Bash regular expression match with groups including example to parse http_proxy environment variable

bash_regex_match_groups.md

The newer versions of bash (>= 3.0) include a regex operator =~

Simple example

$ re='t(es)t'
$ [[ "test" =~ $re ]]
$ echo $?
0
$ echo ${BASH_REMATCH[1]}
es

Note

• operator returns true if it's able to match
• nested groups are possible (example below shows ordering)
  • 0 is the whole match
  • optional groups are counted even if not present and will be indexed, but be empty/null
• global match isn't suported, so it only matches once
• the regex must be provided as an unquoted variable reference to the re var

Match fails if re specified directly as string. Seems to want to be unquoted...

[[ "test" =~ 't(es)t' ]]; echo $?;
1

More complex example to parse the http_proxy env var

http_proxy_re='^https?://(([^:]{1,128}):([^@]{1,256})@)?([^:/]{1,255})(:([0-9]{1,5}))?/?'
if [[ "$http_proxy" =~ $http_proxy_re ]]; then
  # care taken to skip parent nesting groups 1 and 5
  user=${BASH_REMATCH[2]}
  pass=${BASH_REMATCH[3]}
  host=${BASH_REMATCH[4]}
  port=${BASH_REMATCH[6]}
fi

Examples of tricky issues with limitations:

• Pattern backreference to an optional capturing subexpression
• Multiple matches in a string using regex in bash

seem to be first time in v3.0... precisly in bash-3.0-alpha...

• https://git.savannah.gnu.org/cgit/bash.git/tree/CHANGES#n6829

Thanks @mcarbonneaux, updated to reflect exact version of bash the operator is supported with.