [0-9]         ) echo "Digit";;
  *             ) echo "Punctuation, whitespace, or other";;
esac      #  Allows ranges of characters in [square brackets],
          #+ or POSIX ranges in [[double square brackets.

#  In the first version of this example,
#+ the tests for lowercase and uppercase characters were
#+ [a-z] and [A-Z].
#  This no longer works in certain locales and/or Linux distros.
#  Thanks to Frank Wang for pointing this out.

#  Exercise:
#  --------
#  As the script stands, it accepts a single keystroke, then terminates.
#  Change the script so it accepts continuous input,
#+ reports on each keystroke, and terminates only when "X" is hit.
#  Hint: enclose everything in a "while" loop.

exit 0

Example 10-25. Creating menus using case


# Crude address database

clear # Clear the screen.

echo "          Contact List"
echo "          ------- ----"
echo "Choose one of the following persons:" 
echo "[E]vans, Roland"
echo "[J]ones, Mildred"
echo "[S]mith, Julie"
echo "[Z]ane, Morris"

read person

case "$person" in
# Note variable is quoted.

  "E" | "e" )
  # Accept upper or lowercase input.
  echo "Roland Evans"
  echo "4321 Floppy Dr."
  echo "Hardscrabble, CO 80753"
  echo "(303) 734-9874"
  echo "(303) 734-9892 fax"
  echo ""
  echo "Business partner & old friend"
# Note double semicolon to terminate each option.

  "J" | "j" )
  echo "Mildred Jones"
  echo "249 E. 7th St., Apt. 19"
  echo "New York, NY 10009"
  echo "(212) 533-2814"
  echo "(212) 533-9972 fax"
  echo ""
  echo "Girlfriend"
  echo "Birthday: Feb. 11"

# Add info for Smith & Zane later.

          * )
   # Default option.	  
   # Empty input (hitting RETURN) fits here, too.
   echo "Not yet in database."



#  Exercise:
#  --------
#  Change the script so it accepts continuous input,
#+ instead of terminating after displaying just one address.

exit 0

An exceptionally clever use of case involves testing for command-line parameters.
#! /bin/bash

case "$1" in
"") echo "Usage: ${0##*/} <filename>"; exit 65;;  # No command-line parameters,
                                                  # or first parameter empty.
# Note that ${0##*/} is ${var##pattern} param substitution. Net result is $0.

-*) FILENAME=./$1;;   # If filename passed as argument ($1) starts with a dash,
                      # replace it with ./$1
                      # so further commands don't interpret it as an option.

* ) FILENAME=$1;;     # Otherwise, $1.

Example 10-26. Using command substitution to generate the case variable

# Using command substitution to generate a "case" variable.

case $( arch ) in   # "arch" returns machine architecture.
i386 ) echo "80386-based machine";;
i486 ) echo "80486-based machine";;
i586 ) echo "Pentium-based machine";;
i686 ) echo "Pentium2+-based machine";;
*    ) echo "Other type of machine";;

exit 0

A case construct can filter strings for globbing patterns.

Example 10-27. Simple string matching

# simple string matching

match_string ()
  PARAMS=2     # Function requires 2 arguments.

  [ $# -eq $PARAMS ] || return $BAD_PARAMS

  case "$1" in
  "$2") return $MATCH;;
  *   ) return $NOMATCH;;



match_string $a     # wrong number of parameters
echo $?             # 91

match_string $a $b  # no match
echo $?             # 90

match_string $b $d  # match
echo $?             # 0

exit 0		    

Example 10-28. Checking for alphabetic input

# Using a "case" structure to filter a string.


isalpha ()  # Tests whether *first character* of input string is alphabetic.
if [ -z "$1" ]                # No argument passed?
  return $FAILURE

case "$1" in
[a-zA-Z]*) return $SUCCESS;;  # Begins with a letter?
*        ) return $FAILURE;;
}             # Compare this with "isalpha ()" function in C.

isalpha2 ()   # Tests whether *entire string* is alphabetic.
  [ $# -eq 1 ] || return $FAILURE

  case $1 in
  *[!a-zA-Z]*|"") return $FAILURE;;
               *) return $SUCCESS;;

isdigit ()    # Tests whether *entire string* is numerical.
{             # In other words, tests for integer variable.
  [ $# -eq 1 ] || return $FAILURE

  case $1 in
  *[!0-9]*|"") return $FAILURE;;
            *) return $SUCCESS;;

check_var ()  # Front-end to isalpha ().
if isalpha "$@"
  echo "\"$*\" begins with an alpha character."
  if isalpha2 "$@"
  then        # No point in testing if first char is non-alpha.
    echo "\"$*\" contains only alpha characters."
    echo "\"$*\" contains at least one non-alpha character."
  echo "\"$*\" begins with a non-alpha character."
              # Also "non-alpha" if no argument passed.



digit_check ()  # Front-end to isdigit ().
if isdigit "$@"
  echo "\"$*\" contains only digits [0 - 9]."
  echo "\"$*\" has at least one non-digit character."



e=`echo $b`   # Command substitution.

check_var $a
check_var $b
check_var $c
check_var $d
check_var $e
check_var $f
check_var     # No argument passed, so what happens?
digit_check $g
digit_check $h
digit_check $i

exit 0        # Script improved by S.C.

# Exercise:
# --------
#  Write an 'isfloat ()' function that tests for floating point numbers.
#  Hint: The function duplicates 'isdigit ()',
#+ but adds a test for a mandatory decimal point.

The select construct, adopted from the Korn Shell, is yet another tool for building menus.

select variable [in list]

This prompts the user to enter one of the choices presented in the variable list. Note that select uses the PS3 prompt (#? ) by default, but that this may be changed.

Example 10-29. Creating menus using select


PS3='Choose your favorite vegetable: ' # Sets the prompt string.


select vegetable in "beans" "carrots" "potatoes" "onions" "rutabagas"
  echo "Your favorite veggie is $vegetable."
  echo "Yuck!"
  break  # What happens if there is no 'break' here?

exit 0

If in list is omitted, then select uses the list of command line arguments ($@) passed to the script or to the function in which the select construct is embedded.

Compare this to the behavior of a

for variable [in list]

construct with the in list omitted.

Example 10-30. Creating menus using select in a function


PS3='Choose your favorite vegetable: '


select vegetable
# [in list] omitted, so 'select' uses arguments passed to function.
  echo "Your favorite veggie is $vegetable."
  echo "Yuck!"

choice_of beans rice carrots radishes tomatoes spinach
#         $1    $2   $3      $4       $5       $6
#         passed to choice_of() function

exit 0

See also Example 35-3.

Chapter 11. Internal Commands and Builtins

A builtin is a command contained within the Bash tool set, literally built in. This is either for performance reasons -- builtins execute faster than external commands, which usually require forking off a separate process -- or because a particular builtin needs direct access to the shell internals.

A builtin may be a synonym to a system command of the same name, but Bash reimplements it internally. For example, the Bash echo command is not the same as /bin/echo, although their behavior is almost identical.

echo "This line uses the \"echo\" builtin."
/bin/echo "This line uses the /bin/echo system command."

A keyword is a reserved word, token or operator. Keywords have a special meaning to the shell, and indeed are the building blocks of the shell's syntax. As examples, "for", "while", "do", and "!" are keywords. Similar to a builtin, a keyword is hard-coded into Bash, but unlike a builtin, a keyword is not by itself a command, but part of a larger command structure. [24]



prints (to stdout) an expression or variable (see Example 4-1).
echo Hello
echo $a

An echo requires the -e option to print escaped characters. See Example 5-2.

Normally, each echo command prints a terminal newline, but the -n option suppresses this.


An echo can be used to feed a sequence of commands down a pipe.

if echo "$VAR" | grep -q txt   # if [[ $VAR = *txt* ]]
  echo "$VAR contains the substring sequence \"txt\""


An echo, in combination with command substitution can set a variable.

a=`echo "HELLO" | tr A-Z a-z`

See also Example 12-18, Example 12-3, Example 12-38, and Example 12-39.

Be aware that echo `command` deletes any linefeeds that the output of command generates.

The $IFS (internal field separator) variable normally contains \n (linefeed) as one of its set of whitespace characters. Bash therefore splits the output of command at linefeeds into arguments to echo. Then echo outputs these arguments, separated by spaces.

bash$ ls -l /usr/share/apps/kjezz/sounds
-rw-r--r--    1 root     root         1407 Nov  7  2000
 -rw-r--r--    1 root     root          362 Nov  7  2000

bash$ echo `ls -l /usr/share/apps/kjezz/sounds`
total 40 -rw-r--r-- 1 root root 716 Nov 7 2000 -rw-r--r-- 1 root root 362 Nov 7 2000

So, how can we embed a linefeed within an echoed character string?
# Embedding a linefeed?
echo "Why doesn't this string \n split on two lines?"
# Doesn't split.

# Let's try something else.

echo $"A line of text containing
a linefeed."
# Prints as two distinct lines (embedded linefeed).
# But, is the "$" variable prefix really necessary?


echo "This string splits
on two lines."
# No, the "$" is not needed.

echo "---------------"

echo -n $"Another line of text containing
a linefeed."
# Prints as two distinct lines (embedded linefeed).
# Even the -n option fails to suppress the linefeed here.

echo "---------------"

# However, the following doesn't work as expected.
# Why not?
string1=$"Yet another line of text containing
a linefeed (maybe)."

echo $string1

# Thanks, Steve Parker, for pointing this out.


This command is a shell builtin, and not the same as /bin/echo, although its behavior is similar.

bash$ type -a echo
echo is a shell builtin
 echo is /bin/echo


The printf, formatted print, command is an enhanced echo. It is a limited variant of the C language printf() library function, and its syntax is somewhat different.

printf format-string... parameter...

This is the Bash builtin version of the /bin/printf or /usr/bin/printf command. See the printf manpage (of the system command) for in-depth coverage.


Older versions of Bash may not support printf.

Example 11-1. printf in action

# printf demo



printf "Pi to 2 decimal places = %1.2f" $PI
printf "Pi to 9 decimal places = %1.9f" $PI  # It even rounds off correctly.

printf "\n"                                  # Prints a line feed,
                                             # equivalent to 'echo'.

printf "Constant = \t%d\n" $DecimalConstant  # Inserts tab (\t)

printf "%s %s \n" $Message1 $Message2


# ==========================================#
# Simulation of C function, 'sprintf'.
# Loading a variable with a formatted string.


Pi12=$(printf "%1.12f" $PI)
echo "Pi to 12 decimal places = $Pi12"

Msg=`printf "%s %s \n" $Message1 $Message2`
echo $Msg; echo $Msg

# As it happens, the 'sprintf' function can now be accessed
# as a loadable module to Bash, but this is not portable.

exit 0

Formatting error messages is a useful application of printf



  printf "$@" >&2
  # Formats positional params passed, and sents them to stderr.
  exit $E_BADDIR

cd $var || error $"Can't cd to %s." "$var"

# Thanks, S.C.


"Reads" the value of a variable from stdin, that is, interactively fetches input from the keyboard. The -a option lets read get array variables (see Example 26-6).

Example 11-2. Variable assignment, using read


echo -n "Enter the value of variable 'var1': "
# The -n option to echo suppresses newline.

read var1
# Note no '$' in front of var1, since it is being set.

echo "var1 = $var1"


# A single 'read' statement can set multiple variables.
echo -n "Enter the values of variables 'var2' and 'var3' (separated by a space or tab): "
read var2 var3
echo "var2 = $var2      var3 = $var3"
# If you input only one value, the other variable(s) will remain unset (null).

exit 0

A read without an associated variable assigns its input to the dedicated variable $REPLY.

Example 11-3. What happens when read has no variable



# -------------------------- #
echo -n "Enter a value: "
read var
echo "\"var\" = "$var""
# Everything as expected here.
# -------------------------- #


# ------------------------------------------------------------------- #
echo -n "Enter another value: "
read           #  No variable supplied for 'read', therefore...
               #+ Input to 'read' assigned to default variable, $REPLY.
echo "\"var\" = "$var""
# This is equivalent to the first code block.
# ------------------------------------------------------------------- #


exit 0

Normally, inputting a \ suppresses a newline during input to a read. The -r option causes an inputted \ to be interpreted literally.

Example 11-4. Multi-line input to read



echo "Enter a string terminated by a \\, then press <ENTER>."
echo "Then, enter a second string, and again press <ENTER>."
read var1     # The "\" suppresses the newline, when reading "var1".
              #     first line \
              #     second line

echo "var1 = $var1"
#     var1 = first line second line

# For each line terminated by a "\",
# you get a prompt on the next line to continue feeding characters into var1.

echo; echo

echo "Enter another string terminated by a \\ , then press <ENTER>."
read -r var2  # The -r option causes the "\" to be read literally.
              #     first line \

echo "var2 = $var2"
#     var2 = first line \

# Data entry terminates with the first <ENTER>.


exit 0

The read command has some interesting options that permit echoing a prompt and even reading keystrokes without hitting ENTER.

# Read a keypress without hitting ENTER.

read -s -n1 -p "Hit a key " keypress
echo; echo "Keypress was "\"$keypress\""."

# -s option means do not echo input.
# -n N option means accept only N characters of input.
# -p option means echo the following prompt before reading input.

# Using these options is tricky, since they need to be in the correct order.

The -n option to read also allows detection of the arrow keys and certain of the other unusual keys.

Example 11-5. Detecting the arrow keys

# Detects the arrow keys, and a few more.
# Thank you, Sandro Magi, for showing me how.

# --------------------------------------------
# Character codes generated by the keypresses.
# --------------------------------------------


echo -n "Press a key...  "
# May need to also press ENTER if a key not listed above pressed.
read -n3 key                      # Read 3 characters.

echo -n "$key" | grep "$arrowup"  #Check if character code detected.
if [ "$?" -eq $SUCCESS ]
  echo "Up-arrow key pressed."
  exit $SUCCESS

echo -n "$key" | grep "$arrowdown"
if [ "$?" -eq $SUCCESS ]
  echo "Down-arrow key pressed."
  exit $SUCCESS

echo -n "$key" | grep "$arrowrt"
if [ "$?" -eq $SUCCESS ]
  echo "Right-arrow key pressed."
  exit $SUCCESS

echo -n "$key" | grep "$arrowleft"
if [ "$?" -eq $SUCCESS ]
  echo "Left-arrow key pressed."
  exit $SUCCESS

echo -n "$key" | grep "$insert"
if [ "$?" -eq $SUCCESS ]
  echo "\"Insert\" key pressed."
  exit $SUCCESS

echo -n "$key" | grep "$delete"
if [ "$?" -eq $SUCCESS ]
  echo "\"Delete\" key pressed."
  exit $SUCCESS

echo " Some other key pressed."

exit $OTHER

#  Exercises:
#  ---------
#  1) Simplify this script by rewriting the multiple "if" tests
#+    as a 'case' construct.
#  2) Add detection of the "Home," "End," "PgUp," and "PgDn" keys.


The -n option to read will not detect the ENTER (newline) key.

The -t option to read permits timed input (see Example 9-4).

The read command may also "read" its variable value from a file redirected to stdin. If the file contains more than one line, only the first line is assigned to the variable. If read has more than one parameter, then each of these variables gets assigned a successive whitespace-delineated string. Caution!

Example 11-6. Using read with file redirection


read var1 <data-file
echo "var1 = $var1"
# var1 set to the entire first line of the input file "data-file"

read var2 var3 <data-file
echo "var2 = $var2   var3 = $var3"
# Note non-intuitive behavior of "read" here.
# 1) Rewinds back to the beginning of input file.
# 2) Each variable is now set to a corresponding string,
#    separated by whitespace, rather than to an entire line of text.
# 3) The final variable gets the remainder of the line.
# 4) If there are more variables to be set than whitespace-terminated strings
#    on the first line of the file, then the excess variables remain empty.

echo "------------------------------------------------"

# How to resolve the above problem with a loop:
while read line
  echo "$line"
done <data-file
# Thanks, Heiner Steven for pointing this out.

echo "------------------------------------------------"

# Use $IFS (Internal Field Separator variable) to split a line of input to
# "read", if you do not want the default to be whitespace.

echo "List of all users:"
OIFS=$IFS; IFS=:       # /etc/passwd uses ":" for field separator.
while read name passwd uid gid fullname ignore
  echo "$name ($fullname)"
done </etc/passwd   # I/O redirection.
IFS=$OIFS              # Restore originial $IFS.
# This code snippet also by Heiner Steven.

#  Setting the $IFS variable within the loop itself
#+ eliminates the need for storing the original $IFS
#+ in a temporary variable.
#  Thanks, Dim Segebart, for pointing this out.
echo "------------------------------------------------"
echo "List of all users:"

while IFS=: read name passwd uid gid fullname ignore
  echo "$name ($fullname)"
done </etc/passwd   # I/O redirection.

echo "\$IFS still $IFS"

exit 0


Piping output to a read, using echo to set variables will fail.

Yet, piping the output of cat seems to work.

cat file1 file2 |
while read line
echo $line

However, as Bjön Eriksson shows:

Example 11-7. Problems reading from a pipe

# This example contributed by Bjon Eriksson.

cat $0 |
while read line
    echo "{$line}"
printf "\nAll done, last:$last\n"

exit 0  # End of code.
        # (Partial) output of script follows.
        # The 'echo' supplies extra brackets.



{cat $0 |}
{while read line}
{echo "{$line}"}
{printf "nAll done, last:$lastn"}

All done, last:(null)

The variable (last) is set within the subshell but unset outside.

The gendiff script, usually found in /usr/bin on many Linux distros, pipes the output of find to a while read construct.
find $1 \( -name "*$2" -o -name ".*$2" \) -print |
while read f; do
. . .



The familiar cd change directory command finds use in scripts where execution of a command requires being in a specified directory.

(cd /source/directory && tar cf - . ) | (cd /dest/directory && tar xpvf -)
[from the previously cited example by Alan Cox]

The -P (physical) option to cd causes it to ignore symbolic links.

cd - changes to $OLDPWD, the previous working directory.


The cd command does not function as expected when presented with two forward slashes.
bash$ cd //
bash$ pwd
The output should, of course, be /. This is a problem both from the command line and in a script.


Print Working Directory. This gives the user's (or script's) current directory (see Example 11-8). The effect is identical to reading the value of the builtin variable $PWD.

pushd, popd, dirs

This command set is a mechanism for bookmarking working directories, a means of moving back and forth through directories in an orderly manner. A pushdown stack is used to keep track of directory names. Options allow various manipulations of the directory stack.

pushd dir-name pushes the path dir-name onto the directory stack and simultaneously changes the current working directory to dir-name

popd removes (pops) the top directory path name off the directory stack and simultaneously changes the current working directory to that directory popped from the stack.

dirs lists the contents of the directory stack (compare this with the $DIRSTACK variable). A successful pushd or popd will automatically invoke dirs.

Scripts that require various changes to the current working directory without hard-coding the directory name changes can make good use of these commands. Note that the implicit $DIRSTACK array variable, accessible from within a script, holds the contents of the directory stack.

Example 11-8. Changing the current working directory



pushd $dir1
# Will do an automatic 'dirs' (list directory stack to stdout).
echo "Now in directory `pwd`." # Uses back-quoted 'pwd'.

# Now, do some stuff in directory 'dir1'.
pushd $dir2
echo "Now in directory `pwd`."

# Now, do some stuff in directory 'dir2'.
echo "The top entry in the DIRSTACK array is $DIRSTACK."
echo "Now back in directory `pwd`."

# Now, do some more stuff in directory 'dir1'.
echo "Now back in original working directory `pwd`."

exit 0



The let command carries out arithmetic operations on variables. In many cases, it functions as a less complex version of expr.

Example 11-9. Letting "let" do arithmetic.



let a=11            # Same as 'a=11'
let a=a+5           # Equivalent to  let "a = a + 5"
                    # (Double quotes and spaces make it more readable.)
echo "11 + 5 = $a"  # 16

let "a <<= 3"       # Equivalent to  let "a = a << 3"
echo "\"\$a\" (=16) left-shifted 3 places = $a"
                    # 128

let "a /= 4"        # Equivalent to  let "a = a / 4"
echo "128 / 4 = $a" # 32

let "a -= 5"        # Equivalent to  let "a = a - 5"
echo "32 - 5 = $a"  # 27

let "a *=  10"      # Equivalent to  let "a = a * 10"
echo "27 * 10 = $a" # 270

let "a %= 8"        # Equivalent to  let "a = a % 8"
echo "270 modulo 8 = $a  (270 / 8 = 33, remainder $a)"
                    # 6


exit 0

eval arg1 [arg2] ... [argN]

Combines the arguments in an expression or list of expressions and evaluates them. Any variables contained within the expression are expanded. The result translates into a command. This can be useful for code generation from the command line or within a script.

bash$ process=xterm
bash$ show_process="eval ps ax | grep $process"
bash$ $show_process
1867 tty1     S      0:02 xterm
 2779 tty1     S      0:00 xterm
 2886 pts/1    S      0:00 grep xterm

Example 11-10. Showing the effect of eval


y=`eval ls -l`  #  Similar to y=`ls -l`
echo $y         #+ but linefeeds removed because "echoed" variable is unquoted.
echo "$y"       #  Linefeeds preserved when variable is quoted.

echo; echo

y=`eval df`     #  Similar to y=`df`
echo $y         #+ but linefeeds removed.

#  When LF's not preserved, it may make it easier to parse output,
#+ using utilities such as "awk".

echo "==========================================================="

# Now, showing how to "expand" a variable using "eval" . . .

for i in 1 2 3 4 5; do
  eval value=$i
  #  value=$i has same effect. The "eval" is not necessary here.
  #  A variable lacking a meta-meaning evaluates to itself --
  #+ it can't expand to anything other than its literal self.
  echo $value

echo "---"

for i in ls df; do
  value=eval $i
  #  value=$i has an entirely different effect here.
  #  The "eval" evaluates the commands "ls" and "df" . . .
  #  The terms "ls" and "df" have a meta-meaning,
  #+ since they are interpreted as commands,
  #+ rather than just character strings.
  echo $value

exit 0

Example 11-11. Forcing a log-off

# Killing ppp to force a log-off.

# Script should be run as root user.

killppp="eval kill -9 `ps ax | awk '/ppp/ { print $1 }'`"
#                     -------- process ID of ppp -------  

$killppp                  # This variable is now a command.

# The following operations must be done as root user.

chmod 666 /dev/ttyS3      # Must be read+write permissions, or else what?
#  Since doing a SIGKILL on ppp changed the permissions on the serial port,
#+ we restore permissions to previous state.

rm /var/lock/LCK..ttyS3   # Remove the serial port lock file. Why?

exit 0

# Exercises:
# ---------
# 1) Have script check whether root user is invoking it.
# 2) Do a check on whether the process to be killed
#+   is actually running before attempting to kill it.   

Example 11-12. A version of "rot13"

# A version of "rot13" using 'eval'.
# Compare to "" example.

setvar_rot_13()              # "rot13" scrambling
  local varname=$1 varvalue=$2
  eval $varname='$(echo "$varvalue" | tr a-z n-za-m)'

setvar_rot_13 var "foobar"   # Run "foobar" through rot13.
echo $var                    # sbbone

setvar_rot_13 var "$var"     # Run "sbbone" through rot13.
                             # Back to original variable.
echo $var                    # foobar

# This example by Stephane Chazelas.
# Modified by document author.

exit 0

Rory Winston contributed the following instance of how useful eval can be.

Example 11-13. Using eval to force variable substitution in a Perl script

In the Perl script "":
        my $WEBROOT = <WEBROOT_PATH>;

To force variable substitution try:
        $export WEBROOT_PATH=/usr/local/webroot
        $sed 's/<WEBROOT_PATH>/$WEBROOT_PATH/' < > out

But this just gives:
        my $WEBROOT = $WEBROOT_PATH;

        $export WEBROOT_PATH=/usr/local/webroot
        $eval sed 's%\<WEBROOT_PATH\>%$WEBROOT_PATH%' < > out
#        ====

That works fine, and gives the expected substitution:
        my $WEBROOT = /usr/local/webroot;

### Correction applied to original example by Paulo Marcel Coelho Aragao.


The eval command can be risky, and normally should be avoided when there exists a reasonable alternative. An eval $COMMANDS executes the contents of COMMANDS, which may contain such unpleasant surprises as rm -rf *. Running an eval on unfamiliar code written by persons unknown is living dangerously.


The set command changes the value of internal script variables. One use for this is to toggle option flags which help determine the behavior of the script. Another application for it is to reset the positional parameters that a script sees as the result of a command (set `command`). The script can then parse the fields of the command output.

Example 11-14. Using set with positional parameters


# script "set-test"

# Invoke this script with three command line parameters,
# for example, "./set-test one two three".

echo "Positional parameters before  set \`uname -a\` :"
echo "Command-line argument #1 = $1"
echo "Command-line argument #2 = $2"
echo "Command-line argument #3 = $3"

set `uname -a` # Sets the positional parameters to the output
               # of the command `uname -a`

echo $_        # unknown
# Flags set in script.

echo "Positional parameters after  set \`uname -a\` :"
# $1, $2, $3, etc. reinitialized to result of `uname -a`
echo "Field #1 of 'uname -a' = $1"
echo "Field #2 of 'uname -a' = $2"
echo "Field #3 of 'uname -a' = $3"
echo ---
echo $_        # ---

exit 0

Invoking set without any options or arguments simply lists all the environmental and other variables that have been initialized.
bash$ set

Using set with the -- option explicitly assigns the contents of a variable to the positional parameters. When no variable follows the --, it unsets the positional parameters.

Example 11-15. Reassigning the positional parameters


variable="one two three four five"

set -- $variable
# Sets positional parameters to the contents of "$variable".

shift; shift        # Shift past first two positional params.

echo "first parameter = $first_param"             # one
echo "second parameter = $second_param"           # two
echo "remaining parameters = $remaining_params"   # three four five

echo; echo

# Again.
set -- $variable
echo "first parameter = $first_param"             # one
echo "second parameter = $second_param"           # two

# ======================================================

set --
# Unsets positional parameters if no variable specified.

echo "first parameter = $first_param"             # (null value)
echo "second parameter = $second_param"           # (null value)

exit 0

See also Example 10-2 and Example 12-46.


The unset command deletes a shell variable, effectively setting it to null. Note that this command does not affect positional parameters.

bash$ unset PATH

bash$ echo $PATH


Example 11-16. "Unsetting" a variable

# Unsetting a variable.

variable=hello                       # Initialized.
echo "variable = $variable"

unset variable                       # Unset.
                                     # Same effect as   variable=
echo "(unset) variable = $variable"  # $variable is null.

exit 0

The export command makes available variables to all child processes of the running script or shell. Unfortunately, there is no way to export variables back to the parent process, to the process that called or invoked the script or shell. One important use of export command is in startup files, to initialize and make accessible environmental variables to subsequent user processes.

Example 11-17. Using export to pass a variable to an embedded awk script


#  Yet another version of the "column totaler" script (
#+ that adds up a specified column (of numbers) in the target file.
#  This uses the environment to pass a script variable to 'awk'.


if [ $# -ne "$ARGS" ] # Check for proper no. of command line args.
   echo "Usage: `basename $0` filename column-number"
   exit $E_WRONGARGS


#===== Same as original script, up to this point =====#

export column_number
# Export column number to environment, so it's available for retrieval.

# Begin awk script.
# ------------------------------------------------
awk '{ total += $ENVIRON["column_number"]
END { print total }' $filename
# ------------------------------------------------
# End awk script.

# Thanks, Stephane Chazelas.

exit 0


It is possible to initialize and export variables in the same operation, as in export var1=xxx.

However, as Greg Keraunen points out, in certain situations this may have a different effect than setting a variable, then exporting it.

bash$ export var=(a b); echo ${var[0]}
(a b)

bash$ var=(a b); export var; echo ${var[0]}

declare, typeset

The declare and typeset commands specify and/or restrict properties of variables.


Same as declare -r, sets a variable as read-only, or, in effect, as a constant. Attempts to change the variable fail with an error message. This is the shell analog of the C language const type qualifier.


This powerful tool parses command-line arguments passed to the script. This is the Bash analog of the getopt external command and the getopt library function familiar to C programmers. It permits passing and concatenating multiple options [25] and associated arguments to a script (for example scriptname -abc -e /usr/local).

The getopts construct uses two implicit variables. $OPTIND is the argument pointer (OPTion INDex) and $OPTARG (OPTion ARGument) the (optional) argument attached to an option. A colon following the option name in the declaration tags that option as having an associated argument.

A getopts construct usually comes packaged in a while loop, which processes the options and arguments one at a time, then decrements the implicit $OPTIND variable to step to the next.


  1. The arguments passed from the command line to the script must be preceded by a minus (-) or a plus (+). It is the prefixed - or + that lets getopts recognize command-line arguments as options. In fact, getopts will not process arguments without the prefixed - or +, and will terminate option processing at the first argument encountered lacking them.

  2. The getopts template differs slightly from the standard while loop, in that it lacks condition brackets.

  3. The getopts construct replaces the deprecated getopt external command.

while getopts ":abcde:fg" Option
# Initial declaration.
# a, b, c, d, e, f, and g are the options (flags) expected.
# The : after option 'e' shows it will have an argument passed with it.
  case $Option in
    a ) # Do something with variable 'a'.
    b ) # Do something with variable 'b'.
    e)  # Do something with 'e', and also with $OPTARG,
        # which is the associated argument passed with option 'e'.
    g ) # Do something with variable 'g'.
shift $(($OPTIND - 1))
# Move argument pointer to next.

# All this is not nearly as complicated as it looks <grin>.

Example 11-18. Using getopts to read the options/arguments passed to a script

# Exercising getopts and OPTIND
# Script modified 10/09/03 at the suggestion of Bill Gradwohl.

# Here we observe how 'getopts' processes command line arguments to script.
# The arguments are parsed as "options" (flags) and associated arguments.

# Try invoking this script with
# 'scriptname -mn'
# 'scriptname -oq qOption' (qOption can be some arbitrary string.)
# 'scriptname -qXXX -r'
# 'scriptname -qr'    - Unexpected result, takes "r" as the argument to option "q"
# 'scriptname -q -r'  - Unexpected result, same as above
# 'scriptname -mnop -mnop'  - Unexpected result
# (OPTIND is unreliable at stating where an option came from).
#  If an option expects an argument ("flag:"), then it will grab
#+ whatever is next on the command line.


if [ $# -eq "$NO_ARGS" ]  # Script invoked with no command-line args?
  echo "Usage: `basename $0` options (-mnopqrs)"
  exit $E_OPTERROR        # Exit and explain usage, if no argument(s) given.
# Usage: scriptname -options
# Note: dash (-) necessary

while getopts ":mnopq:rs" Option
  case $Option in
    m     ) echo "Scenario #1: option -m-   [OPTIND=${OPTIND}]";;
    n | o ) echo "Scenario #2: option -$Option-   [OPTIND=${OPTIND}]";;
    p     ) echo "Scenario #3: option -p-   [OPTIND=${OPTIND}]";;
    q     ) echo "Scenario #4: option -q-\
 with argument \"$OPTARG\"   [OPTIND=${OPTIND}]";;
    #  Note that option 'q' must have an associated argument,
    #+ otherwise it falls through to the default.
    r | s ) echo "Scenario #5: option -$Option-";;
    *     ) echo "Unimplemented option chosen.";;   # DEFAULT

shift $(($OPTIND - 1))
#  Decrements the argument pointer so it points to next argument.
#  $1 now references the first non option item supplied on the command line
#+ if one exists.

exit 0

#   As Bill Gradwohl states,
#  "The getopts mechanism allows one to specify:  scriptname -mnop -mnop
#+  but there is no reliable way to differentiate what came from where
#+  by using OPTIND."

Script Behavior

source, . (dot command)

This command, when invoked from the command line, executes a script. Within a script, a source file-name loads the file file-name. Sourcing a file (dot-command) imports code into the script, appending to the script (same effect as the #include directive in a C program). The net result is the same as if the "sourced" lines of code were in the body of the script. This is useful in situations when multiple scripts use a common data file or function library.

Example 11-19. "Including" a data file


. data-file    # Load a data file.
# Same effect as "source data-file", but more portable.

#  The file "data-file" must be present in current working directory,
#+ since it is referred to by its 'basename'.

# Now, reference some data from that file.

echo "variable1 (from data-file) = $variable1"
echo "variable3 (from data-file) = $variable3"

let "sum = $variable2 + $variable4"
echo "Sum of variable2 + variable4 (from data-file) = $sum"
echo "message1 (from data-file) is \"$message1\""
# Note:                            escaped quotes

print_message This is the message-print function in the data-file.

exit 0

File data-file for Example 11-19, above. Must be present in same directory.

# This is a data file loaded by a script.
# Files of this type may contain variables, functions, etc.
# It may be loaded with a 'source' or '.' command by a shell script.

# Let's initialize some variables.


message1="Hello, how are you?"
message2="Enough for now. Goodbye."

print_message ()
# Echoes any message passed to it.

  if [ -z "$1" ]
    return 1
    # Error, if argument missing.


  until [ -z "$1" ]
    # Step through arguments passed to function.
    echo -n "$1"
    # Echo args one at a time, suppressing line feeds.
    echo -n " "
    # Insert spaces between words.
    # Next one.


  return 0

It is even possible for a script to source itself, though this does not seem to have any practical applications.

Example 11-20. A (useless) script that sources itself

# a script sourcing itself "recursively."
# From "Stupid Script Tricks," Volume II.

MAXPASSCNT=100    # Maximum number of execution passes.

echo -n  "$pass_count  "
#  At first execution pass, this just echoes two blank spaces,
#+ since $pass_count still uninitialized.

let "pass_count += 1"
#  Assumes the uninitialized variable $pass_count
#+ can be incremented the first time around.
#  This works with Bash and pdksh, but
#+ it relies on non-portable (and possibly dangerous) behavior.
#  Better would be to set $pass_count to 0 if non-initialized.

while [ "$pass_count" -le $MAXPASSCNT ]
  . $0   # Script "sources" itself, rather than calling itself.
         # ./$0 (which would be true recursion) doesn't work here.

#  What occurs here is not actually recursion,
#+ since the script effectively "expands" itself
#+ (generates a new section of code)
#+ with each pass throught the 'while' loop',
#  with each 'source' in line 20.
#  Of course, the script interprets each newly 'sourced' "#!" line
#+ as a comment, and not as the start of a new script.


exit 0   # The net effect is counting from 1 to 100.
         # Very impressive.

# Exercise:
# --------
# Write a script that uses this trick to do something useful.

Unconditionally terminates a script. The exit command may optionally take an integer argument, which is returned to the shell as the exit status of the script. It is good practice to end all but the simplest scripts with an exit 0, indicating a successful run.


If a script terminates with an exit lacking an argument, the exit status of the script is the exit status of the last command executed in the script, not counting the exit.


This shell builtin replaces the current process with a specified command. Normally, when the shell encounters a command, it forks off a child process to actually execute the command. Using the exec builtin, the shell does not fork, and the command exec'ed replaces the shell. When used in a script, therefore, it forces an exit from the script when the exec'ed command terminates. For this reason, if an exec appears in a script, it would probably be the final command.

Example 11-21. Effects of exec


exec echo "Exiting \"$0\"."   # Exit from script here.

# ----------------------------------
# The following lines never execute.

echo "This echo will never echo."

exit 99                       #  This script will not exit here.
                              #  Check exit value after script terminates
                              #+ with an 'echo $?'.
                              #  It will *not* be 99.

Example 11-22. A script that exec's itself



echo "This line appears ONCE in the script, yet it keeps echoing."
echo "The PID of this instance of the script is still $$."
#     Demonstrates that a subshell is not forked off.

echo "==================== Hit Ctl-C to exit ===================="

sleep 1

exec $0   #  Spawns another instance of this same script
          #+ that replaces the previous one.

echo "This line will never echo!"  # Why not?

exit 0

An exec also serves to reassign file descriptors. exec <zzz-file replaces stdin with the file zzz-file (see Example 16-1).


The -exec option to find is not the same as the exec shell builtin.


This command permits changing shell options on the fly (see Example 24-1 and Example 24-2). It often appears in the Bash startup files, but also has its uses in scripts. Needs version 2 or later of Bash.
shopt -s cdspell
# Allows minor misspelling of directory names with 'cd'

cd /hpme  # Oops! Mistyped '/home'.
pwd       # /home
          # The shell corrected the misspelling.



A command that returns a successful (zero) exit status, but does nothing else.

# Endless loop
while true   # alias for ":"
   # Need a way to break out of loop or script will hang.


A command that returns an unsuccessful exit status, but does nothing else.

# Null loop
while false
   # The following code will not execute.
   # Nothing happens!

type [cmd]

Similar to the which external command, type cmd gives the full pathname to "cmd". Unlike which, type is a Bash builtin. The useful -a option to type identifies keywords and builtins, and also locates system commands with identical names.

bash$ type '['
[ is a shell builtin
bash$ type -a '['
[ is a shell builtin
 [ is /usr/bin/[

hash [cmds]

Record the path name of specified commands (in the shell hash table), so the shell or script will not need to search the $PATH on subsequent calls to those commands. When hash is called with no arguments, it simply lists the commands that have been hashed. The -r option resets the hash table.


The bind builtin displays or modifies readline [26] key bindings.


Gets a short usage summary of a shell builtin. This is the counterpart to whatis, but for builtins.

bash$ help exit
exit: exit [n]
    Exit the shell with a status of N.  If N is omitted, the exit status
    is that of the last command executed.

11.1. Job Control Commands

Certain of the following job control commands take a "job identifier" as an argument. See the table at end of the chapter.


Lists the jobs running in the background, giving the job number. Not as useful as ps.


It is all too easy to confuse jobs and processes. Certain builtins, such as kill, disown, and wait accept either a job number or a process number as an argument. The fg, bg and jobs commands accept only a job number.

bash$ sleep 100 &
[1] 1384

bash $ jobs
[1]+  Running                 sleep 100 &

"1" is the job number (jobs are maintained by the current shell), and "1384" is the process number (processes are maintained by the system). To kill this job/process, either a kill %1 or a kill 1384 works.

Thanks, S.C.


Remove job(s) from the shell's table of active jobs.

fg, bg

The fg command switches a job running in the background into the foreground. The bg command restarts a suspended job, and runs it in the background. If no job number is specified, then the fg or bg command acts upon the currently running job.


Stop script execution until all jobs running in background have terminated, or until the job number or process ID specified as an option terminates. Returns the exit status of waited-for command.

You may use the wait command to prevent a script from exiting before a background job finishes executing (this would create a dreaded orphan process).

Example 11-23. Waiting for a process to finish before proceeding


ROOT_UID=0   # Only users with $UID 0 have root privileges.

if [ "$UID" -ne "$ROOT_UID" ]
  echo "Must be root to run this script."
  # "Run along kid, it's past your bedtime."
  exit $E_NOTROOT

if [ -z "$1" ]
  echo "Usage: `basename $0` find-string"
  exit $E_NOPARAMS

echo "Updating 'locate' database..."
echo "This may take a while."
updatedb /usr &     # Must be run as root.

# Don't run the rest of the script until 'updatedb' finished.
# You want the the database updated before looking up the file name.

locate $1

# Without the wait command, in the worse case scenario,
# the script would exit while 'updatedb' was still running,
# leaving it as an orphan process.

exit 0

Optionally, wait can take a job identifier as an argument, for example, wait%1 or wait $PPID. See the job id table.


Within a script, running a command in the background with an ampersand (&) may cause the script to hang until ENTER is hit. This seems to occur with commands that write to stdout. It can be a major annoyance.

ls -l &
echo "Done."
bash$ ./
 [bozo@localhost test-scripts]$ total 1
 -rwxr-xr-x    1 bozo     bozo           34 Oct 11 15:09

Placing a wait after the background command seems to remedy this.

ls -l &
echo "Done."
bash$ ./
 [bozo@localhost test-scripts]$ total 1
 -rwxr-xr-x    1 bozo     bozo           34 Oct 11 15:09
Redirecting the output of the command to a file or even to /dev/null also takes care of this problem.


This has a similar effect to Control-Z, but it suspends the shell (the shell's parent process should resume it at an appropriate time).


Exit a login shell, optionally specifying an exit status.


Gives statistics on the system time used in executing commands, in the following form:
0m0.020s 0m0.020s
This capability is of very limited value, since it is uncommon to profile and benchmark shell scripts.


Forcibly terminate a process by sending it an appropriate terminate signal (see Example 13-5).

Example 11-24. A script that kills itself


kill $$  # Script kills its own process here.
         # Recall that "$$" is the script's PID.

echo "This line will not echo."
# Instead, the shell sends a "Terminated" message to stdout.

exit 0

#  After this script terminates prematurely,
#+ what exit status does it return?
# sh
# echo $?
# 143
# 143 = 128 + 15
#             TERM signal


kill -l lists all the signals. A kill -9 is a "sure kill", which will usually terminate a process that stubbornly refuses to die with a plain kill. Sometimes, a kill -15 works. A "zombie process", that is, a process whose parent has terminated, cannot be killed (you can't kill something that is already dead), but init will generally clean it up sooner or later.


The command COMMAND directive disables aliases and functions for the command "COMMAND".


This is one of three shell directives that effect script command processing. The others are builtin and enable.


Invoking builtin BUILTIN_COMMAND runs the command "BUILTIN_COMMAND" as a shell builtin, temporarily disabling both functions and external system commands with the same name.


This either enables or disables a shell builtin command. As an example, enable -n kill disables the shell builtin kill, so that when Bash subsequently encounters kill, it invokes /bin/kill.

The -a option to enable lists all the shell builtins, indicating whether or not they are enabled. The -f filename option lets enable load a builtin as a shared library (DLL) module from a properly compiled object file. [27].


This is a port to Bash of the ksh autoloader. With autoload in place, a function with an "autoload" declaration will load from an external file at its first invocation. [28] This saves system resources.

Note that autoload is not a part of the core Bash installation. It needs to be loaded in with enable -f (see above).

Table 11-1. Job identifiers

%NJob number [N]
%SInvocation (command line) of job begins with string S
%?SInvocation (command line) of job contains within it string S
%%"current" job (last job stopped in foreground or started in background)
%+"current" job (last job stopped in foreground or started in background)
%-Last job
$!Last background process

Chapter 12. External Filters, Programs and Commands

Standard UNIX commands make shell scripts more versatile. The power of scripts comes from coupling system commands and shell directives with simple programming constructs.

12.1. Basic Commands

The first commands a novice learns


The basic file "list" command. It is all too easy to underestimate the power of this humble command. For example, using the -R, recursive option, ls provides a tree-like listing of a directory structure. Other useful options are -S, sort listing by file size, -t, sort by file modification time, and -i, show file inodes (see Example 12-4).

Example 12-1. Using ls to create a table of contents for burning a CDR disk

# Script to automate burning a CDR.

SPEED=2          # May use higher speed if your hardware supports it.
# DEVICE="0,0"     for older versions of cdrecord
DEFAULTDIR=/opt  # This is the directory containing the data to be burned.
                 # Make sure it exists.
                 # Exercise: Add a test for this.

# Uses Joerg Schilling's "cdrecord" package:

#  If this script invoked as an ordinary user, need to suid cdrecord
#+ chmod u+s /usr/bin/cdrecord, as root.
#  Of course, this creates a security hole, though a relatively minor one.

if [ -z "$1" ]
  # Default directory, if not specified on command line.

# Create a "table of contents" file.
# The "l" option gives a "long" file listing.
# The "R" option makes the listing recursive.
# The "F" option marks the file types (directories get a trailing /).
echo "Creating table of contents."

# Create an image file preparatory to burning it onto the CDR.
echo "Creating ISO9660 file system image ($IMAGEFILE)."

# Burn the CDR.
cdrecord -v -isosize speed=$SPEED dev=$DEVICE $IMAGEFILE
echo "Burning the disk."
echo "Please be patient, this will take a while."

exit 0
cat, tac

cat, an acronym for concatenate, lists a file to stdout. When combined with redirection (> or >>), it is commonly used to concatenate files.
# Uses of 'cat'
cat filename                          # Lists the file.

cat file.1 file.2 file.3 > file.123   # Combines three files into one.
The -n option to cat inserts consecutive numbers before all lines of the target file(s). The -b option numbers only the non-blank lines. The -v option echoes nonprintable characters, using ^ notation. The -s option squeezes multiple consecutive blank lines into a single blank line.

See also Example 12-24 and Example 12-20.


In a pipe, it may be more efficient to redirect the stdin to a file, rather than to cat the file.

cat filename | tr a-z A-Z

tr a-z A-Z < filename   #  Same effect, but starts one less process,
                        #+ and also dispenses with the pipe.

tac, is the inverse of cat, listing a file backwards from its end.


reverses each line of a file, and outputs to stdout. This is not the same effect as tac, as it preserves the order of the lines, but flips each one around.

bash$ cat file1.txt
This is line 1.
 This is line 2.

bash$ tac file1.txt
This is line 2.
 This is line 1.

bash$ rev file1.txt
.1 enil si sihT
 .2 enil si sihT


This is the file copy command. cp file1 file2 copies file1 to file2, overwriting file2 if it already exists (see Example 12-6).


