This is a riveting series:

On the LPIC-1 Exam 102: Shells and Shell Scripting
On the LPIC-1 Exam 102: User Interfaces and Desktops
On the LPIC-1 Exam 102: Administrative Tasks
On the LPIC-1 Exam 102: Essential System Services
On the LPIC-1 Exam 102: Networking Fundamentals
On the LPIC-1 Exam 102: Security

And, so is this one!

On the LPIC-1 Exam 101 (101-500), Part One

When studying for the Linux Professional Institute LPIC-1 certification, I took a bunch of notes when reading the docs and doing an online course. Note that this is not exhaustive, so do not depend on this article to get you prepared for the exam.

The menu items below are not in any order.

This roughly covers Topic 105: Shells and Shell Scripting.

Caveat emptor.

Exam Details
Topic 105: Shells and Shell Scripting
Summary
References

Exam Details

LPIC-1 Exam 102

Exam Objectives Version: 5.0
Exam Code: 102-500

Topic 105: Shells and Shell Scripting

Terminals

tty - tele-typewriter
pts - pseudo-terminal slave

Shell Types

A great way to see which initialization files are sourced by bash for a particular invocation of a shell is to append an echo statement to each global file:

echo "echo hello from /etc/profile" | sudo tee -a /etc/profile
echo "echo hello from /etc/bash.bashrc" | sudo tee -a /etc/bash.bashrc

Examples include:

Invoke a login shell:
- bash -l or bash -login
Drop to a tty on your local machine.
- Ctrl+Alt+F2, for example
Logging into a remote machine via ssh.

The following files are sourced, in order:

/etc/profile
- the system-wide profile file for the Bourne shell and Bourne compatible shells
- sets PATH and PS1
/etc/bash.bashrc
if they exist, scripts in /etc/profile.d/ get executed by /etc/profile
$HOME/{.bash_profile,.bash_login,.profile}
- this will source $HOME/.bashrc
- appends $HOME/bin to PATH
$HOME/.bash_logout

To start an interactive login shell that doesn’t source any configuration files, use the --noprofile switch (will still source /etc/bash.bashrc and $HOME/.bashrc, though).

So, it appears to behave like an interactive non-login shell.

A general definition of an interactive shell is one that reads from and writes to a user’s terminal:

An interactive shell is one started without non-option arguments (unless -s is specified) and without specifying the -c option, whose input and error output are both connected to terminals (as determined by isatty(3)), or one started with the -i option.

From 6.3.1 What is an Interactive Shell?

Here are some fine examples of interactive non-login shells:

Subshell:
```
$ bash
```
Subshell with -i (interactive) switch:
```
$ bash -i
```

With the -s switch and positional parameters:

$ cat test.sh
#!/usr/bin/bash

echo "Today's message is: $1 $2"
$ < test.sh bash -s -- hello world
Today's message is: hello world
$ cat test.sh | bash -s -- hello world
Today's message is: hello world

Note that the second example suffers from UUOC.

The following files are sourced, in order:

/etc/bash.bashrc
$HOME/.bashrc

Optionally, you can use the --rcfile option to have bash skip any initialization from the user $HOME/.bashrc file and instead configure the shell’s environment from the file specified as its value:

$ bash --rcfile alternate.bashrc

This will still source the system-wide /etc/bash.bashrc file (at least, it did on my machine using Debian bullseye).

To start an interactive non-login shell that doesn’t source any configuration files, use the --norc switch.

These are rare.

Some examples:

/bin/bash --login <some_script>
<some_command> | ssh <some_user>@<some_server>

Running a script will give you a bash shell of this type. Every script will run in its own subshell, opening the shell on execution and closing it on exit.

Here’s an interesting property of these types of shells. There is a variable used by the shell that’s named BASH_ENV, and its purpose is to contain a filename that should be sourced to initialize the shell.

Recall that non-interactive non-login shells will not have any initialization files run by the shell when it’s launched to customize its environment, because these types of shells are primarily for shells running scripts.

So, if you need the shell to have a custom environment, put it in BASH_ENV. Note, however, that it should be an absolute path to the file because the shell doesn’t consult the PATH variable for any lookups.

