Losing the plot: Basic grep

Khannie · 17-02-2010 12:12PM #1

Losing the plot here. Trying to match on any capital letters.

Ok, first off we have the working one:

echo Test | grep -E -e "[tT]est"
Test
echo $?
0

Perfect.

echo est | grep -E -e "[tT]est"
echo $?
1

Perfect also.

Now I do:

echo test | grep -E -e "[A-Z]"
test
echo $?
0

What in the name of god is happening here? I have tried many many variations of this and they all match.

Khannie · 17-02-2010 12:49PM

Slight update. I am running gentoo. It works as expected on an ubuntu box.

alias grep says:
alias grep='grep --colour=auto'

I just installed the latest repository version of grep. Same rubbish.

Skrynesaver · 17-02-2010 01:11PM

A lot of my scripts rely on

grep "thing" && doSomeMagicIncation.

This is quite weird RHEL has the same issue

grep --version
grep (GNU grep) 2.5.1

$ echo test | grep -E   '^te[A-Z]t$'
test
$ echo test | grep -E   '^te[0-9]t$'
$

Seems as though -i is set by default !!

Khannie · 17-02-2010 01:35PM

That looks like a bug to me to be honest. My head is melted here. I'm glad it's not just me though.

Skrynesaver · 17-02-2010 02:40PM

Seems to be only in character ranges that it happens

$ echo test | grep -E   '^teSt$'
$ echo test | grep -E   '^te[S]t$'
$ echo test | grep -E   '^te[R-T]t$'
test
$

So are you reporting it or am I?
PS. Which release of grep are you on?

Skrynesaver · 17-02-2010 02:49PM

Aha, found it.

 export LC_COLLATE=POSIX

and try again. Seems if you have dictionary sort order rather than posix sort order the range

[a-d]

is actually

[aBbCcDd]

.
Obvious really

from grep bug search

Khannie · 18-02-2010 11:54AM

Stuff like that really irritates me about Linux. Never in the history of the universe has a-d ever meant aAbBcCdD. Never!

Nice catch btw.

Skrynesaver · 18-02-2010 01:13PM

I think it's an aspect of catering to usability, people aren't used to the idea that Z comes before a and so a "friendly" locale setting fixes it for you.

If, as a result, some "obscure" thing like regex gets broken, well the geeks should be able to fix it themselves, seems to be the attitude.

niallb · 19-02-2010 11:44AM

Skrynesaver wrote: »

So are you reporting it or am I?
PS. Which release of grep are you on?

Real nice catch.
Who to report it to though?
It's really just a LOCALE issue.

Lets get started on en_IE.UTF8.pedantic :-)
To paraphrase sendmail: Don't blame grep!

Of course, it would be far more productive to fix the collate order in ga_IE,
and wait for bewilderment further down the line when only people from gaelscoileanna can pass a regex exam

Losing the plot: Basic grep

Comments