Particularly useful are the -a archive flag (for copying an entire directory tree) and the -r and -R recursive flags.


This is the file move command. It is equivalent to a combination of cp and rm. It may be used to move multiple files to a directory, or even to rename a directory. For some examples of using mv in a script, see Example 9-17 and Example A-3.


When used in a non-interactive script, mv takes the -f (force) option to bypass user input.

When a directory is moved to a preexisting directory, it becomes a subdirectory of the destination directory.

bash$ mv source_directory target_directory

bash$ ls -lF target_directory
total 1
 drwxrwxr-x    2 bozo  bozo      1024 May 28 19:20 source_directory/


Delete (remove) a file or files. The -f option forces removal of even readonly files, and is useful for bypassing user input in a script.


The rm command will, by itself, fail to remove filenames beginning with a dash.

bash$ rm -badname
rm: invalid option -- b
 Try `rm --help' for more information.

The way to accomplish this is to preface the filename to be removed with a dot-slash .

bash$ rm ./-badname


When used with the recursive flag -r, this command removes files all the way down the directory tree from the current directory.


Remove directory. The directory must be empty of all files, including invisible "dotfiles", [29] for this command to succeed.


Make directory, creates a new directory. For example, mkdir -p project/programs/December creates the named directory. The -p option automatically creates any necessary parent directories.


Changes the attributes of an existing file (see Example 11-11).

chmod +x filename
# Makes "filename" executable for all users.

chmod u+s filename
# Sets "suid" bit on "filename" permissions.
# An ordinary user may execute "filename" with same privileges as the file's owner.
# (This does not apply to shell scripts.)

chmod 644 filename
# Makes "filename" readable/writable to owner, readable to
# others
# (octal mode).

chmod 1777 directory-name
# Gives everyone read, write, and execute permission in directory,
# however also sets the "sticky bit".
# This means that only the owner of the directory,
# owner of the file, and, of course, root
# can delete any particular file in that directory.


Change file attributes. This has the same effect as chmod above, but with a different invocation syntax, and it works only on an ext2 filesystem.


Creates links to pre-existings files. A "link" is a reference to a file, an alternate name for it. The ln command permits referencing the linked file by more than one name and is a superior alternative to aliasing (see Example 4-6).

The ln creates only a reference, a pointer to the file only a few bytes in size.

The ln command is most often used with the -s, symbolic or "soft" link flag. An advantage of using the -s flag is that it permits linking across file systems.

The syntax of the command is a bit tricky. For example: ln -s oldfile newfile links the previously existing oldfile to the newly created link, newfile.


If a file named newfile has previously existed, it will be deleted when the filename newfile is preempted as the name for a link.

Links give the ability to invoke a script (or any other type of executable) with multiple names, and having that script behave according to how it was invoked.

Example 12-2. Hello or Good-bye

# Saying "hello" or "goodbye"
#+          depending on how script is invoked.

# Make a link in current working directory ($PWD) to this script:
#    ln -s goodbye
# Now, try invoking this script both ways:
# ./
# ./goodbye


if [ $0 = "./goodbye" ]
  echo "Good-bye!"
  # Some other goodbye-type commands, as appropriate.

echo "Hello!"
# Some other hello-type commands, as appropriate.
man, info

These commands access the manual and information pages on system commands and installed utilities. When available, the info pages usually contain a more detailed description than do the man pages.

12.2. Complex Commands

Commands for more advanced users


-exec COMMAND \;

Carries out COMMAND on each file that find matches. The command sequence terminates with \; (the ";" is escaped to make certain the shell passes it to find literally).

bash$ find ~/ -name '*.txt'

If COMMAND contains {}, then find substitutes the full path name of the selected file for "{}".

find ~/ -name 'core*' -exec rm {} \;
# Removes all core dump files from user's home directory.

find /home/bozo/projects -mtime 1
#  Lists all files in /home/bozo/projects directory tree
#+ that were modified within the last day.
#  mtime = last modification time of the target file
#  ctime = last status change time (via 'chmod' or otherwise)
#  atime = last access time

find "$DIR" -type f -atime +5 -exec rm {} \;
#                                      ^^
#  Curly brackets are placeholder for the pathname output by "find."
#  Deletes all files in "/home/bozo/junk_files"
#+ that have not been accessed in at least 5 days.
#  "-type filetype", where
#  f = regular file
#  d = directory, etc.
#  (The 'find' manpage has a complete listing.)

find /etc -exec grep '[0-9][0-9]*[.][0-9][0-9]*[.][0-9][0-9]*[.][0-9][0-9]*' {} \;

# Finds all IP addresses ( in /etc directory files.
# There a few extraneous hits - how can they be filtered out?

# Perhaps by:

find /etc -type f -exec cat '{}' \; | tr -c '.[:digit:]' '\n' \
 | grep '^[^.][^.]*\.[^.][^.]*\.[^.][^.]*\.[^.][^.]*$'
# [:digit:] is one of the character classes
# introduced with the POSIX 1003.2 standard. 

# Thanks, S.C. 


The -exec option to find should not be confused with the exec shell builtin.

Example 12-3. Badname, eliminate file names in current directory containing bad characters and whitespace.


# Delete filenames in current directory containing bad characters.

for filename in *
  badname=`echo "$filename" | sed -n /[\+\{\;\"\\\=\?~\(\)\<\>\&\*\|\$]/p`
# badname=`echo "$filename" | sed -n '/[+{;"\=?~()<>&*|$]/p'`  also works.
# Deletes files containing these nasties:     + { ; " \ = ? ~ ( ) < > & * | $
  rm $badname 2>/dev/null
#             ^^^^^^^^^^^ Error messages deep-sixed.

# Now, take care of files containing all manner of whitespace.
find . -name "* *" -exec rm -f {} \;
# The path name of the file that "find" finds replaces the "{}".
# The '\' ensures that the ';' is interpreted literally, as end of command.

exit 0

# Commands below this line will not execute because of "exit" command.

# An alternative to the above script:
find . -name '*[+{;"\\=?~()<>&*|$ ]*' -exec rm -f '{}' \;
# (Thanks, S.C.)

Example 12-4. Deleting a file by its inode number

# Deleting a file by its inode number.

#  This is useful when a filename starts with an illegal character,
#+ such as ? or -.

ARGCOUNT=1                      # Filename arg must be passed to script.

if [ $# -ne "$ARGCOUNT" ]
  echo "Usage: `basename $0` filename"

if [ ! -e "$1" ]
  echo "File \""$1"\" does not exist."

inum=`ls -i | grep "$1" | awk '{print $1}'`
# inum = inode (index node) number of file
# Every file has an inode, a record that hold its physical address info.

echo; echo -n "Are you absolutely sure you want to delete \"$1\" (y/n)? "
# The '-v' option to 'rm' also asks this.
read answer
case "$answer" in
[nN]) echo "Changed your mind, huh?"
      exit $E_CHANGED_MIND
*)    echo "Deleting file \"$1\".";;

find . -inum $inum -exec rm {} \;
#                           ^^
#        Curly brackets are placeholder
#+       for text output by "find."
echo "File "\"$1"\" deleted!"

exit 0

See Example 12-25, Example 3-4, and Example 10-9 for scripts using find. Its manpage provides more detail on this complex and powerful command.


A filter for feeding arguments to a command, and also a tool for assembling the commands themselves. It breaks a data stream into small enough chunks for filters and commands to process. Consider it as a powerful replacement for backquotes. In situations where backquotes fail with a too many arguments error, substituting xargs often works. Normally, xargs reads from stdin or from a pipe, but it can also be given the output of a file.

The default command for xargs is echo. This means that input piped to xargs may have linefeeds and other whitespace characters stripped out.
bash$ ls -l
total 0
 -rw-rw-r--    1 bozo  bozo         0 Jan 29 23:58 file1
 -rw-rw-r--    1 bozo  bozo         0 Jan 29 23:58 file2

bash$ ls -l | xargs
total 0 -rw-rw-r-- 1 bozo bozo 0 Jan 29 23:58 file1 -rw-rw-r-- 1 bozo bozo 0 Jan 29 23:58 file2

ls | xargs -p -l gzip gzips every file in current directory, one at a time, prompting before each operation.


An interesting xargs option is -n NN, which limits to NN the number of arguments passed.

ls | xargs -n 8 echo lists the files in the current directory in 8 columns.


Another useful option is -0, in combination with find -print0 or grep -lZ. This allows handling arguments containing whitespace or quotes.

find / -type f -print0 | xargs -0 grep -liwZ GUI | xargs -0 rm -f

grep -rliwZ GUI / | xargs -0 rm -f

Either of the above will remove any file containing "GUI". (Thanks, S.C.)

Example 12-5. Logfile: Using xargs to monitor system log


# Generates a log file in current directory
# from the tail end of /var/log/messages.

# Note: /var/log/messages must be world readable
# if this script invoked by an ordinary user.
#         #root chmod 644 /var/log/messages


( date; uname -a ) >>logfile
# Time and machine name
echo --------------------------------------------------------------------- >>logfile
tail -$LINES /var/log/messages | xargs |  fmt -s >>logfile
echo >>logfile
echo >>logfile

exit 0

#  Note:
#  ----
#  As Frank Wang points out,
#+ unmatched quotes (either single or double quotes) in the source file
#+ may give xargs indigestion.
#  He suggests the following substitution for line 15:
#     tail -$LINES /var/log/messages | tr -d "\"'" | xargs | fmt -s >>logfile

#  Exercise:
#  --------
#  Modify this script to track changes in /var/log/messages at intervals
#+ of 20 minutes.
#  Hint: Use the "watch" command. 

As in find, a curly bracket pair serves as a placeholder for replacement text.

Example 12-6. Copying files in current directory to another


#  Copy (verbose) all files in current directory ($PWD)
#+ to directory specified on command line.


if [ -z "$1" ]   # Exit if no argument given.
  echo "Usage: `basename $0` directory-to-copy-to"
  exit $E_NOARGS

ls . | xargs -i -t cp ./{} $1
#            ^^ ^^      ^^
#  -t is "verbose" (output command line to stderr) option.
#  -i is "replace strings" option.
#  {} is a placeholder for output text.
#  This is similar to the use of a curly bracket pair in "find."
#  List the files in current directory (ls .),
#+ pass the output of "ls" as arguments to "xargs" (-i -t options),
#+ then copy (cp) these arguments ({}) to new directory ($1).  
#  The net result is the exact equivalent of
#+   cp * $1
#+ unless any of the filenames has embedded "whitespace" characters.

exit 0

Example 12-7. Killing processes by name

# Killing processes by name.
# Compare this script with

#  For instance,
#+ try "./ xterm" --
#+ and watch all the xterms on your desktop disappear.

#  Warning:
#  -------
#  This is a fairly dangerous script.
#  Running it carelessly (especially as root)
#+ can cause data loss and other undesirable effects.


if test -z "$1"  # No command line arg supplied?
  echo "Usage: `basename $0` Process(es)_to_kill"
  exit $E_BADARGS

ps ax | grep "$PROCESS_NAME" | awk '{print $1}' | xargs -i kill {} 2&>/dev/null
#                                                       ^^      ^^

# -----------------------------------------------------------
# Notes:
# -i is the "replace strings" option to xargs.
# The curly brackets are the placeholder for the replacement.
# 2&>/dev/null suppresses unwanted error messages.
# -----------------------------------------------------------

exit $?

Example 12-8. Word frequency analysis using xargs

# Crude word frequency analysis on a text file.

# Uses 'xargs' to decompose lines of text into single words.
# Compare this example to the "" script later on.

# Check for input file on command line.

if [ $# -ne "$ARGS" ]
# Correct number of arguments passed to script?
  echo "Usage: `basename $0` filename"
  exit $E_BADARGS

if [ ! -f "$1" ]       # Check if file exists.
  echo "File \"$1\" does not exist."
  exit $E_NOFILE

cat "$1" | xargs -n1 | \
#  List the file, one word per line. 
tr A-Z a-z | \
#  Shift characters to lowercase.
sed -e 's/\.//g'  -e 's/\,//g' -e 's/ /\
/g' | \
#  Filter out periods and commas, and
#+ change space between words to linefeed,
sort | uniq -c | sort -nr
#  Finally prefix occurrence count and sort numerically.

#  This does the same job as the "" example,
#+ but a bit more ponderously, and it runs more slowly.

exit 0

All-purpose expression evaluator: Concatenates and evaluates the arguments according to the operation given (arguments must be separated by spaces). Operations may be arithmetic, comparison, string, or logical.

expr 3 + 5

returns 8

expr 5 % 3

returns 2

expr 5 \* 3

returns 15

The multiplication operator must be escaped when used in an arithmetic expression with expr.

y=`expr $y + 1`

Increment a variable, with the same effect as let y=y+1 and y=$(($y+1)). This is an example of arithmetic expansion.

z=`expr substr $string $position $length`

Extract substring of $length characters, starting at $position.

Example 12-9. Using expr


# Demonstrating some of the uses of 'expr'
# =======================================


# Arithmetic Operators
# ---------- ---------

echo "Arithmetic Operators"
a=`expr 5 + 3`
echo "5 + 3 = $a"

a=`expr $a + 1`
echo "a + 1 = $a"
echo "(incrementing a variable)"

a=`expr 5 % 3`
# modulo
echo "5 mod 3 = $a"


# Logical Operators
# ------- ---------

#  Returns 1 if true, 0 if false,
#+ opposite of normal Bash convention.

echo "Logical Operators"

b=`expr $x = $y`         # Test equality.
echo "b = $b"            # 0  ( $x -ne $y )

b=`expr $a \> 10`
echo 'b=`expr $a \> 10`, therefore...'
echo "If a > 10, b = 0 (false)"
echo "b = $b"            # 0  ( 3 ! -gt 10 )

b=`expr $a \< 10`
echo "If a < 10, b = 1 (true)"
echo "b = $b"            # 1  ( 3 -lt 10 )
# Note escaping of operators.

b=`expr $a \<= 3`
echo "If a <= 3, b = 1 (true)"
echo "b = $b"            # 1  ( 3 -le 3 )
# There is also a "\>=" operator (greater than or equal to).


# String Operators
# ------ ---------

echo "String Operators"

echo "The string being operated upon is \"$a\"."

# length: length of string
b=`expr length $a`
echo "Length of \"$a\" is $b."

# index: position of first character in substring
#        that matches a character in string
b=`expr index $a 23`
echo "Numerical position of first \"2\" in \"$a\" is \"$b\"."

# substr: extract substring, starting position & length specified
b=`expr substr $a 2 6`
echo "Substring of \"$a\", starting at position 2,\
and 6 chars long is \"$b\"."

#  The default behavior of the 'match' operations is to
#+ search for the specified match at the ***beginning*** of the string.
#        uses Regular Expressions
b=`expr match "$a" '[0-9]*'`               #  Numerical count.
echo Number of digits at the beginning of \"$a\" is $b.
b=`expr match "$a" '\([0-9]*\)'`           #  Note that escaped parentheses
#                   ==      ==              + trigger substring match.
echo "The digits at the beginning of \"$a\" are \"$b\"."


exit 0


The : operator can substitute for match. For example, b=`expr $a : [0-9]*` is the exact equivalent of b=`expr match $a [0-9]*` in the above listing.


echo "String operations using \"expr \$string : \" construct"
echo "==================================================="


echo "The string being operated upon is \"`expr "$a" : '\(.*\)'`\"."
#     Escaped parentheses grouping operator.            ==  ==

#       ***************************
#+          Escaped parentheses
#+           match a substring
#       ***************************

#  If no escaped parentheses...
#+ then 'expr' converts the string operand to an integer.

echo "Length of \"$a\" is `expr "$a" : '.*'`."   # Length of string

echo "Number of digits at the beginning of \"$a\" is `expr "$a" : '[0-9]*'`."

# ------------------------------------------------------------------------- #


echo "The digits at the beginning of \"$a\" are `expr "$a" : '\([0-9]*\)'`."
#                                                             ==      ==
echo "The first 7 characters of \"$a\" are `expr "$a" : '\(.......\)'`."
#         =====                                          ==       ==
# Again, escaped parentheses force a substring match.
echo "The last 7 characters of \"$a\" are `expr "$a" : '.*\(.......\)'`."
#         ====                  end of string operator  ^^
#  (actually means skip over one or more of any characters until specified
#+  substring)


exit 0

This example illustrates how expr uses the escaped parentheses -- \( ... \) -- grouping operator in tandem with regular expression parsing to match a substring.

Perl, sed, and awk have far superior string parsing facilities. A short sed or awk "subroutine" within a script (see Section 34.2) is an attractive alternative to using expr.

See Section 9.2 for more on string operations.

12.3. Time / Date Commands

Time/date and timing


Simply invoked, date prints the date and time to stdout. Where this command gets interesting is in its formatting and parsing options.

Example 12-10. Using date

# Exercising the 'date' command

echo "The number of days since the year's beginning is `date +%j`."
# Needs a leading '+' to invoke formatting.
# %j gives day of year.

echo "The number of seconds elapsed since 01/01/1970 is `date +%s`."
#  %s yields number of seconds since "UNIX epoch" began,
#+ but how is this useful?

suffix=$(date +%s)  # The "+%s" option to 'date' is GNU-specific.
echo $filename
#  It's great for creating "unique" temp filenames,
#+ even better than using $$.

# Read the 'date' man page for more formatting options.

exit 0

The -u option gives the UTC (Universal Coordinated Time).

bash$ date
Fri Mar 29 21:07:39 MST 2002

bash$ date -u
Sat Mar 30 04:07:42 UTC 2002

The date command has quite a number of output options. For example %N gives the nanosecond portion of the current time. One interesting use for this is to generate six-digit random integers.
date +%N | sed -e 's/000$//' -e 's/^0//'
# Strip off leading and trailing zeroes, if present.

There are many more options (try man date).
date +%j
# Echoes day of the year (days elapsed since January 1).

date +%k%M
# Echoes hour and minute in 24-hour format, as a single digit string.

See also Example 3-4.


Time zone dump: echoes the time in a specified time zone.

bash$ zdump EST
EST  Tue Sep 18 22:09:22 2001 EST


Outputs very verbose timing statistics for executing a command.

time ls -l / gives something like this:
0.00user 0.01system 0:00.05elapsed 16%CPU (0avgtext+0avgdata 0maxresident)k
 0inputs+0outputs (149major+27minor)pagefaults 0swaps

See also the very similar times command in the previous section.


As of version 2.0 of Bash, time became a shell reserved word, with slightly altered behavior in a pipeline.


Utility for updating access/modification times of a file to current system time or other specified time, but also useful for creating a new file. The command touch zzz will create a new file of zero length, named zzz, assuming that zzz did not previously exist. Time-stamping empty files in this way is useful for storing date information, for example in keeping track of modification times on a project.


The touch command is equivalent to : >> newfile or >> newfile (for ordinary files).


The at job control command executes a given set of commands at a specified time. Superficially, it resembles cron, however, at is chiefly useful for one-time execution of a command set.

at 2pm January 15 prompts for a set of commands to execute at that time. These commands should be shell-script compatible, since, for all practical purposes, the user is typing in an executable shell script a line at a time. Input terminates with a Ctl-D.

Using either the -f option or input redirection (<), at reads a command list from a file. This file is an executable shell script, though it should, of course, be noninteractive. Particularly clever is including the run-parts command in the file to execute a different set of scripts.

bash$ at 2:30 am Friday < at-jobs.list
job 2 at 2000-10-27 02:30


The batch job control command is similar to at, but it runs a command list when the system load drops below .8. Like at, it can read commands from a file with the -f option.


Prints a neatly formatted monthly calendar to stdout. Will do current year or a large range of past and future years.


This is the shell equivalent of a wait loop. It pauses for a specified number of seconds, doing nothing. It can be useful for timing or in processes running in the background, checking for a specific event every so often (polling), as in Example 30-6.
sleep 3     # Pauses 3 seconds.


The sleep command defaults to seconds, but minute, hours, or days may also be specified.
sleep 3 h   # Pauses 3 hours!


The watch command may be a better choice than sleep for running commands at timed intervals.


Microsleep (the "u" may be read as the Greek "mu", or micro- prefix). This is the same as sleep, above, but "sleeps" in microsecond intervals. It can be used for fine-grain timing, or for polling an ongoing process at very frequent intervals.

usleep 30     # Pauses 30 microseconds.

This command is part of the Red Hat initscripts / rc-scripts package.


The usleep command does not provide particularly accurate timing, and is therefore unsuitable for critical timing loops.

hwclock, clock

The hwclock command accesses or adjusts the machine's hardware clock. Some options require root privileges. The /etc/rc.d/rc.sysinit startup file uses hwclock to set the system time from the hardware clock at bootup.

The clock command is a synonym for hwclock.

12.4. Text Processing Commands

Commands affecting text and text files


File sorter, often used as a filter in a pipe. This command sorts a text stream or file forwards or backwards, or according to various keys or character positions. Using the -m option, it merges presorted input files. The info page lists its many capabilities and options. See Example 10-9, Example 10-10, and Example A-9.


Topological sort, reading in pairs of whitespace-separated strings and sorting according to input patterns.


This filter removes duplicate lines from a sorted file. It is often seen in a pipe coupled with sort.
cat list-1 list-2 list-3 | sort | uniq > final.list
# Concatenates the list files,
# sorts them,
# removes duplicate lines,
# and finally writes the result to an output file.

The useful -c option prefixes each line of the input file with its number of occurrences.

bash$ cat testfile
This line occurs only once.
 This line occurs twice.
 This line occurs twice.
 This line occurs three times.
 This line occurs three times.
 This line occurs three times.

bash$ uniq -c testfile
      1 This line occurs only once.
       2 This line occurs twice.
       3 This line occurs three times.

bash$ sort testfile | uniq -c | sort -nr
      3 This line occurs three times.
       2 This line occurs twice.
       1 This line occurs only once.

The sort INPUTFILE | uniq -c | sort -nr command string produces a frequency of occurrence listing on the INPUTFILE file (the -nr options to sort cause a reverse numerical sort). This template finds use in analysis of log files and dictionary lists, and wherever the lexical structure of a document needs to be examined.

Example 12-11. Word Frequency Analysis

# Crude word frequency analysis on a text file.
# This is a more efficient version of the "" script.

# Check for input file on command line.

if [ $# -ne "$ARGS" ]  # Correct number of arguments passed to script?
  echo "Usage: `basename $0` filename"
  exit $E_BADARGS

if [ ! -f "$1" ]       # Check if file exists.
  echo "File \"$1\" does not exist."
  exit $E_NOFILE

# main ()
sed -e 's/\.//g'  -e 's/\,//g' -e 's/ /\
/g' "$1" | tr 'A-Z' 'a-z' | sort | uniq -c | sort -nr
#                           =========================
#                            Frequency of occurrence

#  Filter out periods and commas, and
#+ change space between words to linefeed,
#+ then shift characters to lowercase, and
#+ finally prefix occurrence count and sort numerically.

#  Arun Giridhar suggests modifying the above to:
#  . . . | sort | uniq -c | sort +1 [-f] | sort +0 -nr
#  This adds a secondary sort key, so instances of
#+ equal occurrence are sorted alphabetically.
#  As he explains it:
#  "This is effectively a radix sort, first on the
#+ least significant column
#+ (word or string, optionally case-insensitive)
#+ and last on the most significant column (frequency)."

exit 0

# Exercises:
# ---------
# 1) Add 'sed' commands to filter out other punctuation,
#+   such as semicolons.
# 2) Modify to also filter out multiple spaces and other whitespace.

bash$ cat testfile
This line occurs only once.
 This line occurs twice.
 This line occurs twice.
 This line occurs three times.
 This line occurs three times.
 This line occurs three times.

bash$ ./ testfile
      6 this
       6 occurs
       6 line
       3 times
       3 three
       2 twice
       1 only
       1 once

expand, unexpand

The expand filter converts tabs to spaces. It is often used in a pipe.

The unexpand filter converts spaces to tabs. This reverses the effect of expand.


A tool for extracting fields from files. It is similar to the print $N command set in awk, but more limited. It may be simpler to use cut in a script than awk. Particularly important are the -d (delimiter) and -f (field specifier) options.

Using cut to obtain a listing of the mounted filesystems:
cat /etc/mtab | cut -d ' ' -f1,2

Using cut to list the OS and kernel version:
uname -a | cut -d" " -f1,3,11,12

Using cut to extract message headers from an e-mail folder:
bash$ grep '^Subject:' read-messages | cut -c10-80
Re: Linux suitable for mission-critical apps?
 Spam complaint
 Re: Spam complaint

Using cut to parse a file:
# List all the users in /etc/passwd.


for user in $(cut -d: -f1 $FILENAME)
  echo $user

# Thanks, Oleg Philon for suggesting this.

cut -d ' ' -f2,3 filename is equivalent to awk -F'[ ]' '{ print $2, $3 }' filename

See also Example 12-39.


Tool for merging together different files into a single, multi-column file. In combination with cut, useful for creating system log files.


Consider this a special-purpose cousin of paste. This powerful utility allows merging two files in a meaningful fashion, which essentially creates a simple version of a relational database.

The join command operates on exactly two files, but pastes together only those lines with a common tagged field (usually a numerical label), and writes the result to stdout. The files to be joined should be sorted according to the tagged field for the matchups to work properly.


100 Shoes
200 Laces
300 Socks


100 $40.00
200 $1.00
300 $2.00

bash$ join

 100 Shoes $40.00
 200 Laces $1.00
 300 Socks $2.00


The tagged field appears only once in the output.


lists the beginning of a file to stdout (the default is 10 lines, but this can be changed). It has a number of interesting options.

Example 12-12. Which files are scripts?

# Detects scripts within a directory.

TESTCHARS=2    # Test first 2 characters.
SHABANG='#!'   # Scripts begin with a "sha-bang."

for file in *  # Traverse all the files in current directory.
  if [[ `head -c$TESTCHARS "$file"` = "$SHABANG" ]]
  #      head -c2                      #!
  #  The '-c' option to "head" outputs a specified
  #+ number of characters, rather than lines (the default).
    echo "File \"$file\" is a script."
    echo "File \"$file\" is *not* a script."
exit 0

Example 12-13. Generating 10-digit random numbers

# Outputs a 10-digit random number

# Script by Stephane Chazelas.

head -c4 /dev/urandom | od -N4 -tu4 | sed -ne '1s/.* //p'

# =================================================================== #

# Analysis
# --------

# head:
# -c4 option takes first 4 bytes.

# od:
# -N4 option limits output to 4 bytes.
# -tu4 option selects unsigned decimal format for output.

# sed: 
# -n option, in combination with "p" flag to the "s" command,
# outputs only matched lines.

# The author of this script explains the action of 'sed', as follows.

# head -c4 /dev/urandom | od -N4 -tu4 | sed -ne '1s/.* //p'
# ----------------------------------> |

# Assume output up to "sed" --------> |
# is 0000000 1198195154\n

# sed begins reading characters: 0000000 1198195154\n.
# Here it finds a newline character,
# so it is ready to process the first line (0000000 1198195154).
# It looks at its <range><action>s. The first and only one is

#   range     action
#   1         s/.* //p

# The line number is in the range, so it executes the action:
# tries to substitute the longest string ending with a space in the line
# ("0000000 ") with nothing (//), and if it succeeds, prints the result
# ("p" is a flag to the "s" command here, this is different from the "p" command).

# sed is now ready to continue reading its input. (Note that before
# continuing, if -n option had not been passed, sed would have printed
# the line once again).

# Now, sed reads the remainder of the characters, and finds the end of the file.
# It is now ready to process its 2nd line (which is also numbered '$' as
# it's the last one).
# It sees it is not matched by any <range>, so its job is done.

# In few word this sed commmand means:
# "On the first line only, remove any character up to the right-most space,
# then print it."

# A better way to do this would have been:
#           sed -e 's/.* //;q'

# Here, two <range><action>s (could have been written
#           sed -e 's/.* //' -e q):

#   range                    action
#   nothing (matches line)   s/.* //
#   nothing (matches line)   q (quit)

# Here, sed only reads its first line of input.
# It performs both actions, and prints the line (substituted) before quitting
# (because of the "q" action) since the "-n" option is not passed.

# =================================================================== #

# A simpler altenative to the above 1-line script would be:
#           head -c4 /dev/urandom| od -An -tu4

exit 0
See also Example 12-33.


lists the end of a file to stdout (the default is 10 lines). Commonly used to keep track of changes to a system logfile, using the -f option, which outputs lines appended to the file.

Example 12-14. Using tail to monitor the system log



cat /dev/null > $filename; echo "Creating / cleaning out file."
#  Creates file if it does not already exist,
#+ and truncates it to zero length if it does.
#  : > filename   and   > filename also work.

tail /var/log/messages > $filename  
# /var/log/messages must have world read permission for this to work.

echo "$filename contains tail end of system log."

exit 0

See also Example 12-5, Example 12-33 and Example 30-6.


A multi-purpose file search tool that uses Regular Expressions. It was originally a command/filter in the venerable ed line editor: g/re/p -- global - regular expression - print.

grep pattern [file...]

Search the target file(s) for occurrences of pattern, where pattern may be literal text or a Regular Expression.

bash$ grep '[rst]ystem.$' osinfo.txt
The GPL governs the distribution of the Linux operating system.

If no target file(s) specified, grep works as a filter on stdout, as in a pipe.

bash$ ps ax | grep clock
765 tty1     S      0:00 xclock
 901 pts/1    S      0:00 grep clock

The -i option causes a case-insensitive search.

The -w option matches only whole words.

The -l option lists only the files in which matches were found, but not the matching lines.

The -r (recursive) option searches files in the current working directory and all subdirectories below it.

The -n option lists the matching lines, together with line numbers.

bash$ grep -n Linux osinfo.txt
2:This is a file containing information about Linux.
 6:The GPL governs the distribution of the Linux operating system.

The -v (or --invert-match) option filters out matches.
grep pattern1 *.txt | grep -v pattern2

# Matches all lines in "*.txt" files containing "pattern1",
# but ***not*** "pattern2".	      

The -c (--count) option gives a numerical count of matches, rather than actually listing the matches.
grep -c txt *.sgml   # (number of occurrences of "txt" in "*.sgml" files)

#   grep -cz .
#            ^ dot
# means count (-c) zero-separated (-z) items matching "."
# that is, non-empty ones (containing at least 1 character).
printf 'a b\nc  d\n\n\n\n\n\000\n\000e\000\000\nf' | grep -cz .     # 4
printf 'a b\nc  d\n\n\n\n\n\000\n\000e\000\000\nf' | grep -cz '$'   # 5
printf 'a b\nc  d\n\n\n\n\n\000\n\000e\000\000\nf' | grep -cz '^'   # 5
printf 'a b\nc  d\n\n\n\n\n\000\n\000e\000\000\nf' | grep -c '$'    # 9
# By default, newline chars (\n) separate items to match. 

# Note that the -z option is GNU "grep" specific.

# Thanks, S.C.

When invoked with more than one target file given, grep specifies which file contains matches.

bash$ grep Linux osinfo.txt misc.txt
osinfo.txt:This is a file containing information about Linux.
 osinfo.txt:The GPL governs the distribution of the Linux operating system.
 misc.txt:The Linux operating system is steadily gaining in popularity.


To force grep to show the filename when searching only one target file, simply give /dev/null as the second file.

bash$ grep Linux osinfo.txt /dev/null
osinfo.txt:This is a file containing information about Linux.
 osinfo.txt:The GPL governs the distribution of the Linux operating system.

If there is a successful match, grep returns an exit status of 0, which makes it useful in a condition test in a script, especially in combination with the -q option to suppress output.
SUCCESS=0                      # if grep lookup succeeds

grep -q "$word" "$filename"    # The "-q" option causes nothing to echo to stdout.

if [ $? -eq $SUCCESS ]
# if grep -q "$word" "$filename"   can replace lines 5 - 7.
  echo "$word found in $filename"
  echo "$word not found in $filename"

Example 30-6 demonstrates how to use grep to search for a word pattern in a system logfile.

Example 12-15. Emulating "grep" in a script

# Very crude reimplementation of 'grep'.


if [ -z "$1" ]    # Check for argument to script.
  echo "Usage: `basename $0` pattern"
  exit $E_BADARGS


for file in *     # Traverse all files in $PWD.
  output=$(sed -n /"$1"/p $file)  # Command substitution.

  if [ ! -z "$output" ]           # What happens if "$output" is not quoted?
    echo -n "$file: "
    echo $output
  fi              #  sed -ne "/$1/s|^|${file}: |p"  is equivalent to above.



exit 0

# Exercises:
# ---------
# 1) Add newlines to output, if more than one match in any given file.
# 2) Add features.

How can grep search for two separate patterns? What if you want grep to display all lines in a file or files that contain both "pattern1" and "pattern2"?

One method of accomplishing this is to pipe the result of grep pattern1 to grep pattern2.

# tstfile

This is a sample file.
This is an ordinary text file.
This file does not contain any unusual text.
This file is not unusual.
Here is some text.

bash$ grep file tstfile
# tstfile
 This is a sample file.
 This is an ordinary text file.
 This file does not contain any unusual text.
 This file is not unusual.

bash$ grep file tstfile | grep text
This is an ordinary text file.
 This file does not contain any unusual text.


egrep (extended grep) is the same as grep -E. This uses a somewhat different, extended set of Regular Expressions, which can make the search a bit more flexible.

fgrep (fast grep) is the same as grep -F. It does a literal string search (no regular expressions), which allegedly speeds things up a bit.

agrep (approximate grep) extends the capabilities of grep to approximate matching. The search string may differ by a specified number of characters from the resulting matches. This utility is not part of the core Linux distribution.


To search compressed files, use zgrep, zegrep, or zfgrep. These also work on non-compressed files, though slower than plain grep, egrep, fgrep. They are handy for searching through a mixed set of files, some compressed, some not.

To search bzipped files, use bzgrep.


The command look works like grep, but does a lookup on a "dictionary", a sorted word list. By default, look searches for a match in /usr/dict/words, but a different dictionary file may be specified.

Example 12-16. Checking words in a list for validity

# lookup: Does a dictionary lookup on each word in a data file.  # Data file from which to read words to test.


while [ "$word" != end ]  # Last word in data file.
  read word      # From data file, because of redirection at end of loop.
  look $word > /dev/null  # Don't want to display lines in dictionary file.
  lookup=$?      # Exit status of 'look' command.

  if [ "$lookup" -eq 0 ]
    echo "\"$word\" is valid."
    echo "\"$word\" is invalid."

done <"$file"    # Redirects stdin to $file, so "reads" come from there.


exit 0

# ----------------------------------------------------------------
# Code below line will not execute because of "exit" command above.

# Stephane Chazelas proposes the following, more concise alternative:

while read word && [[ $word != end ]]
do if look "$word" > /dev/null
   then echo "\"$word\" is valid."
   else echo "\"$word\" is invalid."
done <"$file"

exit 0
sed, awk

Scripting languages especially suited for parsing text files and command output. May be embedded singly or in combination in pipes and shell scripts.


Non-interactive "stream editor", permits using many ex commands in batch mode. It finds many uses in shell scripts.


Programmable file extractor and formatter, good for manipulating and/or extracting fields (columns) in structured text files. Its syntax is similar to C.


wc gives a "word count" on a file or I/O stream:
bash $ wc /usr/doc/sed-3.02/README
20     127     838 /usr/doc/sed-3.02/README
[20 lines  127 words  838 characters]

wc -w gives only the word count.

wc -l gives only the line count.

wc -c gives only the character count.

wc -L gives only the length of the longest line.

Using wc to count how many .txt files are in current working directory:
$ ls *.txt | wc -l
# Will work as long as none of the "*.txt" files have a linefeed in their name.

# Alternative ways of doing this are:
#      find . -maxdepth 1 -name \*.txt -print0 | grep -cz .
#      (shopt -s nullglob; set -- *.txt; echo $#)

# Thanks, S.C.

Using wc to total up the size of all the files whose names begin with letters in the range d - h
bash$ wc [d-h]* | grep total | awk '{print $3}'

Using wc to count the instances of the word "Linux" in the main source file for this book.
bash$ grep Linux abs-book.sgml | wc -l

See also Example 12-33 and Example 16-8.

Certain commands include some of the functionality of wc as options.
... | grep foo | wc -l
# This frequently used construct can be more concisely rendered.

... | grep -c foo
# Just use the "-c" (or "--count") option of grep.

# Thanks, S.C.


character translation filter.


Must use quoting and/or brackets, as appropriate. Quotes prevent the shell from reinterpreting the special characters in tr command sequences. Brackets should be quoted to prevent expansion by the shell.

Either tr "A-Z" "*" <filename or tr A-Z \* <filename changes all the uppercase letters in filename to asterisks (writes to stdout). On some systems this may not work, but tr A-Z '[**]' will.

The -d option deletes a range of characters.
echo "abcdef"                 # abcdef
echo "abcdef" | tr -d b-d     # aef

tr -d 0-9 <filename
# Deletes all digits from the file "filename".

The --squeeze-repeats (or -s) option deletes all but the first instance of a string of consecutive characters. This option is useful for removing excess whitespace.
bash$ echo "XXXXX" | tr --squeeze-repeats 'X'

The -c "complement" option inverts the character set to match. With this option, tr acts only upon those characters not matching the specified set.

bash$ echo "acfdeb123" | tr -c b-d +

Note that tr recognizes POSIX character classes. [30]

bash$ echo "abcd2ef1" | tr '[:alpha:]' -

Example 12-17. toupper: Transforms a file to all uppercase.

# Changes a file to all uppercase.


if [ -z "$1" ]  # Standard check for command line arg.
  echo "Usage: `basename $0` filename"
  exit $E_BADARGS

tr a-z A-Z <"$1"

# Same effect as above, but using POSIX character set notation:
#        tr '[:lower:]' '[:upper:]' <"$1"
# Thanks, S.C.

exit 0

Example 12-18. lowercase: Changes all filenames in working directory to lowercase.

#! /bin/bash
# Changes every filename in working directory to all lowercase.
# Inspired by a script of John Dubois,
# which was translated into into Bash by Chet Ramey,
# and considerably simplified by Mendel Cooper, author of this document.

for filename in *                # Traverse all files in directory.
   fname=`basename $filename`
   n=`echo $fname | tr A-Z a-z`  # Change name to lowercase.
   if [ "$fname" != "$n" ]       # Rename only files not already lowercase.
     mv $fname $n

exit 0

# Code below this line will not execute because of "exit".
# To run it, delete script above line.

# The above script will not work on filenames containing blanks or newlines.

# Stephane Chazelas therefore suggests the following alternative:

for filename in *    # Not necessary to use basename,
                     # since "*" won't return any file containing "/".
do n=`echo "$filename/" | tr '[:upper:]' '[:lower:]'`
#                             POSIX char set notation.
#                    Slash added so that trailing newlines are not
#                    removed by command substitution.
   # Variable substitution:
   n=${n%/}          # Removes trailing slash, added above, from filename.
   [[ $filename == $n ]] || mv "$filename" "$n"
                     # Checks if filename already lowercase.

exit 0

Example 12-19. Du: DOS to UNIX text file conversion.

# DOS to UNIX text file converter.


if [ -z "$1" ]
  echo "Usage: `basename $0` filename-to-convert"


CR='\015'  # Carriage return.
           # 015 is octal ASCII code for CR.
           # Lines in a DOS text file end in CR-LF.
           # Lines in a UNIX text file end in LF only.

tr -d $CR < $1 > $NEWFILENAME
# Delete CR's and write to new file.

echo "Original DOS text file is \"$1\"."
echo "Converted UNIX text file is \"$NEWFILENAME\"."

exit 0

# Exercise:
# --------
# Change the above script to convert from UNIX to DOS.

Example 12-20. rot13: rot13, ultra-weak encryption.

# Classic rot13 algorithm,
#           encryption that might fool a 3-year old.

# Usage: ./ filename
# or     ./ <filename
# or     ./ and supply keyboard input (stdin)

cat "$@" | tr 'a-zA-Z' 'n-za-mN-ZA-M'   # "a" goes to "n", "b" to "o", etc.
#  The 'cat "$@"' construction
#+ permits getting input either from stdin or from files.

exit 0

Example 12-21. Generating "Crypto-Quote" Puzzles

# Encrypt quotes

#  Will encrypt famous quotes in a simple monoalphabetic substitution.
#  The result is similar to the "Crypto Quote" puzzles
#+ seen in the Op Ed pages of the Sunday paper.

# The "key" is nothing more than a scrambled alphabet.
# Changing the "key" changes the encryption.

# The 'cat "$@"' construction gets input either from stdin or from files.
# If using stdin, terminate input with a Control-D.
# Otherwise, specify filename as command-line parameter.

cat "$@" | tr "a-z" "A-Z" | tr "A-Z" "$key"
#        |  to uppercase  |     encrypt       
# Will work on lowercase, uppercase, or mixed-case quotes.
# Passes non-alphabetic characters through unchanged.

# Try this script with something like:
# "Nothing so needs reforming as other people's habits."
# --Mark Twain
# Output is:

# To reverse the encryption:
# cat "$@" | tr "$key" "A-Z"

#  This simple-minded cipher can be broken by an average 12-year old
#+ using only pencil and paper.

exit 0

A filter that wraps lines of input to a specified width. This is especially useful with the -s option, which breaks lines at word spaces (see Example 12-22 and Example A-2).


Simple-minded file formatter, used as a filter in a pipe to "wrap" long lines of text output.

Example 12-22. Formatted file listing.


WIDTH=40                    # 40 columns wide.

b=`ls /usr/local/bin`       # Get a file listing...

echo $b | fmt -w $WIDTH

# Could also have been done by
#    echo $b | fold - -s -w $WIDTH
exit 0

See also Example 12-5.


A powerful alternative to fmt is Kamil Toman's par utility, available from


This deceptively named filter removes reverse line feeds from an input stream. It also attempts to replace whitespace with equivalent tabs. The chief use of col is in filtering the output from certain text processing utilities, such as groff and tbl.


Column formatter. This filter transforms list-type text output into a "pretty-printed" table by inserting tabs at appropriate places.

Example 12-23. Using column to format a directory listing

# This is a slight modification of the example file in the "column" man page.

; ls -l | sed 1d) | column -t

#  The "sed 1d" in the pipe deletes the first line of output,
#+ which would be "total        N",
#+ where "N" is the total number of files found by "ls -l".

# The -t option to "column" pretty-prints a table.

exit 0

Column removal filter. This removes columns (characters) from a file and writes the file, lacking the range of specified columns, back to stdout. colrm 2 4 <filename removes the second through fourth characters from each line of the text file filename.


If the file contains tabs or nonprintable characters, this may cause unpredictable behavior. In such cases, consider using expand and unexpand in a pipe preceding colrm.


Line numbering filter. nl filename lists filename to stdout, but inserts consecutive numbers at the beginning of each non-blank line. If filename omitted, operates on stdin.

The output of nl is very similar to cat -n, however, by default nl does not list blank lines.

Example 12-24. nl: A self-numbering script.


# This script echoes itself twice to stdout with its lines numbered.

# 'nl' sees this as line 3 since it does not number blank lines.
# 'cat -n' sees the above line as number 5.

nl `basename $0`

echo; echo  # Now, let's try it with 'cat -n'

cat -n `basename $0`
# The difference is that 'cat -n' numbers the blank lines.
# Note that 'nl -ba' will also do so.

exit 0
# -----------------------------------------------------------------

Print formatting filter. This will paginate files (or stdout) into sections suitable for hard copy printing or viewing on screen. Various options permit row and column manipulation, joining lines, setting margins, numbering lines, adding page headers, and merging files, among other things. The pr command combines much of the functionality of nl, paste, fold, column, and expand.

pr -o 5 --width=65 fileZZZ | more gives a nice paginated listing to screen of fileZZZ with margins set at 5 and 65.

A particularly useful option is -d, forcing double-spacing (same effect as sed -G).


The GNU gettext package is a set of utilities for localizing and translating the text output of programs into foreign languages. While originally intended for C programs, it now supports quite a number of programming and scripting languages.

The gettext program works on shell scripts. See the info page.


A program for generating binary message catalogs. It is used for localization.


A utility for converting file(s) to a different encoding (character set). Its chief use is for localization.


Consider this a fancier version of iconv, above. This very versatile utility for converting a file to a different encoding is not part of the standard Linux installation.

TeX, gs

TeX and Postscript are text markup languages used for preparing copy for printing or formatted video display.

TeX is Donald Knuth's elaborate typsetting system. It is often convenient to write a shell script encapsulating all the options and arguments passed to one of these markup languages.

Ghostscript (gs) is a GPL-ed Postscript interpreter.

groff, tbl, eqn

Yet another text markup and display formatting language is groff. This is the enhanced GNU version of the venerable UNIX roff/troff display and typesetting package. Manpages use groff (see Example A-1).

The tbl table processing utility is considered part of groff, as its function is to convert table markup into groff commands.

The eqn equation processing utility is likewise part of groff, and its function is to convert equation markup into groff commands.

lex, yacc

The lex lexical analyzer produces programs for pattern matching. This has been replaced by the nonproprietary flex on Linux systems.