Here’s an example that’s currently running in production at benjamintoll.com:

scripting.rc

$ cat scripting.rc
export COOTIES="yes, i have them"

c.sh

#!/bin/bash

test -n "$COOTIES" && echo "$COOTIES"

Let’s first run without setting the variable:

$ ./c.sh
$

Now, let’s augment the subshell’s environment:

$ env BASH_ENV=$(pwd)/scripting.rc ./c.sh
yes, i have them

weeeeeeeeeeeeeeeeeee

Determine the Shell Type

To determine what type of shell you have, use the special parameter $0 or $BASH_ARGV0. If the result is prepended by a hyphen (-), then it’s a login shell. If it prints bash without a leading hyphen, then it’s not:

$ echo $0
-bash

In addition, there is the Bash specific, non-POSIX login_shell option.

Let’s take a look at two examples. First, an interactive login shell, and second, an interactive non-login shell.

Interactive login shell:

$ echo $0
-bash
$ shopt login_shell
login_shell     on
$ echo $-
himBCHs

Interactive non-login shell:

$ bash
$ echo $0
bash
$ shopt login_shell
login_shell     off
$ echo $-
himBCHs

The special parameter $- displays the shell options that have been set for the session. Since both of the previous examples are interactive, they both report the -i switch.

Interestingly, the ${PS1-} bash shell variable will also tell you if the shell is interactive or not:

$ echo ${PS1-}
$(tput bold)$(tput setaf 4)\h $(tput setaf 2)|$SHLVL:$0| $(tput setaf 3)~~> \[$(tput bold)\]\[$(tput setaf 6)\]\w\[\]:\[$(tput bold)\]\[$(tput setaf 8)\]$(git branch 2> /dev/null | grep "^*" | colrm 1 2)\[$(tput sgr0)\]\n$

However, if you run it in a script, it won’t echo anything. This is another way of demonstrating that running bash shell scripts are non-interactive non-login.

Note that the PS* variables are shell variables and not environment variables. In other words, they are explicitly set by a script and are not inherited by any child processes.

For instance, you’ll never see any PS* variables exported.

I stumbled across this when viewing the system’s /etc/profile script:

$ head -25 /etc/profile



# /etc/profile: system-wide .profile file for the Bourne shell (sh(1))
# and Bourne compatible shells (bash(1), ksh(1), ash(1), ...).

if [ "$(id -u)" -eq 0 ]; then
  PATH="/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin"
else
  PATH="/usr/local/bin:/usr/bin:/bin:/usr/local/games:/usr/games"
fi
export PATH

if [ "${PS1-}" ]; then
  if [ "${BASH-}" ] && [ "$BASH" != "/bin/sh" ]; then
    # The file bash.bashrc already sets the default PS1.
    # PS1='\h:\w\$ '
    if [ -f /etc/bash.bashrc ]; then
      . /etc/bash.bashrc
    fi
  else
    if [ "$(id -u)" -eq 0 ]; then
      PS1='# '
    else
      PS1='$ '

Interesting, no?

Also, notice that the PS1 shell variable is explicitly set.

`su` and `sudo`

With su, you can run a command with substitute user and group ID or become the superuser root. Both login and non-login shells can be invoked.

For example:

to start an interactive login shell as kilgore:
- su - kilgore
- su -l kilgore
- su --login kilgore
to start an interactive non-login shell as kilgore:
- su kilgore
to start an interactive login shell as root:
- su - root
- su -
to start an interactive non-login shell as root:
- su root
- su

With sudo, you execute command(s) as another user, commonly root. The default sudo security policy plugin sudoers is driven by the /etc/sudoers file, which determines a user’s sudo privileges.

This is especially useful when you can’t login as the root account.

It is important to only modify the /etc/sudoers file through the visudo utility, which will lock the file to prevent concurrent writes.

As with su, sudo allows you to create both login and non-login shells.

For example:

to start an interactive login shell as kilgore:
- sudo su - kilgore
- sudo su -l kilgore
- sudo su --login kilgore
to start an interactive non-login shell as kilgore:
- sudo su kilgore
- sudo -u kilgore -s
  - If no command is specified for the -s (shell) switch, an interactive shell is executed.
to start an interactive login shell as root:
- sudo su - root
- sudo su -
- sudo -i
  - The -i switch tells it to run as an interactive login shell.
to start an interactive login shell as root and return to the calling user after execution of the command:
- sudo -i <some_command>
  - The -i switch tells it to run as an interactive login shell.
to start an interactive non-login shell as root:
- sudo su root
- sudo su
- sudo -s
- sudo -u root -s
  - If no command is specified for the -s (shell) switch, an interactive shell is executed.

