Skip to content

POSIXShell.md

Latest commit

 

History

History
255 lines (179 loc) · 6.35 KB

POSIXShell.md

File metadata and controls

255 lines (179 loc) · 6.35 KB
Raw

Debian Almquist Shell (Dash)

While Bash is very popular (and so is its syntax), Dash replaced Bash as non-interactive shell in 2011 with the release of Debian 6 (Squeeze). Dash is POSIX-compliant, small and fast. For the interactive shell Bash is still used – which is absolutely fine.

However, Bash is not POSIX-compliant which can be a portability problem, especially when shared with the open-source community. Instead of memorizing different syntaxes and features between different shells – which can be dangerously confusing – I decided to use exactly one shell. Since Dash is POSIX-compliant, the choice was clear and I stopped writing shell scripts in Bash.

All of the code above is POSIX-compliant. No language based special features are used ("bashism").

History

Debian Almquist Shell (dash) ‹ Almquist Shell (ash) ‹ System V Shell Release 4 (SVR4) ‹ Bourne Shell (sh)

Manual

See Chapter 2. Shell Command Language of the The Open Group Base Specifications (Issue 7, Edition 2018).

Helpful Tips

1.0 STDERR and STDOUT

A redirection instruction from standard output (STDOUT = file descriptor 1) or error output (STDERR = file descriptor 2) of a command can be placed either at the end or at the beginning to improve readabilty, so these both are equivalent:

foobar 1>/dev/null
1>/dev/null foobar

The instruction >/dev/null means: Redirect standard output (STDOUT) to /dev/null which is a pseudo-device file which just discards the input like a black hole.

For STDOUT you can simply omit the 1, so this is equivalent to 1>/dev/null foobar:

foobar >/dev/null

If you only want to suppress errors, you would replace the 1 with a 2. This would redirect the error output from the command to /dev/null:

foobar 2>/dev/null

To redirect both STDOUT and STDERR you would need to combine them with an additional instruction (2>&1):

foobar >/dev/null 2>&1

The 2>&1 here means: Redirect STDERR to STDOUT. Since STDOUT is redirected to /dev/null, both are suppressed.

If you want to redirect something to STDERR, e.g. an error message in your script, you can use >&2 and place it at the beginning for improved readabilty:

>&2 printf '%s\n' "This is an error message."

The >&2 here is equivalent to 1>&2.

2.0 Return does not return boolean

Bear in mind that return does not return a boolean, but an exit code which is a positive integer between 0 and 255. While 0 stands for success, any other value indicates an error:

# Check if string is an unsigned integer (also +1 is not considered as be valid)
func_IS_UNSIGNED_INTEGER() {
	NUMBER="$1"
	
	case "$NUMBER" in
		*[!0-9]* | '')
			# False
			return 1
		;;
	esac
	
	# True
	return 0
}

3.0 printf vs echo

You should use printf instead of echo, because it is more portable:

echo

echo "Text with a new line after it."

printf

printf '%s\n' "Text with a new line after it."
References
- https://askubuntu.com/questions/467747/which-is-better-printf-or-echo
- https://unix.stackexchange.com/questions/65803/why-is-printf-better-than-echo

4.0 Quoted Variables

In the most cases, you want to place double-quotes around variables ("$VAR"). This also applies to command substitution. Otherwise not the whole content of the variable – including whitespace – is passed, because the shell will split it at the whitespace.

Correct

RESULT="$(dpkg-query -s "$PACKAGE_NAME" >/dev/null 2>&1)"

Wrong

RESULT=$(dpkg-query -s "$PACKAGE_NAME" >/dev/null 2>&1)
RESULT=$(dpkg-query -s $PACKAGE_NAME >/dev/null 2>&1)
References
- https://unix.stackexchange.com/questions/171346/security-implications-of-forgetting-to-quote-a-variable-in-bash-posix-shells

Discouraged

While backticks (`) – or also called backquotes – are not officially deprecated, their use is discouraged. Instead use $(...).

RESULT="`dpkg-query -s "$PACKAGE_NAME"` >/dev/null 2>&1"
References
- https://unix.stackexchange.com/questions/126927/have-backticks-i-e-cmd-in-sh-shells-been-deprecated
- https://unix.stackexchange.com/questions/118433/quoting-within-command-substitution-in-bash

5.0 Comments

# Single line comment
: '''
Comment
over
multiple
lines.
'''

6.0 Strings

Length

VAR="This string is 34 characters long."
printf '%s\n' "${#VAR}"

Remove from the beginning

VAR="/foo/bar/"
printf '%s\n' "${VAR#/}"

Remove from the end

VAR="/foo/bar/"
printf '%s\n' "${VAR%/}"
References
- https://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html

7.0 Assignment of Multiple Variables

Let's say you have a configuration file with following content:

# ID     Name      Country
01       Morton    US
02       Matthew   CA

One way to assign these variables would be using awk '{print $...}':

PARAM_ID="$(printf '%s' "$LINE" | awk '{print $1}')"
PARAM_NAME="$(printf '%s' "$LINE" | awk '{print $2}')"
PARAM_COUNTRY="$(printf '%s' "$LINE" | awk '{print $3}')"

However, this will create one subshell for each parameter. What you could also do is following:

read -r \
PARAM_ID \
PARAM_NAME \
PARAM_COUNTRY \
__TRAILING \
<<-EOF
$(printf '%s' "$LINE")
EOF

The variable __TRAILING at the end covers the rest of the line in case it has more words then expected.

8.0 Looping Through a Range

In many examples you would see following code if you would want to loop through a range:

START=1
END=5

for i in $(seq $START $END)
do
	printf '%d\n' "$i"
done

Note that $END is inclusive (1, 2, 3, 4, 5). However, this is not POSIX-compliant due to the use of seq. Instead, you need to use a while loop. With -le $END is inclusive, with -lt exclusive:

START=1
END=5

i=$START

while [ $i -le $END ]
do
    printf '%d\n' "$i"
    i=$(( $i + 1 ))
done