The yacc utility creates a parser based on a set of specifications. This has been replaced by the nonproprietary bison on Linux systems.

12.5. File and Archiving Commands



The standard UNIX archiving utility. Originally a Tape ARchiving program, it has developed into a general purpose package that can handle all manner of archiving with all types of destination devices, ranging from tape drives to regular files to even stdout (see Example 3-4). GNU tar has been patched to accept various compression filters, such as tar czvf archive_name.tar.gz *, which recursively archives and gzips all files in a directory tree except dotfiles in the current working directory ($PWD). [31]

Some useful tar options:

  1. -c create (a new archive)

  2. -x extract (files from existing archive)

  3. --delete delete (files from existing archive)


    This option will not work on magnetic tape devices.

  4. -r append (files to existing archive)

  5. -A append (tar files to existing archive)

  6. -t list (contents of existing archive)

  7. -u update archive

  8. -d compare archive with specified filesystem

  9. -z gzip the archive

    (compress or uncompress, depending on whether combined with the -c or -x) option

  10. -j bzip2 the archive


It may be difficult to recover data from a corrupted gzipped tar archive. When archiving important files, make multiple backups.


Shell archiving utility. The files in a shell archive are concatenated without compression, and the resultant archive is essentially a shell script, complete with #!/bin/sh header, and containing all the necessary unarchiving commands. Shar archives still show up in Internet newsgroups, but otherwise shar has been pretty well replaced by tar/gzip. The unshar command unpacks shar archives.


Creation and manipulation utility for archives, mainly used for binary object file libraries.


The Red Hat Package Manager, or rpm utility provides a wrapper for source or binary archives. It includes commands for installing and checking the integrity of packages, among other things.

A simple rpm -i package_name.rpm usually suffices to install a package, though there are many more options available.


An rpm -qa gives a complete list of all installed rpm packages on a given system. An rpm -qa package_name lists only the package(s) corresponding to package_name.

bash$ rpm -qa

bash$ rpm -qa docbook-utils

bash$ rpm -qa docbook | grep docbook


This specialized archiving copy command (copy input and output) is rarely seen any more, having been supplanted by tar/gzip. It still has its uses, such as moving a directory tree.

Example 12-25. Using cpio to move a directory tree


# Copying a directory tree using 'cpio.'


if [ $# -ne "$ARGS" ]
  echo "Usage: `basename $0` source destination"
  exit $E_BADARGS


find "$source" -depth | cpio -admvp "$destination"
#               ^^^^^         ^^^^^
# Read the 'find' and 'cpio' man page to decipher these options.

#  It may be useful to check the exit status ($?) here
#+ to see if everything worked all right.

exit 0

This command extracts a cpio archive from an rpm one.

Example 12-26. Unpacking an rpm archive

# Unpack an 'rpm' archive

: ${1?"Usage: `basename $0` target-file"}
# Must specify 'rpm' archive name as an argument.

TEMPFILE=$$.cpio                         # Tempfile with "unique" name.
                                         # $$ is process ID of script.

rpm2cpio < $1 > $TEMPFILE                # Converts rpm archive into cpio archive.
cpio --make-directories -F $TEMPFILE -i  # Unpacks cpio archive.
rm -f $TEMPFILE                          # Deletes cpio archive.

exit 0

#  Exercise:
#  Add check for whether 1) "target-file" exists and
#+                       2) it is really an rpm archive.
#  Hint:                    parse output of 'file' command.



The standard GNU/UNIX compression utility, replacing the inferior and proprietary compress. The corresponding decompression command is gunzip, which is the equivalent of gzip -d.

The zcat filter decompresses a gzipped file to stdout, as possible input to a pipe or redirection. This is, in effect, a cat command that works on compressed files (including files processed with the older compress utility). The zcat command is equivalent to gzip -dc.


On some commercial UNIX systems, zcat is a synonym for uncompress -c, and will not work on gzipped files.

See also Example 7-7.


An alternate compression utility, usually more efficient (but slower) than gzip, especially on large files. The corresponding decompression command is bunzip2.


Newer versions of tar have been patched with bzip2 support.

compress, uncompress

This is an older, proprietary compression utility found in commercial UNIX distributions. The more efficient gzip has largely replaced it. Linux distributions generally include a compress workalike for compatibility, although gunzip can unarchive files treated with compress.


The znew command transforms compressed files into gzipped ones.


Yet another compression utility, a filter that works only on sorted ASCII word lists. It uses the standard invocation syntax for a filter, sq < input-file > output-file. Fast, but not nearly as efficient as gzip. The corresponding uncompression filter is unsq, invoked like sq.


The output of sq may be piped to gzip for further compression.

zip, unzip

Cross-platform file archiving and compression utility compatible with DOS pkzip.exe. "Zipped" archives seem to be a more acceptable medium of exchange on the Internet than "tarballs".

unarc, unarj, unrar

These Linux utilities permit unpacking archives compressed with the DOS arc.exe, arj.exe, and rar.exe programs.

File Information


A utility for identifying file types. The command file file-name will return a file specification for file-name, such as ascii text or data. It references the magic numbers found in /usr/share/magic, /etc/magic, or /usr/lib/magic, depending on the Linux/UNIX distribution.

The -f option causes file to run in batch mode, to read from a designated file a list of filenames to analyze. The -z option, when used on a compressed target file, forces an attempt to analyze the uncompressed file type.

bash$ file test.tar.gz
test.tar.gz: gzip compressed data, deflated, last modified: Sun Sep 16 13:34:51 2001, os: Unix

bash file -z test.tar.gz
test.tar.gz: GNU tar archive (gzip compressed data, deflated, last modified: Sun Sep 16 13:34:51 2001, os: Unix)

Example 12-27. Stripping comments from C program files

# Strips out the comments (/* COMMENT */) in a C program.


if [ $# -eq "$E_NOARGS" ]
  echo "Usage: `basename $0` C-program-file" >&2 # Error message to stderr.
  exit $E_ARGERROR

# Test for correct file type.
type=`file $1 | awk '{ print $2, $3, $4, $5 }'`
# "file $1" echoes file type . . .
# Then awk removes the first field of this, the filename . . .
# Then the result is fed into the variable "type".
correct_type="ASCII C program text"

if [ "$type" != "$correct_type" ]
  echo "This script works on C program files only."

# Rather cryptic sed script:
sed '
' $1
# Easy to understand if you take several hours to learn sed fundamentals.

#  Need to add one more line to the sed script to deal with
#+ case where line of code has a comment following it on same line.
#  This is left as a non-trivial exercise.

#  Also, the above code deletes lines with a "*/" or "/*",
#+ not a desirable result.

exit 0

# ----------------------------------------------------------------
# Code below this line will not execute because of 'exit 0' above.

# Stephane Chazelas suggests the following alternative:

usage() {
  echo "Usage: `basename $0` C-program-file" >&2
  exit 1

WEIRD=`echo -n -e '\377'`   # or WEIRD=$'\377'
[[ $# -eq 1 ]] || usage
case `file "$1"` in
  *"C program text"*) sed -e "s%/\*%${WEIRD}%g;s%\*/%${WEIRD}%g" "$1" \
     | tr '\377\n' '\n\377' \
     | sed -ne 'p;n' \
     | tr -d '\n' | tr '\377' '\n';;
  *) usage;;

#  This is still fooled by things like:
#  printf("/*");
#  or
#  /*  /* buggy embedded comment */
#  To handle all special cases (comments in strings, comments in string
#+ where there is a \", \\" ...) the only way is to write a C parser
#+ (using lex or yacc perhaps?).

exit 0

which command-xxx gives the full path to "command-xxx". This is useful for finding out whether a particular command or utility is installed on the system.

$bash which rm


Similar to which, above, whereis command-xxx gives the full path to "command-xxx", but also to its manpage.

$bash whereis rm
rm: /bin/rm /usr/share/man/man1/rm.1.bz2


whatis filexxx looks up "filexxx" in the whatis database. This is useful for identifying system commands and important configuration files. Consider it a simplified man command.

$bash whatis whatis
whatis               (1)  - search the whatis database for complete words

Example 12-28. Exploring /usr/X11R6/bin


# What are all those mysterious binaries in /usr/X11R6/bin?

# Try also "/bin", "/usr/bin", "/usr/local/bin", etc.

for file in $DIRECTORY/*
  whatis `basename $file`   # Echoes info about the binary.

exit 0
# You may wish to redirect output of this script, like so:
# ./ >>whatis.db
# or view it a page at a time on stdout,
# ./ | less

See also Example 10-3.


Show a detailed directory listing. The effect is similar to ls -l.

This is one of the GNU fileutils.

bash$ vdir
total 10
 -rw-r--r--    1 bozo  bozo      4034 Jul 18 22:04 data1.xrolo
 -rw-r--r--    1 bozo  bozo      4602 May 25 13:58 data1.xrolo.bak
 -rw-r--r--    1 bozo  bozo       877 Dec 17  2000 employment.xrolo

bash ls -l
total 10
 -rw-r--r--    1 bozo  bozo      4034 Jul 18 22:04 data1.xrolo
 -rw-r--r--    1 bozo  bozo      4602 May 25 13:58 data1.xrolo.bak
 -rw-r--r--    1 bozo  bozo       877 Dec 17  2000 employment.xrolo

locate, slocate

The locate command searches for files using a database stored for just that purpose. The slocate command is the secure version of locate (which may be aliased to slocate).

$bash locate hickson


Disclose the file that a symbolic link points to.

bash$ readlink /usr/bin/awk


Use the strings command to find printable strings in a binary or data file. It will list sequences of printable characters found in the target file. This might be handy for a quick 'n dirty examination of a core dump or for looking at an unknown graphic image file (strings image-file | more might show something like JFIF, which would identify the file as a jpeg graphic). In a script, you would probably parse the output of strings with grep or sed. See Example 10-7 and Example 10-9.

Example 12-29. An "improved" strings command

# "word-strings" (enhanced "strings" command)
#  This script filters the output of "strings" by checking it
#+ against a standard word list file.
#  This effectively eliminates gibberish and noise,
#+ and outputs only recognized words.

# =================================================================
#                 Standard Check for Script Argument(s)

if [ $# -ne $ARGS ]
  echo "Usage: `basename $0` filename"
  exit $E_BADARGS

if [ ! -f "$1" ]                      # Check if file exists.
    echo "File \"$1\" does not exist."
    exit $E_NOFILE
# =================================================================

MINSTRLEN=3                           #  Minimum string length.
WORDFILE=/usr/share/dict/linux.words  #  Dictionary file.
                                      #  May specify a different
                                      #+ word list file
                                      #+ of one-word-per-line format.

wlist=`strings "$1" | tr A-Z a-z | tr '[:space:]' Z | \
tr -cs '[:alpha:]' Z | tr -s '\173-\377' Z | tr Z ' '`

# Translate output of 'strings' command with multiple passes of 'tr'.
#  "tr A-Z a-z"  converts to lowercase.
#  "tr '[:space:]'"  converts whitespace characters to Z's.
#  "tr -cs '[:alpha:]' Z"  converts non-alphabetic characters to Z's,
#+ and squeezes multiple consecutive Z's.
#  "tr -s '\173-\377' Z"  converts all characters past 'z' to Z's
#+ and squeezes multiple consecutive Z's,
#+ which gets rid of all the weird characters that the previous
#+ translation failed to deal with.
#  Finally, "tr Z ' '" converts all those Z's to whitespace,
#+ which will be seen as word separators in the loop below.

#  Note the technique of feeding the output of 'tr' back to itself,
#+ but with different arguments and/or options on each pass.

for word in $wlist                    # Important:
                                      # $wlist must not be quoted here.
                                      # "$wlist" does not work.
                                      # Why?

  strlen=${#word}                     # String length.
  if [ "$strlen" -lt "$MINSTRLEN" ]   # Skip over short strings.

  grep -Fw $word "$WORDFILE"          #  Match whole words only.
#      ^^^                            #  "Fixed strings" and
                                      #+ "whole words" options. 


exit 0


diff, patch

diff: flexible file comparison utility. It compares the target files line-by-line sequentially. In some applications, such as comparing word dictionaries, it may be helpful to filter the files through sort and uniq before piping them to diff. diff file-1 file-2 outputs the lines in the files that differ, with carets showing which file each particular line belongs to.

The --side-by-side option to diff outputs each compared file, line by line, in separate columns, with non-matching lines marked. The -c and -u options likewise make the output of the command easier to interpret.

There are available various fancy frontends for diff, such as spiff, wdiff, xdiff, and mgdiff.


The diff command returns an exit status of 0 if the compared files are identical, and 1 if they differ. This permits use of diff in a test construct within a shell script (see below).

A common use for diff is generating difference files to be used with patch The -e option outputs files suitable for ed or ex scripts.

patch: flexible versioning utility. Given a difference file generated by diff, patch can upgrade a previous version of a package to a newer version. It is much more convenient to distribute a relatively small "diff" file than the entire body of a newly revised package. Kernel "patches" have become the preferred method of distributing the frequent releases of the Linux kernel.

patch -p1 <patch-file
# Takes all the changes listed in 'patch-file'
# and applies them to the files referenced therein.
# This upgrades to a newer version of the package.

Patching the kernel:

cd /usr/src
gzip -cd patchXX.gz | patch -p0
# Upgrading kernel source using 'patch'.
# From the Linux kernel docs "README",
# by anonymous author (Alan Cox?).


The diff command can also recursively compare directories (for the filenames present).

bash$ diff -r ~/notes1 ~/notes2
Only in /home/bozo/notes1: file02
 Only in /home/bozo/notes1: file03
 Only in /home/bozo/notes2: file04


Use zdiff to compare gzipped files.


An extended version of diff that compares three files at a time. This command returns an exit value of 0 upon successful execution, but unfortunately this gives no information about the results of the comparison.

bash$ diff3 file-1 file-2 file-3
   This is line 1 of "file-1".
   This is line 1 of "file-2".
   This is line 1 of "file-3"


Compare and/or edit two files in order to merge them into an output file. Because of its interactive nature, this command would find little use in a script.


The cmp command is a simpler version of diff, above. Whereas diff reports the differences between two files, cmp merely shows at what point they differ.


Like diff, cmp returns an exit status of 0 if the compared files are identical, and 1 if they differ. This permits use in a test construct within a shell script.

Example 12-30. Using cmp to compare two files within a script.


ARGS=2  # Two args to script expected.

if [ $# -ne "$ARGS" ]
  echo "Usage: `basename $0` file1 file2"
  exit $E_BADARGS

if [[ ! -r "$1" || ! -r "$2" ]]
  echo "Both files to be compared must exist and be readable."

cmp $1 $2 &> /dev/null  # /dev/null buries the output of the "cmp" command.
#   cmp -s $1 $2  has same result ("-s" silent flag to "cmp")
#   Thank you  Anders Gustavsson for pointing this out.
# Also works with 'diff', i.e.,   diff $1 $2 &> /dev/null

if [ $? -eq 0 ]         # Test exit status of "cmp" command.
  echo "File \"$1\" is identical to file \"$2\"."
  echo "File \"$1\" differs from file \"$2\"."

exit 0


Use zcmp on gzipped files.


Versatile file comparison utility. The files must be sorted for this to be useful.

comm -options first-file second-file

comm file-1 file-2 outputs three columns:

  • column 1 = lines unique to file-1

  • column 2 = lines unique to file-2

  • column 3 = lines common to both.

The options allow suppressing output of one or more columns.

  • -1 suppresses column 1

  • -2 suppresses column 2

  • -3 suppresses column 3

  • -12 suppresses both columns 1 and 2, etc.



Strips the path information from a file name, printing only the file name. The construction basename $0 lets the script know its name, that is, the name it was invoked by. This can be used for "usage" messages if, for example a script is called with missing arguments:
echo "Usage: `basename $0` arg1 arg2 ... argn"


Strips the basename from a filename, printing only the path information.


basename and dirname can operate on any arbitrary string. The argument does not need to refer to an existing file, or even be a filename for that matter (see Example A-8).

Example 12-31. basename and dirname



echo "Basename of /home/bozo/daily-journal.txt = `basename $a`"
echo "Dirname of /home/bozo/daily-journal.txt = `dirname $a`"
echo "My own home is `basename ~/`."         # Also works with just ~.
echo "The home of my home is `dirname ~/`."  # Also works with just ~.

exit 0

Utility for splitting a file into smaller chunks. Usually used for splitting up large files in order to back them up on floppies or preparatory to e-mailing or uploading them.

sum, cksum, md5sum

These are utilities for generating checksums. A checksum is a number mathematically calculated from the contents of a file, for the purpose of checking its integrity. A script might refer to a list of checksums for security purposes, such as ensuring that the contents of key system files have not been altered or corrupted. For security applications, use the 128-bit md5sum (message digest checksum) command.

bash$ cksum /boot/vmlinuz
1670054224 804083 /boot/vmlinuz

bash$ echo -n "Top Secret" | cksum
3391003827 10

bash$ md5sum /boot/vmlinuz
0f43eccea8f09e0a0b2b5cf1dcf333ba  /boot/vmlinuz

bash$ echo -n "Top Secret" | md5sum
8babc97a6f62a4649716f4df8d61728f  -


The cksum command shows the size, in bytes, of its target, whether file or stdout.

The md5sum command displays a dash when it receives its input from stdout.

Example 12-32. Checking file integrity

# Checking whether files in a given directory
#                    have been tampered with.


# Filename for storing records (database file).

set_up_database ()
  echo ""$directory"" > "$dbfile"
  # Write directory name to first line of file.
  md5sum "$directory"/* >> "$dbfile"
  # Append md5 checksums and filenames.

check_database ()
  local n=0
  local filename
  local checksum

  # ------------------------------------------- #
  #  This file check should be unnecessary,
  #+ but better safe than sorry.

  if [ ! -r "$dbfile" ]
    echo "Unable to read checksum database file!"
    exit $E_BAD_DBFILE
  # ------------------------------------------- #

  while read record[n]

    if [ "$directory_checked" != "$directory" ]
      echo "Directories do not match up!"
      # Tried to use file for a different directory.
      exit $E_DIR_NOMATCH

    if [ "$n" -gt 0 ]   # Not directory name.
      filename[n]=$( echo ${record[$n]} | awk '{ print $2 }' )
      #  md5sum writes records backwards,
      #+ checksum first, then filename.
      checksum[n]=$( md5sum "${filename[n]}" )

      if [ "${record[n]}" = "${checksum[n]}" ]
        echo "${filename[n]} unchanged."

      elif [ "`basename ${filename[n]}`" != "$dbfile" ]
             #  Skip over checksum database file,
             #+ as it will change with each invocation of script.
	     #  ---
	     #  This unfortunately means that when running
	     #+ this script on $PWD, tampering with the
	     #+ checksum database file will not be detected.
	     #  Exercise: Fix this.
          echo "${filename[n]} : CHECKSUM ERROR!"
        # File has been changed since last checked.


    let "n+=1"
  done <"$dbfile"       # Read from checksum database file. 


# =================================================== #
# main ()

if [ -z  "$1" ]
  directory="$PWD"      #  If not specified,
else                    #+ use current working directory.

clear                   # Clear screen.
echo " Running file integrity check on $directory"

# ------------------------------------------------------------------ #
  if [ ! -r "$dbfile" ] # Need to create database file?
    echo "Setting up database file, \""$directory"/"$dbfile"\"."; echo
# ------------------------------------------------------------------ #

check_database          # Do the actual work.


#  You may wish to redirect the stdout of this script to a file,
#+ especially if the directory checked has many files in it.

exit 0

#  For a much more thorough file integrity check,
#+ consider the "Tripwire" package,

See also Example A-20 for a creative use of the md5sum command.


Securely erase a file by overwriting it multiple times with random bit patterns before deleting it. This command has the same effect as Example 12-50, but does it in a more thorough and elegant manner.

This is one of the GNU fileutils.


Advanced forensic technology may still be able to recover the contents of a file, even after application of shred.

Encoding and Encryption


This utility encodes binary files into ASCII characters, making them suitable for transmission in the body of an e-mail message or in a newsgroup posting.


This reverses the encoding, decoding uuencoded files back into the original binaries.

Example 12-33. Uudecoding encoded files


lines=35        # Allow 35 lines for the header (very generous).

for File in *   # Test all the files in the current working directory...
  search1=`head -$lines $File | grep begin | wc -w`
  search2=`tail -$lines $File | grep end | wc -w`
  #  Uuencoded files have a "begin" near the beginning,
  #+ and an "end" near the end.
  if [ "$search1" -gt 0 ]
    if [ "$search2" -gt 0 ]
      echo "uudecoding - $File -"
      uudecode $File

#  Note that running this script upon itself fools it
#+ into thinking it is a uuencoded file,
#+ because it contains both "begin" and "end".

# Exercise:
# Modify this script to check for a newsgroup header.

exit 0


The fold -s command may be useful (possibly in a pipe) to process long uudecoded text messages downloaded from Usenet newsgroups.

mimencode, mmencode

The mimencode and mmencode commands process multimedia-encoded e-mail attachments. Although mail user agents (such as pine or kmail) normally handle this automatically, these particular utilities permit manipulating such attachments manually from the command line or in a batch by means of a shell script.


At one time, this was the standard UNIX file encryption utility. [32] Politically motivated government regulations prohibiting the export of encryption software resulted in the disappearance of crypt from much of the UNIX world, and it is still missing from most Linux distributions. Fortunately, programmers have come up with a number of decent alternatives to it, among them the author's very own cruft (see Example A-5).



Create a temporary file with a "unique" filename.

tempfile=`mktemp $PREFIX.XXXXXX`
#                        ^^^^^^ Need at least 6 placeholders
#+                              in the filename template.
echo "tempfile name = $tempfile"
# tempfile name = filename.QA2ZpY
#                 or something similar...


Utility for building and compiling binary packages. This can also be used for any set of operations that is triggered by incremental changes in source files.

The make command checks a Makefile, a list of file dependencies and operations to be carried out.


Special purpose file copying command, similar to cp, but capable of setting permissions and attributes of the copied files. This command seems tailormade for installing software packages, and as such it shows up frequently in Makefiles (in the make install : section). It could likewise find use in installation scripts.


This utility, written by Benjamin Lin and collaborators, converts DOS-formatted text files (lines terminated by CR-LF) to UNIX format (lines terminated by LF only), and vice-versa.


The ptx [targetfile] command outputs a permuted index (cross-reference list) of the targetfile. This may be further filtered and formatted in a pipe, if necessary.

more, less

Pagers that display a text file or stream to stdout, one screenful at a time. These may be used to filter the output of a script.

12.6. Communications Commands

Certain of the following commands find use in chasing spammers, as well as in network data transfer and analysis.

Information and Statistics


Searches for information about an Internet host by name or IP address, using DNS.

bash$ host has address


Displays IP information for a host. With the -h option, ipcalc does a reverse DNS lookup, finding the name of the host (server) from the IP address.

bash$ ipcalc -h


Do an Internet "name server lookup" on a host by IP address. This is essentially equivalent to ipcalc -h or dig -x . The command may be run either interactively or noninteractively, i.e., from within a script.

The nslookup command has allegedly been "deprecated," but it still has its uses.

bash$ nslookup -sil

 Non-authoritative answer:


Domain Information Groper. Similar to nslookup, dig does an Internet "name server lookup" on a host. May be run either interactively or noninteractively, i.e., from within a script.

Some interesting options to dig are +time=N for setting a query timeout to N seconds, +nofail for continuing to query servers until a reply is received, and -x for doing a reverse address lookup.

Compare the output of dig -x with ipcalc -h and nslookup.

bash$ dig -x
;; Got answer:
 ;; ->>HEADER<<- opcode: QUERY, status: NXDOMAIN, id: 11649
 ;; flags: qr rd ra; QUERY: 1, ANSWER: 0, AUTHORITY: 1, ADDITIONAL: 0

 ;         IN      PTR

 ;; AUTHORITY SECTION:    3600    IN      SOA
 2002031705 900 600 86400 3600

 ;; Query time: 537 msec
 ;; WHEN: Wed Jun 26 08:35:24 2002
 ;; MSG SIZE  rcvd: 91

Example 12-34. Checking a spam domain

#! /bin/bash
# Identifying spam domains
#  This is a simplified version of the "is_spammer.bash
#+ script in the Contributed Scripts appendix.

# is-spammer <>
# $Id: is-spammer, v 1.4 2004/09/01 19:37:52 mszick Exp $

# Uses an external program: 'dig'
# Tested with version: 9.2.4rc5

# Uses functions.
# Uses IFS to parse strings by assignment into arrays.
# And even does something useful: checks e-mail blacklists.

# Use the from the text body:
#                       ^^^^^^^^^^^
# Or the from any e-mail address:
#                   ^^^^^^^^^^^
# as the only argument to this script.
#(PS: have your Inet connection running)
# So, to invoke this script in the above two instances:

# Whitespace == :Space:Tab:Line Feed:Carriage Return:

# No Whitespace == Line Feed:Carriage Return

# Field separator for dotted decimal ip addresses

# Get the dns text resource record
# get_txt <error_code> <list_query>
get_txt() {

    # parse $1 by assignment at the dots
    local -a dns
    dns=( $1 )
    if [ "${dns[0]}" == '127' ]
        # see if there is a reason
        echo $(dig +short $2 -t txt)

# Get the dns address resource record
# chk_adr <rev_dns> <list_server>
chk_adr() {
    local reply
    local server
    local reason

    reply=$( dig +short ${server} )

    # if reply might be an error code . . .
    if [ ${#reply} -gt 6 ]
        reason=$(get_txt ${reply} ${server} )
    echo ${reason:-' not blacklisted.'}

# Need to get the IP address from the name.
echo 'Get address of: '$1
ip_adr=$(dig +short $1)
dns_reply=${ip_adr:-' no answer '}
echo ' Found address: '${dns_reply}

# A valid reply is at least 4 digits plus 3 dots.
if [ ${#ip_adr} -gt 6 ]
    declare query

    # Parse by assignment at the dots.
    declare -a dns
    dns=( ${ip_adr} )

    # Reorder octets into dns query order.

# See: (Conservative, well maintained)
    echo -n ' says: '
    echo $(chk_adr ${rev_dns} '')

# See: (Open mail relays)
    echo -n '  says: '
    echo $(chk_adr ${rev_dns} '')

# See: (You can report spammers here)
    echo -n ' says: '
    echo $(chk_adr ${rev_dns} '')

# # # other blacklist operations # # #

# See:
    echo -n ' says: '
    echo $(chk_adr ${rev_dns} '')

# See: (Various mail relays)
    echo 'Distributed Server Listings'
    echo -n ' says: '
    echo $(chk_adr ${rev_dns} '')

    echo -n ' says: '
    echo $(chk_adr ${rev_dns} '')

    echo -n ' says: '
    echo $(chk_adr ${rev_dns} '')

    echo 'Could not use that address.'

exit 0

# Exercises:
# --------

# 1) Check arguments to script,
#    and exit with appropriate error message if necessary.

# 2) Check if on-line at invocation of script,
#    and exit with appropriate error message if necessary.

# 3) Substitute generic variables for "hard-coded" BHL domains.

# 4) Set a time-out for the script using the "+time=" option
     to the 'dig' command.

For a much more elaborate version of the above script, see Example A-26.


Trace the route taken by packets sent to a remote host. This command works within a LAN, WAN, or over the Internet. The remote host may be specified by an IP address. The output of this command may be filtered by grep or sed in a pipe.

bash$ traceroute
traceroute to (, 30 hops max, 38 byte packets
 1 (  191.303 ms  179.400 ms  179.767 ms
 2 (  179.536 ms  179.534 ms  169.685 ms
 3 (  189.471 ms  189.556 ms *


Broadcast an "ICMP ECHO_REQUEST" packet to another machine, either on a local or remote network. This is a diagnostic tool for testing network connections, and it should be used with caution.

A successful ping returns an exit status of 0. This can be tested for in a script.

bash$ ping localhost
PING localhost.localdomain ( from : 56(84) bytes of data.
 Warning: time of day goes back, taking countermeasures.
 64 bytes from localhost.localdomain ( icmp_seq=0 ttl=255 time=709 usec
 64 bytes from localhost.localdomain ( icmp_seq=1 ttl=255 time=286 usec

 --- localhost.localdomain ping statistics ---
 2 packets transmitted, 2 packets received, 0% packet loss
 round-trip min/avg/max/mdev = 0.286/0.497/0.709/0.212 ms


Perform a DNS (Domain Name System) lookup. The -h option permits specifying which particular whois server to query. See Example 4-6.


Retrieve information about users on a network. Optionally, this command can display a user's ~/.plan, ~/.project, and ~/.forward files, if present.

bash$ finger
Login  Name           Tty      Idle  Login Time   Office     Office Phone
 bozo   Bozo Bozeman   tty1        8  Jun 25 16:59
 bozo   Bozo Bozeman   ttyp0          Jun 25 16:59
 bozo   Bozo Bozeman   ttyp1          Jun 25 17:07

bash$ finger bozo
Login: bozo                             Name: Bozo Bozeman
 Directory: /home/bozo                   Shell: /bin/bash
 Office: 2355 Clown St., 543-1234
 On since Fri Aug 31 20:13 (MST) on tty1    1 hour 38 minutes idle
 On since Fri Aug 31 20:13 (MST) on pts/0   12 seconds idle
 On since Fri Aug 31 20:13 (MST) on pts/1
 On since Fri Aug 31 20:31 (MST) on pts/2   1 hour 16 minutes idle
 No mail.
 No Plan.

Out of security considerations, many networks disable finger and its associated daemon. [33]


Change information disclosed by the finger command.


Verify an Internet e-mail address.

Remote Host Access

sx, rx

The sx and rx command set serves to transfer files to and from a remote host using the xmodem protocol. These are generally part of a communications package, such as minicom.

sz, rz

The sz and rz command set serves to transfer files to and from a remote host using the zmodem protocol. Zmodem has certain advantages over xmodem, such as faster transmission rate and resumption of interrupted file transfers. Like sx and rx, these are generally part of a communications package.


Utility and protocol for uploading / downloading files to or from a remote host. An ftp session can be automated in a script (see Example 17-6, Example A-5, and Example A-14).


UNIX to UNIX copy. This is a communications package for transferring files between UNIX servers. A shell script is an effective way to handle a uucp command sequence.

Since the advent of the Internet and e-mail, uucp seems to have faded into obscurity, but it still exists and remains perfectly workable in situations where an Internet connection is not available or appropriate.


Call Up a remote system and connect as a simple terminal. This command is part of the uucp package. It is a sort of dumbed-down version of telnet.


Utility and protocol for connecting to a remote host.


The telnet protocol contains security holes and should therefore probably be avoided.


The wget utility non-interactively retrieves or downloads files from a Web or ftp site. It works well in a script.
wget -p
wget -r -O $SAVEFILE

Example 12-35. Getting a stock quote

# Download a stock quote.


if [ -z "$1" ]  # Must specify a stock (symbol) to fetch.
  then echo "Usage: `basename $0` stock-symbol"
  exit $E_NOPARAMS


# Fetches an HTML file, so name it appropriately.
# Yahoo finance board, with stock query suffix.

# -----------------------------------------------------------
wget -O ${stock_symbol}${file_suffix} "${URL}${stock_symbol}"
# -----------------------------------------------------------

exit $?

# Exercises:
# ---------
# 1) Add a test to ensure the user running the script is on-line.
#    (Hint: parse the output of 'ps -ax' for "ppp" or "connect."
# 2) Modify this script to fetch the local weather report,
#+   taking the user's zip code as an argument.

The lynx Web and file browser can be used inside a script (with the -dump option) to retrieve a file from a Web or ftp site non-interactively.
lynx -dump >$SAVEFILE


Remote login, initates a session on a remote host. This command has security issues, so use ssh instead.


Remote shell, executes command(s) on a remote host. This has security issues, so use ssh instead.


Remote copy, copies files between two different networked machines. Using rcp and similar utilities with security implications in a shell script may not be advisable. Consider, instead, using ssh or an expect script.


Secure shell, logs onto a remote host and executes commands there. This secure replacement for telnet, rlogin, rcp, and rsh uses identity authentication and encryption. See its manpage for details.

Example 12-36. Using ssh

# remote.bash: Using ssh.

# This example by Michael Zick.
# Used with permission.

#   Presumptions:
#   ------------
#   fd-2 isn't being captured ( '2>/dev/null' ).
#   ssh/sshd presumes stderr ('2') will display to user.
#   sshd is running on your machine.
#   For any 'standard' distribution, it probably is,
#+  and without any funky ssh-keygen having been done.

# Try ssh to your machine from the command line:
# $ ssh $HOSTNAME
# Without extra set-up you'll be asked for your password.
#   enter password
#   when done,  $ exit
# Did that work? If so, you're ready for more fun.

# Try ssh to your machine as 'root':
#   $  ssh -l root $HOSTNAME
#   When asked for password, enter root's, not yours.
#          Last login: Tue Aug 10 20:25:49 2004 from localhost.localdomain
#   Enter 'exit' when done.

#  The above gives you an interactive shell.
#  It is possible for sshd to be set up in a 'single command' mode,
#+ but that is beyond the scope of this example.
#  The only thing to note is that the following will work in
#+ 'single command' mode.

# A basic, write stdout (local) command.

ls -l

# Now the same basic command on a remote machine.
# Pass a different 'USERNAME' 'HOSTNAME' if desired:

#  Now excute the above command line on the remote host,
#+ with all transmissions encrypted.

ssh -l ${USER} ${HOST} " ls -l "

#  The expected result is a listing of your username's home
#+ directory on the remote machine.
#  To see any difference, run this script from somewhere
#+ other than your home directory.

#  In other words, the Bash command is passed as a quoted line
#+ to the remote shell, which executes it on the remote machine.
#  In this case, sshd does  ' bash -c "ls -l" '   on your behalf.

#  For information on topics such as not having to enter a
#+ password/passphrase for every command line, see
#+    man ssh
#+    man ssh-keygen
#+    man sshd_config.

exit 0

Local Network


This is a utility for terminal-to-terminal communication. It allows sending lines from your terminal (console or xterm) to that of another user. The mesg command may, of course, be used to disable write access to a terminal

Since write is interactive, it would not normally find use in a script.



Send or read e-mail messages.

This stripped-down command-line mail client works fine as a command embedded in a script.

Example 12-37. A script that mails itself

# Self-mailing script

adr=${1:-`whoami`}     # Default to current user, if not specified.
#  Typing ''
#+ sends this script to that addressee.
#  Just '' (no argument) sends the script
#+ to the person invoking it, for example, bozo@localhost.localdomain.
#  For more on the ${parameter:-default} construct,
#+ see the "Parameter Substitution" section
#+ of the "Variables Revisited" chapter.

# ============================================================================
  cat $0 | mail -s "Script \"`basename $0`\" has mailed itself to you." "$adr"
# ============================================================================

# --------------------------------------------
#  Greetings from the self-mailing script.
#  A mischievous person has run this script,
#+ which has caused it to mail itself to you.
#  Apparently, some people have nothing better
#+ to do with their time.
# --------------------------------------------

echo "At `date`, script \"`basename $0`\" mailed to "$adr"."

exit 0

Similar to the mail command, mailto sends e-mail messages from the command line or in a script. However, mailto also permits sending MIME (multimedia) messages.


This utility automatically replies to e-mails that the intended recipient is on vacation and temporarily unavailable. This runs on a network, in conjunction with sendmail, and is not applicable to a dial-up POPmail account.

12.7. Terminal Control Commands

Command affecting the console or terminal


Initialize terminal and/or fetch information about it from terminfo data. Various options permit certain terminal operations. tput clear is the equivalent of clear, below. tput reset is the equivalent of reset, below. tput sgr0 also resets the terminal, but without clearing the screen.

bash$ tput longname
xterm terminal emulator (XFree86 4.0 Window System)

Issuing a tput cup X Y moves the cursor to the (X,Y) coordinates in the current terminal. A clear to erase the terminal screen would normally precede this.

Note that stty offers a more powerful command set for controlling a terminal.


This command prints out extensive information about the current terminal. It references the terminfo database.

bash$ infocmp
#       Reconstructed via infocmp from file:
 rxvt|rxvt terminal emulator (X Window System), 
         am, bce, eo, km, mir, msgr, xenl, xon, 
         colors#8, cols#80, it#8, lines#24, pairs#64, 
         bel=^G, blink=\E[5m, bold=\E[1m,
         clear=\E[H\E[2J, cnorm=\E[?25h, cr=^M, 


Reset terminal parameters and clear text screen. As with clear, the cursor and prompt reappear in the upper lefthand corner of the terminal.


The clear command simply clears the text screen at the console or in an xterm. The prompt and cursor reappear at the upper lefthand corner of the screen or xterm window. This command may be used either at the command line or in a script. See Example 10-25.


This utility records (saves to a file) all the user keystrokes at the command line in a console or an xterm window. This, in effect, creates a record of a session.

12.8. Math Commands

"Doing the numbers"


Decompose an integer into prime factors.

bash$ factor 27417
27417: 3 13 19 37


Bash can't handle floating point calculations, and it lacks operators for certain important mathematical functions. Fortunately, bc comes to the rescue.

Not just a versatile, arbitrary precision calculation utility, bc offers many of the facilities of a programming language.

bc has a syntax vaguely resembling C.

Since it is a fairly well-behaved UNIX utility, and may therefore be used in a pipe, bc comes in handy in scripts.

Here is a simple template for using bc to calculate a script variable. This uses command substitution.

	      variable=$(echo "OPTIONS; OPERATIONS" | bc)

Example 12-38. Monthly Payment on a Mortgage

# Calculates monthly payment on a mortgage.

#  This is a modification of code in the "mcalc" (mortgage calculator) package,
#+ by Jeff Schmidt and Mendel Cooper (yours truly, the author of this document).
#  [15k]

echo "Given the principal, interest rate, and term of a mortgage,"
echo "calculate the monthly payment."


echo -n "Enter principal (no commas) "
read principal
echo -n "Enter interest rate (percent) "  # If 12%, enter "12", not ".12".
read interest_r
echo -n "Enter term (months) "
read term

 interest_r=$(echo "scale=9; $interest_r/100.0" | bc) # Convert to decimal.
                 # "scale" determines how many decimal places.

 interest_rate=$(echo "scale=9; $interest_r/12 + 1.0" | bc)

 top=$(echo "scale=9; $principal*$interest_rate^$term" | bc)

 echo; echo "Please be patient. This may take a while."

 let "months = $term - 1"
# ==================================================================== 
 for ((x=$months; x > 0; x--))
   bot=$(echo "scale=9; $interest_rate^$x" | bc)
   bottom=$(echo "scale=9; $bottom+$bot" | bc)
#  bottom = $(($bottom + $bot"))
# -------------------------------------------------------------------- 
#  Rick Boivie pointed out a more efficient implementation
#+ of the above loop, which decreases computation time by 2/3.

# for ((x=1; x <= $months; x++))
# do
#   bottom=$(echo "scale=9; $bottom * $interest_rate + 1" | bc)
# done

#  And then he came up with an even more efficient alternative,
#+ one that cuts down the run time by about 95%!

# bottom=`{
#     echo "scale=9; bottom=$bottom; interest_rate=$interest_rate"
#     for ((x=1; x <= $months; x++))
#     do
#          echo 'bottom = bottom * interest_rate + 1'
#     done
#     echo 'bottom'
#     } | bc`       # Embeds a 'for loop' within command substitution.

# ==================================================================== 

 # let "payment = $top/$bottom"
 payment=$(echo "scale=2; $top/$bottom" | bc)
 # Use two decimal places for dollars and cents.
 echo "monthly payment = \$$payment"  # Echo a dollar sign in front of amount.

 exit 0

 # Exercises:
 #   1) Filter input to permit commas in principal amount.
 #   2) Filter input to permit interest to be entered as percent or decimal.
 #   3) If you are really ambitious,
 #      expand this script to print complete amortization tables.

Example 12-39. Base Conversion

# Shellscript: - print number to different bases (Bourne Shell)
# Author     :	Heiner Steven (
# Date       :	07-03-95
# Category   :	Desktop
# $Id:,v 1.2 2000/02/06 19:55:35 heiner Exp $
# Description
# Changes
# 21-03-95 stv	fixed error occuring with 0xb as input (0.2)

# ==> Used in this document with the script author's permission.
# ==> Comments added by document author.

PN=`basename "$0"`			       # Program name
VER=`echo '$Revision: 1.2 $' | cut -d' ' -f2`  # ==> VER=1.2

Usage () {
    echo "$PN - print number to different bases, $VER (stv '95)
usage: $PN [number ...]

If no number is given, the numbers are read from standard input.
A number may be
    binary (base 2)		starting with 0b (i.e. 0b1100)
    octal (base 8)		starting with 0  (i.e. 014)
    hexadecimal (base 16)	starting with 0x (i.e. 0xc)
    decimal			otherwise (i.e. 12)" >&2
    exit $NOARGS 
}   # ==> Function to print usage message.

Msg () {
    for i   # ==> in [list] missing.
    do echo "$PN: $i" >&2

Fatal () { Msg "$@"; exit 66; }

PrintBases () {
    # Determine base of the number
    for i      # ==> in [list] missing...
    do         # ==> so operates on command line arg(s).
	case "$i" in
	    0b*)		ibase=2;;	# binary
	    0x*|[a-f]*|[A-F]*)	ibase=16;;	# hexadecimal
	    0*)			ibase=8;;	# octal
	    [1-9]*)		ibase=10;;	# decimal
		Msg "illegal number $i - ignored"

	# Remove prefix, convert hex digits to uppercase (bc needs this)
	number=`echo "$i" | sed -e 's:^0[bBxX]::' | tr '[a-f]' '[A-F]'`
	# ==> Uses ":" as sed separator, rather than "/".

	# Convert number to decimal
	dec=`echo "ibase=$ibase; $number" | bc`  # ==> 'bc' is calculator utility.
	case "$dec" in
	    [0-9]*)	;;			 # number ok
	    *)		continue;;		 # error: ignore

	# Print all conversions in one line.
	# ==> 'here document' feeds command list to 'bc'.
	echo `bc <<!
	    obase=16; "hex="; $dec
	    obase=10; "dec="; $dec
	    obase=8;  "oct="; $dec
	    obase=2;  "bin="; $dec
    ` | sed -e 's: :	:g'


while [ $# -gt 0 ]
# ==>  Is a "while loop" really necessary here,
# ==>+ since all the cases either break out of the loop
# ==>+ or terminate the script.
# ==> (Thanks, Paulo Marcel Coelho Aragao.)
    case "$1" in
	--)     shift; break;;
	-h)     Usage;;                 # ==> Help message.
	-*)     Usage;;
         *)     break;;			# first number
    esac   # ==> More error checking for illegal input might be useful.

if [ $# -gt 0 ]
    PrintBases "$@"
else					# read from stdin
    while read line
	PrintBases $line

exit 0

An alternate method of invoking bc involves using a here document embedded within a command substitution block. This is especially appropriate when a script needs to pass a list of options and commands to bc.

variable=`bc << LIMIT_STRING


variable=$(bc << LIMIT_STRING

Example 12-40. Invoking bc using a "here document"

# Invoking 'bc' using command substitution
# in combination with a 'here document'.

var1=`bc << EOF
18.33 * 19.78
echo $var1       # 362.56

#  $( ... ) notation also works.

var2=$(bc << EOF
scale = 4
a = ( $v1 + $v2 )
b = ( $v3 * $v4 )
a * b + 15.35
echo $var2       # 593487.8452

var3=$(bc -l << EOF
scale = 9
s ( 1.7 )
# Returns the sine of 1.7 radians.
# The "-l" option calls the 'bc' math library.
echo $var3       # .991664810

# Now, try it in a function...
hyp=             # Declare global variable.
hypotenuse ()    # Calculate hypotenuse of a right triangle.
hyp=$(bc -l << EOF
scale = 9
sqrt ( $1 * $1 + $2 * $2 )
# Unfortunately, can't return floating point values from a Bash function.

hypotenuse 3.68 7.31
echo "hypotenuse = $hyp"    # 8.184039344

exit 0

Example 12-41. Calculating PI

# Approximating PI by firing cannonballs.

# This is a very simple instance of a "Monte Carlo" simulation,
#+ a mathematical model of a real-life event,
#+ using pseudorandom numbers to emulate random chance.

#  Consider a perfectly square plot of land, 10000 units on a side.
#  This land has a perfectly circular lake in its center,
#+ with a diameter of 10000 units.
#  The plot is actually all water, except for the four corners.
#  (Think of it as a square with an inscribed circle.)
#  Let us fire iron cannonballs from an old-style cannon
#+ at the square of land.
#  All the shots impact somewhere on the plot of land,
#+ either in the lake or on the dry corners.
#  Since the lake takes up most of the land area,
#+ most of the shots will SPLASH! into the water.
#  Just a few shots will THUD! into solid ground
#+ in the four corners of the land.
#  If we take enough random, unaimed shots at the plot of land,
#+ Then the ratio of SPLASHES to total shots will approximate
#+ the value of PI/4.
#  The reason for this is that the cannon is actually shooting
#+ only at the upper right-hand quadrant of the square,
#+ i.e., Quadrant I of the Cartesian coordinate plane.
#  (The previous explanation was a simplification.)
#  Theoretically, the more shots taken, the better the fit.
#  However, a shell script, as opposed to a compiled language
#+ with floating-point math built in, requires a few compromises.
#  This tends to lower the accuracy of the simulation, unfortunately.