The decision to use a login or non-login shell comes down to the need and the use case. If you want to have /etc/profile executed and anything else that it sources (such as anything in /etc/profile.d/ and /etc/.bash_profile in the user’s home directory), then choose an option that invokes a login shell.

What’s the difference between sudo su - and su -?

With sudo su - you will be asked to authenticate with your own user password, while you will be expected to know the root user’s password for the latter command.

Recall that a login shell will start sourcing with the system-wide /etc/profile file, while a non-login shell will not, only starting its sourcing at the system-wide /etc/bash.bashrc file.

Add the kilgore user to the sudo group by issuing the following command:
$ sudo usermod -aG sudo kilore

`SKEL`

If set, the SKEL environment variable holds the location of the skel directory. This is the location that holds the (probably hidden) files that are copied to every user’s home directory when their account is created. They must be a regular user (as opposed to a system account) that needs a home directory and is specified by the system administrator at the time the user account is created (keep in mind that not every new account creation needs a home directory, like system accounts).

SKEL is defined in /etc/adduser.conf (default values are in /etc/default/useradd):

$ grep SKEL /etc/adduser.conf
# The SKEL variable specifies the directory containing "skeletal" user
SKEL=/etc/skel
# If SKEL_IGNORE_REGEX is set, adduser will ignore files matching this
SKEL_IGNORE_REGEX="dpkg-(old|new|dist|save)"

The location of SKEL can be changed on user account creation by the value of the -k or --skel option (as mentioned earlier, only applicable if -m or --create-home is specified).

Here are the default contents on my Debian bullseye machine:

$ ls -A /etc/skel/
.bash_logout  .bashrc  .profile

Any directories and files created in this location (or any location specified in the -k option to useradd) will also get these custom entries when their account is created with a home directory.

Why is this environment variable (probably) not set in your environment? Because the env var is only created when an account is created and doesn’t outlive the subprocess.

Shell Variables

Variables can contain the following characters:

[a-z][A-Z]
[0-9]
_ (underscore)

They cannot start with a number.

Shell variables can be created as immutable by prefacing it with the readonly builtin (inherited from the Bourne shell):

$ readonly fudge=original
$ fudge=changed
-bash: fudge: readonly variable

Or, mark it as such after creation:

$ fudge=original
$ readonly fudge

Print all read-only variables by invoking either readonly or readonly -p.

You can also create a read-only variable using the declare builtin:

$ declare -r chicken=maybe
$ echo $chicken
maybe
$ chicken=never
-bash: chicken: readonly variable

Variables become environment variables by exporting them. They are then part of the environment that is inherited by subshells.

Turn an environment variable back into a local variable:

$ export TICKLE=my_fancy
$ bash
$ echo $TICKLE
my_fancy
$ exit
$ export -n TICKLE
$ bash
$ echo $TICKLE
$

export or export -p will print all environment variables.

To turn the TICKLE local variable from the previous example back into an environment variable, you can use either of the following commands:

$ export TICKLE
$ declare -x TICKLE

Quoting

Single and double-quotes are not interchangeable (however, they sometimes are equivalent in what they do, i.e., ‘i am a string with no special characters’ == “i am a string with no special characters”).

Follow these rules, and you’ll be as safe as houses:

single quotes are for literal interpretation
double quotes allow for variable substitution

Let’s see some examples:

$ maga="make attorneys get attorneys"
$ echo '$maga'
$maga
$ echo "$maga"
make attorneys get attorneys

In addition, you’ll want to use double-quotes to avoid word splitting and pathname expansion:

$ hello="|    hello,    it's  me    |"
$ echo '$hello'
$hello
$ echo $hello
| hello, it's me |
$ echo "$hello"
|    hello,    it's  me    |

It’s almost always a good idea to double-quote any variables in scripts so variable substitution occurs and word splitting is avoided. Further, if a variable is empty, double-quoting will prevent a syntax error when evaluated.

