doubt in grep regular expression

[sathiya]

what does the following command sequence means ?

grep "d^c"

What does it will get matched ?

[secondmouse]

It matches any strings containing "d^c", for example:
echo "d^c" | grep "d^c"
echo "ABCd^cMM" | grep "d^c"

When ^ is not the first character in the expression, it is just a literal ^, otherwise it means the beginning of the line (when there are no square brackets around).

Read GREP for more examples.

[sathiya]

Thanks for the same.


I understood that, if the ^ does not occurs at first it will be taken literal..


But now i have strange problem that, it works like what you said in one server, but does not works in another server. I checked the version, and both are same 2.5.1. So what could be the issue.

[secondmouse]

Actually I got similar results after you mentioned in your last post. I had it working on
OpenSuSE 10.3 (grep version 2.5.2) and Ubuntu 8.04 (grep 2.5.3)
but not OpenSuSE 10.2. (grep 2.5.1)
Now I wonder the same question.

[burschik]

I suspect that behaviour does not depend on the grep version, but rather on the locale settings:

$ export LC_ALL=en_US
$ grep "b^gg" testinput
>
$ export LC_ALL=en_US.UTF-8
$ grep "b^gg" testinput
> b^gg

[secondmouse]

I suspect that behaviour does not depend on the grep version, but rather on the locale settings:

$ export LC_ALL=en_US
$ grep "b^gg" testinput
>
$ export LC_ALL=en_US.UTF-8
$ grep "b^gg" testinput
> b^gg

Not really. On my Ubuntu 8.04. grep matches no matter whatever LC_ALL setting is. Same story with my OpenSUSE 10.3.

[burschik]

Alright then. It is this bug:

grep - Bugs: bug #9519, echo do^re | grep do^re; fails... [Savannah]

[drl]

Hi.

Good research on the reported bug.

I would consider character caret (^) to be special enough that I would escape it. In doing so, you have portability among programs and versions. Here are a few trial variations:

#!/bin/bash -

# @(#) s1       Demonstrate handling of caret (^) in [e]grep, perl.

LC_ALL=C ; LANG=C ; export LC_ALL LANG
echo
echo "Environment: LC_ALL = $LC_ALL, LANG = $LANG"
echo "(Versions displayed with local utility \"version\")"
version >/dev/null 2>&1 && version "=o" $(_eat $0 $1) grep perl
set -o nounset
echo

echo " Results with grep:"
echo "1d^c" | grep "d^c"
echo "2d^c" | grep "d\^c"
echo "3d^c" | grep "d[^]c"
echo "4d^c" | grep -U "d^c"

echo
echo " Results with egrep:"
echo "1d^c" | egrep "d^c"
echo "2d^c" | egrep "d\^c"
echo "3d^c" | egrep "d[^]c"
echo "4d^c" | egrep -U "d^c"

echo
echo " Results with perl:"
echo "1d^c" | perl -wp -e '/d^c/;'
echo "2d^c" | perl -wp -e '/d\^c/;'
echo "3d^c" | perl -wp -e '/d[^]c/;'

exit 0

Producing:

% ./s1

Environment: LC_ALL = C, LANG = C
(Versions displayed with local utility "version")
Linux 2.6.11-x1
GNU bash 2.05b.0
grep (GNU grep) 2.5.1
perl 5.8.4

 Results with grep:
2d^c
grep: Invalid regular expression

 Results with egrep:
2d^c
grep: Invalid regular expression

 Results with perl:
1d^c
2d^c
Unmatched [ in regex; marked by <-- HERE in m/d[ <-- HERE ^]c/ at -e line 1.

I also ran this script with GNU grep 2.5.3 (on GNU/Debian lenny) and the line "2d^c" was matched (some others as well, showing the bug to have been fixed there).

So if you cannot change grep, then I would recommend using the \^ sequence or use a one-liner perl script, at least for the near future ... cheers, drl