DIMENSION=10000  # Length of each side of the plot of land.
                 # Also sets ceiling for random integers generated.

MAXSHOTS=1000    # Fire this many shots.
                 # 10000 or more would be better, but would take too long.
PMULTIPLIER=4.0  # Scaling factor to approximate PI.

get_random ()
SEED=$(head -1 /dev/urandom | od -N 1 | awk '{ print $2 }')
RANDOM=$SEED                                  #  From ""
                                              #+ example script.
let "rnum = $RANDOM % $DIMENSION"             #  Range less than 10000.
echo $rnum

distance=        # Declare global variable.
hypotenuse ()    # Calculate hypotenuse of a right triangle.
{                # From "" example.
distance=$(bc -l << EOF
scale = 0
sqrt ( $1 * $1 + $2 * $2 )
#  Setting "scale" to zero rounds down result to integer value,
#+ a necessary compromise in this script.
#  This diminshes the accuracy of the simulation, unfortunately.

# main() {

# Initialize variables.

while [ "$shots" -lt  "$MAXSHOTS" ]           # Main loop.

  xCoord=$(get_random)                        # Get random X and Y coords.
  hypotenuse $xCoord $yCoord                  #  Hypotenuse of right-triangle =
                                              #+ distance.

  printf "#%4d   " $shots
  printf "Xc = %4d  " $xCoord
  printf "Yc = %4d  " $yCoord
  printf "Distance = %5d  " $distance         #  Distance from 
                                              #+ center of lake,
                                              #  the "origin,"
                                              #+ coordinate (0,0).

  if [ "$distance" -le "$DIMENSION" ]
    echo -n "SPLASH!  "
    echo -n "THUD!    "

  Pi=$(echo "scale=9; $PMULTIPLIER*$splashes/$shots" | bc)
  # Multiply ratio by 4.0.
  echo -n "PI ~ $Pi"


echo "After $shots shots, PI looks like approximately $Pi."
# Tends to run a bit high . . . 
# Probably due to round-off error and imperfect randomness of $RANDOM.

# }

exit 0

#  One might well wonder whether a shell script is appropriate for
#+ an application as complex and computation-intensive as a simulation.
#  There are at least two justifications.
#  1) As a proof of concept: to show it can be done.
#  2) To prototype and test the algorithms before rewriting
#+    it in a compiled high-level language.

The dc (desk calculator) utility is stack-oriented and uses RPN ("Reverse Polish Notation"). Like bc, it has much of the power of a programming language.

Most persons avoid dc, since it requires non-intuitive RPN input. Yet it has its uses.

Example 12-42. Converting a decimal number to hexadecimal

# Convert a decimal number to hexadecimal.

BASE=16     # Hexadecimal.

if [ -z "$1" ]
  echo "Usage: $0 number"
  exit $E_NOARGS
  # Need a command line argument.
# Exercise: add argument validity checking.

hexcvt ()
if [ -z "$1" ]
  echo 0
  return    # "Return" 0 if no arg passed to function.

echo ""$1" "$BASE" o p" | dc
#                 "o" sets radix (numerical base) of output.
#                   "p" prints the top of stack.
# See 'man dc' for other options.

hexcvt "$1"

exit 0

Studying the info page for dc gives some insight into its intricacies. However, there seems to be a small, select group of dc wizards who delight in showing off their mastery of this powerful, but arcane utility.

Example 12-43. Factoring

# Factor a number

MIN=2       # Will not work for number smaller than this.

if [ -z $1 ]
  echo "Usage: $0 number"
  exit $E_NOARGS

if [ "$1" -lt "$MIN" ]
  echo "Number to factor must be $MIN or greater."
  exit $E_TOOSMALL

# Exercise: Add type checking (to reject non-integer arg).

echo "Factors of $1:"
# ---------------------------------------------------------------------------------
echo "$1[p]s2[lip/dli%0=1dvsr]s12sid2%0=13sidvsr[dli%0=1lrli2+dsi!>.]ds.xd1<2" | dc
# ---------------------------------------------------------------------------------
# Above line of code written by Michel Charpentier <>.
# Used with permission (thanks).

 exit 0

Yet another way of doing floating point math in a script is using awk's built-in math functions in a shell wrapper.

Example 12-44. Calculating the hypotenuse of a triangle

# Returns the "hypotenuse" of a right triangle.
#               ( square root of sum of squares of the "legs")

ARGS=2                # Script needs sides of triangle passed.
E_BADARGS=65          # Wrong number of arguments.

if [ $# -ne "$ARGS" ] # Test number of arguments to script.
  echo "Usage: `basename $0` side_1 side_2"
  exit $E_BADARGS

AWKSCRIPT=' { printf( "%3.7f\n", sqrt($1*$1 + $2*$2) ) } '
#            command(s) / parameters passed to awk

echo -n "Hypotenuse of $1 and $2 = "
echo $1 $2 | awk "$AWKSCRIPT"

exit 0

12.9. Miscellaneous Commands

Command that fit in no special category

jot, seq

These utilities emit a sequence of integers, with a user-selected increment.

The normal separator character between each integer is a newline, but this can be changed with the -s option.

bash$ seq 5

bash$ seq -s : 5

Both jot and seq come in handy in a for loop.

Example 12-45. Using seq to generate loop arguments

# Using "seq"


for a in `seq 80`  # or   for a in $( seq 80 )
# Same as   for a in 1 2 3 4 5 ... 80   (saves much typing!).
# May also use 'jot' (if present on system).
  echo -n "$a "
done      # 1 2 3 4 5 ... 80
# Example of using the output of a command to generate 
# the [list] in a "for" loop.

echo; echo

COUNT=80  # Yes, 'seq' may also take a replaceable parameter.

for a in `seq $COUNT`  # or   for a in $( seq $COUNT )
  echo -n "$a "
done      # 1 2 3 4 5 ... 80

echo; echo


for a in `seq $BEGIN $END`
#  Giving "seq" two arguments starts the count at the first one,
#+ and continues until it reaches the second.
  echo -n "$a "
done      # 75 76 77 78 79 80

echo; echo


for a in `seq $BEGIN $INTERVAL $END`
#  Giving "seq" three arguments starts the count at the first one,
#+ uses the second for a step interval,
#+ and continues until it reaches the third.
  echo -n "$a "
done      # 45 50 55 60 65 70 75 80

echo; echo

exit 0

The getopt command parses command-line options preceded by a dash. This external command corresponds to the getopts Bash builtin. The getopt permits handling long options by use of the -l flag, and it also allows parameter reshuffling.

Example 12-46. Using getopt to parse command-line options

# Using getopt.

# Try the following when invoking this script:
#   sh -a
#   sh -abc
#   sh -a -b -c
#   sh -d
#   sh -dXYZ
#   sh -d XYZ
#   sh -abcd
#   sh -abcdZ
#   sh -z
#   sh a
# Explain the results of each of the above.


if [ "$#" -eq 0 ]
then   # Script needs at least one command-line argument.
  echo "Usage $0 -[options a,b,c]"
  exit $E_OPTERR

set -- `getopt "abcd:" "$@"`
# Sets positional parameters to command-line arguments.
# What happens if you use "$*" instead of "$@"?

while [ ! -z "$1" ]
  case "$1" in
    -a) echo "Option \"a\"";;
    -b) echo "Option \"b\"";;
    -c) echo "Option \"c\"";;
    -d) echo "Option \"d\" $2";;
     *) break;;


#  It is usually better to use the 'getopts' builtin in a script,
#+ rather than 'getopt'.
#  See "".

exit 0

The run-parts command [34] executes all the scripts in a target directory, sequentially in ASCII-sorted filename order. Of course, the scripts need to have execute permission.

The cron daemon invokes run-parts to run the scripts in the /etc/cron.* directories.


In its default behavior the yes command feeds a continuous string of the character y followed by a line feed to stdout. A control-c terminates the run. A different output string may be specified, as in yes different string, which would continually output different string to stdout. One might well ask the purpose of this. From the command line or in a script, the output of yes can be redirected or piped into a program expecting user input. In effect, this becomes a sort of poor man's version of expect.

yes | fsck /dev/hda1 runs fsck non-interactively (careful!).

yes | rm -r dirname has same effect as rm -rf dirname (careful!).


Caution advised when piping yes to a potentially dangerous system command, such as fsck or fdisk. It may have unintended side-effects.


Prints arguments as a large vertical banner to stdout, using an ASCII character (default '#'). This may be redirected to a printer for hardcopy.


Show all the environmental variables set for a particular user.

bash$ printenv | grep HOME


The lp and lpr commands send file(s) to the print queue, to be printed as hard copy. [35] These commands trace the origin of their names to the line printers of another era.

bash$ lp file1.txt or bash lp <file1.txt

It is often useful to pipe the formatted output from pr to lp.

bash$ pr -options file1.txt | lp

Formatting packages, such as groff and Ghostscript may send their output directly to lp.

bash$ groff -Tascii | lp

bash$ gs -options | lp

Related commands are lpq, for viewing the print queue, and lprm, for removing jobs from the print queue.


[UNIX borrows an idea here from the plumbing trade.]

This is a redirection operator, but with a difference. Like the plumber's "tee," it permits "siponing off" to a file the output of a command or commands within a pipe, but without affecting the result. This is useful for printing an ongoing process to a file or paper, perhaps to keep track of it for debugging purposes.

                 |------> to file
  command--->----|-operator-->---> result of command(s)

cat listfile* | sort | tee check.file | uniq > result.file
(The file check.file contains the concatenated sorted "listfiles", before the duplicate lines are removed by uniq.)


This obscure command creates a named pipe, a temporary first-in-first-out buffer for transferring data between processes. [36] Typically, one process writes to the FIFO, and the other reads from it. See Example A-16.


This command checks the validity of a filename. If the filename exceeds the maximum allowable length (255 characters) or one or more of the directories in its path is not searchable, then an error message results.

Unfortunately, pathchk does not return a recognizable error code, and it is therefore pretty much useless in a script. Consider instead the file test operators.


This is the somewhat obscure and much feared "data duplicator" command. Originally a utility for exchanging data on magnetic tapes between UNIX minicomputers and IBM mainframes, this command still has its uses. The dd command simply copies a file (or stdin/stdout), but with conversions. Possible conversions are ASCII/EBCDIC, [37] upper/lower case, swapping of byte pairs between input and output, and skipping and/or truncating the head or tail of the input file. A dd --help lists the conversion and other options that this powerful utility takes.

Example 12-47. A script that copies itself


# This script copies itself.


dd if=$0 of=$0.$file_subscript 2>/dev/null
# Suppress messages from dd:   ^^^^^^^^^^^

exit $?

Example 12-48. Exercising dd


# Script by Stephane Chazelas.
# Somewhat modified by document author.

input_file=$0   # This script.

dd if=$input_file of=$output_file bs=1 skip=$((n-1)) count=$((p-n+1)) 2> /dev/null
# Extracts characters n to p from this script.

# -------------------------------------------------------

echo -n "hello world" | dd cbs=1 conv=unblock 2> /dev/null
# Echoes "hello world" vertically.

exit 0

To demonstrate just how versatile dd is, let's use it to capture keystrokes.

Example 12-49. Capturing Keystrokes

# Capture keystrokes without needing to press ENTER.

keypresses=4                      # Number of keypresses to capture.

old_tty_setting=$(stty -g)        # Save old terminal settings.

echo "Press $keypresses keys."
stty -icanon -echo                # Disable canonical mode.
                                  # Disable local echo.
keys=$(dd bs=1 count=$keypresses 2> /dev/null)
# 'dd' uses stdin, if "if" not specified.

stty "$old_tty_setting"           # Restore old terminal settings.

echo "You pressed the \"$keys\" keys."

# Thanks, S.C. for showing the way.
exit 0

The dd command can do random access on a data stream.
echo -n . | dd bs=1 seek=4 of=file conv=notrunc
# The "conv=notrunc" option means that the output file will not be truncated.		

# Thanks, S.C.

The dd command can copy raw data and disk images to and from devices, such as floppies and tape drives (Example A-6). A common use is creating boot floppies.

dd if=kernel-image of=/dev/fd0H1440

Similarly, dd can copy the entire contents of a floppy, even one formatted with a "foreign" OS, to the hard drive as an image file.

dd if=/dev/fd0 of=/home/bozo/projects/floppy.img

Other applications of dd include initializing temporary swap files (Example 29-2) and ramdisks (Example 29-3). It can even do a low-level copy of an entire hard drive partition, although this is not necessarily recommended.

People (with presumably nothing better to do with their time) are constantly thinking of interesting applications of dd.

Example 12-50. Securely deleting a file

# Erase all traces of a file.

#  This script overwrites a target file alternately
#+ with random bytes, then zeros before finally deleting it.
#  After that, even examining the raw disk sectors
#+ will not reveal the original file data.

PASSES=7         #  Number of file-shredding passes.
                 #  Increasing this slows script execution,
                 #+ especially on large target files.
BLOCKSIZE=1      #  I/O with /dev/urandom requires unit block size,
                 #+ otherwise you get weird results.
E_BADARGS=70     #  Various error exit codes.

if [ -z "$1" ]   # No filename specified.
  echo "Usage: `basename $0` filename"
  exit $E_BADARGS


if [ ! -e "$file" ]
  echo "File \"$file\" not found."
  exit $E_NOT_FOUND

echo; echo -n "Are you absolutely sure you want to blot out \"$file\" (y/n)? "
read answer
case "$answer" in
[nN]) echo "Changed your mind, huh?"
      exit $E_CHANGED_MIND
*)    echo "Blotting out file \"$file\".";;

flength=$(ls -l "$file" | awk '{print $5}')  # Field 5 is file length.

chmod u+w "$file"   # Allow overwriting/deleting the file.


while [ "$pass_count" -le "$PASSES" ]
  echo "Pass #$pass_count"
  sync         # Flush buffers.
  dd if=/dev/urandom of=$file bs=$BLOCKSIZE count=$flength
               # Fill with random bytes.
  sync         # Flush buffers again.
  dd if=/dev/zero of=$file bs=$BLOCKSIZE count=$flength
               # Fill with zeros.
  sync         # Flush buffers yet again.
  let "pass_count += 1"

rm -f $file    # Finally, delete scrambled and shredded file.
sync           # Flush buffers a final time.

echo "File \"$file\" blotted out and deleted."; echo

exit 0

#  This is a fairly secure, if inefficient and slow method
#+ of thoroughly "shredding" a file.
#  The "shred" command, part of the GNU "fileutils" package,
#+ does the same thing, although more efficiently.

#  The file cannot not be "undeleted" or retrieved by normal methods.
#  However . . .
#+ this simple method would *not* likely withstand forensic analysis.

#  This script may not play well with a journaled file system.
#  Exercise: Fix it so it does.

#  Tom Vier's "wipe" file-deletion package does a much more thorough job
#+ of file shredding than this simple script.

#  For an in-depth analysis on the topic of file deletion and security,
#+ see Peter Gutmann's paper,
#+     "Secure Deletion of Data From Magnetic and Solid-State Memory".

The od, or octal dump filter converts input (or files) to octal (base-8) or other bases. This is useful for viewing or processing binary data files or otherwise unreadable system device files, such as /dev/urandom, and as a filter for binary data. See Example 9-27 and Example 12-13.


Performs a hexadecimal, octal, decimal, or ASCII dump of a binary file. This command is the rough equivalent of od, above, but not nearly as useful.


Displays information about an object file or binary executable in either hexadecimal form or as a disassembled listing (with the -d option).

bash$ objdump -d /bin/ls
/bin/ls:     file format elf32-i386

 Disassembly of section .init:

 080490bc <.init>:
  80490bc:       55                      push   %ebp
  80490bd:       89 e5                   mov    %esp,%ebp
  . . .


This command generates a "magic cookie", a 128-bit (32-character) pseudorandom hexadecimal number, normally used as an authorization "signature" by the X server. This also available for use in a script as a "quick 'n dirty" random number.

Of course, a script could use md5 for the same purpose.
# Generate md5 checksum on the script itself.
random001=`md5sum $0 | awk '{print $1}'`
# Uses 'awk' to strip off the filename.

The mcookie command gives yet another way to generate a "unique" filename.

Example 12-51. Filename generator

#  temp filename generator

BASE_STR=`mcookie`   # 32-character magic cookie.
POS=11               # Arbitrary position in magic cookie string.
LEN=5                # Get $LEN consecutive characters.

prefix=temp          #  This is, after all, a "temp" file.
                     #  For more "uniqueness," generate the filename prefix
                     #+ using the same method as the suffix, below.

                     # Extract a 5-character string, starting at position 11.

                     # Construct the filename.

echo "Temp filename = "$temp_filename""

# sh
# Temp filename = temp.e19ea

exit 0

This utility converts between different units of measure. While normally invoked in interactive mode, units may find use in a script.

Example 12-52. Converting meters to miles


convert_units ()  # Takes as arguments the units to convert.
  cf=$(units "$1" "$2" | sed --silent -e '1p' | awk '{print $2}')
  # Strip off everything except the actual conversion factor.
  echo "$cf"

cfactor=`convert_units $Unit1 $Unit2`

result=$(echo $quantity*$cfactor | bc)

echo "There are $result $Unit2 in $quantity $Unit1."

#  What happens if you pass incompatible units,
#+ such as "acres" and "miles" to the function?

exit 0

A hidden treasure, m4 is a powerful macro processing filter, [38] virtually a complete language. Although originally written as a pre-processor for RatFor, m4 turned out to be useful as a stand-alone utility. In fact, m4 combines some of the functionality of eval, tr, and awk, in addition to its extensive macro expansion facilities.

The April, 2002 issue of Linux Journal has a very nice article on m4 and its uses.

Example 12-53. Using m4

# Using the m4 macro processor

# Strings
echo "len($string)" | m4                           # 7
echo "substr($string,4)" | m4                      # A01
echo "regexp($string,[0-1][0-1],\&Z)" | m4         # 01Z

# Arithmetic
echo "incr(22)" | m4                               # 23
echo "eval(99 / 3)" | m4                           # 33

exit 0

The doexec command enables passing an arbitrary list of arguments to a binary executable. In particular, passing argv[0] (which corresponds to $0 in a script) lets the executable be invoked by various names, and it can then carry out different sets of actions, according to the name by which it was called. What this amounts to is roundabout way of passing options to an executable.

For example, the /usr/local/bin directory might contain a binary called "aaa". Invoking doexec /usr/local/bin/aaa list would list all those files in the current working directory beginning with an "a", while invoking (the same executable with) doexec /usr/local/bin/aaa delete would delete those files.


The various behaviors of the executable must be defined within the code of the executable itself, analogous to something like the following in a shell script:
case `basename $0` in
"name1" ) do_something;;
"name2" ) do_something_else;;
"name3" ) do_yet_another_thing;;
*       ) bail_out;;


The dialog family of tools provide a method of calling interactive "dialog" boxes from a script. The more elaborate variations of dialog -- gdialog, Xdialog, and kdialog -- actually invoke X-Windows widgets. See Example 34-16.


The sox, or "sound exchange" command plays and performs transformations on sound files.

For example, sox soundfile.wav changes a WAV sound file into a (Sun audio format) AU sound file.

Shell scripts are ideally suited for batch processing sox operations on sound files.

Chapter 13. System and Administrative Commands

The startup and shutdown scripts in /etc/rc.d illustrate the uses (and usefulness) of many of these comands. These are usually invoked by root and used for system maintenance or emergency filesystem repairs. Use with caution, as some of these commands may damage your system if misused.

Users and Groups


Show all logged on users. This is the approximate equivalent of who -q.


Lists the current user and the groups she belongs to. This corresponds to the $GROUPS internal variable, but gives the group names, rather than the numbers.

bash$ groups
bozita cdrom cdwriter audio xgrp

bash$ echo $GROUPS
chown, chgrp

The chown command changes the ownership of a file or files. This command is a useful method that root can use to shift file ownership from one user to another. An ordinary user may not change the ownership of files, not even her own files. [39]

root# chown bozo *.txt


The chgrp command changes the group ownership of a file or files. You must be owner of the file(s) as well as a member of the destination group (or root) to use this operation.
chgrp --recursive dunderheads *.data
#  The "dunderheads" group will now own all the "*.data" files
#+ all the way down the $PWD directory tree (that's what "recursive" means).

useradd, userdel

The useradd administrative command adds a user account to the system and creates a home directory for that particular user, if so specified. The corresponding userdel command removes a user account from the system [40] and deletes associated files.


The adduser command is a synonym for useradd and is usually a symbolic link to it.


Modify a user account. Changes may be made to the password, group membership, expiration date, and other attributes of a given user's account. With this command, a user's password may be locked, which has the effect of disabling the account.


Modify a given group. The group name and/or ID number may be changed using this command.


The id command lists the real and effective user IDs and the group IDs of the user associated with the current process. This is the counterpart to the $UID, $EUID, and $GROUPS internal Bash variables.

bash$ id
uid=501(bozo) gid=501(bozo) groups=501(bozo),22(cdrom),80(cdwriter),81(audio)

bash$ echo $UID


The id command shows the effective IDs only when they differ from the real ones.

Also see Example 9-5.


Show all users logged on to the system.

bash$ who
bozo  tty1     Apr 27 17:45
 bozo  pts/0    Apr 27 17:46
 bozo  pts/1    Apr 27 17:47
 bozo  pts/2    Apr 27 17:49

The -m gives detailed information about only the current user. Passing any two arguments to who is the equivalent of who -m, as in who am i or who The Man.

bash$ who -m
localhost.localdomain!bozo  pts/2    Apr 27 17:49

whoami is similar to who -m, but only lists the user name.

bash$ whoami


Show all logged on users and the processes belonging to them. This is an extended version of who. The output of w may be piped to grep to find a specific user and/or process.

bash$ w | grep startx
bozo  tty1     -                 4:22pm  6:41   4.47s  0.45s  startx

Show current user's login name (as found in /var/run/utmp). This is a near-equivalent to whoami, above.

bash$ logname

bash$ whoami


bash$ su
Password: ......

bash# whoami
bash# logname


While logname prints the name of the logged in user, whoami gives the name of the user attached to the current process. As we have just seen, sometimes these are not the same.


Runs a program or script as a substitute user. su rjones starts a shell as user rjones. A naked su defaults to root. See Example A-16.


Runs a command as root (or another user). This may be used in a script, thus permitting a regular user to run the script.


# Some commands.
sudo cp /root/secretfile /home/bozo/secret
# Some more commands.

The file /etc/sudoers holds the names of users permitted to invoke sudo.


Sets, changes, or manages a user's password.

The passwd command can be used in a script, but should not be.

Example 13-1. Setting a new password

# For demonstration purposes only.
#                      Not a good idea to actually run this script.
#  This script must be run as root.

ROOT_UID=0         # Root has $UID 0.
E_WRONG_USER=65    # Not root?


if [ "$UID" -ne "$ROOT_UID" ]
  echo; echo "Only root can run this script."; echo
  exit $E_WRONG_USER
  echo "You should know better than to run this script, root."
  echo "Even root users get the blues... "


# Check if bozo lives here.
grep -q "$username" /etc/passwd
if [ $? -ne $SUCCESS ]
  echo "User $username does not exist."
  echo "No password changed."

echo "$NEWPASSWORD" | passwd --stdin "$username"
#  The '--stdin' option to 'passwd' permits
#+ getting a new password from stdin (or a pipe).

echo; echo "User $username's password changed!"

# Using the 'passwd' command in a script is dangerous.

exit 0

The passwd command's -l, -u, and -d options permit locking, unlocking, and deleting a user's password. Only root may use these options.


Show users' logged in time, as read from /var/log/wtmp. This is one of the GNU accounting utilities.

bash$ ac
        total       68.08

List last logged in users, as read from /var/log/wtmp. This command can also show remote logins.


Change user's group ID without logging out. This permits access to the new group's files. Since users may be members of multiple groups simultaneously, this command finds little use.



Echoes the name of the current user's terminal. Note that each separate xterm window counts as a different terminal.

bash$ tty

Shows and/or changes terminal settings. This complex command, used in a script, can control terminal behavior and the way output displays. See the info page, and study it carefully.

Example 13-2. Setting an erase character

# Using "stty" to set an erase character when reading input.

echo -n "What is your name? "
read name                      #  Try to backspace
                               #+ to erase characters of input.
                               #  Problems?
echo "Your name is $name."

stty erase '#'                 #  Set "hashmark" (#) as erase character.
echo -n "What is your name? "
read name                      #  Use # to erase last character typed.
echo "Your name is $name."

# Warning: Even after the script exits, the new key value remains set.

exit 0

Example 13-3. secret password: Turning off terminal echoing


echo -n "Enter password "
read passwd
echo "password is $passwd"
echo -n "If someone had been looking over your shoulder, "
echo "your password would have been compromised."

echo && echo  # Two line-feeds in an "and list".

stty -echo    # Turns off screen echo.

echo -n "Enter password again "
read passwd
echo "password is $passwd"

stty echo     # Restores screen echo.

exit 0

A creative use of stty is detecting a user keypress (without hitting ENTER).

Example 13-4. Keypress detection

# Detect a user keypress ("hot keyboard").


old_tty_settings=$(stty -g)   # Save old settings.
stty -icanon
Keypress=$(head -c1)          # or $(dd bs=1 count=1 2> /dev/null)
                              # on non-GNU systems

echo "Key pressed was \""$Keypress"\"."

stty "$old_tty_settings"      # Restore old settings.

# Thanks, Stephane Chazelas.

exit 0

Also see Example 9-3.


Set certain terminal attributes. This command writes to its terminal's stdout a string that changes the behavior of that terminal.

bash$ setterm -cursor off

The setterm command can be used within a script to change the appearance of text written to stdout, although there are certainly better tools available for this purpose.

setterm -bold on
echo bold hello

setterm -bold off
echo normal hello


Show or initialize terminal settings. This is a less capable version of stty.

bash$ tset -r
Terminal type is xterm-xfree86.
 Kill is control-U (^U).
 Interrupt is control-C (^C).


Set or display serial port parameters. This command must be run by root user and is usually found in a system setup script.

# From /etc/pcmcia/serial script:

IRQ=`setserial /dev/$DEVICE | sed -e 's/.*IRQ: //'`
setserial /dev/$DEVICE irq 0 ; setserial /dev/$DEVICE irq $IRQ

getty, agetty

The initialization process for a terminal uses getty or agetty to set it up for login by a user. These commands are not used within user shell scripts. Their scripting counterpart is stty.


Enables or disables write access to the current user's terminal. Disabling access would prevent another user on the network to write to the terminal.


It can be very annoying to have a message about ordering pizza suddenly appear in the middle of the text file you are editing. On a multi-user network, you might therefore wish to disable write access to your terminal when you need to avoid interruptions.


This is an acronym for "write all", i.e., sending a message to all users at every terminal logged into the network. It is primarily a system administrator's tool, useful, for example, when warning everyone that the system will shortly go down due to a problem (see Example 17-2).

bash$ wall System going down for maintenance in 5 minutes!
Broadcast message from bozo (pts/1) Sun Jul  8 13:53:27 2001...

 System going down for maintenance in 5 minutes!


If write access to a particular terminal has been disabled with mesg, then wall cannot send a message to it.


Lists all system bootup messages to stdout. Handy for debugging and ascertaining which device drivers were installed and which system interrupts in use. The output of dmesg may, of course, be parsed with grep, sed, or awk from within a script.

bash$ dmesg | grep hda
Kernel command line: ro root=/dev/hda2
 hda: IBM-DLGA-23080, ATA DISK drive
 hda: 6015744 sectors (3080 MB) w/96KiB Cache, CHS=746/128/63
 hda: hda1 hda2 hda3 < hda5 hda6 hda7 > hda4

Information and Statistics


Output system specifications (OS, kernel version, etc.) to stdout. Invoked with the -a option, gives verbose system info (see Example 12-5). The -s option shows only the OS type.

bash$ uname -a
Linux localhost.localdomain 2.2.15-2.5.0 #1 Sat Feb 5 00:13:43 EST 2000 i686 unknown

bash$ uname -s

Show system architecture. Equivalent to uname -m. See Example 10-26.

bash$ arch

bash$ uname -m

Gives information about previous commands, as stored in the /var/account/pacct file. Command name and user name can be specified by options. This is one of the GNU accounting utilities.


List the last login time of all system users. This references the /var/log/lastlog file.

bash$ lastlog
root          tty1                      Fri Dec  7 18:43:21 -0700 2001
 bin                                     **Never logged in**
 daemon                                  **Never logged in**
 bozo          tty1                      Sat Dec  8 21:14:29 -0700 2001

bash$ lastlog | grep root
root          tty1                      Fri Dec  7 18:43:21 -0700 2001


This command will fail if the user invoking it does not have read permission for the /var/log/lastlog file.


List open files. This command outputs a detailed table of all currently open files and gives information about their owner, size, the processes associated with them, and more. Of course, lsof may be piped to grep and/or awk to parse and analyze its results.

bash$ lsof
 init         1    root  mem    REG        3,5   30748    30303 /sbin/init
 init         1    root  mem    REG        3,5   73120     8069 /lib/
 init         1    root  mem    REG        3,5  931668     8075 /lib/
 cardmgr    213    root  mem    REG        3,5   36956    30357 /sbin/cardmgr


Diagnostic and debugging tool for tracing system calls and signals. The simplest way of invoking it is strace COMMAND.

bash$ strace df
execve("/bin/df", ["df"], [/* 45 vars */]) = 0
 uname({sys="Linux", node="bozo.localdomain", ...}) = 0
 brk(0)                                  = 0x804f5e4

This is the Linux equivalent of truss.


Network port scanner. This command scans a server to locate open ports and the services associated with those ports. It is an important security tool for locking down a network against hacking attempts.


SERVER=$HOST                           # localhost.localdomain (
PORT_NUMBER=25                         # SMTP port.

nmap $SERVER | grep -w "$PORT_NUMBER"  # Is that particular port open?
#              grep -w matches whole words only,
#+             so this wouldn't match port 1025, for example.

exit 0

# 25/tcp     open        smtp


Shows memory and cache usage in tabular form. The output of this command lends itself to parsing, using grep, awk or Perl. The procinfo command shows all the information that free does, and much more.

bash$ free
                total       used       free     shared    buffers     cached
   Mem:         30504      28624       1880      15820       1608       16376
   -/+ buffers/cache:      10640      19864
   Swap:        68540       3128      65412

To show unused RAM memory:

bash$ free | grep Mem | awk '{ print $4 }'

Extract and list information and statistics from the /proc pseudo-filesystem. This gives a very extensive and detailed listing.

bash$ procinfo | grep Bootup
Bootup: Wed Mar 21 15:15:50 2001    Load average: 0.04 0.21 0.34 3/47 6829

List devices, that is, show installed hardware.

bash$ lsdev
Device            DMA   IRQ  I/O Ports
 cascade             4     2 
 dma                          0080-008f
 dma1                         0000-001f
 dma2                         00c0-00df
 fpu                          00f0-00ff
 ide0                     14  01f0-01f7 03f6-03f6


Show (disk) file usage, recursively. Defaults to current working directory, unless otherwise specified.

bash$ du -ach
1.0k    ./
 1.0k    ./
 1.0k    ./random.file
 6.0k    .
 6.0k    total

Shows filesystem usage in tabular form.

bash$ df
Filesystem           1k-blocks      Used Available Use% Mounted on
 /dev/hda5               273262     92607    166547  36% /
 /dev/hda8               222525    123951     87085  59% /home
 /dev/hda7              1408796   1075744    261488  80% /usr

Gives detailed and verbose statistics on a given file (even a directory or device file) or set of files.

bash$ stat test.cru
  File: "test.cru"
   Size: 49970        Allocated Blocks: 100          Filetype: Regular File
   Mode: (0664/-rw-rw-r--)         Uid: (  501/ bozo)  Gid: (  501/ bozo)
 Device:  3,8   Inode: 18185     Links: 1    
 Access: Sat Jun  2 16:40:24 2001
 Modify: Sat Jun  2 16:40:24 2001
 Change: Sat Jun  2 16:40:24 2001

If the target file does not exist, stat returns an error message.

bash$ stat nonexistent-file
nonexistent-file: No such file or directory


Display virtual memory statistics.

bash$ vmstat
   procs                      memory    swap          io system         cpu
 r  b  w   swpd   free   buff  cache  si  so    bi    bo   in    cs  us  sy id
 0  0  0      0  11040   2636  38952   0   0    33     7  271    88   8   3 89


Show current network statistics and information, such as routing tables and active connections. This utility accesses information in /proc/net (Chapter 28). See Example 28-3.

netstat -r is equivalent to route.

bash$ netstat
Active Internet connections (w/o servers)
 Proto Recv-Q Send-Q Local Address           Foreign Address         State      
 Active UNIX domain sockets (w/o servers)
 Proto RefCnt Flags       Type       State         I-Node Path
 unix  11     [ ]         DGRAM                    906    /dev/log
 unix  3      [ ]         STREAM     CONNECTED     4514   /tmp/.X11-unix/X0
 unix  3      [ ]         STREAM     CONNECTED     4513
 . . .

Shows how long the system has been running, along with associated statistics.

bash$ uptime
10:28pm  up  1:57,  3 users,  load average: 0.17, 0.34, 0.27

Lists the system's host name. This command sets the host name in an /etc/rc.d setup script (/etc/rc.d/rc.sysinit or similar). It is equivalent to uname -n, and a counterpart to the $HOSTNAME internal variable.

bash$ hostname

bash$ echo $HOSTNAME

Similar to the hostname command are the domainname, dnsdomainname, nisdomainname, and ypdomainname commands. Use these to display or set the system DNS or NIS/YP domain name. Various options to hostname also perform these functions.


Echo a 32-bit hexadecimal numerical identifier for the host machine.

bash$ hostid


This command allegedly fetches a "unique" serial number for a particular system. Certain product registration procedures use this number to brand a particular user license. Unfortunately, hostid only returns the machine network address in hexadecimal, with pairs of bytes transposed.

The network address of a typical non-networked Linux machine, is found in /etc/hosts.

bash$ cat /etc/hosts               localhost.localdomain localhost

As it happens, transposing the bytes of, we get, which translates in hex to 007f0100, the exact equivalent of what hostid returns, above. There exist only a few million other Linux machines with this identical hostid.


Invoking sar (System Activity Reporter) gives a very detailed rundown on system statistics. The Santa Cruz Operation (SCO) released sar as Open Source in June, 1999.

This command is not part of the base Linux distribution, but may be obtained as part of the sysstat utilities package, written by Sebastien Godard.

bash$ sar
Linux 2.4.9 ( 	09/26/03

10:30:00          CPU     %user     %nice   %system   %iowait     %idle
10:40:00          all      2.21     10.90     65.48      0.00     21.41
10:50:00          all      3.36      0.00     72.36      0.00     24.28
11:00:00          all      1.12      0.00     80.77      0.00     18.11
Average:          all      2.23      3.63     72.87      0.00     21.27

14:32:30          LINUX RESTART

15:00:00          CPU     %user     %nice   %system   %iowait     %idle
15:10:00          all      8.59      2.40     17.47      0.00     71.54
15:20:00          all      4.07      1.00     11.95      0.00     82.98
15:30:00          all      0.79      2.94      7.56      0.00     88.71
Average:          all      6.33      1.70     14.71      0.00     77.26

Show information and statistics about a designated elf binary. This is part of the binutils package.

bash$ readelf -h /bin/bash
ELF Header:
   Magic:   7f 45 4c 46 01 01 01 00 00 00 00 00 00 00 00 00 
   Class:                             ELF32
   Data:                              2's complement, little endian
   Version:                           1 (current)
   OS/ABI:                            UNIX - System V
   ABI Version:                       0
   Type:                              EXEC (Executable file)
   . . .

The size [/path/to/binary] command gives the segment sizes of a binary executable or archive file. This is mainly of use to programmers.

bash$ size /bin/bash
   text    data     bss     dec     hex filename
  495971   22496   17392  535859   82d33 /bin/bash

System Logs


Appends a user-generated message to the system log (/var/log/messages). You do not have to be root to invoke logger.
logger Experiencing instability in network connection at 23:10, 05/21.
# Now, do a 'tail /var/log/messages'.

By embedding a logger command in a script, it is possible to write debugging information to /var/log/messages.
logger -t $0 -i Logging at line "$LINENO".
# The "-t" option specifies the tag for the logger entry.
# The "-i" option records the process ID.

# tail /var/log/message
# ...
# Jul  7 20:48:58 localhost ./[1712]: Logging at line 3.


This utility manages the system log files, rotating, compressing, deleting, and/or mailing them, as appropriate. Usually cron runs logrotate on a daily basis.

Adding an appropriate entry to /etc/logrotate.conf makes it possible to manage personal log files, as well as system-wide ones.

Job Control


Process Statistics: lists currently executing processes by owner and PID (process ID). This is usually invoked with ax options, and may be piped to grep or sed to search for a specific process (see Example 11-11 and Example 28-2).

bash$  ps ax | grep sendmail
295 ?	   S	  0:00 sendmail: accepting connections on port 25

Lists currently executing processes in "tree" format. The -p option shows the PIDs, as well as the process names.


Continuously updated display of most cpu-intensive processes. The -b option displays in text mode, so that the output may be parsed or accessed from a script.

bash$ top -b
  8:30pm  up 3 min,  3 users,  load average: 0.49, 0.32, 0.13
 45 processes: 44 sleeping, 1 running, 0 zombie, 0 stopped
 CPU states: 13.6% user,  7.3% system,  0.0% nice, 78.9% idle
 Mem:    78396K av,   65468K used,   12928K free,       0K shrd,    2352K buff
 Swap:  157208K av,       0K used,  157208K free                   37244K cached

   848 bozo      17   0   996  996   800 R     5.6  1.2   0:00 top
     1 root       8   0   512  512   444 S     0.0  0.6   0:04 init
     2 root       9   0     0    0     0 SW    0.0  0.0   0:00 keventd


Run a background job with an altered priority. Priorities run from 19 (lowest) to -20 (highest). Only root may set the negative (higher) priorities. Related commands are renice, snice, and skill.


Keeps a command running even after user logs off. The command will run as a foreground process unless followed by &. If you use nohup within a script, consider coupling it with a wait to avoid creating an orphan or zombie process.


Identifies process ID (PID) of a running job. Since job control commands, such as kill and renice act on the PID of a process (not its name), it is sometimes necessary to identify that PID. The pidof command is the approximate counterpart to the $PPID internal variable.

bash$ pidof xclock

Example 13-5. pidof helps kill a process



process=xxxyyyzzz  # Use nonexistent process.
# For demo purposes only...
# ... don't want to actually kill any actual process with this script.
# If, for example, you wanted to use this script to logoff the Internet,
#     process=pppd

t=`pidof $process`       # Find pid (process id) of $process.
# The pid is needed by 'kill' (can't 'kill' by program name).

if [ -z "$t" ]           # If process not present, 'pidof' returns null.
  echo "Process $process was not running."
  echo "Nothing killed."

kill $t                  # May need 'kill -9' for stubborn process.

# Need a check here to see if process allowed itself to be killed.
# Perhaps another " t=`pidof $process` ".

# This entire script could be replaced by
#    kill $(pidof -x process_name)
# but it would not be as instructive.

exit 0

Identifies the processes (by PID) that are accessing a given file, set of files, or directory. May also be invoked with the -k option, which kills those processes. This has interesting implications for system security, especially in scripts preventing unauthorized users from accessing system services.


Administrative program scheduler, performing such duties as cleaning up and deleting system log files and updating the slocate database. This is the superuser version of at (although each user may have their own crontab file which can be changed with the crontab command). It runs as a daemon and executes scheduled entries from /etc/crontab.


Some flavors of Linux run crond, Matthew Dillon's version of cron.

Process Control and Booting


The init command is the parent of all processes. Called in the final step of a bootup, init determines the runlevel of the system from /etc/inittab. Invoked by its alias telinit, and by root only.


Symlinked to init, this is a means of changing the system runlevel, usually done for system maintenance or emergency filesystem repairs. Invoked only by root. This command can be dangerous - be certain you understand it well before using!


Shows the current and last runlevel, that is, whether the system is halted (runlevel 0), in single-user mode (1), in multi-user mode (2 or 3), in X Windows (5), or rebooting (6). This command accesses the /var/run/utmp file.

halt, shutdown, reboot

Command set to shut the system down, usually just prior to a power down.



Network interface configuration and tuning utility.

bash$ ifconfig -a
lo        Link encap:Local Loopback
           inet addr:  Mask:
           UP LOOPBACK RUNNING  MTU:16436  Metric:1
           RX packets:10 errors:0 dropped:0 overruns:0 frame:0
           TX packets:10 errors:0 dropped:0 overruns:0 carrier:0
           collisions:0 txqueuelen:0 
           RX bytes:700 (700.0 b)  TX bytes:700 (700.0 b)

The ifconfig command is most often used at bootup to set up the interfaces, or to shut them down when rebooting.

# Code snippets from /etc/rc.d/init.d/network

# ...

# Check that networking is up.
[ ${NETWORKING} = "no" ] && exit 0

[ -x /sbin/ifconfig ] || exit 0

# ...

for i in $interfaces ; do
  if ifconfig $i 2>/dev/null | grep -q "UP" >/dev/null 2>&1 ; then
    action "Shutting down interface $i: " ./ifdown $i boot
# The GNU-specific "-q" option to "grep" means "quiet", i.e., producing no output.
# Redirecting output to /dev/null is therefore not strictly necessary.
# ...

echo "Currently active devices:"
echo `/sbin/ifconfig | grep ^[a-z] | awk '{print $1}'`
#                            ^^^^^  should be quoted to prevent globbing.
#  The following also work.
#    echo $(/sbin/ifconfig | awk '/^[a-z]/ { print $1 })'
#    echo $(/sbin/ifconfig | sed -e 's/ .*//')
#  Thanks, S.C., for additional comments.

See also Example 30-6.


This is the command set for configuring a wireless network. It is the wireless equivalent of ifconfig, above.


Show info about or make changes to the kernel routing table.

bash$ route
Destination     Gateway         Genmask         Flags   MSS Window  irtt Iface
 pm3-67.bozosisp *      UH       40 0          0 ppp0       *            U        40 0          0 lo
 default         pm3-67.bozosisp         UG       40 0          0 ppp0


Check network configuration. This command lists and manages the network services started at bootup in the /etc/rc?.d directory.

Originally a port from IRIX to Red Hat Linux, chkconfig may not be part of the core installation of some Linux flavors.

bash$ chkconfig --list
atd             0:off   1:off   2:off   3:on    4:on    5:on    6:off
 rwhod           0:off   1:off   2:off   3:off   4:off   5:off   6:off


Network packet "sniffer". This is a tool for analyzing and troubleshooting traffic on a network by dumping packet headers that match specified criteria.

Dump ip packet traffic between hosts bozoville and caduceus:
bash$ tcpdump ip host bozoville and caduceus

Of course, the output of tcpdump can be parsed, using certain of the previously discussed text processing utilities.



Mount a filesystem, usually on an external device, such as a floppy or CDROM. The file /etc/fstab provides a handy listing of available filesystems, partitions, and devices, including options, that may be automatically or manually mounted. The file /etc/mtab shows the currently mounted filesystems and partitions (including the virtual ones, such as /proc).

mount -a mounts all filesystems and partitions listed in /etc/fstab, except those with a noauto option. At bootup, a startup script in /etc/rc.d (rc.sysinit or something similar) invokes this to get everything mounted.

mount -t iso9660 /dev/cdrom /mnt/cdrom
# Mounts CDROM
mount /mnt/cdrom
# Shortcut, if /mnt/cdrom listed in /etc/fstab

This versatile command can even mount an ordinary file on a block device, and the file will act as if it were a filesystem. Mount accomplishes that by associating the file with a loopback device. One application of this is to mount and examine an ISO9660 image before burning it onto a CDR. [41]

Example 13-6. Checking a CD image

# As root...

mkdir /mnt/cdtest  # Prepare a mount point, if not already there.

mount -r -t iso9660 -o loop cd-image.iso /mnt/cdtest   # Mount the image.
#                  "-o loop" option equivalent to "losetup /dev/loop0"
cd /mnt/cdtest     # Now, check the image.
ls -alR            # List the files in the directory tree there.
                   # And so forth.

Unmount a currently mounted filesystem. Before physically removing a previously mounted floppy or CDROM disk, the device must be umounted, else filesystem corruption may result.
umount /mnt/cdrom
# You may now press the eject button and safely remove the disk.


The automount utility, if properly installed, can mount and unmount floppies or CDROM disks as they are accessed or removed. On laptops with swappable floppy and CDROM drives, this can cause problems, though.


Forces an immediate write of all updated data from buffers to hard drive (synchronize drive with buffers). While not strictly necessary, a sync assures the sys admin or user that the data just changed will survive a sudden power failure. In the olden days, a sync; sync (twice, just to make absolutely sure) was a useful precautionary measure before a system reboot.