Running a Program in a Modified Environment

To start a subshell with a limited environment, preface the shell command with env and the -i or --ignore-environment switch:

$ env -i bash
$ env
PWD=/home/btoll/projects/benjamintoll.com
LS_COLORS=
LESSCLOSE=/usr/bin/lesspipe %s %s
LESSOPEN=| /usr/bin/lesspipe %s
SHLVL=1
_=/usr/bin/env

You can also use the --ignore-environment option and also just -, which implies -i.

You can also augment the environment inherited by a subshell by prefacing the bash command again with env, but this time specifying the new variables to inherit:

$ env ZOMBIES=are_real ./run_away.sh

You can omit the env command in the previous example, as it’s implied.

Common Environment Variables

DISPLAY
HISTCONTROL
- ignorespace
  - don’t save commands starting with a space
- ignoredups
  - repeated, consecutive commands will not be saved
- ignoreboth
  - a command that falls into the previous two categories won’t be saved
HISTSIZE
HISTFILESIZE
HISTFILE
- defaults to .bash_history
HOME
HOSTNAME
HOSTTYPE
- the CPU architecture (x86_64)
LANG
- the locale of the system
LD_LIBRARY_PATH
- the location of shared libraries
MAIL
MAILCHECK
- the frequency in seconds that mail is checked
PATH
PS1
- the prompt
PS2
- the continuation prompt for multi-line commands (defaults to >)
PS3
- the select command prompt
PS4
- for debugging (defaults to +)
SHELL
USER

The following commands will all print the shell’s environment variables:

export
printenv
env

Sourcing

Sourcing a bash script will not create a subprocess. This means that all of the commands will execute within the context (i.e,. environment) of the current shell.

Why is this useful? Well, this method is used a lot when creating environment variables that are needed for a particular session. By sourcing the file instead of executing it, when the file is done being read the variables are still part of the session, whereas they wouldn’t be if executing the file (recall that a child cannot modify a parent’s environment).

Seeing this in action will make more sense.

The first thing we’ll do is show the current values of the number of bash instances, the PID of the current bash process and the status of the FOO environment variable:

$ echo $$ ; echo $SHLVL ; if [ -z $FOO ]; then echo unset; else echo set; fi
25350
3
unset

Here is our little source.sh bash script:

#!/bin/bash

echo "Number of bash instances: $SHLVL"
echo "PID of current bash instance: $$"
FOO=bar

First, let’s source it:

$ . source.sh
Number of bash instances: 3
PID of current bash instance: 25350

Ok, this demonstrates that the script did indeed run in the same shell. The number of bash instances is the same (as shown by the value of the SHLVL variable), as well as the PID.

Let’s see if the FOO variable is on the stack:

$ echo $FOO
bar

Kool Moe Dee.

Before we execute the script, unset that variable:

$ unset FOO ; if [ -z $FOO ]; then echo unset; else echo set; fi
unset

Second, we’ll execute the same file:

$ bash source.sh
Number of bash instances: 4
PID of current bash instance: 35246

It (the subprocess) created another bash instance, thus incrementing SHLVL as we’d expect. Also, the PID is that of the subprocess.

Did it set the FOO variable in the current shell?

$ if [ -z $FOO ]; then echo unset; else echo set; fi
unset

Nope.

Weeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee

Aliases

To unmask a command that has been masked by an alias, preface the command with a backslash (\l):

$ alias ls
alias ls='ls -F'
$ ls
archetypes/           config.toml  Dockerfile  k8s/      public/    resources/
build_and_deploy.sh*  content/     env.sh*     Makefile  README.md  static/
$ \ls
archetypes           config.toml  Dockerfile  k8s       public     resources
build_and_deploy.sh  content      env.sh      Makefile  README.md  static

Note that the ls alias has added the -F to classify the directory entries (this appends the indicators to the appropriate entries). This is a common alias that is provided out-of-the-box, so to speak, by many distributions.

Use the backslash to temporarily “turn off” (i.e., just for that command) the alias to allow the original command, now unshadowed, to run.

However, if there is no underlying command that is being shadowed, you will get an error. Observe:

$ type xkcd
xkcd is aliased to `open https://c.xkcd.com/random/comic/'
$ alias xkcd
alias xkcd='open https://c.xkcd.com/random/comic/'
$ xkcd
( opened the page in a browser )
$ \xkcd
-bash: xkcd: command not found

If for some reason you wish to remove all of the aliases from your session:

$ unalias -a

Let’s take a look at the difference between single and double quotes in the definition of an alias.

