Shellscript match regex

Question

I am trying to check if a variable matches a regex in POSIX shell. However, this seems more difficult than I thought. I am trying to check if the variable is a valid size string (e.g. 10M or 6G etc)

if echo $var | grep -Eq '^\d+[MG]$';
then
    echo "match"
else
    echo "no match"
fi

That's what I tried, but for some reason I never get a match, even if the variable contains the correct string? Any idea why?

Always paste your script into https://shellcheck.net, a syntax checker, or install shellcheck locally. Make using shellcheck part of your development process. — waltinator, May 10 '21 at 00:23
Alright. It told me to add doublequotes, however, the problem remains. Will use shellcheck from now on tho — user4042470, May 10 '21 at 00:26
Related: grep not working as expected and Why does my regular expression work in X but not in Y? — steeldriver, May 10 '21 at 00:36

score 7 · Accepted Answer · answered May 10 '21 at 00:42

GNU grep supports three types of regular expressions: Basic, Extended (ERE) and Perl (PCRE). In GNU grep, EREs don't provide functionality over basic ones, but some characters have special meanings, such as the plus sign. Source: man page.

\d does not mean anything special in an ERE; it's just the character d. To express digits, use [[:digit:]] or the old [0-9] (I wonder if there are character encodings where [0-9] is not the same as [[:digit:]]).

Your expression works as a PCRE, though:

if echo $var | grep -Pq '^\d+[MG]$';
then
    echo "match"
else
    echo "no match"
fi

Note the -P option instead of -E.

It would seem that POSIX grep does not support PCREs. I have not read the POSIX ERE definition, though.

Does every country use Arabic numerals?. I wonder if there are character encodings where [0-9] is not the same as [0123456789] — glenn jackman, May 10 '21 at 12:11
Ironically, I guess Arabic countries don't, or they do plus they use their own fonts. — berndbausch, May 10 '21 at 12:18

Shellscript match regex

1 Answers1