At times, you may wish to force an immediate buffer flush, as when securely deleting a file (see Example 12-50) or when the lights begin to flicker.


Sets up and configures loopback devices.

Example 13-7. Creating a filesystem in a file

SIZE=1000000  # 1 meg

head -c $SIZE < /dev/zero > file  # Set up file of designated size.
losetup /dev/loop0 file           # Set it up as loopback device.
mke2fs /dev/loop0                 # Create filesystem.
mount -o loop /dev/loop0 /mnt     # Mount it.

# Thanks, S.C.

Creates a swap partition or file. The swap area must subsequently be enabled with swapon.

swapon, swapoff

Enable / disable swap partitition or file. These commands usually take effect at bootup and shutdown.


Create a Linux ext2 filesystem. This command must be invoked as root.

Example 13-8. Adding a new hard drive


# Adding a second hard drive to system.
# Software configuration. Assumes hardware already mounted.
# From an article by the author of this document.
# In issue #38 of "Linux Gazette",

ROOT_UID=0     # This script must be run as root.
E_NOTROOT=67   # Non-root exit error.

if [ "$UID" -ne "$ROOT_UID" ]
  echo "Must be root to run this script."
  exit $E_NOTROOT

# Use with extreme caution!
# If something goes wrong, you may wipe out your current filesystem.

NEWDISK=/dev/hdb         # Assumes /dev/hdb vacant. Check!
MOUNTPOINT=/mnt/newdisk  # Or choose another mount point.

fdisk $NEWDISK
mke2fs -cv $NEWDISK1   # Check for bad blocks & verbose output.
#  Note:    /dev/hdb1, *not* /dev/hdb!
chmod 777 $MOUNTPOINT  # Makes new drive accessible to all users.

# Now, test...
# mount -t ext2 /dev/hdb1 /mnt/newdisk
# Try creating a directory.
# If it works, umount it, and proceed.

# Final step:
# Add the following line to /etc/fstab.
# /dev/hdb1  /mnt/newdisk  ext2  defaults  1 1

exit 0

See also Example 13-7 and Example 29-3.


Tune ext2 filesystem. May be used to change filesystem parameters, such as maximum mount count. This must be invoked as root.


This is an extremely dangerous command. Use it at your own risk, as you may inadvertently destroy your filesystem.


Dump (list to stdout) very verbose filesystem info. This must be invoked as root.

root# dumpe2fs /dev/hda7 | grep 'ount count'
dumpe2fs 1.19, 13-Jul-2000 for EXT2 FS 0.5b, 95/08/09
 Mount count:              6
 Maximum mount count:      20

List or change hard disk parameters. This command must be invoked as root, and it may be dangerous if misused.


Create or change a partition table on a storage device, usually a hard drive. This command must be invoked as root.


Use this command with extreme caution. If something goes wrong, you may destroy an existing filesystem.

fsck, e2fsck, debugfs

Filesystem check, repair, and debug command set.

fsck: a front end for checking a UNIX filesystem (may invoke other utilities). The actual filesystem type generally defaults to ext2.

e2fsck: ext2 filesystem checker.

debugfs: ext2 filesystem debugger. One of the uses of this versatile, but dangerous command is to (attempt to) recover deleted files. For advanced users only!


All of these should be invoked as root, and they can damage or destroy a filesystem if misused.


Checks for bad blocks (physical media flaws) on a storage device. This command finds use when formatting a newly installed hard drive or testing the integrity of backup media. [42] As an example, badblocks /dev/fd0 tests a floppy disk.

The badblocks command may be invoked destructively (overwrite all data) or in non-destructive read-only mode. If root user owns the device to be tested, as is generally the case, then root must invoke this command.

lsusb, usbmodules

The lsusb command lists all USB (Universal Serial Bus) buses and the devices hooked up to them.

The usbmodules command outputs information about the driver modules for connected USB devices.

root# lsusb
Bus 001 Device 001: ID 0000:0000  
 Device Descriptor:
   bLength                18
   bDescriptorType         1
   bcdUSB               1.00
   bDeviceClass            9 Hub
   bDeviceSubClass         0 
   bDeviceProtocol         0 
   bMaxPacketSize0         8
   idVendor           0x0000 
   idProduct          0x0000
   . . .


Creates a boot floppy which can be used to bring up the system if, for example, the MBR (master boot record) becomes corrupted. The mkbootdisk command is actually a Bash script, written by Erik Troan, in the /sbin directory.


CHange ROOT directory. Normally commands are fetched from $PATH, relative to /, the default root directory. This changes the root directory to a different one (and also changes the working directory to there). This is useful for security purposes, for instance when the system administrator wishes to restrict certain users, such as those telnetting in, to a secured portion of the filesystem (this is sometimes referred to as confining a guest user to a "chroot jail"). Note that after a chroot, the execution path for system binaries is no longer valid.

A chroot /opt would cause references to /usr/bin to be translated to /opt/usr/bin. Likewise, chroot /aaa/bbb /bin/ls would redirect future instances of ls to /aaa/bbb as the base directory, rather than / as is normally the case. An alias XX 'chroot /aaa/bbb ls' in a user's ~/.bashrc effectively restricts which portion of the filesystem she may run command "XX" on.

The chroot command is also handy when running from an emergency boot floppy (chroot to /dev/fd0), or as an option to lilo when recovering from a system crash. Other uses include installation from a different filesystem (an rpm option) or running a readonly filesystem from a CD ROM. Invoke only as root, and use with care.


It might be necessary to copy certain system files to a chrooted directory, since the normal $PATH can no longer be relied upon.


This utility is part of the procmail package ( It creates a lock file, a semaphore file that controls access to a file, device, or resource. The lock file serves as a flag that this particular file, device, or resource is in use by a particular process ("busy"), and this permits only restricted access (or no access) to other processes.

Lock files are used in such applications as protecting system mail folders from simultaneously being changed by multiple users, indicating that a modem port is being accessed, and showing that an instance of Netscape is using its cache. Scripts may check for the existence of a lock file created by a certain process to check if that process is running. Note that if a script attempts to create a lock file that already exists, the script will likely hang.

Normally, applications create and check for lock files in the /var/lock directory. A script can test for the presence of a lock file by something like the following.
# Application "xyzip" created lock file "/var/lock/xyzip.lock".

if [ -e "/var/lock/$appname.lock ]


Creates block or character device files (may be necessary when installing new hardware on the system). The MAKEDEV utility has virtually all of the functionality of mknod, and is easier to use.


Utility for creating device files. It must be run as root, and in the /dev directory.
root# ./MAKEDEV
This is a sort of advanced version of mknod.


Automatically deletes files which have not been accessed within a specified period of time. Usually invoked by cron to remove stale log files.


dump, restore

The dump command is an elaborate filesystem backup utility, generally used on larger installations and networks. [43] It reads raw disk partitions and writes a backup file in a binary format. Files to be backed up may be saved to a variety of storage media, including disks and tape drives. The restore command restores backups made with dump.


Perform a low-level format on a floppy disk.

System Resources


Sets an upper limit on use of system resources. Usually invoked with the -f option, which sets a limit on file size (ulimit -f 1000 limits files to 1 meg maximum). The -t option limits the coredump size (ulimit -c 0 eliminates coredumps). Normally, the value of ulimit would be set in /etc/profile and/or ~/.bash_profile (see Chapter 27).


Judicious use of ulimit can protect a system against the dreaded fork bomb.

# This script is for illustrative purposes only.
# Run it at your own peril -- it *will* freeze your system.

while true  #  Endless loop.
  $0 &      #  This script invokes itself . . .
            #+ forks an infinite number of times . . .
            #+ until the system freezes up because all resources exhausted.
done        #  This is the notorious "sorcerer's appentice" scenario.	   

exit 0      #  Will not exit here, because this script will never terminate.

A ulimit -Hu XX (where XX is the user process limit) in /etc/profile would abort this script when it exceeds the preset limit.


Set user or group disk quotas from the command line.


User file creation MASK. Limit the default file attributes for a particular user. All files created by that user take on the attributes specified by umask. The (octal) value passed to umask defines the file permissions disabled. For example, umask 022 ensures that new files will have at most 755 permissions (777 NAND 022). [44] Of course, the user may later change the attributes of particular files with chmod. The usual practice is to set the value of umask in /etc/profile and/or ~/.bash_profile (see Chapter 27).


Get info about or make changes to root device, swap space, or video mode. The functionality of rdev has generally been taken over by lilo, but rdev remains useful for setting up a ram disk. This is a dangerous command, if misused.



List installed kernel modules.

bash$ lsmod
Module                  Size  Used by
 autofs                  9456   2 (autoclean)
 opl3                   11376   0
 serial_cs               5456   0 (unused)
 sb                     34752   0
 uart401                 6384   0 [sb]
 sound                  58368   0 [opl3 sb uart401]
 soundlow                 464   0 [sound]
 soundcore               2800   6 [sb sound]
 ds                      6448   2 [serial_cs]
 i82365                 22928   2
 pcmcia_core            45984   0 [serial_cs ds i82365]


Doing a cat /proc/modules gives the same information.


Force installation of a kernel module (use modprobe instead, when possible). Must be invoked as root.


Force unloading of a kernel module. Must be invoked as root.


Module loader that is normally invoked automatically in a startup script. Must be invoked as root.


Creates module dependency file, usually invoked from startup script.


Output information about a loadable module.

bash$ modinfo hid
filename:    /lib/modules/2.4.20-6/kernel/drivers/usb/hid.o
 description: "USB HID support drivers"
 author:      "Andreas Gal, Vojtech Pavlik <>"
 license:     "GPL"



Runs a program or script with certain environmental variables set or changed (without changing the overall system environment). The [varname=xxx] permits changing the environmental variable varname for the duration of the script. With no options specified, this command lists all the environmental variable settings.


In Bash and other Bourne shell derivatives, it is possible to set variables in a single command's environment.
var1=value1 var2=value2 commandXXX
# $var1 and $var2 set in the environment of 'commandXXX' only.


The first line of a script (the "sha-bang" line) may use env when the path to the shell or interpreter is unknown.
#! /usr/bin/env perl

print "This Perl script will run,\n";
print "even when I don't know where to find Perl.\n";

# Good for portable cross-platform scripts,
# where the Perl binaries may not be in the expected place.
# Thanks, S.C.


Show shared lib dependencies for an executable file.

bash$ ldd /bin/ls => /lib/ (0x4000c000)
/lib/ => /lib/ (0x80000000)

Run a command repeatedly, at specified time intervals.

The default is two-second intervals, but this may be changed with the -n option.

watch -n 5 tail /var/log/messages
# Shows tail end of system log, /var/log/messages, every five seconds.


Remove the debugging symbolic references from an executable binary. This decreases its size, but makes debugging it impossible.

This command often occurs in a Makefile, but rarely in a shell script.


List symbols in an unstripped compiled binary.


Remote distribution client: synchronizes, clones, or backs up a file system on a remote server.

Using our knowledge of administrative commands, let us examine a system script. One of the shortest and simplest to understand scripts is killall, used to suspend running processes at system shutdown.

Example 13-9. killall, from /etc/rc.d/init.d


# --> Comments added by the author of this document marked by "# -->".

# --> This is part of the 'rc' script package
# --> by Miquel van Smoorenburg, <>

# --> This particular script seems to be Red Hat specific
# --> (may not be present in other distributions).

# Bring down all unneeded services that are still running (there shouldn't 
# be any, so this is just a sanity check)

for i in /var/lock/subsys/*; do
        # --> Standard for/in loop, but since "do" is on same line,
        # --> it is necessary to add ";".
	# Check if the script is there.
	[ ! -f $i ] && continue
	# --> This is a clever use of an "and list", equivalent to:
	# --> if [ ! -f "$i" ]; then continue

	# Get the subsystem name.
	# --> Match variable name, which, in this case, is the file name.
	# --> This is the exact equivalent of subsys=`basename $i`.
	# -->  It gets it from the lock file name (if there is a lock file,
	# -->+ that's proof the process has been running).
	# -->  See the "lockfile" entry, above.

	# Bring the subsystem down.
	if [ -f /etc/rc.d/init.d/$subsys.init ]; then
	    /etc/rc.d/init.d/$subsys.init stop
	    /etc/rc.d/init.d/$subsys stop
        # -->  Suspend running jobs and daemons.
        # -->  Note that "stop" is a positional parameter,
        # -->+ not a shell builtin.

That wasn't so bad. Aside from a little fancy footwork with variable matching, there is no new material there.

Exercise 1. In /etc/rc.d/init.d, analyze the halt script. It is a bit longer than killall, but similar in concept. Make a copy of this script somewhere in your home directory and experiment with it (do not run it as root). Do a simulated run with the -vn flags (sh -vn scriptname). Add extensive comments. Change the "action" commands to "echos".

Exercise 2. Look at some of the more complex scripts in /etc/rc.d/init.d. See if you can understand parts of them. Follow the above procedure to analyze them. For some additional insight, you might also examine the file sysvinitfiles in /usr/share/doc/initscripts-?.??, which is part of the "initscripts" documentation.

Chapter 14. Command Substitution

Command substitution reassigns the output of a command [45] or even multiple commands; it literally plugs the command output into another context. [46]

The classic form of command substitution uses backquotes (`...`). Commands within backquotes (backticks) generate command line text.
script_name=`basename $0`
echo "The name of this script is $script_name."

The output of commands can be used as arguments to another command, to set a variable, and even for generating the argument list in a for loop.

rm `cat filename`   # "filename" contains a list of files to delete.
# S. C. points out that "arg list too long" error might result.
# Better is              xargs rm -- < filename 
# ( -- covers those cases where "filename" begins with a "-" )

textfile_listing=`ls *.txt`
# Variable contains names of all *.txt files in current working directory.
echo $textfile_listing

textfile_listing2=$(ls *.txt)   # The alternative form of command substitution.
echo $textfile_listing2
# Same result.

# A possible problem with putting a list of files into a single string
# is that a newline may creep in.
# A safer way to assign a list of files to a parameter is with an array.
#      shopt -s nullglob    # If no match, filename expands to nothing.
#      textfile_listing=( *.txt )
# Thanks, S.C.


Command substitution invokes a subshell.


Command substitution may result in word splitting.
COMMAND `echo a b`     # 2 args: a and b

COMMAND "`echo a b`"   # 1 arg: "a b"

COMMAND `echo`         # no arg

COMMAND "`echo`"       # one empty arg

# Thanks, S.C.

Even when there is no word splitting, command substitution can remove trailing newlines.
# cd "`pwd`"  # This should always work.
# However...

mkdir 'dir with trailing newline

cd 'dir with trailing newline

cd "`pwd`"  # Error message:
# bash: cd: /tmp/file with trailing newline: No such file or directory

cd "$PWD"   # Works fine.

old_tty_setting=$(stty -g)   # Save old terminal setting.
echo "Hit a key "
stty -icanon -echo           # Disable "canonical" mode for terminal.
                             # Also, disable *local* echo.
key=$(dd bs=1 count=1 2> /dev/null)   # Using 'dd' to get a keypress.
stty "$old_tty_setting"      # Restore old setting. 
echo "You hit ${#key} key."  # ${#variable} = number of characters in $variable
# Hit any key except RETURN, and the output is "You hit 1 key."
# Hit RETURN, and it's "You hit 0 key."
# The newline gets eaten in the command substitution.

Thanks, S.C.


Using echo to output an unquoted variable set with command substitution removes trailing newlines characters from the output of the reassigned command(s). This can cause unpleasant surprises.
dir_listing=`ls -l`
echo $dir_listing     # unquoted

# Expecting a nicely ordered directory listing.

# However, what you get is:
# total 3 -rw-rw-r-- 1 bozo bozo 30 May 13 17:15 1.txt -rw-rw-r-- 1 bozo
# bozo 51 May 15 20:57 -rwxr-xr-x 1 bozo bozo 217 Mar 5 21:13

# The newlines disappeared.

echo "$dir_listing"   # quoted
# -rw-rw-r--    1 bozo       30 May 13 17:15 1.txt
# -rw-rw-r--    1 bozo       51 May 15 20:57
# -rwxr-xr-x    1 bozo      217 Mar  5 21:13

Command substitution even permits setting a variable to the contents of a file, using either redirection or the cat command.

variable1=`<file1`      #  Set "variable1" to contents of "file1".
variable2=`cat file2`   #  Set "variable2" to contents of "file2".
                        #  This, however, forks a new process,
                        #+ so the line of code executes slower than the above version.

#  Note:
#  The variables may contain embedded whitespace,
#+ or even (horrors), control characters.

#  Excerpts from system file, /etc/rc.d/rc.sysinit
#+ (on a Red Hat Linux installation)

if [ -f /fsckoptions ]; then
        fsckoptions=`cat /fsckoptions`
if [ -e "/proc/ide/${disk[$device]}/media" ] ; then
             hdmedia=`cat /proc/ide/${disk[$device]}/media`
if [ ! -n "`uname -r | grep -- "-"`" ]; then
       ktag="`cat /proc/version`"
if [ $usb = "1" ]; then
    sleep 5
    mouseoutput=`cat /proc/bus/usb/devices 2>/dev/null|grep -E "^I.*Cls=03.*Prot=02"`
    kbdoutput=`cat /proc/bus/usb/devices 2>/dev/null|grep -E "^I.*Cls=03.*Prot=01"`


Do not set a variable to the contents of a long text file unless you have a very good reason for doing so. Do not set a variable to the contents of a binary file, even as a joke.

Example 14-1. Stupid script tricks

# Don't try this at home, folks.
# From "Stupid Script Tricks," Volume I.

dangerous_variable=`cat /boot/vmlinuz`   # The compressed Linux kernel itself.

echo "string-length of \$dangerous_variable = ${#dangerous_variable}"
# string-length of $dangerous_variable = 794151
# (Does not give same count as 'wc -c /boot/vmlinuz'.)

# echo "$dangerous_variable"
# Don't try this! It would hang the script.

#  The document author is aware of no useful applications for
#+ setting a variable to the contents of a binary file.

exit 0

Notice that a buffer overrun does not occur. This is one instance where an interpreted language, such as Bash, provides more protection from programmer mistakes than a compiled language.

Command substitution permits setting a variable to the output of a loop. The key to this is grabbing the output of an echo command within the loop.

Example 14-2. Generating a variable from a loop

# Setting a variable to the output of a loop.

variable1=`for i in 1 2 3 4 5
  echo -n "$i"                 #  The 'echo' command is critical
done`                          #+ to command substitution here.

echo "variable1 = $variable1"  # variable1 = 12345

variable2=`while [ "$i" -lt 10 ]
  echo -n "$i"                 # Again, the necessary 'echo'.
  let "i += 1"                 # Increment.

echo "variable2 = $variable2"  # variable2 = 0123456789

#  Demonstrates that it's possible to embed a loop
#+ within a variable declaration.

exit 0


The $(COMMAND) form has superseded backticks for command substitution.

output=$(sed -n /"$1"/p $file)   # From ""	example.
# Setting a variable to the contents of a text file.
File_contents1=$(cat $file1)      
File_contents2=$(<$file2)        # Bash permits this also.

The $(...) form of command substitution treats a double backslash in a different way than `...`.

bash$ echo `echo \\`

bash$ echo $(echo \\)

Examples of command substitution in shell scripts:

  1. Example 10-7

  2. Example 10-26

  3. Example 9-27

  4. Example 12-3

  5. Example 12-18

  6. Example 12-15

  7. Example 12-45

  8. Example 10-13

  9. Example 10-10

  10. Example 12-27

  11. Example 16-8

  12. Example A-18

  13. Example 28-2

  14. Example 12-38

  15. Example 12-39

  16. Example 12-40

Chapter 15. Arithmetic Expansion

Arithmetic expansion provides a powerful tool for performing arithmetic operations in scripts. Translating a string into a numerical expression is relatively straightforward using backticks, double parentheses, or let.


Arithmetic expansion with backticks (often used in conjunction with expr)

z=`expr $z + 3`          # The 'expr' command performs the expansion.

Arithmetic expansion with double parentheses, and using let

The use of backticks in arithmetic expansion has been superseded by double parentheses -- ((...)) and $((...)) -- and also by the very convenient let construction.

z=$((z+3))                                  #  Also correct.
                                            #  Within double parentheses,
                                            #+ parameter dereferencing
                                            #+ is optional.

# $((EXPRESSION)) is arithmetic expansion.  #  Not to be confused with
                                            #+ command substitution.

# You may also use operations within double parentheses without assignment.

  echo "n = $n"                             # n = 0

  (( n += 1 ))                              # Increment.
# (( $n += 1 )) is incorrect!
  echo "n = $n"                             # n = 1

let z=z+3
let "z += 3"  #  Quotes permit the use of spaces.
              #  The 'let' operator actually performs arithmetic evaluation,
              #+ rather than expansion.

Examples of arithmetic expansion in scripts:

  1. Example 12-9

  2. Example 10-14

  3. Example 26-1

  4. Example 26-11

  5. Example A-18

Chapter 16. I/O Redirection

There are always three default "files" open, stdin (the keyboard), stdout (the screen), and stderr (error messages output to the screen). These, and any other open files, can be redirected. Redirection simply means capturing output from a file, command, program, script, or even code block within a script (see Example 3-1 and Example 3-2) and sending it as input to another file, command, program, or script.

Each open file gets assigned a file descriptor. [47] The file descriptors for stdin, stdout, and stderr are 0, 1, and 2, respectively. For opening additional files, there remain descriptors 3 to 9. It is sometimes useful to assign one of these additional file descriptors to stdin, stdout, or stderr as a temporary duplicate link. [48] This simplifies restoration to normal after complex redirection and reshuffling (see Example 16-1).

      # Redirect stdout to a file.
      # Creates the file if not present, otherwise overwrites it.

      ls -lR > dir-tree.list
      # Creates a file containing a listing of the directory tree.

   : > filename
      # The > truncates file "filename" to zero length.
      # If file not present, creates zero-length file (same effect as 'touch').
      # The : serves as a dummy placeholder, producing no output.

   > filename    
      # The > truncates file "filename" to zero length.
      # If file not present, creates zero-length file (same effect as 'touch').
      # (Same result as ": >", above, but this does not work with some shells.)

      # Redirect stdout to a file.
      # Creates the file if not present, otherwise appends to it.

      # Single-line redirection commands (affect only the line they are on):
      # --------------------------------------------------------------------

      # Redirect stdout to file "filename".
      # Redirect and append stdout to file "filename".
      # Redirect stderr to file "filename".
      # Redirect and append stderr to file "filename".
      # Redirect both stdout and stderr to file "filename".

      # Redirecting stdout, one line at a time.

      echo "This statement is sent to the log file, \"$LOGFILE\"." 1>$LOGFILE
      echo "This statement is appended to \"$LOGFILE\"." 1>>$LOGFILE
      echo "This statement is also appended to \"$LOGFILE\"." 1>>$LOGFILE
      echo "This statement is echoed to stdout, and will not appear in \"$LOGFILE\"."
      # These redirection commands automatically "reset" after each line.

      # Redirecting stderr, one line at a time.

      bad_command1 2>$ERRORFILE       #  Error message sent to $ERRORFILE.
      bad_command2 2>>$ERRORFILE      #  Error message appended to $ERRORFILE.
      bad_command3                    #  Error message echoed to stderr,
                                      #+ and does not appear in $ERRORFILE.
      # These redirection commands also automatically "reset" after each line.

      # Redirects stderr to stdout.
      # Error messages get sent to same place as standard output.

      # Redirects file descriptor i to j.
      # All output of file pointed to by i gets sent to file pointed to by j.

      # Redirects, by default, file descriptor 1 (stdout) to j.
      # All stdout gets sent to file pointed to by j.

      # Accept input from a file.
      # Companion command to ">", and often used in combination with it.
      # grep search-word <filename

      # Open file "filename" for reading and writing, and assign file descriptor "j" to it.
      # If "filename" does not exist, create it.
      # If file descriptor "j" is not specified, default to fd 0, stdin.
      # An application of this is writing at a specified place in a file. 
      echo 1234567890 > File    # Write string to "File".
      exec 3<> File             # Open "File" and assign fd 3 to it.
      read -n 4 <&3             # Read only 4 characters.
      echo -n . >&3             # Write a decimal point there.
      exec 3>&-                 # Close fd 3.
      cat File                  # ==> 1234.67890
      # Random access, by golly.

      # Pipe.
      # General purpose process and command chaining tool.
      # Similar to ">", but more general in effect.
      # Useful for chaining commands, scripts, files, and programs together.
      cat *.txt | sort | uniq > result-file
      # Sorts the output of all the .txt files and deletes duplicate lines,
      # finally saves results to "result-file".

Multiple instances of input and output redirection and/or pipes can be combined in a single command line.
command < input-file > output-file

command1 | command2 | command3 > output-file
See Example 12-26 and Example A-16.

Multiple output streams may be redirected to one file.
ls -yz >> command.log 2>&1
#  Capture result of illegal options "yz" in file "command.log."
#  Because stderr is redirected to the file,
#+ any error messages will also be there.

#  Note, however, that the following does *not* give the same result.
ls -yz 2>&1 >> command.log
#  Outputs an error message and does not write to file.

#  If redirecting both stdout and stderr,
#+ the order of the commands makes a difference.

Closing File Descriptors


Close input file descriptor n.

0<&-, <&-

Close stdin.


Close output file descriptor n.

1>&-, >&-

Close stdout.

Child processes inherit open file descriptors. This is why pipes work. To prevent an fd from being inherited, close it.
# Redirecting only stderr to a pipe.

exec 3>&1                              # Save current "value" of stdout.
ls -l 2>&1 >&3 3>&- | grep bad 3>&-    # Close fd 3 for 'grep' (but not 'ls').
#              ^^^^   ^^^^
exec 3>&-                              # Now close it for the remainder of the script.

# Thanks, S.C.

For a more detailed introduction to I/O redirection see Appendix E.

16.1. Using exec

An exec <filename command redirects stdin to a file. From that point on, all stdin comes from that file, rather than its normal source (usually keyboard input). This provides a method of reading a file line by line and possibly parsing each line of input using sed and/or awk.

Example 16-1. Redirecting stdin using exec

# Redirecting stdin using 'exec'.

exec 6<&0          # Link file descriptor #6 with stdin.
                   # Saves stdin.

exec < data-file   # stdin replaced by file "data-file"

read a1            # Reads first line of file "data-file".
read a2            # Reads second line of file "data-file."

echo "Following lines read from file."
echo "-------------------------------"
echo $a1
echo $a2

echo; echo; echo

exec 0<&6 6<&-
#  Now restore stdin from fd #6, where it had been saved,
#+ and close fd #6 ( 6<&- ) to free it for other processes to use.
# <&6 6<&-    also works.

echo -n "Enter data  "
read b1  # Now "read" functions as expected, reading from normal stdin.
echo "Input read from stdin."
echo "----------------------"
echo "b1 = $b1"


exit 0

Similarly, an exec >filename command redirects stdout to a designated file. This sends all command output that would normally go to stdout to that file.

Example 16-2. Redirecting stdout using exec



exec 6>&1           # Link file descriptor #6 with stdout.
                    # Saves stdout.

exec > $LOGFILE     # stdout replaced with file "logfile.txt".

# ----------------------------------------------------------- #
# All output from commands in this block sent to file $LOGFILE.

echo -n "Logfile: "
echo "-------------------------------------"

echo "Output of \"ls -al\" command"
ls -al
echo; echo
echo "Output of \"df\" command"

# ----------------------------------------------------------- #

exec 1>&6 6>&-      # Restore stdout and close file descriptor #6.

echo "== stdout now restored to default == "
ls -al

exit 0

Example 16-3. Redirecting both stdin and stdout in the same script with exec

# Converts a specified input file to uppercase.


if [ ! -r "$1" ]     # Is specified input file readable?
  echo "Can't read from input file!"
  echo "Usage: $0 input-file output-file"
fi                   #  Will exit with same error
                     #+ even if input file ($1) not specified.

if [ -z "$2" ]
  echo "Need to specify output file."
  echo "Usage: $0 input-file output-file"
  exit $E_WRONG_ARGS

exec 4<&0
exec < $1            # Will read from input file.

exec 7>&1
exec > $2            # Will write to output file.
                     # Assumes output file writable (add check?).

# -----------------------------------------------
    cat - | tr a-z A-Z   # Uppercase conversion.
#   ^^^^^                # Reads from stdin.
#           ^^^^^^^^^^   # Writes to stdout.
# However, both stdin and stdout were redirected.
# -----------------------------------------------

exec 1>&7 7>&-       # Restore stout.
exec 0<&4 4<&-       # Restore stdin.

# After restoration, the following line prints to stdout as expected.
echo "File \"$1\" written to \"$2\" as uppercase conversion."

exit 0

I/O redirection is a clever way of avoiding the dreaded inaccessible variables within a subshell problem.

Example 16-4. Avoiding a subshell

# Suggested by Matthew Walker.



cat myfile.txt | while read line;
                 do {
                   echo $line
                   (( Lines++ ));  #  Incremented values of this variable
                                   #+ inaccessible outside loop.
                                   #  Subshell problem.

echo "Number of lines read = $Lines"     # 0
                                         # Wrong!

echo "------------------------"

exec 3<> myfile.txt
while read line <&3
do {
  echo "$line"
  (( Lines++ ));                   #  Incremented values of this variable
                                   #+ accessible outside loop.
                                   #  No subshell, no problem.
exec 3>&-

echo "Number of lines read = $Lines"     # 8


exit 0

# Lines below not seen by script.

$ cat myfile.txt

Line 1.
Line 2.
Line 3.
Line 4.
Line 5.
Line 6.
Line 7.
Line 8.

16.2. Redirecting Code Blocks

Blocks of code, such as while, until, and for loops, even if/then test blocks can also incorporate redirection of stdin. Even a function may use this form of redirection (see Example 23-10). The < operator at the end of the code block accomplishes this.

Example 16-5. Redirected while loop


if [ -z "$1" ]
then       # Default, if no filename specified.
#+ Filename=${}
#  can replace the above test (parameter substitution).



while [ "$name" != Smith ]  # Why is variable $name in quotes?
  read name                 # Reads from $Filename, rather than stdin.
  echo $name
  let "count += 1"
done <"$Filename"           # Redirects stdin to file $Filename. 
#    ^^^^^^^^^^^^

echo; echo "$count names read"; echo

#  Note that in some older shell scripting languages,
#+ the redirected loop would run as a subshell.
# Therefore, $count would return 0, the initialized value outside the loop.
#  Bash and ksh avoid starting a subshell whenever possible,
# +so that this script, for example, runs correctly.
# Thanks to Heiner Steven for pointing this out.

exit 0

Example 16-6. Alternate form of redirected while loop


# This is an alternate form of the preceding script.

#  Suggested by Heiner Steven
#+ as a workaround in those situations when a redirect loop
#+ runs as a subshell, and therefore variables inside the loop
# +do not keep their values upon loop termination.

if [ -z "$1" ]
then     # Default, if no filename specified.

exec 3<&0                 # Save stdin to file descriptor 3.
exec 0<"$Filename"        # Redirect standard input.


while [ "$name" != Smith ]
  read name               # Reads from redirected stdin ($Filename).
  echo $name
  let "count += 1"
done                      #  Loop reads from file $Filename
                          #+ because of line 20.

#  The original version of this script terminated the "while" loop with
#+      done <"$Filename" 
#  Exercise:
#  Why is this unnecessary?

exec 0<&3                 # Restore old stdin.
exec 3<&-                 # Close temporary fd 3.

echo; echo "$count names read"; echo

exit 0

Example 16-7. Redirected until loop

# Same as previous example, but with "until" loop.

if [ -z "$1" ]
then         # Default, if no filename specified.

# while [ "$name" != Smith ]
until [ "$name" = Smith ]     # Change  !=  to =.
  read name                   # Reads from $Filename, rather than stdin.
  echo $name
done <"$Filename"             # Redirects stdin to file $Filename. 
#    ^^^^^^^^^^^^

# Same results as with "while" loop in previous example.

exit 0

Example 16-8. Redirected for loop


if [ -z "$1" ]
then          # Default, if no filename specified.

line_count=`wc $Filename | awk '{ print $1 }'`
#           Number of lines in target file.
#  Very contrived and kludgy, nevertheless shows that
#+ it's possible to redirect stdin within a "for" loop...
#+ if you're clever enough.
# More concise is     line_count=$(wc -l < "$Filename")

for name in `seq $line_count`  # Recall that "seq" prints sequence of numbers.
# while [ "$name" != Smith ]   --   more complicated than a "while" loop   --
  read name                    # Reads from $Filename, rather than stdin.
  echo $name
  if [ "$name" = Smith ]       # Need all this extra baggage here.
done <"$Filename"              # Redirects stdin to file $Filename. 
#    ^^^^^^^^^^^^

exit 0

We can modify the previous example to also redirect the output of the loop.

Example 16-9. Redirected for loop (both stdin and stdout redirected)


if [ -z "$1" ]
then          # Default, if no filename specified.

Savefile=$         # Filename to save results in.
FinalName=Jonah                # Name to terminate "read" on.

line_count=`wc $Filename | awk '{ print $1 }'`  # Number of lines in target file.

for name in `seq $line_count`
  read name
  echo "$name"
  if [ "$name" = "$FinalName" ]
done < "$Filename" > "$Savefile"     # Redirects stdin to file $Filename,
#    ^^^^^^^^^^^^^^^^^^^^^^^^^^^       and saves it to backup file.

exit 0

Example 16-10. Redirected if/then test


if [ -z "$1" ]
then   # Default, if no filename specified.


if [ "$TRUE" ]          # if true    and   if :   also work.
 read name
 echo $name
fi <"$Filename"
#  ^^^^^^^^^^^^

# Reads only first line of file.
# An "if/then" test has no way of iterating unless embedded in a loop.

exit 0

Example 16-11. Data file "" for above examples


#  This is a data file for
#+ "", "", "", "", "".

Redirecting the stdout of a code block has the effect of saving its output to a file. See Example 3-2.

Here documents are a special case of redirected code blocks.

16.3. Applications

Clever use of I/O redirection permits parsing and stitching together snippets of command output (see Example 11-6). This permits generating report and log files.

Example 16-12. Logging events

#, by Stephane Chazelas.

# Event logging to a file.
# Must be run as root (for write access in /var/log).

ROOT_UID=0     # Only users with $UID 0 have root privileges.
E_NOTROOT=67   # Non-root exit error.

if [ "$UID" -ne "$ROOT_UID" ]
  echo "Must be root to run this script."
  exit $E_NOTROOT


# Uncomment one of the two lines below to activate script.

log()  # Writes time and date to log file.
echo "$(date)  $*" >&7     # This *appends* the date to the file.
                              # See below.

case $LOG_LEVEL in
 1) exec 3>&2         4> /dev/null 5> /dev/null;;
 2) exec 3>&2         4>&2         5> /dev/null;;
 3) exec 3>&2         4>&2         5>&2;;
 *) exec 3> /dev/null 4> /dev/null 5> /dev/null;;

if [[ $LOG_VARS ]]
then exec 6>> /var/log/vars.log
else exec 6> /dev/null               # Bury output.

if [[ $LOG_EVENTS ]]
  # then exec 7 >(exec gawk '{print strftime(), $0}' >> /var/log/event.log)
  # Above line will not work in Bash, version 2.04.
  exec 7>> /var/log/event.log        # Append to "event.log".
  log                                      # Write time and date.
else exec 7> /dev/null                  # Bury output.

echo "DEBUG3: beginning" >&${FD_DEBUG3}

ls -l >&5 2>&4                       # command1 >&5 2>&4

echo "Done"                                # command2 

echo "sending mail" >&${FD_LOGEVENTS}   # Writes "sending mail" to fd #7.

exit 0

Chapter 17. Here Documents

A here document is a special-purpose code block. It uses a form of I/O redirection to feed a command list to an interactive program or command, such as ftp, telnet, or ex.

COMMAND <<InputComesFromHERE

A "limit string" delineates (frames) the command list. The special symbol << designates the limit string. This has the effect of redirecting the output of a file into the stdin of the program or command. It is similar to interactive-program < command-file, where command-file contains
command #1
command #2

The "here document" alternative looks like this:

interactive-program <<LimitString
command #1
command #2

Choose a limit string sufficiently unusual that it will not occur anywhere in the command list and confuse matters.

Note that here documents may sometimes be used to good effect with non-interactive utilities and commands.

Example 17-1. dummyfile: Creates a 2-line dummy file


# Non-interactive use of 'vi' to edit a file.
# Emulates 'sed'.


if [ -z "$1" ]
  echo "Usage: `basename $0` filename"
  exit $E_BADARGS


# Insert 2 lines in file, then save.
#--------Begin here document-----------#
vi $TARGETFILE <<x23LimitStringx23
This is line 1 of the example file.
This is line 2 of the example file.
#----------End here document-----------#

#  Note that ^[ above is a literal escape
#+ typed by Control-V <Esc>.

#  Bram Moolenaar points out that this may not work with 'vim',
#+ because of possible problems with terminal interaction.

exit 0

The above script could just as effectively have been implemented with ex, rather than vi. Here documents containing a list of ex commands are common enough to form their own category, known as ex scripts.

Example 17-2. broadcast: Sends message to everyone logged in


wall <<zzz23EndOfMessagezzz23
E-mail your noontime orders for pizza to the system administrator.
    (Add an extra dollar for anchovy or mushroom topping.)
# Additional message text goes here.
# Note: 'wall' prints comment lines.

# Could have been done more efficiently by
#         wall <message-file
# However, saving a message template in a script saves work.

exit 0

Example 17-3. Multi-line message using cat


#  'echo' is fine for printing single line messages,
#+  but somewhat problematic for for message blocks.
#   A 'cat' here document overcomes this limitation.

cat <<End-of-message
This is line 1 of the message.
This is line 2 of the message.
This is line 3 of the message.
This is line 4 of the message.
This is the last line of the message.

#  Replacing line 7, above, with
#+   cat > $Newfile <<End-of-message
#+       ^^^^^^^^^^
#+ writes the output to the file $Newfile, rather than to stdout.

exit 0

# Code below disabled, due to "exit 0" above.

# S.C. points out that the following also works.
echo "-------------------------------------
This is line 1 of the message.
This is line 2 of the message.
This is line 3 of the message.
This is line 4 of the message.
This is the last line of the message.
# However, text may not include double quotes unless they are escaped.

The - option to mark a here document limit string (<<-LimitString) suppresses leading tabs (but not spaces) in the output. This may be useful in making a script more readable.

Example 17-4. Multi-line message, with tabs suppressed

# Same as previous example, but...

#  The - option to a here document <<-
#+ suppresses leading tabs in the body of the document,
#+ but *not* spaces.

	This is line 1 of the message.
	This is line 2 of the message.
	This is line 3 of the message.
	This is line 4 of the message.
	This is the last line of the message.
# The output of the script will be flush left.
# Leading tab in each line will not show.

# Above 5 lines of "message" prefaced by a tab, not spaces.
# Spaces not affected by   <<-  .

# Note that this option has no effect on *embedded* tabs.

exit 0

A here document supports parameter and command substitution. It is therefore possible to pass different parameters to the body of the here document, changing its output accordingly.

Example 17-5. Here document with parameter substitution

# Another 'cat' here document, using parameter substitution.

# Try it with no command line parameters,   ./scriptname
# Try it with one command line parameter,   ./scriptname Mortimer
# Try it with one two-word quoted command line parameter,
#                           ./scriptname "Mortimer Jones"

CMDLINEPARAM=1     # Expect at least command line parameter.

if [ $# -ge $CMDLINEPARAM ]
  NAME=$1          # If more than one command line param,
                   # then just take the first.
  NAME="John Doe"  # Default, if no command line parameter.

RESPONDENT="the author of this fine script"  

cat <<Endofmessage

Hello, there, $NAME.
Greetings to you, $NAME, from $RESPONDENT.

# This comment shows up in the output (why?).


# Note that the blank lines show up in the output.
# So does the "comment".

exit 0

This is a useful script containing a here document with parameter substitution.

Example 17-6. Upload a file pair to "Sunsite" incoming directory


#  Upload file pair (Filename.lsm, Filename.tar.gz)
#+ to incoming directory at Sunsite/UNC (
#  Filename.tar.gz is the tarball itself.
#  Filename.lsm is the descriptor file.


if [ -z "$1" ]
  echo "Usage: `basename $0` Filename-to-upload"
  exit $E_ARGERROR

Filename=`basename $1`           # Strips pathname out of file name.

#  These need not be hard-coded into script,
#+ but may instead be changed to command line argument.

Password="your.e-mail.address"   # Change above to suit.

ftp -n $Server <<End-Of-Session
# -n option disables auto-logon

user anonymous "$Password"
bell                # Ring 'bell' after each file transfer
cd $Directory
put "$Filename.lsm"
put "$Filename.tar.gz"

exit 0

Quoting or escaping the "limit string" at the head of a here document disables parameter substitution within its body.

Example 17-7. Parameter substitution turned off

#  A 'cat' here document, but with parameter substitution disabled.

NAME="John Doe"
RESPONDENT="the author of this fine script"  

cat <<'Endofmessage'

Hello, there, $NAME.
Greetings to you, $NAME, from $RESPONDENT.


#  No parameter substitution when the "limit string" is quoted or escaped.
#  Either of the following at the head of the here document would have the same effect.
#  cat <<"Endofmessage"
#  cat <<\Endofmessage

exit 0

Disabling parameter substitution permits outputting literal text. Generating scripts or even program code is one use for this.

Example 17-8. A script that generates another script

# Based on an idea by Albert Reiner.         # Name of the file to generate.

# -----------------------------------------------------------
# 'Here document containing the body of the generated script.
cat <<'EOF'

echo "This is a generated shell script."
#  Note that since we are inside a subshell,
#+ we can't access variables in the "outside" script.

echo "Generated file will be named: $OUTFILE"
#  Above line will not work as normally expected
#+ because parameter expansion has been disabled.
#  Instead, the result is literal output.


let "c = $a * $b"
echo "c = $c"

exit 0
# -----------------------------------------------------------

#  Quoting the 'limit string' prevents variable expansion
#+ within the body of the above 'here document.'
#  This permits outputting literal strings in the output file.

if [ -f "$OUTFILE" ]
  chmod 755 $OUTFILE
  # Make the generated file executable.
  echo "Problem in creating file: \"$OUTFILE\""

#  This method can also be used for generating
#+ C programs, Perl programs, Python programs, Makefiles,
#+ and the like.

exit 0

It is possible to set a variable from the output of a here document.
variable=$(cat <<SETVAR
This variable
runs over multiple lines.

echo "$variable"

A here document can supply input to a function in the same script.

Example 17-9. Here documents and functions


GetPersonalData ()
  read firstname
  read lastname
  read address
  read city 
  read state 
  read zipcode
} # This certainly looks like an interactive function, but...

# Supply input to the above function.
GetPersonalData <<RECORD001
2726 Nondescript Dr.

echo "$firstname $lastname"
echo "$address"
echo "$city, $state $zipcode"

exit 0

It is possible to use : as a dummy command accepting output from a here document. This, in effect, creates an "anonymous" here document.

Example 17-10. "Anonymous" Here Document


${HOSTNAME?}${USER?}${MAIL?}  # Print error message if one of the variables not set.

exit 0


A variation of the above technique permits "commenting out" blocks of code.

Example 17-11. Commenting out a block of code


echo "This line will not echo."
This is a comment line missing the "#" prefix.
This is another comment line missing the "#" prefix.

The above line will cause no error message,
because the Bash interpreter will ignore it.

echo "Exit value of above \"COMMENTBLOCK\" is $?."   # 0
# No error shown.

#  The above technique also comes in useful for commenting out
#+ a block of working code for debugging purposes.
#  This saves having to put a "#" at the beginning of each line,
#+ then having to go back and delete each "#" later.

for file in *
 cat "$file"

exit 0


Yet another twist of this nifty trick makes "self-documenting" scripts possible.

Example 17-12. A self-documenting script

# self-documenting script
# Modification of "".


if [ "$1" = "-h"  -o "$1" = "--help" ]     # Request help.
  echo; echo "Usage: $0 [directory-name]"; echo
  sed --silent -e '/DOCUMENTATIONXX$/,/^DOCUMENTATIONXX$/p' "$0" |
  sed -e '/DOCUMENTATIONXX$/d'; exit $DOC_REQUEST; fi

List the statistics of a specified directory in tabular format.
The command line parameter gives the directory to be listed.
If no directory specified or directory specified cannot be read,
then list the current working directory.


if [ -z "$1" -o ! -r "$1" ]

echo "Listing of "$directory":"; echo
; ls -l "$directory" | sed 1d) | column -t

exit 0

See also Example A-26 for an excellent example of a self-documenting script.


Here documents create temporary files, but these files are deleted after opening and are not accessible to any other process.

bash$ bash -c 'lsof -a -p $$ -d0' << EOF
lsof    1213 bozo    0r   REG    3,5    0 30386 /tmp/t1213-0-sh (deleted)


Some utilities will not work inside a here document.