With single quotes, the shell variable expansion is dynamic:

$ alias you_are_here='echo $PWD'
$ you_are_here
/home/btoll/projects/benjamintoll.com
$ cd
$ you_are_here
/home/btoll

Now, redefine the alias using double quotes. Notice now that the expansion is static and will always refer to the location in which the alias was defined:

$ alias you_are_here="echo $PWD"
$ you_are_here
/home/btoll
$ cd -
/home/btoll/projects/benjamintoll.com
$ !-2
you_are_here
/home/btoll

Functions

Syntax can be either:

function FUNC_NAME {
    ...
}

or:

FUNC_NAME() {
    ...
}

To create local variables scoped to the function:

local VAR
declare -A|-a|-i|-n VAR

Otherwise, the variable will be global and then part of the calling environment.

Let’s see an example of this.

When I was a history major at university, my ancient Roman professor introduced us to a Greek warrior that fought the Romans. His name was Testicles (pronounced test-ee-klees).

Let’s dedicate this bash function to Testicles.

Here is its definition:

$ type testicles
testicles is a function
testicles ()
{
    local roman=0;
    greek=1
}

Note that there is a roman variable prefaced by the local keyword and a greek variable that is not.

And now we’ll use an interactive login shell to test the known variables before and after explicitly calling the function:

So far, neither variable is known to the shell.

$ set | grep -E "^(roman|greek)"
$ echo $roman

$ echo $greek

Now, we’ll call the function and see if anything has changed:

$ testicles
$ echo $roman

$ echo $greek
1
$ set | grep -E "^(roman|greek)"
greek=1

Uh oh, the greek variable, unprefaced by the local keyword and thus a non-local (global) variable, is now present in the shell. This is not good.

Always use local in your bash functions, children.

Builtin Variables

These special variables are known as parameters:

Variable	Description
`$?`	the return value of the last command
`$$`	the `PID` of the shell
`$!`	the `PID` of the last background job
`$0`-`$9`	positional parameters
`$#`	the number of arguments passed to the command
`$@` or `$*`	the arguments passed to the command
`$_`	the last parameter or the name of the script
`$-`	the shell options that have been set for the session

Unsetting

unset -v
- for variables
unset -f
- for functions
unset (with no option)
- first tries to unset as a variable and then failing that as a function

Testing

It is highly recommended to use double quotes around any variables in case the variable is empty, in which case any command expecting an operand would throw an error without the double quotes.

Variables

One operand:

-n - the length of STRING is nonzero
-z - the length of STRING is zero

Two operands:

Operands	Description
`STRING1` = `STRING2`	the strings are equal (can also use double equal sign `==`)
`STRING1` != `STRING2`	the strings are not equal
`INTEGER1 -eq INTEGER2`	`INTEGER1` is equal to `INTEGER2`
`INTEGER1 -ge INTEGER2`	`INTEGER1` is greater than or equal to `INTEGER2`
`INTEGER1 -gt INTEGER2`	`INTEGER1` is greater than `INTEGER2`
`INTEGER1 -le INTEGER2`	`INTEGER1` is less than or equal to `INTEGER2`
`INTEGER1 -lt INTEGER2`	`INTEGER1` is less than `INTEGER2`
`INTEGER1 -ne INTEGER2`	`INTEGER1` is not equal to `INTEGER2`

Distinct languages may have different rules for alphabetical ordering. To obtain consistent results, regardless of the localization settings of the system where the script is being executed, it is recommended to set the environment variable LANG to C, as in LANG=C, before doing operations involving alphabetical ordering.

Files

Here are two utilities for testing at the terminal. However, they’re mostly used in scripts:

test
[ (is a synonym of test)
Yes, that’s right, the character above is an open bracket ([):
```
$ which [
/usr/bin/[
```
It’s mostly seen in conditional checks in shell scripts, i.e.,
```
if [ -d /etc/squid ]
then
...
```
weeeeeeeeeeeeeeeeeeee

Also, see this astounding article on testing that will leave you wanting more.

