Grab ID of OS from /etc/os-release

Question

When I cat /etc/os-release I get the following:

PRETTY_NAME="Kali GNU/Linux Rolling"
NAME="Kali GNU/Linux"
ID=kali
VERSION="2018.1"
VERSION_ID="2018.1"
ID_LIKE=debian
ANSI_COLOR="1;31"
HOME_URL="http://www.kali.org/"
SUPPORT_URL="http://forums.kali.org/"
BUG_REPORT_URL="http://bugs.kali.org/"

How would I grab kali from ID= in bash? How would I grab 2018.1 from VERSION= in bash?

Answer 1

Sourcing and awk are poor ideas

Sourcing these sorts of files with a shell script interpreter is a poor idea, as it introduces yet another place where the malicious can hide shell script to be executed with superuser permissions, or can introduce things like an override for variables like PATH, LANG, and LD_LIBRARY_PATH.

% cat /etc/os-release
PRETTY_NAME="Debian GNU/Linux 9 (stretch with subversive \$PATH)"
NAME="Debian GNU/Linux"
VERSION_ID="9"
VERSION="9 (stretch)"
ID=debian
HOME_URL="https://www.debian.org/"
SUPPORT_URL="https://www.debian.org/support"
BUG_REPORT_URL="https://bugs.debian.org/"
echo 1>&2 'install -m 0644 /etc/shadow /home/malefactor/etc/ 2>/dev/null'
PATH="/home/malefactor/bin:${PATH}"
%
% ( source /etc/os-release ; command -v start-stop-daemon ; )
install -m 0644 /etc/shadow /home/malefactor/etc/ 2>/dev/null
/home/malefactor/bin/start-stop-daemon
%

Sourcing actually violates one of the defined semantics for /etc/os-release, which is that (to quote its manual) "variable expansion is explicitly not supported".

Similarly, awk as supplied in Archemar's answer does not correctly handle quoting; and any value can be quoted. Notice how Archemar's answer does not address the VERSION value in the question, which the given awk script does not provide the correct value for, because it erroneously retains the quoting and does not handle escape sequences.

% awk -F= '$1=="VERSION" { print $2 ;}' /etc/os-release
"9 (stretch)"      
% awk -F= '$1=="PRETTY_NAME" { print $2 ;}' /etc/os-release
"Debian GNU/Linux 9 (stretch with subversive \$PATH)"
%

/etc/os-release is actually permitted a fairly broad latitude in quoting, and also requires that escape sequences be processed properly. awk is not really the right tool for the job at all, here.

There are other, subtler, problems such as leakage of already set variables into the results if the actual key sought is missing from the file.

% ( source /etc/os-release ; echo $LANG ; )
install -m 0644 /etc/shadow /home/malefactor/etc/ 2>/dev/null
en_GB.UTF-8
%

There's also the fact that if ID or PRETTY_NAME are missing, they are defined as having default values. And on top of that, there's the fallback to /usr/lib/os-release to deal with.

Learning from other people

Files that contain key-equals-value assignments, with sh-style quoting, escaping, and commentary, are a fairly common thing. Some programming languages have library functions for dealing with them directly. The lesson of OpenBSD's rules shift on /etc/rc.conf in OpenBSD 5.6 is that it is the wiser course to use such library functions, if available, or dedicated tools that have nowhere near the capabilities of a full shell interpreter, for processing such files.

From shell scripts, I for one use a tool named read-conf to process such files:

% read_os() {
    if test -r /etc/os-release
    then
        clearenv setenv "$1" "$2" read-conf /etc/os-release printenv "$1"
    else
        clearenv setenv "$1" "$2" read-conf /usr/lib/os-release printenv "$1"
    fi
}
% read_os ID linux
debian
% read_os VERSION
9 (stretch)
% read_os PRETTY_NAME Linux
Debian GNU/Linux 9 (stretch with subversive $PATH)
% read_os PATH
/home/malefactor/bin:${PATH}
% read_os LANG
%

The aforegiven relies on setenv, read-conf, and printenv commands being built-in commands in the nosh toolset, so that clearenv, setenv, and read-conf each find the next command in the chain without using the PATH environment variable at all. (Before I added the built-in printenv, the slightly longer original one-liners made careful use of "`command -v printenv`" to avoid searching for printenv after the infected configuration file had maliciously set an altered PATH, and after clearenv had wiped that variable from the environment.)

Further reading

• Theo de Raadt et al. (2014-11-01). OpenBSD 5.6 Changelog. OpenBSD.
• Jonathan de Boyne Pollard (2018). "read-conf". Manual. nosh toolset. Softwares.
• Jonathan de Boyne Pollard (2018). "clearenv". Manual. nosh toolset. Softwares.
• Jonathan de Boyne Pollard (2018). "setenv". Manual. nosh toolset. Softwares.
• Jonathan de Boyne Pollard (2018). "printenv". Manual. nosh toolset. Softwares.

Answer 2

you can source the file and use var's value

. /etc/os-release
echo $ID
echo $VERSION

or try with awk

awk -F= '$1=="ID" { print $2 ;}' /etc/os-release

where

• -F= tell awk to use = as separator
• $1=="ID" filter on ID
• { print $2 ;} print value

Answer 3

ID=$(grep -oP '(?<=^ID=).+' /etc/os-release | tr -d '"')
VERSION=$(grep -oP '(?<=^VERSION_ID=).+' /etc/os-release | tr -d '"')
echo $ID:$VERSION

For the explanation see the other two great answers:

• Can grep output only specified groupings that match?
• Shell Script - remove first and last quote (") from a variable

Answer 4

I recently had to do something similar and found this answer. Maybe this helps:

grep -ioP '^ID=\K.+' /etc/os-release

The very cool part about this is the \K. Simply, ignore it while parsing the regex in your head. Once you see a match in a line, then the result to be outputted on the terminal will start from right after the \K.

Answer 5

Heeding @JdeBP's recommendations not to blindly interpret any shell code in those files, but still handle quoting the way a shell would, in zsh, you could do that using its z and Q parameter expansion flags which are designed to perform shell tokenisation and quote removal:

contents=$(</etc/os-release)
shell_tokens=( ${(zZ[Cn])contents} )

set -o extendedglob
assignment_tokens=( ${(M)shell_tokens:#[[:IDENT:]]##=*} )
typeset -A field=()
for token ($assignment_tokens) field[${token%%=*}]=${(QX)token#*=}

Then, you can use $field[VERSION], $field[VERSION_ID]... in your zsh script.

• $(<file): like in ksh: gets the contents of the file stripped of all trailing newline characters¹
• ${(z)scalar} extracts all the shell tokens².
• Z[Cn]: tune the z flag, here to strip shell comments and treat newline as just whitespace.
• ${array:#pattern}: expand to the array elements except those matching the pattern. With the (M) flag, expands to only the Matching elements instead.
• [[:IDENT:]]: characters allowed in a shell identifier.
• ##: (with extendedglob): one or more of the preceding atom (like + in extended regexps).
• ${(QX)var}: removes one layer of quoting (X to report errors in there).

---

^{¹ beware that contrary to ksh/bash, NUL bytes in there are preserved.}

^{² top-level ones, i.e. not recursively if there are command substitutions for instance.}