The closing limit string, on the final line of a here document, must start in the first character position. There can be no leading whitespace. Trailing whitespace after the limit string likewise causes unexpected behavior. The whitespace prevents the limit string from being recognized.


echo "----------------------------------------------------------------------"

cat <<LimitString
echo "This is line 1 of the message inside the here document."
echo "This is line 2 of the message inside the here document."
echo "This is the final line of the message inside the here document."
#^^^^Indented limit string. Error! This script will not behave as expected.

echo "----------------------------------------------------------------------"

#  These comments are outside the 'here document',
#+ and should not echo.

echo "Outside the here document."

exit 0

echo "This line had better not echo."  # Follows an 'exit' command.

For those tasks too complex for a "here document", consider using the expect scripting language, which is specifically tailored for feeding input into interactive programs.

17.1. Here Strings

A here string can be considered as a stripped-down form of here document. It consists of nothing more than COMMAND <<<$WORD, where $WORD is expanded and fed to the stdin of COMMAND.

Example 17-13. Prepending a line to a file

# Add text at beginning of file.
#  Example contributed by Kenny Stauffer,
#  and slightly modified by document author.


read -p "File: " file   # -p arg to 'read' displays prompt.
if [ ! -e "$file" ]
then   # Bail out if no such file.
  echo "File $file not found."

read -p "Title: " title
cat - $file <<<$title > $

echo "Modified file is $"

exit 0

# from "man bash"
# Here Strings
# 	A variant of here documents, the format is:
# 		<<<word
# 	The word is expanded and supplied to the command on its standard input.

Exercise: Find other uses for here strings.

Chapter 18. Recess Time

  This bizarre little intermission gives the reader a chance to
  relax and maybe laugh a bit.

  Fellow Linux user, greetings!  You are reading something
  which will bring you luck and good fortune.  Just e-mail a
  copy of this document to 10 of your friends. Before you make
  the copies, send a 100-line Bash script to the first person
  on the list given at the bottom of this letter. Then delete
  their name and add yours to the bottom of the list.

  Don't break the chain!  Make the copies within 48 hours.
  Wilfred P. of Brooklyn failed to send out his ten copies and
  woke the next morning to find his job description changed
  to "COBOL programmer."  Howard L. of Newport News sent
  out his ten copies and within a month had enough hardware
  to build a 100-node Beowulf cluster dedicated to playing
  xbill.  Amelia V. of Chicago laughed at this letter and
  broke the chain.  Shortly thereafter, a fire broke out in her
  terminal and she now spends her days writing documentation
  for MS Windows.

  Don't break the chain!  Send out your ten copies today!

Courtesy 'NIX "fortune cookies", with some alterations and many apologies

Part 4. Advanced Topics

At this point, we are ready to delve into certain of the difficult and unusual aspects of scripting. Along the way, we will attempt to "push the envelope" in various ways and examine boundary conditions (what happens when we move into uncharted territory?).

Chapter 19. Regular Expressions

To fully utilize the power of shell scripting, you need to master Regular Expressions. Certain commands and utilities commonly used in scripts, such as grep, expr, sed and awk interpret and use REs.

19.1. A Brief Introduction to Regular Expressions

An expression is a string of characters. Those characters that have an interpretation above and beyond their literal meaning are called metacharacters. A quote symbol, for example, may denote speech by a person, ditto, or a meta-meaning for the symbols that follow. Regular Expressions are sets of characters and/or metacharacters that an operating system endows with special features. [49]

The main uses for Regular Expressions (REs) are text searches and string manipulation. An RE matches a single character or a set of characters -- a string or a part of a string.

  • The asterisk -- * -- matches any number of repeats of the character string or RE preceding it, including zero.

    "1133*" matches 11 + one or more 3's + possibly other characters: 113, 1133, 111312, and so forth.

  • The dot -- . -- matches any one character, except a newline. [50]

    "13." matches 13 + at least one of any character (including a space): 1133, 11333, but not 13 (additional character missing).

  • The caret -- ^ -- matches the beginning of a line, but sometimes, depending on context, negates the meaning of a set of characters in an RE.

  • The dollar sign -- $ -- at the end of an RE matches the end of a line.

    "^$" matches blank lines.


    The ^ and $ are known as anchors, since they indicate or anchor a position within an RE.

  • Brackets -- [...] -- enclose a set of characters to match in a single RE.

    "[xyz]" matches the characters x, y, or z.

    "[c-n]" matches any of the characters in the range c to n.

    "[B-Pk-y]" matches any of the characters in the ranges B to P and k to y.

    "[a-z0-9]" matches any lowercase letter or any digit.

    "[^b-d]" matches all characters except those in the range b to d. This is an instance of ^ negating or inverting the meaning of the following RE (taking on a role similar to ! in a different context).

    Combined sequences of bracketed characters match common word patterns. "[Yy][Ee][Ss]" matches yes, Yes, YES, yEs, and so forth. "[0-9][0-9][0-9]-[0-9][0-9]-[0-9][0-9][0-9][0-9]" matches any Social Security number.

  • The backslash -- \ -- escapes a special character, which means that character gets interpreted literally.

    A "\$" reverts back to its literal meaning of "$", rather than its RE meaning of end-of-line. Likewise a "\\" has the literal meaning of "\".

  • Escaped "angle brackets" -- \<...\> -- mark word boundaries.

    The angle brackets must be escaped, since otherwise they have only their literal character meaning.

    "\<the\>" matches the word "the", but not the words "them", "there", "other", etc.

    bash$ cat textfile
    This is line 1, of which there is only one instance.
     This is the only instance of line 2.
     This is line 3, another line.
     This is line 4.
    bash$ grep 'the' textfile
    This is line 1, of which there is only one instance.
     This is the only instance of line 2.
     This is line 3, another line.
    bash$ grep '\<the\>' textfile
    This is the only instance of line 2.

  • Extended REs. Used in egrep, awk, and Perl

  • The question mark -- ? -- matches zero or one of the previous RE. It is generally used for matching single characters.

  • The plus -- + -- matches one or more of the previous RE. It serves a role similar to the *, but does not match zero occurrences.

    # GNU versions of sed and awk can use "+",
    # but it needs to be escaped.
    echo a111b | sed -ne '/a1\+b/p'
    echo a111b | grep 'a1\+b'
    echo a111b | gawk '/a1+b/'
    # All of above are equivalent.
    # Thanks, S.C.

  • Escaped "curly brackets" -- \{ \} -- indicate the number of occurrences of a preceding RE to match.

    It is necessary to escape the curly brackets since they have only their literal character meaning otherwise. This usage is technically not part of the basic RE set.

    "[0-9]\{5\}" matches exactly five digits (characters in the range of 0 to 9).


    Curly brackets are not available as an RE in the "classic" (non-POSIX compliant) version of awk. However, gawk has the --re-interval option that permits them (without being escaped).

    bash$ echo 2222 | gawk --re-interval '/2{3}/'

    Perl and some egrep versions do not require escaping the curly brackets.

  • Parentheses -- ( ) -- enclose groups of REs. They are useful with the following "|" operator and in substring extraction using expr.

  • The -- | -- "or" RE operator matches any of a set of alternate characters.

    bash$ egrep 're(a|e)d' misc.txt
    People who read seem to be better informed than those who do not.
     The clarinet produces sound by the vibration of its reed.


Some versions of sed, ed, and ex support escaped versions of the extended Regular Expressions described above, as do the GNU utilities.

  • POSIX Character Classes. [:class:]

    This is an alternate method of specifying a range of characters to match.

  • [:alnum:] matches alphabetic or numeric characters. This is equivalent to A-Za-z0-9.

  • [:alpha:] matches alphabetic characters. This is equivalent to A-Za-z.

  • [:blank:] matches a space or a tab.

  • [:cntrl:] matches control characters.

  • [:digit:] matches (decimal) digits. This is equivalent to 0-9.

  • [:graph:] (graphic printable characters). Matches characters in the range of ASCII 33 - 126. This is the same as [:print:], below, but excluding the space character.

  • [:lower:] matches lowercase alphabetic characters. This is equivalent to a-z.

  • [:print:] (printable characters). Matches characters in the range of ASCII 32 - 126. This is the same as [:graph:], above, but adding the space character.

  • [:space:] matches whitespace characters (space and horizontal tab).

  • [:upper:] matches uppercase alphabetic characters. This is equivalent to A-Z.

  • [:xdigit:] matches hexadecimal digits. This is equivalent to 0-9A-Fa-f.


    POSIX character classes generally require quoting or double brackets ([[ ]]).

    bash$ grep [[:digit:]] test.file

    These character classes may even be used with globbing, to a limited extent.

    bash$ ls -l ?[[:digit:]][[:digit:]]?
    -rw-rw-r--    1 bozo  bozo         0 Aug 21 14:47 a33b

    To see POSIX character classes used in scripts, refer to Example 12-17 and Example 12-18.

Sed, awk, and Perl, used as filters in scripts, take REs as arguments when "sifting" or transforming files or I/O streams. See Example A-13 and Example A-18 for illustrations of this.

The standard reference on this complex topic is Friedl's "Mastering Regular Expressions." "Sed & Awk," by Dougherty and Robbins also gives a very lucid treatment of REs. See the Bibliography for more information on these books.

19.2. Globbing

Bash itself cannot recognize Regular Expressions. In scripts, commands and utilities, such as sed and awk, interpret RE's.

Bash does carry out filename expansion, a process known as "globbing," but this does not use the standard RE set. Instead, globbing recognizes and expands wildcards. Globbing interprets the standard wildcard characters, * and ?, character lists in square brackets, and certain other special characters (such as ^ for negating the sense of a match). There are some important limitations on wildcard characters in globbing, however. Strings containing * will not match filenames that start with a dot, as, for example, .bashrc. [51] Likewise, the ? has a different meaning in globbing than as part of an RE.

bash$ ls -l
total 2
 -rw-rw-r--    1 bozo  bozo         0 Aug  6 18:42 a.1
 -rw-rw-r--    1 bozo  bozo         0 Aug  6 18:42 b.1
 -rw-rw-r--    1 bozo  bozo         0 Aug  6 18:42 c.1
 -rw-rw-r--    1 bozo  bozo       466 Aug  6 17:48
 -rw-rw-r--    1 bozo  bozo       758 Jul 30 09:02 test1.txt

bash$ ls -l t?.sh
-rw-rw-r--    1 bozo  bozo       466 Aug  6 17:48

bash$ ls -l [ab]*
-rw-rw-r--    1 bozo  bozo         0 Aug  6 18:42 a.1
 -rw-rw-r--    1 bozo  bozo         0 Aug  6 18:42 b.1

bash$ ls -l [a-c]*
-rw-rw-r--    1 bozo  bozo         0 Aug  6 18:42 a.1
 -rw-rw-r--    1 bozo  bozo         0 Aug  6 18:42 b.1
 -rw-rw-r--    1 bozo  bozo         0 Aug  6 18:42 c.1

bash$ ls -l [^ab]*
-rw-rw-r--    1 bozo  bozo         0 Aug  6 18:42 c.1
 -rw-rw-r--    1 bozo  bozo       466 Aug  6 17:48
 -rw-rw-r--    1 bozo  bozo       758 Jul 30 09:02 test1.txt

bash$ ls -l {b*,c*,*est*}
-rw-rw-r--    1 bozo  bozo         0 Aug  6 18:42 b.1
 -rw-rw-r--    1 bozo  bozo         0 Aug  6 18:42 c.1
 -rw-rw-r--    1 bozo  bozo       758 Jul 30 09:02 test1.txt

Bash performs filename expansion on unquoted command-line arguments. The echo command demonstrates this.

bash$ echo *
a.1 b.1 c.1 test1.txt

bash$ echo t* test1.txt


It is possible to modify the way Bash interprets special characters in globbing. A set -f command disables globbing, and the nocaseglob and nullglob options to shopt change globbing behavior.

See also Example 10-4.

Chapter 20. Subshells

Running a shell script launches another instance of the command processor. Just as your commands are interpreted at the command line prompt, similarly does a script batch process a list of commands in a file. Each shell script running is, in effect, a subprocess of the parent shell, the one that gives you the prompt at the console or in an xterm window.

A shell script can also launch subprocesses. These subshells let the script do parallel processing, in effect executing multiple subtasks simultaneously.

Command List in Parentheses

( command1; command2; command3; ... )

A command list embedded between parentheses runs as a subshell.


Variables in a subshell are not visible outside the block of code in the subshell. They are not accessible to the parent process, to the shell that launched the subshell. These are, in effect, local variables.

Example 20-1. Variable scope in a subshell




echo "From subshell, \"inner_variable\" = $inner_variable"
echo "From subshell, \"outer\" = $outer_variable"


if [ -z "$inner_variable" ]
  echo "inner_variable undefined in main body of shell"
  echo "inner_variable defined in main body of shell"

echo "From main body of shell, \"inner_variable\" = $inner_variable"
# $inner_variable will show as uninitialized because
# variables defined in a subshell are "local variables".


exit 0

See also Example 32-2.


Directory changes made in a subshell do not carry over to the parent shell.

Example 20-2. List User Profiles

# print all user profiles

# This script written by Heiner Steven, and modified by the document author.

FILE=.bashrc  #  File containing user profile,
              #+ was ".profile" in original script.

for home in `awk -F: '{print $6}' /etc/passwd`
  [ -d "$home" ] || continue    # If no home directory, go to next.
  [ -r "$home" ] || continue    # If not readable, go to next.
  (cd $home; [ -e $FILE ] && less $FILE)

#  When script terminates, there is no need to 'cd' back to original directory,
#+ because 'cd $home' takes place in a subshell.

exit 0

A subshell may be used to set up a "dedicated environment" for a command group.
  unset TERMINFO
  set -C
  shift 5
  exit 3 # Only exits the subshell.
# The parent shell has not been affected, and the environment is preserved.
One application of this is testing whether a variable is defined.
if (set -u; : $variable) 2> /dev/null
  echo "Variable is set."
fi     #  Variable has been set in current script,
       #+ or is an an internal Bash variable,
       #+ or is present in environment (has been exported).

# Could also be written [[ ${variable-x} != x || ${variable-y} != y ]]
# or                    [[ ${variable-x} != x$variable ]]
# or                    [[ ${variable+x} = x ]])
# or                    [[ ${variable-x} != x ]])
Another application is checking for a lock file:
if (set -C; : > lock_file) 2> /dev/null
if (set -C; : > lock_file) 2> /dev/null
  :   # lock_file didn't exist: no user running the script
  echo "Another user is already running that script."
exit 65

#  Code snippet by Stephane Chazelas,
#+ with modifications by Paulo Marcel Coelho Aragao.

Processes may execute in parallel within different subshells. This permits breaking a complex task into subcomponents processed concurrently.

Example 20-3. Running parallel processes in subshells

	(cat list1 list2 list3 | sort | uniq > list123) &
	(cat list4 list5 list6 | sort | uniq > list456) &
	# Merges and sorts both sets of lists simultaneously.
	# Running in background ensures parallel execution.
	# Same effect as
	#   cat list1 list2 list3 | sort | uniq > list123 &
	#   cat list4 list5 list6 | sort | uniq > list456 &
	wait   # Don't execute the next command until subshells finish.
	diff list123 list456

Redirecting I/O to a subshell uses the "|" pipe operator, as in ls -al | (command).


A command block between curly braces does not launch a subshell.

{ command1; command2; command3; ... }

Chapter 21. Restricted Shells

Disabled commands in restricted shells

Running a script or portion of a script in restricted mode disables certain commands that would otherwise be available. This is a security measure intended to limit the privileges of the script user and to minimize possible damage from running the script.

Using cd to change the working directory.

Changing the values of the $PATH, $SHELL, $BASH_ENV, or $ENV environmental variables.

Reading or changing the $SHELLOPTS, shell environmental options.

Output redirection.

Invoking commands containing one or more /'s.

Invoking exec to substitute a different process for the shell.

Various other commands that would enable monkeying with or attempting to subvert the script for an unintended purpose.

Getting out of restricted mode within the script.

Example 21-1. Running a script in restricted mode

# Starting the script with "#!/bin/bash -r"
# runs entire script in restricted mode.


echo "Changing directory."
cd /usr/local
echo "Now in `pwd`"
echo "Coming back home."
echo "Now in `pwd`"

# Everything up to here in normal, unrestricted mode.

set -r
# set --restricted    has same effect.
echo "==> Now in restricted mode. <=="


echo "Attempting directory change in restricted mode."
cd ..
echo "Still in `pwd`"


echo "\$SHELL = $SHELL"
echo "Attempting to change shell in restricted mode."
echo "\$SHELL= $SHELL"


echo "Attempting to redirect output in restricted mode."
ls -l /usr/bin > bin.files
ls -l bin.files    # Try to list attempted file creation effort.


exit 0

Chapter 22. Process Substitution

Process substitution is the counterpart to command substitution. Command substitution sets a variable to the result of a command, as in dir_contents=`ls -al` or xref=$( grep word datafile). Process substitution feeds the output of a process to another process (in other words, it sends the results of a command to another command).

Command substitution template

command within parentheses



These initiate process substitution. This uses /dev/fd/<n> files to send the results of the process within parentheses to another process. [52]


There is no space between the the "<" or ">" and the parentheses. Space there would give an error message.

bash$ echo >(true)

bash$ echo <(true)
Bash creates a pipe with two file descriptors, --fIn and fOut--. The stdin of true connects to fOut (dup2(fOut, 0)), then Bash passes a /dev/fd/fIn argument to echo. On systems lacking /dev/fd/<n> files, Bash may use temporary files. (Thanks, S.C.)

Process substitution can compare the output of two different commands, or even the output of different options to the same command.

bash$ comm <(ls -l) <(ls -al)
total 12
-rw-rw-r--    1 bozo bozo       78 Mar 10 12:58 File0
-rw-rw-r--    1 bozo bozo       42 Mar 10 12:58 File2
-rw-rw-r--    1 bozo bozo      103 Mar 10 12:58
        total 20
        drwxrwxrwx    2 bozo bozo     4096 Mar 10 18:10 .
        drwx------   72 bozo bozo     4096 Mar 10 17:58 ..
        -rw-rw-r--    1 bozo bozo       78 Mar 10 12:58 File0
        -rw-rw-r--    1 bozo bozo       42 Mar 10 12:58 File2
        -rw-rw-r--    1 bozo bozo      103 Mar 10 12:58

Using process substitution to compare the contents of two directories (to see which filenames are in one, but not the other):
diff <(ls $first_directory) <(ls $second_directory)

Some other usages and uses of process substitution:

cat <(ls -l)
# Same as     ls -l | cat

sort -k 9 <(ls -l /bin) <(ls -l /usr/bin) <(ls -l /usr/X11R6/bin)
# Lists all the files in the 3 main 'bin' directories, and sorts by filename.
# Note that three (count 'em) distinct commands are fed to 'sort'.

diff <(command1) <(command2)    # Gives difference in command output.

tar cf >(bzip2 -c > file.tar.bz2) $directory_name
# Calls "tar cf /dev/fd/?? $directory_name", and "bzip2 -c > file.tar.bz2".
# Because of the /dev/fd/<n> system feature,
# the pipe between both commands does not need to be named.
# This can be emulated.
bzip2 -c < pipe > file.tar.bz2&
tar cf pipe $directory_name
rm pipe
#        or
exec 3>&1
tar cf /dev/fd/4 $directory_name 4>&1 >&3 3>&- | bzip2 -c > file.tar.bz2 3>&-
exec 3>&-

# Thanks, Stepane Chazelas

A reader sent in the following interesting example of process substitution.

# Script fragment taken from SuSE distribution:

while read  des what mask iface; do
# Some commands ...
done < <(route -n)  

# To test it, let's make it do something.
while read  des what mask iface; do
  echo $des $what $mask $iface
done < <(route -n)  

# Output:
# Kernel IP routing table
# Destination Gateway Genmask Flags Metric Ref Use Iface
# U 0 0 0 lo

# As Stephane Chazelas points out, an easier-to-understand equivalent is:
route -n |
  while read des what mask iface; do   # Variables set from output of pipe.
    echo $des $what $mask $iface
  done  #  This yields the same output as above.
        #  However, as Ulrich Gayer points out . . .
        #+ this simplified equivalent uses a subshell for the while loop,
        #+ and therefore the variables disappear when the pipe terminates.

#  However, Filip Moritz comments that there is a subtle difference
#+ between the above two examples, as the following shows.

route -n | while read x; do ((y++)); done
echo $y # $y is still unset

while read x; do ((y++)); done < <(route -n)
echo $y # $y has the number of lines of output of route -n

More generally spoken
: | x=x
# seems to start a subshell like
: | ( x=x )
# while
x=x < <(:)
# does not

# This is useful, when parsing csv and the like.
# That is, in effect, what the original SuSE code fragment does.

Chapter 23. Functions

Like "real" programming languages, Bash has functions, though in a somewhat limited implementation. A function is a subroutine, a code block that implements a set of operations, a "black box" that performs a specified task. Wherever there is repetitive code, when a task repeats with only slight variations, then consider using a function.

function function_name {


function_name () {

This second form will cheer the hearts of C programmers (and is more portable).

As in C, the function's opening bracket may optionally appear on the second line.

function_name ()

Functions are called, triggered, simply by invoking their names.

Example 23-1. Simple functions



funky ()
{ # This is about as simple as functions get.
  echo "This is a funky function."
  echo "Now exiting funky function."
} # Function declaration must precede call.

fun ()
{ # A somewhat more complex function.

  echo "And now the fun really begins."

  sleep $JUST_A_SECOND    # Hey, wait a second!
  while [ $i -lt $REPEATS ]
    echo "----------FUNCTIONS---------->"
    echo "<------------ARE-------------"
    echo "<------------FUN------------>"
    let "i+=1"

  # Now, call the functions.


exit 0

The function definition must precede the first call to it. There is no method of "declaring" the function, as, for example, in C.
# Will give an error message, since function "f1" not yet defined.

declare -f f1      # This doesn't help either.
f1                 # Still an error message.

# However...

f1 ()
  echo "Calling function \"f2\" from within function \"f1\"."

f2 ()
  echo "Function \"f2\"."

f1  #  Function "f2" is not actually called until this point,
    #+ although it is referenced before its definition.
    #  This is permissable.
    # Thanks, S.C.

It is even possible to nest a function within another function, although this is not very useful.
f1 ()

  f2 () # nested
    echo "Function \"f2\", inside \"f1\"."


f2  #  Gives an error message.
    #  Even a preceding "declare -f f2" wouldn't help.


f1  #  Does nothing, since calling "f1" does not automatically call "f2".
f2  #  Now, it's all right to call "f2",
    #+ since its definition has been made visible by calling "f1".

    # Thanks, S.C.

Function declarations can appear in unlikely places, even where a command would otherwise go.
ls -l | foo() { echo "foo"; }  # Permissable, but useless.

if [ "$USER" = bozo ]
  bozo_greet ()   # Function definition embedded in an if/then construct.
    echo "Hello, Bozo."

bozo_greet        # Works only for Bozo, and other users get an error.

# Something like this might be useful in some contexts.
NO_EXIT=1   # Will enable function definition below.

[[ $NO_EXIT -eq 1 ]] && exit() { true; }     # Function definition in an "and-list".
# If $NO_EXIT is 1, declares "exit ()".
# This disables the "exit" builtin by aliasing it to "true".

exit  # Invokes "exit ()" function, not "exit" builtin.

# Thanks, S.C.

23.1. Complex Functions and Function Complexities

Functions may process arguments passed to them and return an exit status to the script for further processing.

function_name $arg1 $arg2

The function refers to the passed arguments by position (as if they were positional parameters), that is, $1, $2, and so forth.

Example 23-2. Function Taking Parameters

# Functions and parameters

DEFAULT=default                             # Default param value.

func2 () {
   if [ -z "$1" ]                           # Is parameter #1 zero length?
     echo "-Parameter #1 is zero length.-"  # Or no parameter passed.
     echo "-Param #1 is \"$1\".-"

   variable=${1-$DEFAULT}                   #  What does
   echo "variable = $variable"              #+ parameter substitution show?
                                            #  ---------------------------
                                            #  It distinguishes between
                                            #+ no param and a null param.

   if [ "$2" ]
     echo "-Parameter #2 is \"$2\".-"

   return 0

echo "Nothing passed."   
func2                          # Called with no params

echo "Zero-length parameter passed."
func2 ""                       # Called with zero-length param

echo "Null parameter passed."
func2 "$uninitialized_param"   # Called with uninitialized param

echo "One parameter passed."   
func2 first           # Called with one param

echo "Two parameters passed."   
func2 first second    # Called with two params

echo "\"\" \"second\" passed."
func2 "" second       # Called with zero-length first parameter
echo                  # and ASCII string as a second one.

exit 0


The shift command works on arguments passed to functions (see Example 34-12).

In contrast to certain other programming languages, shell scripts normally pass only value parameters to functions. Variable names (which are actually pointers), if passed as parameters to functions, will be treated as string literals. Functions interpret their arguments literally.

Indirect variable references (see Example 35-2) provide a clumsy sort of mechanism for passing variable pointers to functions.

Example 23-3. Passing an indirect reference to a function

# Passing an indirect reference to a function.

echo_var ()
echo "$1"


echo_var "$message"        # Hello
# Now, let's pass an indirect reference to the function.
echo_var "${!message}"     # Goodbye

echo "-------------"

# What happens if we change the contents of "hello" variable?
Hello="Hello, again!"
echo_var "$message"        # Hello
echo_var "${!message}"     # Hello, again!

exit 0

The next logical question is whether parameters can be dereferenced after being passed to a function.

Example 23-4. Dereferencing a parameter passed to a function

# Dereferencing parameter passed to a function.
# Script by Bruce W. Clare.

dereference ()
     y=\$"$1"   # Name of variable.
     echo $y    # $Junk

     x=`eval "expr \"$y\" "`
     echo $1=$x
     eval "$1=\"Some Different Text \""  # Assign new value.

Junk="Some Text"
echo $Junk "before"    # Some Text before

dereference Junk
echo $Junk "after"     # Some Different Text after

exit 0

Example 23-5. Again, dereferencing a parameter passed to a function

# Dereferencing a parameter passed to a function.
#                (Complex Example)

ITERATIONS=3  # How many times to get input.

my_read () {
  #  Called with my_read varname,
  #+ outputs the previous value between brackets as the default value,
  #+ then asks for a new value.

  local local_var

  echo -n "Enter a value "
  eval 'echo -n "[$'$1'] "'  #  Previous value.
# eval echo -n "[\$$1] "     #  Easier to understand,
                             #+ but loses trailing space in user prompt.
  read local_var
  [ -n "$local_var" ] && eval $1=\$local_var

  # "And-list": if "local_var" then set "$1" to its value.


while [ "$icount" -le "$ITERATIONS" ]
  my_read var
  echo "Entry #$icount = $var"
  let "icount += 1"

# Thanks to Stephane Chazelas for providing this instructive example.

exit 0

Exit and Return

exit status

Functions return a value, called an exit status. The exit status may be explicitly specified by a return statement, otherwise it is the exit status of the last command in the function (0 if successful, and a non-zero error code if not). This exit status may be used in the script by referencing it as $?. This mechanism effectively permits script functions to have a "return value" similar to C functions.


Terminates a function. A return command [53] optionally takes an integer argument, which is returned to the calling script as the "exit status" of the function, and this exit status is assigned to the variable $?.

Example 23-6. Maximum of two numbers

# Maximum of two integers.

E_PARAM_ERR=-198    # If less than 2 params passed to function.
EQUAL=-199          # Return value if both params equal.

max2 ()             # Returns larger of two numbers.
{                   # Note: numbers compared must be less than 257.
if [ -z "$2" ]
  return $E_PARAM_ERR

if [ "$1" -eq "$2" ]
  return $EQUAL
  if [ "$1" -gt "$2" ]
    return $1
    return $2

max2 33 34

if [ "$return_val" -eq $E_PARAM_ERR ]
  echo "Need to pass two parameters to the function."
elif [ "$return_val" -eq $EQUAL ]
    echo "The two numbers are equal."
    echo "The larger of the two numbers is $return_val."

exit 0

#  Exercise (easy):
#  ---------------
#  Convert this to an interactive script,
#+ that is, have the script ask for input (two numbers).


For a function to return a string or array, use a dedicated variable.
  [[ -r /etc/passwd ]] && REPLY=$(echo $(wc -l < /etc/passwd))
  #  If /etc/passwd is readable, set REPLY to line count.
  #  Returns both a parameter value and status information.
  #  The 'echo' seems unnecessary, but . . .
  #+ it removes excess whitespace from the output.

if count_lines_in_etc_passwd
  echo "There are $REPLY lines in /etc/passwd."
  echo "Cannot count lines in /etc/passwd."

# Thanks, S.C.

Example 23-7. Converting numbers to Roman numerals


# Arabic number to Roman numeral conversion
# Range: 0 - 200
# It's crude, but it works.

# Extending the range and otherwise improving the script is left as an exercise.

# Usage: roman number-to-convert


if [ -z "$1" ]
  echo "Usage: `basename $0` number-to-convert"
  exit $E_ARG_ERR

if [ "$num" -gt $LIMIT ]
  echo "Out of range!"
  exit $E_OUT_OF_RANGE

to_roman ()   # Must declare function before first call to it.
let "remainder = number - factor"
while [ "$remainder" -ge 0 ]
  echo -n $rchar
  let "number -= factor"
  let "remainder = number - factor"

return $number
       # Exercise:
       # --------
       # Explain how this function works.
       # Hint: division by successive subtraction.

to_roman $num 100 C
to_roman $num 90 LXXXX
to_roman $num 50 L
to_roman $num 40 XL
to_roman $num 10 X
to_roman $num 9 IX
to_roman $num 5 V
to_roman $num 4 IV
to_roman $num 1 I


exit 0

See also Example 10-28.


The largest positive integer a function can return is 255. The return command is closely tied to the concept of exit status, which accounts for this particular limitation. Fortunately, there are various workarounds for those situations requiring a large integer return value from a function.

Example 23-8. Testing large return values in a function


# The largest positive value a function can return is 255.

return_test ()         # Returns whatever passed to it.
  return $1

return_test 27         # o.k.
echo $?                # Returns 27.
return_test 255        # Still o.k.
echo $?                # Returns 255.

return_test 257        # Error!
echo $?                # Returns 1 (return code for miscellaneous error).

# ======================================================
return_test -151896    # Do large negative numbers work?
echo $?                # Will this return -151896?
                       # No! It returns 168.
#  Version of Bash before 2.05b permitted
#+ large negative integer return values.
#  Newer versions of Bash plug this loophole.
#  This may break older scripts.
#  Caution!
# ======================================================

exit 0

A workaround for obtaining large integer "return values" is to simply assign the "return value" to a global variable.
Return_Val=   # Global variable to hold oversize return value of function.

alt_return_test ()
  return   # Returns 0 (success).

alt_return_test 1
echo $?                              # 0
echo "return value = $Return_Val"    # 1

alt_return_test 256
echo "return value = $Return_Val"    # 256

alt_return_test 257
echo "return value = $Return_Val"    # 257

alt_return_test 25701
echo "return value = $Return_Val"    #25701

An even more elegant method is to simply have the function echo its "return value to stdout," and to then capture it by command substitution. See the discussion of this in Section 34.7.

Example 23-9. Comparing two large integers

# Maximum of two LARGE integers.

#  This is the previous "" example,
#+ modified to permit comparing large integers.

EQUAL=0             # Return value if both params equal.
E_PARAM_ERR=-99999  # Not enough params passed to function.

max2 ()             # "Returns" larger of two numbers.
if [ -z "$2" ]
  echo $E_PARAM_ERR

if [ "$1" -eq "$2" ]
  echo $EQUAL
  if [ "$1" -gt "$2" ]

echo $retval        # Echoes (to stdout), rather than returning value.


return_val=$(max2 33001 33997)
#  This is actually a form of command substitution:
#+ treating a function as if it were a command,
#+ and assigning the stdout of the function to the variable 'return_val.'

# ========================= OUTPUT ========================
if [ "$return_val" -eq "$E_PARAM_ERR" ]
  echo "Error in parameters passed to comparison function!"
elif [ "$return_val" -eq "$EQUAL" ]
    echo "The two numbers are equal."
    echo "The larger of the two numbers is $return_val."
# =========================================================
exit 0

#  Exercises:
#  ---------
#  1) Find a more elegant way of testing
#+    the parameters passed to the function.
#  2) Simplify the if/then structure at "OUTPUT."
#  3) Rewrite the script to take input from command-line parameters.

See also Example A-8.

Exercise: Using what we have just learned, extend the previous Roman numerals example to accept arbitrarily large input.


Redirecting the stdin of a function

A function is essentially a code block, which means its stdin can be redirected (as in Example 3-1).

Example 23-10. Real name from username


# From username, gets "real name" from /etc/passwd.

ARGCOUNT=1  # Expect one arg.


if [ $# -ne "$ARGCOUNT" ]
  echo "Usage: `basename $0` USERNAME"

file_excerpt ()  # Scan file for pattern, the print relevant portion of line.
while read line  # while does not necessarily need "[ condition]"
  echo "$line" | grep $1 | awk -F":" '{ print $5 }'  # Have awk use ":" delimiter.
} <$file  # Redirect into function's stdin.

file_excerpt $pattern

# Yes, this entire script could be reduced to
#       grep PATTERN /etc/passwd | awk -F":" '{ print $5 }'
# or
#       awk -F: '/PATTERN/ {print $5}'
# or
#       awk -F: '($1 == "username") { print $5 }' # real name from username
# However, it might not be as instructive.

exit 0

There is an alternative, and perhaps less confusing method of redirecting a function's stdin. This involves redirecting the stdin to an embedded bracketed code block within the function.
# Instead of:
Function ()
 } < file

# Try this:
Function ()
   } < file

# Similarly,

Function ()  # This works.
   echo $*
  } | tr a b

Function ()  # This doesn't work.
  echo $*
} | tr a b   # A nested code block is mandatory here.

# Thanks, S.C.

23.2. Local Variables

What makes a variable "local"?

local variables

A variable declared as local is one that is visible only within the block of code in which it appears. It has local "scope". In a function, a local variable has meaning only within that function block.

Example 23-11. Local variable visibility

# Global and local variables inside a function.

func ()
  local loc_var=23       # Declared as local variable.
  echo                   # Uses the 'local' builtin.
  echo "\"loc_var\" in function = $loc_var"
  global_var=999         # Not declared as local.
                         # Defaults to global. 
  echo "\"global_var\" in function = $global_var"


# Now, to see if local variable "loc_var" exists outside function.

echo "\"loc_var\" outside function = $loc_var"
                                      # $loc_var outside function = 
                                      # No, $loc_var not visible globally.
echo "\"global_var\" outside function = $global_var"
                                      # $global_var outside function = 999
                                      # $global_var is visible globally.

exit 0
#  In contrast to C, a Bash variable declared inside a function
#+ is local *only* if declared as such.


Before a function is called, all variables declared within the function are invisible outside the body of the function, not just those explicitly declared as local.

func ()
global_var=37    #  Visible only within the function block
                 #+ before the function has been called. 
}                # END OF FUNCTION

echo "global_var = $global_var"  # global_var =
                                 #  Function "func" has not yet been called,
                                 #+ so $global_var is not visible here.

echo "global_var = $global_var"  # global_var = 37
                                 # Has been set by function call.

23.2.1. Local variables help make recursion possible.

Local variables permit recursion, [54] but this practice generally involves much computational overhead and is definitely not recommended in a shell script. [55]

Example 23-12. Recursion, using a local variable


#               factorial
#               ---------

# Does bash permit recursion?
# Well, yes, but...
# You gotta have rocks in your head to try it.


if [ -z "$1" ]
  echo "Usage: `basename $0` number"
  exit $E_WRONG_ARGS

if [ "$1" -gt $MAX_ARG ]
  echo "Out of range (5 is maximum)."
  # Let's get real now.
  # If you want greater range than this,
  # rewrite it in a real programming language.
  exit $E_RANGE_ERR

fact ()
  local number=$1
  # Variable "number" must be declared as local,
  # otherwise this doesn't work.
  if [ "$number" -eq 0 ]
    factorial=1    # Factorial of 0 = 1.
    let "decrnum = number - 1"
    fact $decrnum  # Recursive function call.
    let "factorial = $number * $?"

  return $factorial

fact $1
echo "Factorial of $1 is $?."

exit 0

See also Example A-17 for an example of recursion in a script. Be aware that recursion is resource-intensive and executes slowly, and is therefore generally not appropriate to use in a script.

23.3. Recursion Without Local Variables

A function may recursively call itself even without use of local variables.

Example 23-13. The Towers of Hanoi

#! /bin/bash
# The Towers Of Hanoi
# Bash script
# Copyright (C) 2000 Amit Singh. All Rights Reserved.
# Last tested under bash version 2.05b.0(13)-release
#  Used in "Advanced Bash Scripting Guide"
#+ with permission of script author.
#  Slightly modified and commented by ABS author.

#  The Tower of Hanoi is an old mathematical puzzle.
#  There are three vertical posts set in a base.
#  The first post has a set of annular rings stacked on it.
#  These rings are flat disks with a hole drilled out of the center,
#+ so they can slip over the posts.
#  The rings have different diameters, and they stack in descending
#+ order, according to size.
#  The smallest ring is on top, and the largest on the bottom.
#  The task is to transfer the stack of rings
#+ to one of the other posts.
#  You can move only one ring at a time to another post.
#  You are permitted to move rings back to the original post.
#  You may place a smaller ring atop a larger one,
#+ but *not* vice versa.
#  Again, it is forbidden to place a larger ring atop a smaller one.
#  For a small number of rings, only a few moves are required.
#+ For each additional ring,
#+ the required number of moves approximately doubles,
#+ and the "strategy" becomes increasingly complicated.
#  For more information, see
#         ...                   ...                    ...
#         | |                   | |                    | |
#        _|_|_                  | |                    | |
#       |_____|                 | |                    | |
#      |_______|                | |                    | |
#     |_________|               | |                    | |
#    |___________|              | |                    | |
#   |             |             | |                    | |
# .--------------------------------------------------------------.
# |**************************************************************|
#          #1                   #2                      #3

E_NOPARAM=66  # No parameter passed to script.
E_BADPARAM=67 # Illegal number of disks passed to script.
Moves=        # Global variable holding number of moves.
              # Modifications to original script.

dohanoi() {   # Recursive function.
    case $1 in
        dohanoi "$(($1-1))" $2 $4 $3
        echo move $2 "-->" $3
	let "Moves += 1"  # Modification to original script.
        dohanoi "$(($1-1))" $4 $3 $2

case $# in
    case $(($1>0)) in     # Must have at least one disk.
        dohanoi $1 1 3 2
        echo "Total moves = $Moves"
        exit 0;
        echo "$0: illegal value for number of disks";
        exit $E_BADPARAM;
    echo "usage: $0 N"
    echo "       Where \"N\" is the number of disks."
    exit $E_NOPARAM;

# Exercises:
# ---------
# 1) Would commands beyond this point ever be executed?
#    Why not? (Easy)
# 2) Explain the workings of the workings of the "dohanoi" function.
#    (Difficult)

Chapter 24. Aliases

A Bash alias is essentially nothing more than a keyboard shortcut, an abbreviation, a means of avoiding typing a long command sequence. If, for example, we include alias lm="ls -l | more" in the ~/.bashrc file, then each lm typed at the command line will automatically be replaced by a ls -l | more. This can save a great deal of typing at the command line and avoid having to remember complex combinations of commands and options. Setting alias rm="rm -i" (interactive mode delete) may save a good deal of grief, since it can prevent inadvertently losing important files.

In a script, aliases have very limited usefulness. It would be quite nice if aliases could assume some of the functionality of the C preprocessor, such as macro expansion, but unfortunately Bash does not expand arguments within the alias body. [56] Moreover, a script fails to expand an alias itself within "compound constructs", such as if/then statements, loops, and functions. An added limitation is that an alias will not expand recursively. Almost invariably, whatever we would like an alias to do could be accomplished much more effectively with a function.

Example 24-1. Aliases within a script

# Invoke with command line parameter to exercise last section of this script.

shopt -s expand_aliases
# Must set this option, else script will not expand aliases.

# First, some fun.
alias Jesse_James='echo "\"Alias Jesse James\" was a 1959 comedy starring Bob Hope."'

echo; echo; echo;

alias ll="ls -l"
# May use either single (') or double (") quotes to define an alias.

echo "Trying aliased \"ll\":"
ll /usr/X11R6/bin/mk*   #* Alias works.


prefix=mk*  # See if wild-card causes problems.
echo "Variables \"directory\" + \"prefix\" = $directory$prefix"

alias lll="ls -l $directory$prefix"

echo "Trying aliased \"lll\":"
lll         # Long listing of all files in /usr/X11R6/bin stating with mk.
# Alias handles concatenated variables, including wild-card o.k.



if [ TRUE ]
  alias rr="ls -l"
  echo "Trying aliased \"rr\" within if/then statement:"
  rr /usr/X11R6/bin/mk*   #* Error message results!
  # Aliases not expanded within compound statements.
  echo "However, previously expanded alias still recognized:"
  ll /usr/X11R6/bin/mk*


while [ $count -lt 3 ]
  alias rrr="ls -l"
  echo "Trying aliased \"rrr\" within \"while\" loop:"
  rrr /usr/X11R6/bin/mk*   #* Alias will not expand here either.
                           # line 57: rrr: command not found
  let count+=1

echo; echo

alias xyz='cat $0'   # Script lists itself.
                     # Note strong quotes.
#  This seems to work,
#+ although the Bash documentation suggests that it shouldn't.
#  However, as Steve Jacobson points out,
#+ the "$0" parameter expands immediately upon declaration of the alias.

exit 0

The unalias command removes a previously set alias.

Example 24-2. unalias: Setting and unsetting an alias


shopt -s expand_aliases  # Enables alias expansion.

alias llm='ls -al | more'


unalias llm              # Unset alias.
# Error message results, since 'llm' no longer recognized.

exit 0
bash$ ./
total 6
drwxrwxr-x    2 bozo     bozo         3072 Feb  6 14:04 .
drwxr-xr-x   40 bozo     bozo         2048 Feb  6 14:04 ..
-rwxr-xr-x    1 bozo     bozo          199 Feb  6 14:04

./ llm: command not found

Chapter 25. List Constructs

The "and list" and "or list" constructs provide a means of processing a number of commands consecutively. These can effectively replace complex nested if/then or even case statements.

Chaining together commands

and list

command-1 && command-2 && command-3 && ... command-n
Each command executes in turn provided that the previous command has given a return value of true (zero). At the first false (non-zero) return, the command chain terminates (the first command returning false is the last one to execute).

Example 25-1. Using an "and list" to test for command-line arguments

# "and list"

if [ ! -z "$1" ] && echo "Argument #1 = $1" && [ ! -z "$2" ] && echo "Argument #2 = $2"
  echo "At least 2 arguments passed to script."
  # All the chained commands return true.
  echo "Less than 2 arguments passed to script."
  # At least one of the chained commands returns false.
# Note that "if [ ! -z $1 ]" works, but its supposed equivalent,
#  if [ -n $1 ] does not. However, quoting fixes this.
#  if [ -n "$1" ] works.  Careful!
# It is best to always quote tested variables.

# This accomplishes the same thing, using "pure" if/then statements.
if [ ! -z "$1" ]
  echo "Argument #1 = $1"
if [ ! -z "$2" ]
  echo "Argument #2 = $2"
  echo "At least 2 arguments passed to script."
  echo "Less than 2 arguments passed to script."
# It's longer and less elegant than using an "and list".

exit 0

Example 25-2. Another command-line arg test using an "and list"


ARGS=1        # Number of arguments expected.
E_BADARGS=65  # Exit value if incorrect number of args passed.

test $# -ne $ARGS && echo "Usage: `basename $0` $ARGS argument(s)" && exit $E_BADARGS
# If condition-1 true (wrong number of args passed to script),
# then the rest of the line executes, and script terminates.

# Line below executes only if the above test fails.
echo "Correct number of arguments passed to this script."

exit 0

# To check exit value, do a "echo $?" after script termination.

Of course, an and list can also set variables to a default value.
arg1=$@       # Set $arg1 to command line arguments, if any.

[ -z "$arg1" ] && arg1=DEFAULT
              # Set to DEFAULT if not specified on command line.

or list

command-1 || command-2 || command-3 || ... command-n
Each command executes in turn for as long as the previous command returns false. At the first true return, the command chain terminates (the first command returning true is the last one to execute). This is obviously the inverse of the "and list".

Example 25-3. Using "or lists" in combination with an "and list"


#, not-so-cunning file deletion utility.
#  Usage: delete filename


if [ -z "$1" ]
  echo "Usage: `basename $0` filename"
  exit $E_BADARGS  # No arg? Bail out.
  file=$1          # Set filename.

[ ! -f "$file" ] && echo "File \"$file\" not found. \
Cowardly refusing to delete a nonexistent file."
# AND LIST, to give error message if file not present.
# Note echo message continued on to a second line with an escape.

[ ! -f "$file" ] || (rm -f $file; echo "File \"$file\" deleted.")
# OR LIST, to delete file if present.