One operand:

Operand	Description
`-a`	`FILE` exists
`-b`	`FILE` exists and is block special
`-c`	`FILE` exists and is character special
`-d`	`FILE` exists and is a directory
`-e`	`FILE` exists
`-f`	`FILE` exists and is a regular file
`-g`	`FILE` exists and is set-group-ID
`-G`	`FILE` exists and is owned by the effective group ID
`-h`	`FILE` exists and is a symbolic link (same as `-L`)
`-k`	`FILE` exists and has its sticky bit set
`-L`	`FILE` exists and is a symbolic link (same as `-h`)
`-N`	`FILE` exists and has been modified since it was last read
`-O`	`FILE` exists and is owned by the effective user ID
`-p`	`FILE` exists and is a named pipe
`-r`	`FILE` exists and read permission is granted
`-s`	`FILE` exists and has a size greater than zero
`-S`	`FILE` exists and is a socket
`-t`	`FD` is opened on a terminal
`-u`	`FILE` exists and its set-user-ID bit is set
`-w`	`FILE` exists and write permission is granted
`-x`	`FILE` exists and execute (or search) permission is granted

Two operands:

Operands	Description
`FILE1 -ef FILE2`	`FILE1` and `FILE2` have the same device and `inode` numbers
`FILE1 -nt FILE2`	`FILE1` is newer (modification date) than `FILE2`
`FILE1 -ot FILE2`	`FILE1` is older than `FILE2`

Is executable?

$ [ -x /bin/bash ]
$ echo $?
0
$ test -x /bin/bash
$ echo $?
0

Is a soft link?

$ test -L /lib64/ld-linux-x86-64.so.2
$ echo $?
0

Is present and is a directory?

$ [ -d /etc ] ; echo $?
0
$ [ -d /etcy ] ; echo $?
1

Expressions

( EXPRESSION ) - EXPRESSION is true
! EXPRESSION - EXPRESSION is false
EXPRESSION1 -a EXPRESSION2 - both EXPRESSION1 and EXPRESSION2 are true
EXPRESSION1 -o EXPRESSION2 - either EXPRESSION1 or EXPRESSION2 is true

`set -o` vs `shopt`

So, what is the difference between set -o and shopt, anyway?

set -o options are those inherited from Bourne-style shells like ksh, and the shopt options are those specific to `bash.

The help information for shopt is quite telling:

$ help shopt
shopt: shopt [-pqsu] [-o] [optname ...]
    Set and unset shell options.

    Change the setting of each shell option OPTNAME.  Without any option
    arguments, list each supplied OPTNAME, or all shell options if no
    OPTNAMEs are given, with an indication of whether or not each is set.

    Options:
      -o        restrict OPTNAMEs to those defined for use with `set -o'
      -p        print each shell option with an indication of its status
      -q        suppress output
      -s        enable (set) each OPTNAME
      -u        disable (unset) each OPTNAME

    Exit Status:
    Returns success if OPTNAME is enabled; fails if an invalid option is
    given or OPTNAME is disabled.

Let’s create an interactive non-login subshell. Then, we’ll print the total number of shell options from shopt and the number after its been restricted to just those supported by set -o:

$ bash
$ shopt | wc -l
53
$ shopt -o | wc -l
27

So, it looks like the bash shell has added support for many more shell options that aren’t available in older shells. Probably, it’s important to be aware of this when writing shell scripts, and another reason why a static analysis tool like shellcheck is essential and should absolutely be part of your toolchain.

For example, if you use the pipefail shell option (always a good idea) in your shell scripts but then have it interpreted by the Bourne shell, shellcheck will give you the following error:

In POSIX sh, set option pipefail is undefined.

Of course, I’m using shellcheck as a Vim plugin because I’m cool as hell.

In .vimrc:
call plug#begin('~/.vim/plugged')
Plug 'koalaman/shellcheck'
call plug#end()

Lastly, you can quickly find out which shell options are enabled, that is, which ones are reported as “on” by shopt.

$ echo $BASHOPTS
autocd:checkjobs:checkwinsize:cmdhist:complete_fullquote:expand_aliases:extquote:force_fignore:globasciiranges:globstar:histappend:hostcomplete:interactive_comments:login_shell:progcomp:promptvars:sourcepath

This is a read-only variable, and each word in the list is a valid argument for the -s option to shopt.

`IFS`

The IFS environment variable stands for the Internal Field Separator.

