bash: better way to delete files not matching a regex?

pbhj · 10-12-2007, 10:04 AM

I'm trying to delete all non-relevant .po files from a gallery-2.2.3-full release.

I've tried this:

Code:

find ./ -regextype posix-extended ! -regex ^.*\/en_GB\.po$ | grep -e .*\\.po$ >> rm.txt

Which gets me a list of files to remove. But I'm not sure what to do now as I can't seem to pipe that through rm with cat.

How should I do this?

I have actually come up with a fix of:

Code:

find ./ -regextype posix-extended -type f ! -regex ^.*\/en_GB\.po$ | grep po$ | xargs rm

But this seems rather dirty and hackish. Any easier way?

andrews-mark · 10-12-2007, 10:51 AM

maybe I'm misunderstanding (plus I am a bit stupid) but are you saying that you have a list of files you want to delete stored in a text file called "rm.txt". If so, I would do

Code:

for i in `cat rm.txt `; do rm $i;done

am I misunderstanding?

-mark

pixellany · 10-12-2007, 10:53 AM

If it works, it's OK.....

You can also use backtics to enclose the find|grep commands and cause their output to go to "rm":

rm `find.....|grep....`

(backtics, not single quotes)

pbhj · 10-12-2007, 04:50 PM

andrews-mark, thanks, no you didn't misunderstand and yes it's me that's stupid and not you; that would work fine.

pixellany, thanks too, thats the simple solution (or using $() instead of backticks, which I think is the "modern" approach) which I'd forgotten when I posted ... I was kinda hoping someone would mention a simple single command I could use without the arsing around with multiple pipes just to get a negative-match, but at least I have a couple options that work.

Thanks. Clearly not having a good brain day!

druuna · 10-12-2007, 05:13 PM

Hi,

If I understand correctly you are looking for files ending with .po, leaving out the files ending with en_GB.po.

If that is the case, would this help:

find . -type f -name "*.po" -not -name "*en_GB\.po"

If it comes up with the names you want and, more importantly, leaves out all that you don't want you can modify it to:

find . -type f -name "*.po" -not -name "*en_GB\.po" -exec rm {} \;

All in one command.

If the list that needs to be removed is long, you should use | xargs .... it's faster:

find . -type f -name "*.po" -not -name "*en_GB\.po" | xargs rm

Hope this helps.

radoulov · 10-13-2007, 05:28 AM

Or with zsh:

Code:

setopt kshglob
rm -- **/!(*en_GB).po

and if the files are in the same directory:

ksh

Code:

rm -- !(*en_GB).po

bash

Code:

shopt -s extglob
rm -- !(*en_GB).po

gnashley · 10-13-2007, 01:26 PM

If you want to delete language files perhaps you should look into the purgelocales utility.

pbhj · 10-15-2007, 10:31 AM

thanks again all.

I used druuna's

Code:

find . -type f -name "*.po" -not -name "*en_GB\.po" -exec rm {} \;

and saved 21M from an 80M install of gallery2 (http://codex.gallery2.org/Gallery2

ownload)

that's a sizeable chunk!

gnashley · 10-15-2007, 03:05 PM

I used the purgelocales utility on a full install once which was around 800MB IIRC -it removed nearly 150MB of stuff!
i have incorporated into my 'magic' package builder, src2pkg, so that you can delete the language files from packages as you build them.
src2pkg is here:

http://distro.ibiblio.org/pub/linux/...nload/src2pkg/