# Note logic inversion above.
# AND LIST executes on true, OR LIST on false.

exit 0


If the first command in an "or list" returns true, it will execute.

# ==> The following snippets from the /etc/rc.d/init.d/single script by Miquel van Smoorenburg
#+==> illustrate use of "and" and "or" lists.
# ==> "Arrowed" comments added by document author.

[ -x /usr/bin/clear ] && /usr/bin/clear
  # ==> If /usr/bin/clear exists, then invoke it.
  # ==> Checking for the existence of a command before calling it
  #+==> avoids error messages and other awkward consequences.

  # ==> . . .

# If they want to run something in single user mode, might as well run it...
for i in /etc/rc1.d/S[0-9][0-9]* ; do
        # Check if the script is there.
        [ -x "$i" ] || continue
  # ==> If corresponding file in $PWD *not* found,
  #+==> then "continue" by jumping to the top of the loop.

        # Reject backup files and files generated by rpm.
        case "$1" in
        [ "$i" = "/etc/rc1.d/S00single" ] && continue
  # ==> Set script name, but don't execute it yet.
        $i start

  # ==> . . .


The exit status of an and list or an or list is the exit status of the last command executed.

Clever combinations of "and" and "or" lists are possible, but the logic may easily become convoluted and require extensive debugging.
false && true || echo false         # false

# Same result as
( false && true ) || echo false     # false
# But *not*
false && ( true || echo false )     # (nothing echoed)

#  Note left-to-right grouping and evaluation of statements,
#+ since the logic operators "&&" and "||" have equal precedence.

#  It's best to avoid such complexities, unless you know what you're doing.

#  Thanks, S.C.

See Example A-8 and Example 7-4 for illustrations of using an and / or list to test variables.

Chapter 26. Arrays

Newer versions of Bash support one-dimensional arrays. Array elements may be initialized with the variable[xx] notation. Alternatively, a script may introduce the entire array by an explicit declare -a variable statement. To dereference (find the contents of) an array element, use curly bracket notation, that is, ${variable[xx]}.

Example 26-1. Simple array usage



# Array members need not be consecutive or contiguous.

# Some members of the array can be left uninitialized.
# Gaps in the array are o.k.

echo -n "area[11] = "
echo ${area[11]}    #  {curly brackets} needed

echo -n "area[13] = "
echo ${area[13]}

echo "Contents of area[51] are ${area[51]}."

# Contents of uninitialized array variable print blank.
echo -n "area[43] = "
echo ${area[43]}
echo "(area[43] unassigned)"


# Sum of two array variables assigned to third
area[5]=`expr ${area[11]} + ${area[13]}`
echo "area[5] = area[11] + area[13]"
echo -n "area[5] = "
echo ${area[5]}

area[6]=`expr ${area[11]} + ${area[51]}`
echo "area[6] = area[11] + area[51]"
echo -n "area[6] = "
echo ${area[6]}
# This fails because adding an integer to a string is not permitted.

echo; echo; echo

# -----------------------------------------------------------------
# Another array, "area2".
# Another way of assigning array variables...
# array_name=( XXX YYY ZZZ ... )

area2=( zero one two three four )

echo -n "area2[0] = "
echo ${area2[0]}
# Aha, zero-based indexing (first element of array is [0], not [1]).

echo -n "area2[1] = "
echo ${area2[1]}    # [1] is second element of array.
# -----------------------------------------------------------------

echo; echo; echo

# -----------------------------------------------
# Yet another array, "area3".
# Yet another way of assigning array variables...
# array_name=([xx]=XXX [yy]=YYY ...)

area3=([17]=seventeen [24]=twenty-four)

echo -n "area3[17] = "
echo ${area3[17]}

echo -n "area3[24] = "
echo ${area3[24]}
# -----------------------------------------------

exit 0


Bash permits array operations on variables, even if the variables are not explicitly declared as arrays.
echo ${string[@]}               # abcABC123ABCabc
echo ${string[*]}               # abcABC123ABCabc 
echo ${string[0]}               # abcABC123ABCabc
echo ${string[1]}               # No output!
                                # Why?
echo ${#string[@]}              # 1
                                # One element in the array.
                                # The string itself.

# Thank you, Michael Zick, for pointing this out.
Once again this demonstrates that Bash variables are untyped.

Example 26-2. Formatting a poem

# Pretty-prints one of the document author's favorite poems.

# Lines of the poem (single stanza).
Line[1]="I do not know which to prefer,"
Line[2]="The beauty of inflections"
Line[3]="Or the beauty of innuendoes,"
Line[4]="The blackbird whistling"
Line[5]="Or just after."

# Attribution.
Attrib[1]=" Wallace Stevens"
Attrib[2]="\"Thirteen Ways of Looking at a Blackbird\""
# This poem is in the Public Domain (copyright expired).


for index in 1 2 3 4 5    # Five lines.
  printf "     %s\n" "${Line[index]}"

for index in 1 2          # Two attribution lines.
  printf "          %s\n" "${Attrib[index]}"


exit 0

# Exercise:
# --------
# Modify this script to pretty-print a poem from a text data file.

Array variables have a syntax all their own, and even standard Bash commands and operators have special options adapted for array use.

Example 26-3. Various array operations

# More fun with arrays.

array=( zero one two three four five )
# Element 0   1   2    3     4    5

echo ${array[0]}       #  zero
echo ${array:0}        #  zero
                       #  Parameter expansion of first element,
                       #+ starting at position # 0 (1st character).
echo ${array:1}        #  ero
                       #  Parameter expansion of first element,
                       #+ starting at position # 1 (2nd character).

echo "--------------"

echo ${#array[0]}      #  4
                       #  Length of first element of array.
echo ${#array}         #  4
                       #  Length of first element of array.
                       #  (Alternate notation)

echo ${#array[1]}      #  3
                       #  Length of second element of array.
                       #  Arrays in Bash have zero-based indexing.

echo ${#array[*]}      #  6
                       #  Number of elements in array.
echo ${#array[@]}      #  6
                       #  Number of elements in array.

echo "--------------"

array2=( [0]="first element" [1]="second element" [3]="fourth element" )

echo ${array2[0]}      # first element
echo ${array2[1]}      # second element
echo ${array2[2]}      #
                       # Skipped in initialization, and therefore null.
echo ${array2[3]}      # fourth element

exit 0

Many of the standard string operations work on arrays.

Example 26-4. String operations on arrays

# String operations on arrays.
# Script by Michael Zick.
# Used with permission.

#  In general, any string operation in the ${name ... } notation
#+ can be applied to all string elements in an array
#+ with the ${name[@] ... } or ${name[*] ...} notation.

arrayZ=( one two three four five five )


# Trailing Substring Extraction
echo ${arrayZ[@]:0}     # one two three four five five
                        # All elements.

echo ${arrayZ[@]:1}     # two three four five five
                        # All elements following element[0].

echo ${arrayZ[@]:1:2}   # two three
                        # Only the two elements after element[0].

echo "-----------------------"

#  Substring Removal
#  Removes shortest match from front of string(s),
#+ where the substring is a regular expression.

echo ${arrayZ[@]#f*r}   # one two three five five
                        # Applied to all elements of the array.
                        # Matches "four" and removes it.

# Longest match from front of string(s)
echo ${arrayZ[@]##t*e}  # one two four five five
                        # Applied to all elements of the array.
                        # Matches "three" and removes it.

# Shortest match from back of string(s)
echo ${arrayZ[@]%h*e}   # one two t four five five
                        # Applied to all elements of the array.
                        # Matches "hree" and removes it.

# Longest match from back of string(s)
echo ${arrayZ[@]%%t*e}  # one two four five five
                        # Applied to all elements of the array.
                        # Matches "three" and removes it.

echo "-----------------------"

# Substring Replacement

# Replace first occurance of substring with replacement
echo ${arrayZ[@]/fiv/XYZ}   # one two three four XYZe XYZe
                            # Applied to all elements of the array.

# Replace all occurances of substring
echo ${arrayZ[@]//iv/YY}    # one two three four fYYe fYYe
                            # Applied to all elements of the array.

# Delete all occurances of substring
# Not specifing a replacement means 'delete'
echo ${arrayZ[@]//fi/}      # one two three four ve ve
                            # Applied to all elements of the array.

# Replace front-end occurances of substring
echo ${arrayZ[@]/#fi/XY}    # one two three four XYve XYve
                            # Applied to all elements of the array.

# Replace back-end occurances of substring
echo ${arrayZ[@]/%ve/ZZ}    # one two three four fiZZ fiZZ
                            # Applied to all elements of the array.

echo ${arrayZ[@]/%o/XX}     # one twXX three four five five
                            # Why?

echo "-----------------------"

# Before reaching for awk (or anything else) --
# Recall:
#   $( ... ) is command substitution.
#   Functions run as a sub-process.
#   Functions write their output to stdout.
#   Assignment reads the function's stdout.
#   The name[@] notation specifies a "for-each" operation.

newstr() {
    echo -n "!!!"

echo ${arrayZ[@]/%e/$(newstr)}
# on!!! two thre!!! four fiv!!! fiv!!!
# Q.E.D: The replacement action is an 'assignment.'

#  Accessing the "For-Each"
echo ${arrayZ[@]//*/$(newstr optional_arguments)}
#  Now, if Bash would just pass the matched string as $0
#+ to the function being called . . .


exit 0

Command substitution can construct the individual elements of an array.

Example 26-5. Loading the contents of a script into an array

# Loads this script into an array.
# Inspired by an e-mail from Chris Martin (thanks!).

script_contents=( $(cat "$0") )  #  Stores contents of this script ($0)
                                 #+ in an array.

for element in $(seq 0 $((${#script_contents[@]} - 1)))
  do                #  ${#script_contents[@]}
                    #+ gives number of elements in the array.
                    #  Question:
                    #  Why is  seq 0  necessary?
                    #  Try changing it to seq 1.
  echo -n "${script_contents[$element]}"
                    # List each field of this script on a single line.
  echo -n " -- "    # Use " -- " as a field separator.


exit 0

# Exercise:
# --------
#  Modify this script so it lists itself
#+ in its original format,
#+ complete with whitespace, line breaks, etc.

In an array context, some Bash builtins have a slightly altered meaning. For example, unset deletes array elements, or even an entire array.

Example 26-6. Some special properties of arrays


declare -a colors
#  All subsequent commands in this script will treat
#+ the variable "colors" as an array.

echo "Enter your favorite colors (separated from each other by a space)."

read -a colors    # Enter at least 3 colors to demonstrate features below.
#  Special option to 'read' command,
#+ allowing assignment of elements in an array.


# Special syntax to extract number of elements in array.
#     element_count=${#colors[*]} works also.
#  The "@" variable allows word splitting within quotes
#+ (extracts variables separated by whitespace).
#  This corresponds to the behavior of "$@" and "$*"
#+ in positional parameters. 


while [ "$index" -lt "$element_count" ]
do    # List all the elements in the array.
  echo ${colors[$index]}
  let "index = $index + 1"
# Each array element listed on a separate line.
# If this is not desired, use  echo -n "${colors[$index]} "
# Doing it with a "for" loop instead:
#   for i in "${colors[@]}"
#   do
#     echo "$i"
#   done
# (Thanks, S.C.)


# Again, list all the elements in the array, but using a more elegant method.
  echo ${colors[@]}          # echo ${colors[*]} also works.


# The "unset" command deletes elements of an array, or entire array.
unset colors[1]              # Remove 2nd element of array.
                             # Same effect as   colors[1]=
echo  ${colors[@]}           # List array again, missing 2nd element.

unset colors                 # Delete entire array.
                             #  unset colors[*] and
                             #+ unset colors[@] also work.
echo; echo -n "Colors gone."			   
echo ${colors[@]}            # List array again, now empty.

exit 0

As seen in the previous example, either ${array_name[@]} or ${array_name[*]} refers to all the elements of the array. Similarly, to get a count of the number of elements in an array, use either ${#array_name[@]} or ${#array_name[*]}. ${#array_name} is the length (number of characters) of ${array_name[0]}, the first element of the array.

Example 26-7. Of empty arrays and empty elements


#  Thanks to Stephane Chazelas for the original example,
#+ and to Michael Zick for extending it.

# An empty array is not the same as an array with empty elements.

array0=( first second third )
array1=( '' )   # "array1" consists of one empty element.
array2=( )      # No elements . . . "array2" is empty.

echo "Elements in array0:  ${array0[@]}"
echo "Elements in array1:  ${array1[@]}"
echo "Elements in array2:  ${array2[@]}"
echo "Length of first element in array0 = ${#array0}"
echo "Length of first element in array1 = ${#array1}"
echo "Length of first element in array2 = ${#array2}"
echo "Number of elements in array0 = ${#array0[*]}"  # 3
echo "Number of elements in array1 = ${#array1[*]}"  # 1  (Surprise!)
echo "Number of elements in array2 = ${#array2[*]}"  # 0

# ===================================================================


# Try extending those arrays.

# Adding an element to an array.
array0=( "${array0[@]}" "new1" )
array1=( "${array1[@]}" "new1" )
array2=( "${array2[@]}" "new1" )


# or


# When extended as above; arrays are 'stacks'
# The above is the 'push'
# The stack 'height' is:
echo "Stack height for array2 = $height"

# The 'pop' is:
unset array2[${#array2[@]}-1]	#  Arrays are zero-based,
height=${#array2[@]}            #+ which means first element has index 0.
echo "POP"
echo "New stack height for array2 = $height"


# List only 2nd and 3rd elements of array0.
from=1		# Zero-based numbering.
to=2		#
array3=( ${array0[@]:1:2} )
echo "Elements in array3:  ${array3[@]}"

# Works like a string (array of characters).
# Try some other "string" forms.

# Replacement:
array4=( ${array0[@]/second/2nd} )
echo "Elements in array4:  ${array4[@]}"

# Replace all matching wildcarded string.
array5=( ${array0[@]//new?/old} )
echo "Elements in array5:  ${array5[@]}"

# Just when you are getting the feel for this . . .
array6=( ${array0[@]#*new} )
echo # This one might surprise you.
echo "Elements in array6:  ${array6[@]}"

array7=( ${array0[@]#new1} )
echo # After array6 this should not be a surprise.
echo "Elements in array7:  ${array7[@]}"

# Which looks a lot like . . .
array8=( ${array0[@]/new1/} )
echo "Elements in array8:  ${array8[@]}"

#  So what can one say about this?

#  The string operations are performed on
#+ each of the elements in var[@] in succession.
#  Therefore : Bash supports string vector operations
#+ if the result is a zero length string,
#+ that element disappears in the resulting assignment.

#  Question, are those strings hard or soft quotes?

array9=( ${array0[@]/$zap/} )
echo "Elements in array9:  ${array9[@]}"

# Just when you thought you where still in Kansas . . .
array10=( ${array0[@]#$zap} )
echo "Elements in array10:  ${array10[@]}"

# Compare array7 with array10.
# Compare array8 with array9.

# Answer: must be soft quotes.

exit 0

The relationship of ${array_name[@]} and ${array_name[*]} is analogous to that between $@ and $*. This powerful array notation has a number of uses.

# Copying an array.
array2=( "${array1[@]}" )
# or

# Adding an element to an array.
array=( "${array[@]}" "new element" )
# or
array[${#array[*]}]="new element"

# Thanks, S.C.


The array=( element1 element2 ... elementN ) initialization operation, with the help of command substitution, makes it possible to load the contents of a text file into an array.



#            cat sample_file
#            1 a b c
#            2 d e fg

declare -a array1

array1=( `cat "$filename"`)                #  Loads contents
#         List file to stdout              #+ of $filename into array1.
#  array1=( `cat "$filename" | tr '\n' ' '`)
#                            change linefeeds in file to spaces. 
#  Not necessary because Bash does word splitting,
#+ changing linefeeds to spaces.

echo ${array1[@]}            # List the array.
#                              1 a b c 2 d e fg
#  Each whitespace-separated "word" in the file
#+ has been assigned to an element of the array.

echo $element_count          # 8

Clever scripting makes it possible to add array operations.

Example 26-8. Initializing arrays

#! /bin/bash
# array-assign.bash

#  Array operations are Bash specific,
#+ hence the ".bash" in the script name.

# Copyright (c) Michael S. Zick, 2003, All rights reserved.
# License: Unrestricted reuse in any form, for any purpose.
# Version: $ID$
# Clarification and additional comments by William Park.

#  Based on an example provided by Stephane Chazelas
#+ which appeared in the book: Advanced Bash Scripting Guide.

# Output format of the 'times' command:
# User CPU <space> System CPU
# User CPU of dead children <space> System CPU of dead children

#  Bash has two versions of assigning all elements of an array
#+ to a new array variable.
#  Both drop 'null reference' elements
#+ in Bash versions 2.04, 2.05a and 2.05b.
#  An additional array assignment that maintains the relationship of
#+ [subscript]=value for arrays may be added to newer versions.

#  Constructs a large array using an internal command,
#+ but anything creating an array of several thousand elements
#+ will do just fine.

declare -a bigOne=( /dev/* )
echo 'Conditions: Unquoted, default IFS, All-Elements-Of'
echo "Number of elements in array is ${#bigOne[@]}"

# set -vx

echo '- - testing: =( ${array[@]} ) - -'
declare -a bigTwo=( ${bigOne[@]} )
#                 ^              ^

echo '- - testing: =${array[@]} - -'
declare -a bigThree=${bigOne[@]}
# No parentheses this time.

#  Comparing the numbers shows that the second form, pointed out
#+ by Stephane Chazelas, is from three to four times faster.
#  William Park explains:
#+ The bigTwo array assigned as single string, whereas
#+ bigThree assigned element by element.
#  So, in essence, you have:
#                   bigTwo=( [0]="... ... ..." )
#                   bigThree=( [0]="..." [1]="..." [2]="..." ... )

#  I will continue to use the first form in my example descriptions
#+ because I think it is a better illustration of what is happening.

#  The reusable portions of my examples will actual contain
#+ the second form where appropriate because of the speedup.

# MSZ: Sorry about that earlier oversight folks.

#  Note:
#  ----
#  The "declare -a" statements in lines 31 and 43
#+ are not strictly necessary, since it is implicit
#+ in the  Array=( ... )  assignment form.
#  However, eliminating these declarations slows down
#+ the execution of the following sections of the script.
#  Try it, and see what happens.

exit 0


Adding a superfluous declare -a statement to an array declaration may speed up execution of subsequent operations on the array.

Example 26-9. Copying and concatenating arrays

#! /bin/bash
# This script written by Michael Zick.
# Used here with permission.

#  How-To "Pass by Name & Return by Name"
#+ or "Building your own assignment statement".

CpArray_Mac() {

# Assignment Command Statement Builder

    echo -n 'eval '
    echo -n "$2"                    # Destination name
    echo -n '=( ${'
    echo -n "$1"                    # Source name
    echo -n '[@]} )'

# That could all be a single command.
# Matter of style only.

declare -f CopyArray                # Function "Pointer"
CopyArray=CpArray_Mac               # Statement Builder


# Hype the array named $1.
# (Splice it together with array containing "Really Rocks".)
# Return in array named $2.

    local -a TMP
    local -a hype=( Really Rocks )

    $($CopyArray $1 TMP)
    TMP=( ${TMP[@]} ${hype[@]} )
    $($CopyArray TMP $2)

declare -a before=( Advanced Bash Scripting )
declare -a after

echo "Array Before = ${before[@]}"

Hype before after

echo "Array After = ${after[@]}"

# Too much hype?

echo "What ${after[@]:3:2}?"

declare -a modest=( ${after[@]:2:1} ${after[@]:3:2} )
#                    ---- substring extraction ----

echo "Array Modest = ${modest[@]}"

# What happened to 'before' ?

echo "Array Before = ${before[@]}"

exit 0

Example 26-10. More on concatenating arrays

#! /bin/bash
# array-append.bash

# Copyright (c) Michael S. Zick, 2003, All rights reserved.
# License: Unrestricted reuse in any form, for any purpose.
# Version: $ID$
# Slightly modified in formatting by M.C.

# Array operations are Bash-specific.
# Legacy UNIX /bin/sh lacks equivalents.

#  Pipe the output of this script to 'more'
#+ so it doesn't scroll off the terminal.

# Subscript packed.
declare -a array1=( zero1 one1 two1 )
# Subscript sparse ([1] is not defined).
declare -a array2=( [0]=zero2 [2]=two2 [3]=three2 )

echo '- Confirm that the array is really subscript sparse. -'
echo "Number of elements: 4"        # Hard-coded for illustration.
for (( i = 0 ; i < 4 ; i++ ))
    echo "Element [$i]: ${array2[$i]}"
# See also the more general code example in basics-reviewed.bash.

declare -a dest

# Combine (append) two arrays into a third array.
echo 'Conditions: Unquoted, default IFS, All-Elements-Of operator'
echo '- Undefined elements not present, subscripts not maintained. -'
# # The undefined elements do not exist; they are not being dropped.

dest=( ${array1[@]} ${array2[@]} )
# dest=${array1[@]}${array2[@]}     # Strange results, possibly a bug.

# Now, list the result.
echo '- - Testing Array Append - -'

echo "Number of elements: $cnt"
for (( i = 0 ; i < cnt ; i++ ))
    echo "Element [$i]: ${dest[$i]}"

# Assign an array to a single array element (twice).

# List the result.
echo '- - Testing modified array - -'

echo "Number of elements: $cnt"
for (( i = 0 ; i < cnt ; i++ ))
    echo "Element [$i]: ${dest[$i]}"

# Examine the modified second element.
echo '- - Reassign and list second element - -'

declare -a subArray=${dest[1]}

echo "Number of elements: $cnt"
for (( i = 0 ; i < cnt ; i++ ))
    echo "Element [$i]: ${subArray[$i]}"

#  The assignment of an entire array to a single element
#+ of another array using the '=${ ... }' array assignment
#+ has converted the array being assigned into a string,
#+ with the elements separated by a space (the first character of IFS).

# If the original elements didn't contain whitespace . . .
# If the original array isn't subscript sparse . . .
# Then we could get the original array structure back again.

# Restore from the modified second element.
echo '- - Listing restored element - -'

declare -a subArray=( ${dest[1]} )

echo "Number of elements: $cnt"
for (( i = 0 ; i < cnt ; i++ ))
    echo "Element [$i]: ${subArray[$i]}"
echo '- - Do not depend on this behavior. - -'
echo '- - This behavior is subject to change - -'
echo '- - in versions of Bash newer than version 2.05b - -'

# MSZ: Sorry about any earlier confusion folks.

exit 0


Arrays permit deploying old familiar algorithms as shell scripts. Whether this is necessarily a good idea is left to the reader to decide.

Example 26-11. An old friend: The Bubble Sort

# Bubble sort, of sorts.

# Recall the algorithm for a bubble sort. In this particular version...

#  With each successive pass through the array to be sorted,
#+ compare two adjacent elements, and swap them if out of order.
#  At the end of the first pass, the "heaviest" element has sunk to bottom.
#  At the end of the second pass, the next "heaviest" one has sunk next to bottom.
#  And so forth.
#  This means that each successive pass needs to traverse less of the array.
#  You will therefore notice a speeding up in the printing of the later passes.

  # Swaps two members of the array.
  local temp=${Countries[$1]} #  Temporary storage
                              #+ for element getting swapped out.

declare -a Countries  #  Declare array,
                      #+ optional here since it's initialized below.

#  Is it permissable to split an array variable over multiple lines
#+ using an escape (\)?
#  Yes.

Countries=(Netherlands Ukraine Zaire Turkey Russia Yemen Syria \
Brazil Argentina Nicaragua Japan Mexico Venezuela Greece England \
Israel Peru Canada Oman Denmark Wales France Kenya \
Xanadu Qatar Liechtenstein Hungary)

# "Xanadu" is the mythical place where, according to Coleridge,
#+ Kubla Khan did a pleasure dome decree.

clear                      # Clear the screen to start with. 

echo "0: ${Countries[*]}"  # List entire array at pass 0.

let "comparisons = $number_of_elements - 1"

count=1 # Pass number.

while [ "$comparisons" -gt 0 ]          # Beginning of outer loop

  index=0  # Reset index to start of array after each pass.

  while [ "$index" -lt "$comparisons" ] # Beginning of inner loop
    if [ ${Countries[$index]} \> ${Countries[`expr $index + 1`]} ]
    #  If out of order...
    #  Recalling that \> is ASCII comparison operator
    #+ within single brackets.

    #  if [[ ${Countries[$index]} > ${Countries[`expr $index + 1`]} ]]
    #+ also works.
      exchange $index `expr $index + 1`  # Swap.
    let "index += 1"
  done # End of inner loop

# ----------------------------------------------------------------------
# Paulo Marcel Coelho Aragao suggests for-loops as a simpler altenative.
# for (( last = $number_of_elements - 1 ; last > 1 ; last-- ))
# do
#     for (( i = 0 ; i < last ; i++ ))
#     do
#         [[ "${Countries[$i]}" > "${Countries[$((i+1))]}" ]] \
#             && exchange $i $((i+1))
#     done
# done
# ----------------------------------------------------------------------

let "comparisons -= 1" #  Since "heaviest" element bubbles to bottom,
                       #+ we need do one less comparison each pass.

echo "$count: ${Countries[@]}"  # Print resultant array at end of each pass.
let "count += 1"                # Increment pass count.

done                            # End of outer loop
                                # All done.

exit 0


Is it possible to nest arrays within arrays?

# "Nested" array.

#  Michael Zick provided this example,
#+ with corrections and clarifications by William Park.

AnArray=( $(ls --inode --ignore-backups --almost-all \
	--directory --full-time --color=none --time=status \
	--sort=time -l ${PWD} ) )  # Commands and options.

# Spaces are significant . . . and don't quote anything in the above.

SubArray=( ${AnArray[@]:11:1}  ${AnArray[@]:6:5} )
#  This array has six elements:
#+     SubArray=( [0]=${AnArray[11]} [1]=${AnArray[6]} [2]=${AnArray[7]}
#      [3]=${AnArray[8]} [4]=${AnArray[9]} [5]=${AnArray[10]} )
#  Arrays in Bash are (circularly) linked lists
#+ of type string (char *).
#  So, this isn't actually a nested array,
#+ but it's functionally similar.

echo "Current directory and date of last status change:"
echo "${SubArray[@]}"

exit 0


Embedded arrays in combination with indirect references create some fascinating possibilities

Example 26-12. Embedded arrays and indirect references

# Embedded arrays and indirect references.

# This script by Dennis Leeuw.
# Used with permission.
# Modified by document author.


        STRING="VAR1=value1 VAR2=value2 VAR3=value3"
)       # Embed ARRAY1 within this second array.

function print () {
        IFS=$'\n'       #  To print each array element
                        #+ on a separate line.
        local ${!TEST1} # See what happens if you delete this line.
        #  Indirect reference.
	#  This makes the components of $TEST1
	#+ accessible to this function.

        #  Let's see what we've got so far.
        echo "\$TEST1 = $TEST1"       #  Just the name of the variable.
        echo; echo
        echo "{\$TEST1} = ${!TEST1}"  #  Contents of the variable.
                                      #  That's what an indirect
                                      #+ reference does.
        echo "-------------------------------------------"; echo

        # Print variable
        echo "Variable VARIABLE: $VARIABLE"
        # Print a string element
        local ${!TEST2}      # Indirect reference (as above).
        echo "String element VAR2: $VAR2 from STRING"

        # Print an array element
        local ${!TEST2}      # Indirect reference (as above).
        echo "Array element VAR1_1: $VAR1_1 from ARRAY21"


exit 0

#   As the author of the script notes,
#+ "you can easily expand it to create named-hashes in bash."
#   (Difficult) exercise for the reader: implement this.


Arrays enable implementing a shell script version of the Sieve of Eratosthenes. Of course, a resource-intensive application of this nature should really be written in a compiled language, such as C. It runs excruciatingly slowly as a script.

Example 26-13. Complex array application: Sieve of Eratosthenes


# Sieve of Eratosthenes
# Ancient algorithm for finding prime numbers.

# This runs a couple of orders of magnitude
# slower than the equivalent C program.

LOWER_LIMIT=1       # Starting with 1.
UPPER_LIMIT=1000    # Up to 1000.
# (You may set this higher...  if you have time on your hands.)


# Optimization:
# Need to test numbers only halfway to upper limit.

declare -a Primes
# Primes[] is an array.

initialize ()
# Initialize the array.

until [ "$i" -gt "$UPPER_LIMIT" ]
  let "i += 1"
# Assume all array members guilty (prime)
# until proven innocent.

print_primes ()
# Print out the members of the Primes[] array tagged as prime.


until [ "$i" -gt "$UPPER_LIMIT" ]

  if [ "${Primes[i]}" -eq "$PRIME" ]
    printf "%8d" $i
    # 8 spaces per number gives nice, even columns.
  let "i += 1"


sift () # Sift out the non-primes.

let i=$LOWER_LIMIT+1
# We know 1 is prime, so let's start with 2.

until [ "$i" -gt "$UPPER_LIMIT" ]

if [ "${Primes[i]}" -eq "$PRIME" ]
# Don't bother sieving numbers already sieved (tagged as non-prime).


  while [ "$t" -le "$UPPER_LIMIT" ]
    let "t += $i "
    # Tag as non-prime all multiples.


  let "i += 1"


# Invoke the functions sequentially.
# This is what they call structured programming.


exit 0

# ----------------------------------------------- #
# Code below line will not execute.

# This improved version of the Sieve, by Stephane Chazelas,
# executes somewhat faster.

# Must invoke with command-line argument (limit of primes).

UPPER_LIMIT=$1                  # From command line.
let SPLIT=UPPER_LIMIT/2         # Halfway to max number.

Primes=( '' $(seq $UPPER_LIMIT) )

until (( ( i += 1 ) > SPLIT ))  # Need check only halfway.
  if [[ -n $Primes[i] ]]
    until (( ( t += i ) > UPPER_LIMIT ))
echo ${Primes[*]}

exit 0

Compare this array-based prime number generator with an alternative that does not use arrays, Example A-17.


Arrays lend themselves, to some extent, to emulating data structures for which Bash has no native support.

Example 26-14. Emulating a push-down stack

# push-down stack simulation

#  Similar to the CPU stack, a push-down stack stores data items
#+ sequentially, but releases them in reverse order, last-in first-out.

BP=100            #  Base Pointer of stack array.
                  #  Begin at element 100.

SP=$BP            #  Stack Pointer.
                  #  Initialize it to "base" (bottom) of stack.

Data=             #  Contents of stack location.  
                  #  Must use global variable,
                  #+ because of limitation on function return range.

declare -a stack

push()            # Push item on stack.
if [ -z "$1" ]    # Nothing to push?

let "SP -= 1"     # Bump stack pointer.


pop()                    # Pop item off stack.
Data=                    # Empty out data item.

if [ "$SP" -eq "$BP" ]   # Stack empty?
fi                       #  This also keeps SP from getting past 100,
                         #+ i.e., prevents a runaway stack.

let "SP += 1"            # Bump stack pointer.

status_report()          # Find out what's happening.
echo "-------------------------------------"
echo "REPORT"
echo "Stack Pointer = $SP"
echo "Just popped \""$Data"\" off the stack."
echo "-------------------------------------"

# =======================================================
# Now, for some fun.


# See if you can pop anything off empty stack.


push garbage
status_report     # Garbage in, garbage out.      

value1=23; push $value1
value2=skidoo; push $value2
value3=FINAL; push $value3

pop              # FINAL
pop              # skidoo
pop              # 23
status_report    # Last-in, first-out!

#  Notice how the stack pointer decrements with each push,
#+ and increments with each pop.

# =======================================================

# Exercises:
# ---------

# 1)  Modify the "push()" function to permit pushing
#   + multiple element on the stack with a single function call.

# 2)  Modify the "pop()" function to permit popping
#   + multiple element from the stack with a single function call.

# 3)  Using this script as a starting point,
#   + write a stack-based 4-function calculator.

exit 0


Fancy manipulation of array "subscripts" may require intermediate variables. For projects involving this, again consider using a more powerful programming language, such as Perl or C.

Example 26-15. Complex array application: Exploring a weird mathematical series


# Douglas Hofstadter's notorious "Q-series":

# Q(1) = Q(2) = 1
# Q(n) = Q(n - Q(n-1)) + Q(n - Q(n-2)), for n>2

# This is a "chaotic" integer series with strange and unpredictable behavior.
# The first 20 terms of the series are:
# 1 1 2 3 3 4 5 5 6 6 6 8 8 8 10 9 10 11 11 12 

# See Hofstadter's book, "Goedel, Escher, Bach: An Eternal Golden Braid",
# p. 137, ff.

LIMIT=100     # Number of terms to calculate
LINEWIDTH=20  # Number of terms printed per line

Q[1]=1        # First two terms of series are 1.

echo "Q-series [$LIMIT terms]:"
echo -n "${Q[1]} "             # Output first two terms.
echo -n "${Q[2]} "

for ((n=3; n <= $LIMIT; n++))  # C-like loop conditions.
do   # Q[n] = Q[n - Q[n-1]] + Q[n - Q[n-2]]  for n>2
# Need to break the expression into intermediate terms,
# since Bash doesn't handle complex array arithmetic very well.

  let "n1 = $n - 1"        # n-1
  let "n2 = $n - 2"        # n-2
  t0=`expr $n - ${Q[n1]}`  # n - Q[n-1]
  t1=`expr $n - ${Q[n2]}`  # n - Q[n-2]
  T0=${Q[t0]}              # Q[n - Q[n-1]]
  T1=${Q[t1]}              # Q[n - Q[n-2]]

Q[n]=`expr $T0 + $T1`      # Q[n - Q[n-1]] + Q[n - Q[n-2]]
echo -n "${Q[n]} "

if [ `expr $n % $LINEWIDTH` -eq 0 ]    # Format output.
then   #     mod
  echo # Break lines into neat chunks.



exit 0

# This is an iterative implementation of the Q-series.
# The more intuitive recursive implementation is left as an exercise.
# Warning: calculating this series recursively takes a *very* long time.


Bash supports only one-dimensional arrays, however a little trickery permits simulating multi-dimensional ones.

Example 26-16. Simulating a two-dimensional array, then tilting it

# Simulating a two-dimensional array.

# A one-dimensional array consists of a single row.
# A two-dimensional array stores rows sequentially.

# 5 X 5 Array.

declare -a alpha     # char alpha [Rows] [Columns];
                     # Unnecessary declaration. Why?

load_alpha ()
local rc=0
local index

for i in A B C D E F G H I J K L M N O P Q R S T U V W X Y
do     # Use different symbols if you like.
  local row=`expr $rc / $Columns`
  local column=`expr $rc % $Rows`
  let "index = $row * $Rows + $column"
# alpha[$row][$column]
  let "rc += 1"

#  Simpler would be
#+   declare -a alpha=( A B C D E F G H I J K L M N O P Q R S T U V W X Y )
#+ but this somehow lacks the "flavor" of a two-dimensional array.

print_alpha ()
local row=0
local index


while [ "$row" -lt "$Rows" ]   #  Print out in "row major" order:
do                             #+ columns vary,
                               #+ while row (outer loop) remains the same.
  local column=0

  echo -n "       "            #  Lines up "square" array with rotated one.
  while [ "$column" -lt "$Columns" ]
    let "index = $row * $Rows + $column"
    echo -n "${alpha[index]} "  # alpha[$row][$column]
    let "column += 1"

  let "row += 1"


# The simpler equivalent is
#   echo ${alpha[*]} | xargs -n $Columns


filter ()     # Filter out negative array indices.

echo -n "  "  # Provides the tilt.
              # Explain why.

if [[ "$1" -ge 0 &&  "$1" -lt "$Rows" && "$2" -ge 0 && "$2" -lt "$Columns" ]]
    let "index = $1 * $Rows + $2"
    # Now, print it rotated.
    echo -n " ${alpha[index]}"
    #           alpha[$row][$column]


rotate ()  #  Rotate the array 45 degrees --
{          #+ "balance" it on its lower lefthand corner.
local row
local column

for (( row = Rows; row > -Rows; row-- ))
  do       # Step through the array backwards. Why?

  for (( column = 0; column < Columns; column++ ))

    if [ "$row" -ge 0 ]
      let "t1 = $column - $row"
      let "t2 = $column"
      let "t1 = $column"
      let "t2 = $column + $row"

    filter $t1 $t2   # Filter out negative array indices. Why?

  echo; echo


#  Array rotation inspired by examples (pp. 143-146) in
#+ "Advanced C Programming on the IBM PC," by Herbert Mayer
#+ (see bibliography).
#  This just goes to show that much of what can be done in C
#+ can also be done in shell scripting.


#--------------- Now, let the show begin. ------------#
load_alpha     # Load the array.
print_alpha    # Print it out.  
rotate         # Rotate it 45 degrees counterclockwise.

exit 0

# This is a rather contrived, not to mention inelegant simulation.

# Exercises:
# ---------
# 1)  Rewrite the array loading and printing functions
#     in a more intuitive and less kludgy fashion.
# 2)  Figure out how the array rotation functions work.
#     Hint: think about the implications of backwards-indexing an array.
# 3)  Rewrite this script to handle a non-square array,
#     such as a 6 X 4 one.
#     Try to minimize "distortion" when the array is rotated.

A two-dimensional array is essentially equivalent to a one-dimensional one, but with additional addressing modes for referencing and manipulating the individual elements by "row" and "column" position.

For an even more elaborate example of simulating a two-dimensional array, see Example A-11.

Chapter 27. Files

startup files

These files contain the aliases and environmental variables made available to Bash running as a user shell and to all Bash scripts invoked after system initialization.


systemwide defaults, mostly setting the environment (all Bourne-type shells, not just Bash [57])


systemwide functions and aliases for Bash


user-specific Bash environmental default settings, found in each user's home directory (the local counterpart to /etc/profile)


user-specific Bash init file, found in each user's home directory (the local counterpart to /etc/bashrc). Only interactive shells and user scripts read this file. See Appendix J for a sample .bashrc file.

logout file


user-specific instruction file, found in each user's home directory. Upon exit from a login (Bash) shell, the commands in this file execute.

Chapter 28. /dev and /proc

A Linux or UNIX machine typically has the /dev and /proc special-purpose directories.

28.1. /dev

The /dev directory contains entries for the physical devices that may or may not be present in the hardware. [58] The hard drive partitions containing the mounted filesystem(s) have entries in /dev, as a simple df shows.
bash$ df
Filesystem           1k-blocks      Used Available Use%
 Mounted on
 /dev/hda6               495876    222748    247527  48% /
 /dev/hda1                50755      3887     44248   9% /boot
 /dev/hda8               367013     13262    334803   4% /home
 /dev/hda5              1714416   1123624    503704  70% /usr

Among other things, the /dev directory also contains loopback devices, such as /dev/loop0. A loopback device is a gimmick that allows an ordinary file to be accessed as if it were a block device. [59] This enables mounting an entire filesystem within a single large file. See Example 13-7 and Example 13-6.

A few of the pseudo-devices in /dev have other specialized uses, such as /dev/null, /dev/zero, /dev/urandom, /dev/sda1, /dev/udp, and /dev/tcp.

For instance:

To mount a USB flash drive, append the following line to /etc/fstab. [60]
/dev/sda1    /mnt/flashdrive    auto    noauto,user,noatime    0 0
(See also Example A-22.)

When executing a command on a /dev/tcp/$host/$port pseudo-device file, Bash opens a TCP connection to the associated socket. [61]

Getting the time from from

bash$ cat </dev/tcp/
53082 04-03-18 04:26:54 68 0 0 502.3 UTC(NIST) *

[Mark contributed the above example.]

Downloading a URL:

bash$ exec 5<>/dev/tcp/
bash$ echo -e "GET / HTTP/1.0\n" >&5
bash$ cat <&5

[Thanks, Mark and Mihai Maties.]

Example 28-1. Using /dev/tcp for troubleshooting

# /dev/tcp redirection to check Internet connection.

# Script by Troy Engel.
# Used with permission.
TCP_PORT=80   # Port 80 is http.
# Try to connect. (Somewhat similar to a 'ping.') 
echo "HEAD / HTTP/1.0" >/dev/tcp/${TCP_HOST}/${TCP_PORT}

If bash was compiled with --enable-net-redirections, it has the capability of
using a special character device for both TCP and UDP redirections. These
redirections are used identically as STDIN/STDOUT/STDERR. The device entries
are 30,36 for /dev/tcp:

  mknod /dev/tcp c 30 36

>From the bash reference:
    If host is a valid hostname or Internet address, and port is an integer
port number or service name, Bash attempts to open a TCP connection to the
corresponding socket.

if [ "X$MYEXIT" = "X0" ]; then
  echo "Connection successful. Exit code: $MYEXIT"
  echo "Connection unsuccessful. Exit code: $MYEXIT"

exit $MYEXIT

28.2. /proc

The /proc directory is actually a pseudo-filesystem. The files in /proc mirror currently running system and kernel processes and contain information and statistics about them.

bash$ cat /proc/devices
Character devices:
   1 mem
   2 pty
   3 ttyp
   4 ttyS
   5 cua
   7 vcs
  10 misc
  14 sound
  29 fb
  36 netlink
 128 ptm
 136 pts
 162 raw
 254 pcmcia

 Block devices:
   1 ramdisk
   2 fd
   3 ide0
   9 md

bash$ cat /proc/interrupts
   0:      84505          XT-PIC  timer
   1:       3375          XT-PIC  keyboard
   2:          0          XT-PIC  cascade
   5:          1          XT-PIC  soundblaster
   8:          1          XT-PIC  rtc
  12:       4231          XT-PIC  PS/2 Mouse
  14:     109373          XT-PIC  ide0
 NMI:          0 
 ERR:          0

bash$ cat /proc/partitions
major minor  #blocks  name     rio rmerge rsect ruse wio wmerge wsect wuse running use aveq

    3     0    3007872 hda 4472 22260 114520 94240 3551 18703 50384 549710 0 111550 644030
    3     1      52416 hda1 27 395 844 960 4 2 14 180 0 800 1140
    3     2          1 hda2 0 0 0 0 0 0 0 0 0 0 0
    3     4     165280 hda4 10 0 20 210 0 0 0 0 0 210 210

bash$ cat /proc/loadavg
0.13 0.42 0.27 2/44 1119

bash$ cat /proc/apm
1.16 1.2 0x03 0x01 0xff 0x80 -1% -1 ?

Shell scripts may extract data from certain of the files in /proc. [62]

FS=iso                       # ISO filesystem support in kernel?

grep $FS /proc/filesystems   # iso9660

kernel_version=$( awk '{ print $3 }' /proc/version )

CPU=$( awk '/model name/ {print $4}' < /proc/cpuinfo )

if [ $CPU = Pentium ]


bus_speed=$(grep Spd $devfile | awk '{print $9}')

if [ "$bus_speed" = "$USB1" ]
  echo "USB 1.1 port found."
  # Do something appropriate for USB 1.1.

The /proc directory contains subdirectories with unusual numerical names. Every one of these names maps to the process ID of a currently running process. Within each of these subdirectories, there are a number of files that hold useful information about the corresponding process. The stat and status files keep running statistics on the process, the cmdline file holds the command-line arguments the process was invoked with, and the exe file is a symbolic link to the complete path name of the invoking process. There are a few more such files, but these seem to be the most interesting from a scripting standpoint.

Example 28-2. Finding the process associated with a PID

# Gives complete path name to process associated with pid.

ARGNO=1  # Number of arguments the script expects.

if [ $# -ne $ARGNO ]
  echo "Usage: `basename $0` PID-number" >&2  # Error message >stderr.

pidno=$( ps ax | grep $1 | awk '{ print $1 }' | grep $1 )
# Checks for pid in "ps" listing, field #1.
# Then makes sure it is the actual process, not the process invoked by this script.
# The last "grep $1" filters out this possibility.
if [ -z "$pidno" ]  # If, after all the filtering, the result is a zero-length string,
then                # no running process corresponds to the pid given.
  echo "No such process running."

# Alternatively:
#   if ! ps $1 > /dev/null 2>&1
#   then                # no running process corresponds to the pid given.
#     echo "No such process running."
#    fi

# To simplify the entire process, use "pidof".

if [ ! -r "/proc/$1/$PROCFILE" ]  # Check for read permission.
  echo "Process $1 running, but..."
  echo "Can't get read permission on /proc/$1/$PROCFILE."
  exit $E_NOPERMISSION  # Ordinary user can't access some files in /proc.

# The last two tests may be replaced by:
#    if ! kill -0 $1 > /dev/null 2>&1 # '0' is not a signal, but
                                      # this will test whether it is possible
                                      # to send a signal to the process.
#    then echo "PID doesn't exist or you're not its owner" >&2
#    exit $E_BADPID
#    fi

exe_file=$( ls -l /proc/$1 | grep "exe" | awk '{ print $11 }' )
# Or       exe_file=$( ls -l /proc/$1/exe | awk '{print $11}' )
# /proc/pid-number/exe is a symbolic link
# to the complete path name of the invoking process.

if [ -e "$exe_file" ]  # If /proc/pid-number/exe exists...
then                 # the corresponding process exists.
  echo "Process #$1 invoked by $exe_file."
  echo "No such process running."