It is an array of three whitespace characters:

[SPACE]
\t
\n

$ echo ${#IFS}
3
$
$ printf '%q' "$IFS"
$' \t\n'
$
$ for v in "$IFS"; do printf '%q' "$v"; done
$' \t\n'

Scripting

In order be able to execute a script using the filename as the argument to the interpreter (i.e., bash foo.sh), the read permission bit must be set. Otherwise, a Permission Denied error is raised:

$ bash hello.sh
hi
$ chmod 200 foo.sh
$ bash foo.sh
bash: foo.sh: Permission denied

To execute the file by its path, the execute permission bit must be enabled.

The following are all equivalent (just pretend the mclovin.sh script already exists):

$ /home/btoll/mclovin.sh
hi
$ $(pwd)/mclovin.sh
hi
$ ./mclovin.sh
hi

However, the script’s directory must be included in the PATH if the pathname is to be omitted, leaving just the file name:

$ mclovin.sh
-bash: mclovin.sh: command not found

Add the directory to the PATH, and you will be golden, perhaps even the golden god.

Local Variables

The scripts also have the same special parameters as listed in the Builtin Variables section, with the same meaning.

Positional parameters are numeric. The first one is the script name ($0 or $BASH_ARGV0). Any numbers greater than nine need to be enclosed in curly brackets, such as ${10}. However, if your shell script is accepting that many parameters, you need to rethink some things.

Reading Input

Getting input from the command-line is easy: use read.

Examples:

$ echo "Do you want to continue (y/n)?"
$ read ANSWER

If you don’t give a variable name, it will default to REPLY.

$ echo "Type your first name and last name:"
$ read FIRST LAST

This combines getting input from the user and also echoing at the same time:

$ read -p "Type your first name and last name:" FIRST LAST

`case`

case "$foo" in
debian | ubuntu | mint )
    echo -n "uses .deb"
    ;;

centos | fedora | opensuse )
    echo -n "uses .rpm"
    ;;
*)
    echo -n "uses unknown package format"
    ;;
esac

Note that the items to be matched in a case block can employ command subsitition, parameter and arithmetic expansion, etc.

`printf`

Many programming languages have a printf function, or something similarly-named that operates like it, that formats and prints data. You will not be surprised that bash has a shell builtin called printf that allows for the same functionality.

Here is an example from the LPIC-1 documentation:

$ OS=$(uname -o)
$ FREE=$(( 1000 * `sed -nre '2s/[^[:digit:]]//gp' < /proc/meminfo` ))
$ MSG='Operating system:\t%s\nUnallocated RAM:\t%d MB\n'
$ printf "$MSG" $OS $(( $FREE / 1024**2 ))
Operating system:       GNU/Linux
Unallocated RAM:        19375 MB

printf is short for print formatted.

Random Examples

get the length of a variable
```
OS=$(uname -o)
echo "${#OS}"
9
```
create an array (one-dimensional)
- declare -a SIZES
- SIZES=( 5677 77 23 )
  - this both declares and defines
- SIZES=( $(cut -f2 /proc/filesystems | head -3) )
  - use command substitution
  - this works, but it will only create an array of one element
```
echo "${SIZES[0]}"
sysfs
tmpfs
bdev
```
  - to get three elements, use mapfile or its synonym readarray to split at each newline (\n) and generate a full array
```
mapfile -t SIZES <<< "$(cut -f2 /proc/filesystems | head -3 )"
echo "${SIZES[0]}"
sysfs
```
access array elements
- get the first element
  - $SIZES
  - ${SIZES[0]}
- change the first element
  - SIZES[0]=20
- get the length of the third element
  - ${#SIZES[2]}
- get total length
  - ${#SIZES[*]}
  - ${#SIZES[@]}
Trying to access a non-existent array element (i.e., out-of-bounds) does not produce an error.

arithmetic

the following examples use the expr and bc utilities:
```
$ expr 5 + 4
9
$ bc <<< 5+4
9
```

bash

$ sum=$[5+4] ; echo $sum
9
$ _32=$((8*4))
$ echo $_32
32

Summary

Continue your journey with the second installment in this titillating series, On the LPIC-1 Exam 102: User Interfaces and Desktops.

On the LPIC-1 Exam 102: Shells and Shell Scripting