How do I use a regex in a shell script?

To complement the existing helpful answers:

Using Bash’s own regex-matching operator, =~, is a faster alternative in this case, given that you’re only matching a single value already stored in a variable:

set -- '12-34-5678' # set $1 to sample value

kREGEX_DATE='^[0-9]{2}[-/][0-9]{2}[-/][0-9]{4}$' # note use of [0-9] to avoid \d
[[ $1 =~ $kREGEX_DATE ]]
echo $? # 0 with the sample value, i.e., a successful match

Note, however, that the caveat re using flavor-specific regex constructs such as \d equally applies:
While =~ supports EREs (extended regular expressions), it also supports the host platform’s specific extension – it’s a rare case of Bash’s behavior being platform-dependent.

To remain portable (in the context of Bash), stick to the POSIX ERE specification.

Note that =~ even allows you to define capture groups (parenthesized subexpressions) whose matches you can later access through Bash’s special ${BASH_REMATCH[@]} array variable.

Further notes:

  • $kREGEX_DATE is used unquoted, which is necessary for the regex to be recognized as such (quoted parts would be treated as literals).

  • While not always necessary, it is advisable to store the regex in a variable first, because Bash has trouble with regex literals containing \.

    • E.g., on Linux, where \< is supported to match word boundaries, [[ 3 =~ \<3 ]] && echo yes doesn’t work, but re="\<3"; [[ 3 =~ $re ]] && echo yes does.
  • I’ve changed variable name REGEX_DATE to kREGEX_DATE (k signaling a (conceptual) constant), so as to ensure that the name isn’t an all-uppercase name, because all-uppercase variable names should be avoided to prevent conflicts with special environment and shell variables.

Leave a Comment