# This elaborate script can *almost* be replaced by
# ps ax | grep $1 | awk '{ print $5 }'
# However, this will not work...
# because the fifth field of 'ps' is argv[0] of the process,
# not the executable file path.
# However, either of the following would work.
#       find /proc/$1/exe -printf '%l\n'
#       lsof -aFn -p $1 -d txt | sed -ne 's/^n//p'

# Additional commentary by Stephane Chazelas.

exit 0

Example 28-3. On-line connect status


PROCNAME=pppd        # ppp daemon
PROCFILENAME=status  # Where to look.
INTERVAL=2           # Update every 2 seconds.

pidno=$( ps ax | grep -v "ps ax" | grep -v grep | grep $PROCNAME | awk '{ print $1 }' )
# Finding the process number of 'pppd', the 'ppp daemon'.
# Have to filter out the process lines generated by the search itself.
#  However, as Oleg Philon points out,
#+ this could have been considerably simplified by using "pidof".
#  pidno=$( pidof $PROCNAME )
#  Moral of the story:
#+ When a command sequence gets too complex, look for a shortcut.

if [ -z "$pidno" ]   # If no pid, then process is not running.
  echo "Not connected."
  echo "Connected."; echo

while [ true ]       # Endless loop, script can be improved here.

  if [ ! -e "/proc/$pidno/$PROCFILENAME" ]
  # While process running, then "status" file exists.
    echo "Disconnected."

netstat -s | grep "packets received"  # Get some connect statistics.
netstat -s | grep "packets delivered"

  sleep $INTERVAL
  echo; echo


exit 0

# As it stands, this script must be terminated with a Control-C.

#    Exercises:
#    ---------
#    Improve the script so it exits on a "q" keystroke.
#    Make the script more user-friendly in other ways.


In general, it is dangerous to write to the files in /proc, as this can corrupt the filesystem or crash the machine.

Chapter 29. Of Zeros and Nulls

/dev/zero and /dev/null

Uses of /dev/null

Think of /dev/null as a "black hole". It is the nearest equivalent to a write-only file. Everything written to it disappears forever. Attempts to read or output from it result in nothing. Nevertheless, /dev/null can be quite useful from both the command line and in scripts.

Suppressing stdout.
cat $filename >/dev/null
# Contents of the file will not list to stdout.

Suppressing stderr (from Example 12-3).
rm $badname 2>/dev/null
#           So error messages [stderr] deep-sixed.

Suppressing output from both stdout and stderr.
cat $filename 2>/dev/null >/dev/null
# If "$filename" does not exist, there will be no error message output.
# If "$filename" does exist, the contents of the file will not list to stdout.
# Therefore, no output at all will result from the above line of code.
#  This can be useful in situations where the return code from a command
#+ needs to be tested, but no output is desired.
# cat $filename &>/dev/null
#     also works, as Baris Cicek points out.

Deleting contents of a file, but preserving the file itself, with all attendant permissions (from Example 2-1 and Example 2-3):
cat /dev/null > /var/log/messages
#  : > /var/log/messages   has same effect, but does not spawn a new process.

cat /dev/null > /var/log/wtmp

Automatically emptying the contents of a logfile (especially good for dealing with those nasty "cookies" sent by Web commercial sites):

Example 29-1. Hiding the cookie jar

if [ -f ~/.netscape/cookies ]  # Remove, if exists.
  rm -f ~/.netscape/cookies

ln -s /dev/null ~/.netscape/cookies
# All cookies now get sent to a black hole, rather than saved to disk.
Uses of /dev/zero

Like /dev/null, /dev/zero is a pseudo file, but it actually contains nulls (binary zeros, not the ASCII kind). Output written to it disappears, and it is fairly difficult to actually read the nulls in /dev/zero, though it can be done with od or a hex editor. The chief use for /dev/zero is in creating an initialized dummy file of specified length intended as a temporary swap file.

Example 29-2. Setting up a swapfile using /dev/zero

# Creating a swapfile.

ROOT_UID=0         # Root has $UID 0.
E_WRONG_USER=65    # Not root?


# This script must be run as root.
if [ "$UID" -ne "$ROOT_UID" ]
  echo; echo "You must be root to run this script."; echo
  exit $E_WRONG_USER

blocks=${1:-$MINBLOCKS}          #  Set to default of 40 blocks,
                                 #+ if nothing specified on command line.
# This is the equivalent of the command block below.
# --------------------------------------------------
# if [ -n "$1" ]
# then
#   blocks=$1
# else
#   blocks=$MINBLOCKS
# fi
# --------------------------------------------------

if [ "$blocks" -lt $MINBLOCKS ]
  blocks=$MINBLOCKS              # Must be at least 40 blocks long.

echo "Creating swap file of size $blocks blocks (KB)."
dd if=/dev/zero of=$FILE bs=$BLOCKSIZE count=$blocks  # Zero out file.

mkswap $FILE $blocks             # Designate it a swap file.
swapon $FILE                     # Activate swap file.

echo "Swap file created and activated."


Another application of /dev/zero is to "zero out" a file of a designated size for a special purpose, such as mounting a filesystem on a loopback device (see Example 13-7) or securely deleting a file (see Example 12-50).

Example 29-3. Creating a ramdisk


#  A "ramdisk" is a segment of system RAM memory
#+ which acts as if it were a filesystem.
#  Its advantage is very fast access (read/write time).
#  Disadvantages: volatility, loss of data on reboot or powerdown.
#+                less RAM available to system.
#  Of what use is a ramdisk?
#  Keeping a large dataset, such as a table or dictionary on ramdisk,
#+ speeds up data lookup, since memory access is much faster than disk access.

E_NON_ROOT_USER=70             # Must run as root.

SIZE=2000                      # 2K blocks (change as appropriate)
BLOCKSIZE=1024                 # 1K (1024 byte) block size
DEVICE=/dev/ram0               # First ram device

username=`id -nu`
if [ "$username" != "$ROOTUSER_NAME" ]
  echo "Must be root to run \"`basename $0`\"."

if [ ! -d "$MOUNTPT" ]         #  Test whether mount point already there,
then                           #+ so no error if this script is run
  mkdir $MOUNTPT               #+ multiple times.

dd if=/dev/zero of=$DEVICE count=$SIZE bs=$BLOCKSIZE  # Zero out RAM device.
                                                      # Why is this necessary?
mke2fs $DEVICE                 # Create an ext2 filesystem on it.
mount $DEVICE $MOUNTPT         # Mount it.
chmod 777 $MOUNTPT             # Enables ordinary user to access ramdisk.
                               # However, must be root to unmount it.

echo "\"$MOUNTPT\" now available for use."
# The ramdisk is now accessible for storing files, even by an ordinary user.

#  Caution, the ramdisk is volatile, and its contents will disappear
#+ on reboot or power loss.
#  Copy anything you want saved to a regular directory.

# After reboot, run this script to again set up ramdisk.
# Remounting /mnt/ramdisk without the other steps will not work.

#  Suitably modified, this script can by invoked in /etc/rc.d/rc.local,
#+ to set up ramdisk automatically at bootup.
#  That may be appropriate on, for example, a database server.

exit 0

Chapter 30. Debugging

The Bash shell contains no debugger, nor even any debugging-specific commands or constructs. [63] Syntax errors or outright typos in the script generate cryptic error messages that are often of no help in debugging a non-functional script.

Example 30-1. A buggy script


# This is a buggy script.
# Where, oh where is the error?


if [$a -gt 27 ]
  echo $a

exit 0

Output from script:
./ [37: command not found
What's wrong with the above script (hint: after the if)?

Example 30-2. Missing keyword

# What error message will this generate?

for a in 1 2 3
  echo "$a"
# done     # Required keyword 'done' commented out in line 7.

exit 0  

Output from script: line 10: syntax error: unexpected end of file
Note that the error message does not necessarily reference the line in which the error occurs, but the line where the Bash interpreter finally becomes aware of the error.

Error messages may disregard comment lines in a script when reporting the line number of a syntax error.

What if the script executes, but does not work as expected? This is the all too familiar logic error.

Example 30-3. test24, another buggy script


#  This script is supposed to delete all filenames in current directory
#+ containing embedded spaces.
#  It doesn't work.
#  Why not?

badname=`ls | grep ' '`

# Try this:
# echo "$badname"

rm "$badname"

exit 0

Try to find out what's wrong with Example 30-3 by uncommenting the echo "$badname" line. Echo statements are useful for seeing whether what you expect is actually what you get.

In this particular case, rm "$badname" will not give the desired results because $badname should not be quoted. Placing it in quotes ensures that rm has only one argument (it will match only one filename). A partial fix is to remove to quotes from $badname and to reset $IFS to contain only a newline, IFS=$'\n'. However, there are simpler ways of going about it.
# Correct methods of deleting filenames containing spaces.
rm *\ *
rm *" "*
rm *' '*
# Thank you. S.C.

Summarizing the symptoms of a buggy script,

  1. It bombs with a "syntax error" message, or

  2. It runs, but does not work as expected (logic error).

  3. It runs, works as expected, but has nasty side effects (logic bomb).

Tools for debugging non-working scripts include

  1. echo statements at critical points in the script to trace the variables, and otherwise give a snapshot of what is going on.

  2. using the tee filter to check processes or data flows at critical points.

  3. setting option flags -n -v -x

    sh -n scriptname checks for syntax errors without actually running the script. This is the equivalent of inserting set -n or set -o noexec into the script. Note that certain types of syntax errors can slip past this check.

    sh -v scriptname echoes each command before executing it. This is the equivalent of inserting set -v or set -o verbose in the script.

    The -n and -v flags work well together. sh -nv scriptname gives a verbose syntax check.

    sh -x scriptname echoes the result each command, but in an abbreviated manner. This is the equivalent of inserting set -x or set -o xtrace in the script.

    Inserting set -u or set -o nounset in the script runs it, but gives an unbound variable error message at each attempt to use an undeclared variable.

  4. Using an "assert" function to test a variable or condition at critical points in a script. (This is an idea borrowed from C.)

    Example 30-4. Testing a condition with an "assert"

    assert ()                 #  If condition false,
    {                         #+ exit from script with error message.
      if [ -z "$2" ]          # Not enough parameters passed.
        return $E_PARAM_ERR   # No damage done.
      if [ ! $1 ] 
        echo "Assertion failed:  \"$1\""
        echo "File \"$0\", line $lineno"
        exit $E_ASSERT_FAILED
      # else
      #   return
      #   and continue executing script.
    condition="$a -lt $b"     # Error message and exit from script.
                              #  Try setting "condition" to something else,
                              #+ and see what happens.
    assert "$condition" $LINENO
    # The remainder of the script executes only if the "assert" does not fail.
    # Some commands.
    # ...
    echo "This statement echoes only if the \"assert\" does not fail."
    # ...
    # Some more commands.
    exit 0
  5. trapping at exit.

    The exit command in a script triggers a signal 0, terminating the process, that is, the script itself. [64] It is often useful to trap the exit, forcing a "printout" of variables, for example. The trap must be the first command in the script.

Trapping signals


Specifies an action on receipt of a signal; also useful for debugging.


A signal is simply a message sent to a process, either by the kernel or another process, telling it to take some specified action (usually to terminate). For example, hitting a Control-C, sends a user interrupt, an INT signal, to a running program.

trap '' 2
# Ignore interrupt 2 (Control-C), with no action specified. 

trap 'echo "Control-C disabled."' 2
# Message when Control-C pressed.

Example 30-5. Trapping at exit

# Hunting variables with a trap.

trap 'echo Variable Listing --- a = $a  b = $b' EXIT
#  EXIT is the name of the signal generated upon exit from a script.
#  The command specified by the "trap" doesn't execute until
#+ the appropriate signal is sent.

echo "This prints before the \"trap\" --"
echo "even though the script sees the \"trap\" first."



exit 0
#  Note that commenting out the 'exit' command makes no difference,
#+ since the script exits in any case after running out of commands.

Example 30-6. Cleaning up after Control-C

# A quick 'n dirty script to check whether you are on-line yet.

#  Note that $LOGFILE must be readable
#+ (as root, chmod 644 /var/log/messages).
#  Create a "unique" temp file name, using process id of the script.
#  At logon, the line "remote IP address"
#                      appended to /var/log/messages.
#  How many lines in log file to check.

#  Cleans up the temp file if script interrupted by control-c.


while [ $TRUE ]  #Endless loop.
  #  Saves last 100 lines of system log file as temp file.
  #  Necessary, since newer kernels generate many log messages at log on.
  search=`grep $KEYWORD $TEMPFILE`
  #  Checks for presence of the "IP address" phrase,
  #+ indicating a successful logon.

  if [ ! -z "$search" ] #  Quotes necessary because of possible spaces.
     echo "On-line"
     rm -f $TEMPFILE    #  Clean up temp file.
     exit $ONLINE
     echo -n "."        #  The -n option to echo suppresses newline,
                        #+ so you get continuous rows of dots.

  sleep 1  

#  Note: if you change the KEYWORD variable to "Exit",
#+ this script can be used while on-line
#+ to check for an unexpected logoff.

# Exercise: Change the script, per the above note,
#           and prettify it.

exit 0

# Nick Drage suggests an alternate method:

while true
  do ifconfig ppp0 | grep UP 1> /dev/null && echo "connected" && exit 0
  echo -n "."   # Prints dots (.....) until connected.
  sleep 2

# Problem: Hitting Control-C to terminate this process may be insufficient.
#+         (Dots may keep on echoing.)
# Exercise: Fix this.

# Stephane Chazelas has yet another alternative:


while ! tail -1 "$LOGFILE" | grep -q "$KEYWORD"
do echo -n .
echo "On-line"

# Exercise: Discuss the relative strengths and weaknesses
#!          of each of these various approaches.


The DEBUG argument to trap causes a specified action to execute after every command in a script. This permits tracing variables, for example.

Example 30-7. Tracing a variable


trap 'echo "VARIABLE-TRACE> \$variable = \"$variable\""' DEBUG
# Echoes the value of $variable after every command.


echo "Just initialized \"\$variable\" to $variable."

let "variable *= 3"
echo "Just multiplied \"\$variable\" by 3."

# The "trap 'commands' DEBUG" construct would be more useful
# in the context of a complex script,
# where placing multiple "echo $variable" statements might be
# clumsy and time-consuming.

# Thanks, Stephane Chazelas for the pointer.

exit 0

Of course, the trap command has other uses aside from debugging.

Example 30-8. Running multiple processes (on an SMP box)

# Run multiple processes on an SMP box.

# Script written by Vernia Damiano.
# Used with permission.

#  Must call script with at least one integer parameter
#+ (number of concurrent processes).
#  All other parameters are passed through to the processes started.

INDICE=8        # Total number of process to start
TEMPO=5         # Maximum sleep time per process
E_BADARGS=65    # No arg(s) passed to script.

if [ $# -eq 0 ] # Check for at least one argument passed to script.
  echo "Usage: `basename $0` number_of_processes [passed params]"
  exit $E_BADARGS

NUMPROC=$1              # Number of concurrent process
PARAMETRI=( "$@" )      # Parameters of each process

function avvia() {
	local temp
	local index
	let "temp %= $TEMPO"
	let "temp += 1"
	echo "Starting $index Time:$temp" "$@"
	sleep ${temp}
	echo "Ending $index"
	kill -s SIGRTMIN $$

function parti() {
	if [ $INDICE -gt 0 ] ; then
		avvia $INDICE "${PARAMETRI[@]}" &
		let "INDICE--"
		trap : SIGRTMIN

trap parti SIGRTMIN

while [ "$NUMPROC" -gt 0 ]; do
	let "NUMPROC--"


exit $?

I had the need to run a program, with specified options, on a number of
different files, using a SMP machine. So I thought [I'd] keep running
a specified number of processes and start a new one each time . . . one
of these terminates.

The "wait" instruction does not help, since it waits for a given process
or *all* process started in background. So I wrote [this] bash script
that can do the job, using the "trap" instruction.
  --Vernia Damiano


trap '' SIGNAL (two adjacent apostrophes) disables SIGNAL for the remainder of the script. trap SIGNAL restores the functioning of SIGNAL once more. This is useful to protect a critical portion of a script from an undesirable interrupt.

	trap '' 2  # Signal 2 is Control-C, now disabled.
	trap 2     # Reenables Control-C

Chapter 31. Options

Options are settings that change shell and/or script behavior.

The set command enables options within a script. At the point in the script where you want the options to take effect, use set -o option-name or, in short form, set -option-abbrev. These two forms are equivalent.


      set -o verbose
      # Echoes all commands before executing.


      set -v
      # Exact same effect as above.


To disable an option within a script, use set +o option-name or set +option-abbrev.


      set -o verbose
      # Command echoing on.

      set +o verbose
      # Command echoing off.
      # Not echoed.

      set -v
      # Command echoing on.

      set +v
      # Command echoing off.

      exit 0

An alternate method of enabling options in a script is to specify them immediately following the #! script header.

      #!/bin/bash -x
      # Body of script follows.

It is also possible to enable script options from the command line. Some options that will not work with set are available this way. Among these are -i, force script to run interactive.

bash -v script-name

bash -o verbose script-name

The following is a listing of some useful options. They may be specified in either abbreviated form (preceded by a single dash) or by complete name (preceded by a double dash or by -o).

Table 31-1. Bash options

-CnoclobberPrevent overwriting of files by redirection (may be overridden by >|)
-D(none)List double-quoted strings prefixed by $, but do not execute commands in script
-aallexportExport all defined variables
-bnotifyNotify when jobs running in background terminate (not of much use in a script)
-c ...(none)Read commands from ...
-eerrexitAbort script at first error, when a command exits with non-zero status (except in until or while loops, if-tests, list constructs)
-fnoglobFilename expansion (globbing) disabled
-iinteractiveScript runs in interactive mode
-nnoexecRead commands in script, but do not execute them (syntax check)
-o Option-Name(none)Invoke the Option-Name option
-o posixPOSIXChange the behavior of Bash, or invoked script, to conform to POSIX standard.
-pprivilegedScript runs as "suid" (caution!)
-rrestrictedScript runs in restricted mode (see Chapter 21).
-sstdinRead commands from stdin
-t(none)Exit after first command
-unounsetAttempt to use undefined variable outputs error message, and forces an exit
-vverbosePrint each command to stdout before executing it
-xxtraceSimilar to -v, but expands commands
-(none)End of options flag. All other arguments are positional parameters.
--(none)Unset positional parameters. If arguments given (-- arg1 arg2), positional parameters set to arguments.

Chapter 32. Gotchas


Turandot: Gli enigmi sono tre, la morte una!

Caleph: No, no! Gli enigmi sono tre, una la vita!


Assigning reserved words or characters to variable names.
case=value0       # Causes problems.
23skidoo=value1   # Also problems.
# Variable names starting with a digit are reserved by the shell.
# Try _23skidoo=value1. Starting variables with an underscore is o.k.

# However...      using just the underscore will not work.
echo $_           # $_ is a special variable set to last arg of last command.

xyz((!*=value2    # Causes severe problems.

Using a hyphen or other reserved characters in a variable name.
# Use 'var_1' instead.

Using the same name for a variable and a function. This can make a script difficult to understand.
do_something ()
  echo "This function does something with \"$1\"."


do_something do_something

# All this is legal, but highly confusing.

Using whitespace inappropriately. In contrast to other programming languages, Bash can be quite finicky about whitespace.
var1 = 23   # 'var1=23' is correct.
# On line above, Bash attempts to execute command "var1"
# with the arguments "=" and "23".
let c = $a - $b   # 'let c=$a-$b' or 'let "c = $a - $b"' are correct.

if [ $a -le 5]    # if [ $a -le 5 ]   is correct.
# if [ "$a" -le 5 ]   is even better.
# [[ $a -le 5 ]] also works.

Assuming uninitialized variables (variables before a value is assigned to them) are "zeroed out". An uninitialized variable has a value of "null", not zero.

echo "uninitialized_var = $uninitialized_var"
# uninitialized_var =

Mixing up = and -eq in a test. Remember, = is for comparing literal variables and -eq for integers.
if [ "$a" = 273 ]      # Is $a an integer or string?
if [ "$a" -eq 273 ]    # If $a is an integer.

# Sometimes you can mix up -eq and = without adverse consequences.
# However...

a=273.0   # Not an integer.
if [ "$a" = 273 ]
  echo "Comparison works."
  echo "Comparison does not work."
fi    # Comparison does not work.

# Same with   a=" 273"  and a="0273".

# Likewise, problems trying to use "-eq" with non-integer values.
if [ "$a" -eq 273.0 ]
  echo "a = $a'
fi  # Aborts with an error message.  
# [: 273.0: integer expression expected

Misusing string comparison operators.

Example 32-1. Numerical and string comparison are not equivalent

# Trying to use a string comparison on integers.


# The following "while loop" has two errors:
#+ one blatant, and the other subtle.

while [ "$number" < 5 ]    # Wrong! Should be:  while [ "$number" -lt 5 ]
  echo -n "$number "
  let "number += 1"
#  Attempt to run this bombs with the error message:
#+ line 10: 5: No such file or directory
#  Within single brackets, "<" must be escaped,
#+ and even then, it's still wrong for comparing integers.

echo "---------------------"

while [ "$number" \< 5 ]    #  1 2 3 4
do                          #
  echo -n "$number "        #  This *seems to work, but . . .
  let "number += 1"         #+ it  actually does an ASCII comparison,
done                        #+ rather than a numerical one.

echo; echo "---------------------"

# This can cause problems. For example:


if [ "$greater" \< "$lesser" ]
  echo "$greater is less than $lesser"
fi                          # 105 is less than 5
#  In fact, "105" actually is less than "5"
#+ in a string comparison (ASCII sort order).


exit 0

Sometimes variables within "test" brackets ([ ]) need to be quoted (double quotes). Failure to do so may cause unexpected behavior. See Example 7-6, Example 16-5, and Example 9-6.

Commands issued from a script may fail to execute because the script owner lacks execute permission for them. If a user cannot invoke a command from the command line, then putting it into a script will likewise fail. Try changing the attributes of the command in question, perhaps even setting the suid bit (as root, of course).

Attempting to use - as a redirection operator (which it is not) will usually result in an unpleasant surprise.
command1 2> - | command2  # Trying to redirect error output of command1 into a pipe...
#    ...will not work.	

command1 2>& - | command2  # Also futile.

Thanks, S.C.

Using Bash version 2+ functionality may cause a bailout with error messages. Older Linux machines may have version 1.XX of Bash as the default installation.

# Since Chet Ramey is constantly adding features to Bash,
# you may set $minimum_version to 2.XX, or whatever is appropriate.

if [ "$BASH_VERSION" \< "$minimum_version" ]
  echo "This script works only with Bash, version $minimum or greater."
  echo "Upgrade strongly recommended."


Using Bash-specific functionality in a Bourne shell script (#!/bin/sh) on a non-Linux machine may cause unexpected behavior. A Linux system usually aliases sh to bash, but this does not necessarily hold true for a generic UNIX machine.

Using undocumented features in Bash turns out to be a dangerous practice. In previous releases of this book there were several scripts that depended on the "feature" that, although the maximum value of an exit or return value was 255, that limit did not apply to negative integers. Unfortunately, in version 2.05b and later, that loophole disappeared. See Example 23-8.

A script with DOS-type newlines (\r\n) will fail to execute, since #!/bin/bash\r\n is not recognized, not the same as the expected #!/bin/bash\n. The fix is to convert the script to UNIX-style newlines.

echo "Here"

unix2dos $0    # Script changes itself to DOS format.
chmod 755 $0   # Change back to execute permission.
               # The 'unix2dos' command removes execute permission.

./$0           # Script tries to run itself again.
               # But it won't work as a DOS file.

echo "There"

exit 0

A shell script headed by #!/bin/sh will not run in full Bash-compatibility mode. Some Bash-specific functions might be disabled. Scripts that need complete access to all the Bash-specific extensions should start with #!/bin/bash.

Putting whitespace in front of the terminating limit string of a here document will cause unexpected behavior in a script.

A script may not export variables back to its parent process, the shell, or to the environment. Just as we learned in biology, a child process can inherit from a parent, but not vice versa.
exit 0
bash$ echo $WHATEVER

Sure enough, back at the command prompt, $WHATEVER remains unset.

Setting and manipulating variables in a subshell, then attempting to use those same variables outside the scope of the subshell will result an unpleasant surprise.

Example 32-2. Subshell Pitfalls

# Pitfalls of variables in a subshell.

echo "outer_variable = $outer_variable"

# Begin subshell

echo "outer_variable inside subshell = $outer_variable"
inner_variable=inner  # Set
echo "inner_variable inside subshell = $inner_variable"
outer_variable=inner  # Will value change globally?
echo "outer_variable inside subshell = $outer_variable"

# End subshell

echo "inner_variable outside subshell = $inner_variable"  # Unset.
echo "outer_variable outside subshell = $outer_variable"  # Unchanged.

exit 0

Piping echo output to a read may produce unexpected results. In this scenario, the read acts as if it were running in a subshell. Instead, use the set command (as in Example 11-15).

Example 32-3. Piping the output of echo to a read

#  Attempting to use 'echo and 'read'
#+ to assign variables non-interactively.


echo "one two three" | read a b c
# Try to reassign a, b, and c.

echo "a = $a"  # a = aaa
echo "b = $b"  # b = bbb
echo "c = $c"  # c = ccc
# Reassignment failed.

# ------------------------------

# Try the following alternative.

var=`echo "one two three"`
set -- $var
a=$1; b=$2; c=$3

echo "-------"
echo "a = $a"  # a = one
echo "b = $b"  # b = two
echo "c = $c"  # c = three 
# Reassignment succeeded.

# ------------------------------

#  Note also that an echo to a 'read' works within a subshell.
#  However, the value of the variable changes *only* within the subshell.

a=aaa          # Starting all over again.

echo; echo
echo "one two three" | ( read a b c;
echo "Inside subshell: "; echo "a = $a"; echo "b = $b"; echo "c = $c" )
# a = one
# b = two
# c = three
echo "-----------------"
echo "Outside subshell: "
echo "a = $a"  # a = aaa
echo "b = $b"  # b = bbb
echo "c = $c"  # c = ccc

exit 0

In fact, as Anthony Richardson points out, piping to any loop can cause a similar problem.

# Loop piping troubles.
#  This example by Anthony Richardson,
#+ with addendum by Wilbert Berendsen.

find $HOME -type f -atime +30 -size 100k |
while true
   read f
   echo "$f is over 100KB and has not been accessed in over 30 days"
   echo "Consider moving the file to archives."
#  foundone will always be false here since it is
#+ set to true inside a subshell
if [ $foundone = false ]
   echo "No files need archiving."

# =====================Now, here is the correct way:=================

for f in $(find $HOME -type f -atime +30 -size 100k)  # No pipe here.
   echo "$f is over 100KB and has not been accessed in over 30 days"
   echo "Consider moving the file to archives."
if [ $foundone = false ]
   echo "No files need archiving."

# ==================And here is another alternative==================

#  Places the part of the script that reads the variables
#+ within a code block, so they share the same subshell.
#  Thank you, W.B.

find $HOME -type f -atime +30 -size 100k | {
     while read f
       echo "$f is over 100KB and has not been accessed in over 30 days"
       echo "Consider moving the file to archives."

     if ! $foundone
       echo "No files need archiving."

A related problem occurs when trying to write the stdout of a tail -f piped to grep.
tail -f /var/log/messages | grep "$ERROR_MSG" >> error.log
# The "error.log" file will not have anything written to it.


Using "suid" commands within scripts is risky, as it may compromise system security. [65]

Using shell scripts for CGI programming may be problematic. Shell script variables are not "typesafe", and this can cause undesirable behavior as far as CGI is concerned. Moreover, it is difficult to "cracker-proof" shell scripts.

Bash does not handle the double slash (//) string correctly.

Bash scripts written for Linux or BSD systems may need fixups to run on a commercial UNIX machine. Such scripts often employ GNU commands and filters which have greater functionality than their generic UNIX counterparts. This is particularly true of such text processing utilites as tr.


Danger is near thee --

Beware, beware, beware, beware.

Many brave hearts are asleep in the deep.

So beware --


 A.J. Lamb and H.W. Petrie

Chapter 33. Scripting With Style

Get into the habit of writing shell scripts in a structured and systematic manner. Even "on-the-fly" and "written on the back of an envelope" scripts will benefit if you take a few minutes to plan and organize your thoughts before sitting down and coding.

Herewith are a few stylistic guidelines. This is not intended as an Official Shell Scripting Stylesheet.

33.1. Unofficial Shell Scripting Stylesheet

  • Comment your code. This makes it easier for others to understand (and appreciate), and easier for you to maintain.
    # It made perfect sense when you wrote it last year, but now it's a complete mystery.
    # (From Antek Sawicki's "" script.)

    Add descriptive headers to your scripts and functions.
    #                                #
    #           written by Bozo Bozeman              #
    #                July 05, 2001                   #
    #                                                #
    #           Clean up project files.              #
    BADDIR=65                       # No such directory.
    projectdir=/home/bozo/projects  # Directory to clean up.
    # ------------------------------------------------------- #
    # cleanup_pfiles ()                                       #
    # Removes all files in designated directory.              #
    # Parameter: $target_directory                            #
    # Returns: 0 on success, $BADDIR if something went wrong. #
    # ------------------------------------------------------- #
    cleanup_pfiles ()
      if [ ! -d "$1" ]  # Test if target directory exists.
        echo "$1 is not a directory."
        return $BADDIR
      rm -f "$1"/*
      return 0   # Success.
    cleanup_pfiles $projectdir
    exit 0
    Be sure to put the #!/bin/bash at the beginning of the first line of the script, preceding any comment headers.

  • Avoid using "magic numbers", [66] that is, "hard-wired" literal constants. Use meaningful variable names instead. This makes the script easier to understand and permits making changes and updates without breaking the application.
    if [ -f /var/log/messages ]
    # A year later, you decide to change the script to check /var/log/syslog.
    # It is now necessary to manually change the script, instance by instance,
    # and hope nothing breaks.
    # A better way:
    LOGFILE=/var/log/messages  # Only line that needs to be changed.
    if [ -f "$LOGFILE" ]

  • Choose descriptive names for variables and functions.
    fl=`ls -al $dirname`                 # Cryptic.
    file_listing=`ls -al $dirname`       # Better.
    MAXVAL=10   # All caps used for a script constant.
    while [ "$index" -le "$MAXVAL" ]
    E_NOTFOUND=75                        # Uppercase for an errorcode,
                                         # and name begins with "E_".
    if [ ! -e "$filename" ]
      echo "File $filename not found."
      exit $E_NOTFOUND
    MAIL_DIRECTORY=/var/spool/mail/bozo  # Uppercase for an environmental variable.
    GetAnswer ()                         # Mixed case works well for a function.
      echo -n $prompt
      read answer
      return $answer
    GetAnswer "What is your favorite number? "
    echo $favorite_number
    _uservariable=23                     # Permissable, but not recommended.
    # It's better for user-defined variables not to start with an underscore.
    # Leave that for system variables.

  • Use exit codes in a systematic and meaningful way.
    exit $E_WRONG_ARGS
    See also Appendix D.

  • Break complex scripts into simpler modules. Use functions where appropriate. See Example 35-4.

  • Don't use a complex construct where a simpler one will do.
    if [ $? -eq 0 ]
    # Redundant and non-intuitive.
    if COMMAND
    # More concise (if perhaps not quite as legible).


... reading the UNIX source code to the Bourne shell (/bin/sh). I was shocked at how much simple algorithms could be made cryptic, and therefore useless, by a poor choice of code style. I asked myself, "Could someone be proud of this code?"

 Landon Noll

Chapter 34. Miscellany


Nobody really knows what the Bourne shell's grammar is. Even examination of the source code is little help.

 Tom Duff

34.1. Interactive and non-interactive shells and scripts

An interactive shell reads commands from user input on a tty. Among other things, such a shell reads startup files on activation, displays a prompt, and enables job control by default. The user can interact with the shell.

A shell running a script is always a non-interactive shell. All the same, the script can still access its tty. It is even possible to emulate an interactive shell in a script.
while :
  echo -n "$MY_PROMPT"
  read line
  eval "$line"

exit 0

# This example script, and much of the above explanation supplied by
# Stephane Chazelas (thanks again).

Let us consider an interactive script to be one that requires input from the user, usually with read statements (see Example 11-2). "Real life" is actually a bit messier than that. For now, assume an interactive script is bound to a tty, a script that a user has invoked from the console or an xterm.

Init and startup scripts are necessarily non-interactive, since they must run without human intervention. Many administrative and system maintenance scripts are likewise non-interactive. Unvarying repetitive tasks cry out for automation by non-interactive scripts.

Non-interactive scripts can run in the background, but interactive ones hang, waiting for input that never comes. Handle that difficulty by having an expect script or embedded here document feed input to an interactive script running as a background job. In the simplest case, redirect a file to supply input to a read statement (read variable <file). These particular workarounds make possible general purpose scripts that run in either interactive or non-interactive modes.

If a script needs to test whether it is running in an interactive shell, it is simply a matter of finding whether the prompt variable, $PS1 is set. (If the user is being prompted for input, then the script needs to display a prompt.)
if [ -z $PS1 ] # no prompt?
  # non-interactive
  # interactive
Alternatively, the script can test for the presence of option "i" in the $- flag.
case $- in
*i*)    # interactive shell
*)      # non-interactive shell
# (Courtesy of "UNIX F.A.Q.," 1993)


Scripts may be forced to run in interactive mode with the -i option or with a #!/bin/bash -i header. Be aware that this can cause erratic script behavior or show error messages even when no error is present.

34.2. Shell Wrappers

A "wrapper" is a shell script that embeds a system command or utility, that saves a set of parameters passed to that command. Wrapping a script around a complex command line simplifies invoking it. This is expecially useful with sed and awk.

A sed or awk script would normally be invoked from the command line by a sed -e 'commands' or awk 'commands'. Embedding such a script in a Bash script permits calling it more simply, and makes it "reusable". This also enables combining the functionality of sed and awk, for example piping the output of a set of sed commands to awk. As a saved executable file, you can then repeatedly invoke it in its original form or modified, without the inconvenience of retyping it on the command line.

Example 34-1. shell wrapper


# This is a simple script that removes blank lines from a file.
# No argument checking.
# You might wish to add something like:
# if [ -z "$1" ]
# then
#  echo "Usage: `basename $0` target-file"
#  exit 65
# fi

# Same as
#    sed -e '/^$/d' filename
# invoked from the command line.

sed -e /^$/d "$1"
#  The '-e' means an "editing" command follows (optional here).
#  '^' is the beginning of line, '$' is the end.
#  This match lines with nothing between the beginning and the end,
#+ blank lines.
#  The 'd' is the delete command.

#  Quoting the command-line arg permits
#+ whitespace and special characters in the filename.

exit 0

Example 34-2. A slightly more complex shell wrapper


#  "subst", a script that substitutes one pattern for
#+ another in a file,
#+ i.e., "subst Smith Jones letter.txt".

ARGS=3         # Script requires 3 arguments.
E_BADARGS=65   # Wrong number of arguments passed to script.

if [ $# -ne "$ARGS" ]
# Test number of arguments to script (always a good idea).
  echo "Usage: `basename $0` old-pattern new-pattern filename"
  exit $E_BADARGS


if [ -f "$3" ]
    echo "File \"$3\" does not exist."
    exit $E_BADARGS

#  Here is where the heavy work gets done.

# -----------------------------------------------
sed -e "s/$old_pattern/$new_pattern/g" $file_name
# -----------------------------------------------

#  's' is, of course, the substitute command in sed,
#+ and /pattern/ invokes address matching.
#  The "g", or global flag causes substitution for *every*
#+ occurence of $old_pattern on each line, not just the first.
#  Read the literature on 'sed' for an in-depth explanation.

exit 0    # Successful invocation of the script returns 0.

Example 34-3. A shell wrapper around an awk script


# Adds up a specified column (of numbers) in the target file.


if [ $# -ne "$ARGS" ] # Check for proper no. of command line args.
   echo "Usage: `basename $0` filename column-number"
   exit $E_WRONGARGS


# Passing shell variables to the awk part of the script is a bit tricky.
# See the awk documentation for more details.

# A multi-line awk script is invoked by   awk ' ..... '

# Begin awk script.
# -----------------------------
awk '

{ total += $'"${column_number}"'
     print total

' "$filename"
# -----------------------------
# End awk script.

#   It may not be safe to pass shell variables to an embedded awk script,
#   so Stephane Chazelas proposes the following alternative:
#   ---------------------------------------
#   awk -v column_number="$column_number" '
#   { total += $column_number
#   }
#   END {
#       print total
#   }' "$filename"
#   ---------------------------------------

exit 0

For those scripts needing a single do-it-all tool, a Swiss army knife, there is Perl. Perl combines the capabilities of sed and awk, and throws in a large subset of C, to boot. It is modular and contains support for everything ranging from object-oriented programming up to and including the kitchen sink. Short Perl scripts lend themselves to embedding in shell scripts, and there may even be some substance to the claim that Perl can totally replace shell scripting (though the author of this document remains skeptical).

Example 34-4. Perl embedded in a Bash script


# Shell commands may precede the Perl script.
echo "This precedes the embedded Perl script within \"$0\"."
echo "==============================================================="

perl -e 'print "This is an embedded Perl script.\n";'
# Like sed, Perl also uses the "-e" option.

echo "==============================================================="
echo "However, the script may also contain shell and system commands."

exit 0

It is even possible to combine a Bash script and Perl script within the same file. Depending on how the script is invoked, either the Bash part or the Perl part will execute.

Example 34-5. Bash and Perl scripts combined


echo "Greetings from the Bash part of the script."
# More Bash commands may follow here.

exit 0
# End of Bash part of the script.

# =======================================================

# This part of the script must be invoked with -x option.

print "Greetings from the Perl part of the script.\n";
# More Perl commands may follow here.

# End of Perl part of the script.

bash$ bash
Greetings from the Bash part of the script.

bash$ perl -x
Greetings from the Perl part of the script.

34.3. Tests and Comparisons: Alternatives

For tests, the [[ ]] construct may be more appropriate than [ ]. Likewise, arithmetic comparisons might benefit from the (( )) construct.

# All of the comparisons below are equivalent.
test "$a" -lt 16 && echo "yes, $a < 16"         # "and list"
/bin/test "$a" -lt 16 && echo "yes, $a < 16" 
[ "$a" -lt 16 ] && echo "yes, $a < 16" 
[[ $a -lt 16 ]] && echo "yes, $a < 16"          # Quoting variables within
(( a < 16 )) && echo "yes, $a < 16"             # [[ ]] and (( )) not necessary.

city="New York"
# Again, all of the comparisons below are equivalent.
test "$city" \< Paris && echo "Yes, Paris is greater than $city"  # Greater ASCII order.
/bin/test "$city" \< Paris && echo "Yes, Paris is greater than $city" 
[ "$city" \< Paris ] && echo "Yes, Paris is greater than $city" 
[[ $city < Paris ]] && echo "Yes, Paris is greater than $city"    # Need not quote $city.

# Thank you, S.C.

34.4. Recursion

Can a script recursively call itself? Indeed.

Example 34-6. A (useless) script that recursively calls itself


#  Can a script recursively call itself?
#  Yes, but is this of any practical use?
#  (See the following.)


let "i %= $RANGE"  # Generate a random number between 0 and $RANGE - 1.

if [ "$i" -lt "$MAXVAL" ]
  echo "i = $i"
  ./$0             #  Script recursively spawns a new instance of itself.
fi                 #  Each child script does the same, until
                   #+ a generated $i equals $MAXVAL.

#  Using a "while" loop instead of an "if/then" test causes problems.
#  Explain why.

exit 0

# Note:
# ----
# This script must have execute permission for it to work properly.
# This is the case even if it is invoked by an "sh" command.
# Explain why.

Example 34-7. A (useful) script that recursively calls itself

# phone book

# Written by Rick Boivie, and used with permission.
# Modifications by document author.

MINARGS=1     #  Script needs at least one argument.
              #  A data file in current working directory
              #+ named "phonebook" must exist.
E_NOARGS=70   #  No arguments error.

if [ $# -lt $MINARGS ]; then
      echo "Usage: "$PROGNAME" data"
      exit $E_NOARGS

if [ $# -eq $MINARGS ]; then
      grep $1 "$DATAFILE"
      # 'grep' prints an error message if $DATAFILE not present.
      ( shift; "$PROGNAME" $* ) | grep $1
      # Script recursively calls itself.

exit 0        #  Script exits here.
              #  Therefore, it's o.k. to put
	      #+ non-hashmarked comments and data after this point.

# ------------------------------------------------------------------------
Sample "phonebook" datafile:

John Doe        1555 Main St., Baltimore, MD 21228          (410) 222-3333
Mary Moe        9899 Jones Blvd., Warren, NH 03787          (603) 898-3232
Richard Roe     856 E. 7th St., New York, NY 10009          (212) 333-4567
Sam Roe         956 E. 8th St., New York, NY 10009          (212) 444-5678
Zoe Zenobia     4481 N. Baker St., San Francisco, SF 94338  (415) 501-1631
# ------------------------------------------------------------------------

$bash Roe
Richard Roe     856 E. 7th St., New York, NY 10009          (212) 333-4567
Sam Roe         956 E. 8th St., New York, NY 10009          (212) 444-5678

$bash Roe Sam
Sam Roe         956 E. 8th St., New York, NY 10009          (212) 444-5678

#  When more than one argument is passed to this script,
#+ it prints *only* the line(s) containing all the arguments.

Example 34-8. Another (useful) script that recursively calls itself

#, written by Anthony Richardson
# Used with permission.

# usage:
# description: mount device, invoking user must be listed in the
#              MNTUSERS group in the /etc/sudoers file.

# ----------------------------------------------------------
#  This is a usermount script that reruns itself using sudo.
#  A user with the proper permissions only has to type

#   usermount /dev/fd0 /mnt/floppy

# instead of

#   sudo usermount /dev/fd0 /mnt/floppy

#  I use this same technique for all of my
#+ sudo scripts, because I find it convenient.
# ----------------------------------------------------------

#  If SUDO_COMMAND variable is not set we are not being run through
#+ sudo, so rerun ourselves. Pass the user's real and group id . . .

if [ -z "$SUDO_COMMAND" ]
   mntusr=$(id -u) grpusr=$(id -g) sudo $0 $*
   exit 0

# We will only get here if we are being run by sudo.
/bin/mount $* -o uid=$mntusr,gid=$grpusr

exit 0

# Additional notes (from the author of this script): 
# -------------------------------------------------

# 1) Linux allows the "users" option in the /etc/fstab
#    file so that any user can mount removable media.
#    But, on a server, I like to allow only a few
#    individuals access to removable media.
#    I find using sudo gives me more control.

# 2) I also find sudo to be more convenient than
#    accomplishing this task through groups.

# 3) This method gives anyone with proper permissions
#    root access to the mount command, so be careful
#    about who you allow access.
#    You can get finer control over which access can be mounted
#    by using this same technique in separate mntfloppy, mntcdrom,
#    and mntsamba scripts.


Too many levels of recursion can exhaust the script's stack space, causing a segfault.

34.5. "Colorizing" Scripts

The ANSI [67] escape sequences set screen attributes, such as bold text, and color of foreground and background. DOS batch files commonly used ANSI escape codes for color output, and so can Bash scripts.

Example 34-9. A "colorized" address database

# "Colorized" version of
#            Crude address database

clear                                   # Clear the screen.

echo -n "          "
echo -e '\E[37;44m'"\033[1mContact List\033[0m"
                                        # White on blue background
echo; echo
echo -e "\033[1mChoose one of the following persons:\033[0m"
                                        # Bold
tput sgr0
echo "(Enter only the first letter of name.)"
echo -en '\E[47;34m'"\033[1mE\033[0m"   # Blue
tput sgr0                               # Reset colors to "normal."
echo "vans, Roland"                     # "[E]vans, Roland"
echo -en '\E[47;35m'"\033[1mJ\033[0m"   # Magenta
tput sgr0
echo "ones, Mildred"
echo -en '\E[47;32m'"\033[1mS\033[0m"   # Green
tput sgr0
echo "mith, Julie"
echo -en '\E[47;31m'"\033[1mZ\033[0m"   # Red
tput sgr0
echo "ane, Morris"

read person

case "$person" in
# Note variable is quoted.

  "E" | "e" )
  # Accept upper or lowercase input.
  echo "Roland Evans"
  echo "4321 Floppy Dr."
  echo "Hardscrabble, CO 80753"
  echo "(303) 734-9874"
  echo "(303) 734-9892 fax"
  echo ""
  echo "Business partner & old friend"

  "J" | "j" )
  echo "Mildred Jones"
  echo "249 E. 7th St., Apt. 19"
  echo "New York, NY 10009"
  echo "(212) 533-2814"
  echo "(212) 533-9972 fax"
  echo ""
  echo "Girlfriend"
  echo "Birthday: Feb. 11"

# Add info for Smith & Zane later.

          * )
   # Default option.	  
   # Empty input (hitting RETU