bug#38725: Regexp bug

[Davide Brini]

On Mon, 23 Dec 2019 23:09:28 +0100, Chewbaka Roskov <address@hidden>
wrote:

> akem@akem-HP:~$ echo '112233' | sed 's/2*//'  ### DOESN'T WORK
> 112233

You're asking for "zero or more twos". At the beginning of the string (and
between each pair of characters) there is indeed a zero-length match for
your regexp, which sed duly replaces with nothing as you asked. Had you
replaced with something or used the /g switch to the "s" command, you'd
have seen it. For example:

$ echo '112233' | sed 's/2*//g'
1133
$ echo '112233' | sed 's/2*/XXX/'
XXX112233

# Blatant test case

$ echo 'abc' | sed 's/2*/XXX/g'
XXXaXXXbXXXcXXX


> akem@akem-HP:~$ echo '112233' | sed 's/1*//'  ### THIS ONE WORK
> 2233

In this case 1* matches at the beginning of the string as before, but it's
allowed to match more than zero characters, so the "11" (the match for 1*)
is replaced with nothing.

> akem@akem-HP:~$ echo 'ee112233' | sed 's/1*//'  ### BUT THIS ONE DOES NOT
> ee112233

Same as first case.

--
D.