bash regex with quotes?

It was changed between 3.1 and 3.2. Guess the advanced guide needs an update.

This is a terse description of the new features added to bash-3.2 since the release of bash-3.1. As always, the manual page (doc/bash.1) is the place to look for complete descriptions.

1. New Features in Bash

snip

f. Quoting the string argument to the [[ command's =~ operator now forces string matching, as with the other pattern-matching operators.

Sadly this'll break existing quote using scripts unless you had the insight to store patterns in variables and use them instead of the regexes directly. Example below.

$ bash --version
GNU bash, version 3.2.39(1)-release (i486-pc-linux-gnu)
Copyright (C) 2007 Free Software Foundation, Inc.
$ number=2
$ if [[ $number =~ "[0-9]" ]]; then echo match; fi
$ if [[ $number =~ [0-9] ]]; then echo match; fi
match
$ re="[0-9]"
$ if [[ $number =~ $re ]]; then echo MATCH; fi
MATCH


$ bash --version
GNU bash, version 3.00.0(1)-release (i586-suse-linux)
Copyright (C) 2004 Free Software Foundation, Inc.
$ number=2
$ if [[ $number =~ "[0-9]" ]]; then echo match; fi
match
$ if [[ "$number" =~ [0-9] ]]; then echo match; fi
match

Bash 3.2 introduced a compatibility option compat31 which reverts bash regular expression quoting behavior back to 3.1

Without compat31:

$ shopt -u compat31
$ shopt compat31
compat31        off
$ set -x
$ if [[ "9" =~ "[0-9]" ]]; then echo match; else echo no match; fi
+ [[ 9 =~ \[0-9] ]]
+ echo no match
no match

With compat31:

$ shopt -s compat31
+ shopt -s compat31
$ if [[ "9" =~ "[0-9]" ]]; then echo match; else echo no match; fi
+ [[ 9 =~ [0-9] ]]
+ echo match
match

Link to patch: http://ftp.gnu.org/gnu/bash/bash-3.2-patches/bash32-039

GNU bash, version 4.2.25(1)-release (x86_64-pc-linux-gnu)

Some examples of string match and regex match

    $ if [[ 234 =~ "[0-9]" ]]; then echo matches;  fi # string match
$


$ if [[ 234 =~ [0-9] ]]; then echo matches;  fi # regex natch
matches




$ var="[0-9]"


$ if [[ 234 =~ $var ]]; then echo matches;  fi # regex match
matches




$ if [[ 234 =~ "$var" ]]; then echo matches;  fi # string match after substituting $var as [0-9]


$ if [[ 'rss$var919' =~ "$var" ]]; then echo matches;  fi   # string match after substituting $var as [0-9]


$ if [[ 'rss$var919' =~ $var ]]; then echo matches;  fi # regex match after substituting $var as [0-9]
matches




$ if [[ "rss\$var919" =~ "$var" ]]; then echo matches;  fi # string match won't work


$ if [[ "rss\\$var919" =~ "$var" ]]; then echo matches;  fi # string match won't work




$ if [[ "rss'$var'""919" =~ "$var" ]]; then echo matches;  fi # $var is substituted on LHS & RHS and then string match happens
matches


$ if [[ 'rss$var919' =~ "\$var" ]]; then echo matches;  fi # string match !
matches






$ if [[ 'rss$var919' =~ "$var" ]]; then echo matches;  fi # string match failed
$


$ if [[ 'rss$var919' =~ '$var' ]]; then echo matches;  fi # string match
matches






$ echo $var
[0-9]


$


$ if [[ abc123def =~ "[0-9]" ]]; then echo matches;  fi


$ if [[ abc123def =~ [0-9] ]]; then echo matches;  fi
matches


$ if [[ 'rss$var919' =~ '$var' ]]; then echo matches;  fi # string match due to single quotes on RHS $var matches $var
matches




$ if [[ 'rss$var919' =~ $var ]]; then echo matches;  fi # Regex match
matches
$ if [[ 'rss$var' =~ $var ]]; then echo matches;  fi # Above e.g. really is regex match and not string match
$




$ if [[ 'rss$var919[0-9]' =~ "$var" ]]; then echo matches;  fi # string match RHS substituted and then matched
matches


$ if [[ 'rss$var919' =~ "'$var'" ]]; then echo matches;  fi # trying to string match '$var' fails




$ if [[ '$var' =~ "'$var'" ]]; then echo matches;  fi # string match still fails as single quotes are omitted on RHS


$ if [[ \'$var\' =~ "'$var'" ]]; then echo matches;  fi # this string match works as single quotes are included now on RHS
matches

As mentioned in other answers, putting the regular expression in a variable is a general way to achieve compatibility over different bash versions. You may also use this workaround to achieve the same thing, while keeping your regular expression within the conditional expression:

$ number=1
$ if [[ $number =~ $(echo "[0-9]") ]]; then echo matched; fi
matched
$

Using a local variable has slightly better performance than using command substitution.

For larger scripts, or collections of scripts, it might make sense to use a utility to prevent unwanted local variables polluting the code, and to reduce verbosity. This seems to work well:

# Bash's built-in regular expression matching requires the regular expression
# to be unqouted (see https://stackoverflow.com/q/218156), which makes it harder
# to use some special characters, e.g., the dollar sign.
# This wrapper works around the issue by using a local variable, which means the
# quotes are not passed on to the regex engine.
regex_match() {
local string regex
string="${1?}"
regex="${2?}"
# shellcheck disable=SC2046 `regex` is deliberately unquoted, see above.
[[ "${string}" =~ ${regex} ]]
}

Example usage:

if regex_match "${number}" '[0-9]'; then
echo matched
fi