Linux regular expressions

[rogersma]

Ok, I feel like a complete idiot, but I can't figure this out.

I had a directory that contained a bunch of files. Many had names like c1.out, c2.out, ... but there were also some named AS32.out, BT12.out, ...

I didn't need any of these latter files. Since they were all characterized by upper-case letters, I attempted to remove them using:

rm [A-Z]*.out
This removed everything in the directory, which I didn't want. Fortunately I could run the same program to generate the files.

I thought perhaps "[A-Z]*" meant "0 or more", so I tried this instead:

rm [A-Z][A-Z]*.out

This also removed all the files. Now I really feel like an idiot. Can someone spot my error (which I'm sure is painfully obvious)? How do I remove the [A-Z][A-Z]*.out files without also removing the c*.out files? I should mention my shell is bash.

[Bucky Ball]

Thread moved to Programming Talk.

I can't help, and I realise you're relatively new here, but this could be the best place for this support request.

Any commands involving 'rm' should be treated with caution, particularly if proceeded by becoming root, as I'm sure you're aware. ;)

Good luck.

[ofnuts]

Ok, I feel like a complete idiot, but I can't figure this out.

I had a directory that contained a bunch of files. Many had names like c1.out, c2.out, ... but there were also some named AS32.out, BT12.out, ...

I didn't need any of these latter files. Since they were all characterized by upper-case letters, I attempted to remove them using:

rm [A-Z]*.out
This removed everything in the directory, which I didn't want. Fortunately I could run the same program to generate the files.

I thought perhaps "[A-Z]*" meant "0 or more", so I tried this instead:

rm [A-Z][A-Z]*.out

This also removed all the files. Now I really feel like an idiot. Can someone spot my error (which I'm sure is painfully obvious)? How do I remove the [A-Z][A-Z]*.out files without also removing the c*.out files? I should mention my shell is bash.

These aren't regular expressions. These are filename expansions (https://www.gnu.org/software/bash/manual/html_node/Filename-Expansion.html) (that are done by the shell, before calling the command, so you can try with "ls" before using them with "rm"). They use a different syntax (https://www.gnu.org/software/bash/manual/html_node/Pattern-Matching.html#Pattern-Matching). You can also use brace expansion (https://www.gnu.org/software/bash/manual/html_node/Brace-Expansion.html) to enumerate the files.

[steeldriver]

I think the issue here is that the shell interprets commands using glob syntax rather than regex syntax. TBH I'm not sure the right way to do what you want with globs - perhaps



rm +([[:upper:]])*.out


(one or more upper-case letters followed by anything, followed by .out) - but test it first with something innocuous like



ls +([[:upper:]])*.out


or even just



echo +([[:upper:]])*.out

[sisco311]

I think the issue here is that the shell interprets commands using glob syntax rather than regex syntax. TBH I'm not sure the right way to do what you want with globs - perhaps



rm +([[:upper:]])*.out


(one or more upper-case letters followed by anything, followed by .out) - but test it first with something innocuous like



ls +([[:upper:]])*.out


or even just



echo +([[:upper:]])*.out



+(pattern-list) is an extglob in bash. extglobs aren't enabled by default.

@OP check out:

man 7 glob
http://mywiki.wooledge.org/locale
http://mywiki.wooledge.org/glob


In bash the LC_COLLATE variable defines the order in which names are sorted.

[Vaphell]

[A-Z] is not foolproof because it depends on locale, LC_COLLATE to be exact. In some locales chars are ordered aAbBcC so the range catches pretty much everything.

[[:upper:]]* should always match only names with 1st uppercase char.

[steeldriver]

I didn't want to complicate things by mentioning extglob - afaik it *is* enabled by the default for interactive shells

[sisco311]

I didn't want to complicate things by mentioning extglob - afaik it *is* enabled by the default for interactive shells


You are right, steeldriver. I stand corrected.

Thanks!

[rogersma]

[A-Z] is not foolproof because it depends on locale, LC_COLLATE to be exact. In some locales chars are ordered aAbBcC so the range catches pretty much everything.

[[:upper:]]* should always match only names with 1st uppercase char.


I think this may be it. I just tried the exact same same rm [A-Z]*.out on a different machine and it worked just fine. Oddly they both have the exact same Ubuntu installed, but different people set them up.

Thanks for your fast replies -- the answer wasn't quite as obvious as I'd thought!

[gray380]

Hi,

Try to put before rm:
$ export LC_ALL=POSIX

should work.

brg,
Serhiy.