Globbing pattern: exclude a directory

Question

I have the following filepaths

/vol/form/various.txt
/vol/var/sender.py
/vol/var/hello.txt

I would like to get all .txt files that do not have the directory of form in them. For example, something like:

*.txt AND ! */form/*

What would be the correct globbing pattern to do this -- i.e, in a single pattern (or is that not possible)?

With a combination of `globstar` and `extglob`, this glob might work: `**/!(form)/*.txt` – but I've seen `!(...)` be not very reliable; for example, if there is a text file in a subdirectory of `form`, it won't be excluded, and `**/!(form)/**/*.txt` matches everything. But if the `.txt` file is always directly in the `form` directory, it should work. — Benjamin W., Jun 28 '20 at 21:02

Gilles Quénot · Accepted Answer · 2020-06-28T20:19:32.270

1

Like this:

find /vol ! -path '*/form/*' -type f -name '*.txt'

If you need to negate a pattern, like with a regex, AFAIK look around are not supported by find -regex*. So it's not possible only with one find regex expression.

edited Jun 28 '20 at 20:19

answered Jun 28 '20 at 19:49

Gilles Quénot

173,512
41
224
223

@Giles -- sure, is it possible to do it with just one `-path` argument? Or is the chaining necessary, that's basically my question. – David542 Jun 28 '20 at 19:55
You don't show us what you've tried, this is required for good quality question. Have you tested my solution ? It's a full working command. Dunno what you need more. It's exactly the purpose of your question – Gilles Quénot Jun 28 '20 at 19:57
@david542 -- yes I know it works, I was just wondering about the ability to do more complex patterns in a single pattern string. – David542 Jun 28 '20 at 19:59
1

This is not in your original question. Don't change the rules in comments: open a new question if this is not not sufficient – Gilles Quénot Jun 28 '20 at 20:02

dawg · Answer 2 · 2020-06-28T22:57:22.763

1

You can negate a grep:

find /vol -type f -name '*.txt' | grep -v '/form/'

If you want a single regex you need to either find file names with find and feed to GNU grep:

find /vol -type f | ggrep -oP "(?!.*?\/form\/)(^.*\.txt$)"

or use Perl with that same regex:

perl -MFile::Find -e 'find sub {
      print "$File::Find::name\n" if -f && m/(?!.*?\/form\/)(^.*\.txt$)/
      },"/vol/"'

Explanation of the regex:

(?!^.*?\/form\/)(^.*\.txt$)

^              ^                 Negative lookahead fail on /form/
                 ^       ^       Anchors for start and end of string
                   ^             all horizontal characters
                     ^^          literal . 
                                 (if not escaped it would match any character)
                       ^         txt extension

Demo

edited Jun 28 '20 at 22:57

answered Jun 28 '20 at 20:21

dawg

98,345
23
131
206

Yes, but not in a _single expression_ or _single parrern_ like OP wrote – Gilles Quénot Jun 28 '20 at 20:24
1

If you can do it in a single regex expression, show us =) – Gilles Quénot Jun 28 '20 at 21:13
There you go! Possible but not elegant – dawg Jun 28 '20 at 21:23
Yes, not very obvious to re-use. What is `(^)` ? Don't get it – Gilles Quénot Jun 28 '20 at 21:56
@GillesQuenot the `^` would mean start of the string (`$` would mean end). I don't think it's required here since the filepaths are sort of self-contained and anchored by themselves. Is that what you mean? – David542 Jun 28 '20 at 22:01
@GillesQuenot here's an example: https://regex101.com/r/V4ag0W/2. If you remove the `^` it could match starting from halfway into the filepath, rather than from the start of the line. Try removing the `^` in he above regex (I had to do that myself to see what it was needed, or if it was just cosmetic). – David542 Jun 28 '20 at 22:04
@dawg how about this for the regex? `^(?!.*?\/form\/).*\.txt` -- https://regex101.com/r/V4ag0W/3 ? – David542 Jun 28 '20 at 22:04
It doesn't match the second line: https://regex101.com/r/V4ag0W/4 – Gilles Quénot Jun 28 '20 at 22:05
1

@It's not supposed to, the extension there is `.py` instead of `.txt`. – David542 Jun 28 '20 at 22:06
yea I'm not sure, I don't know perl, but using those same inputs in regex101 it looks ok. I would ask @dawg why. – David542 Jun 28 '20 at 22:49
I am not sure what that pastebin is showing. Check the regex on regex101 and it works as stated.... – dawg Jun 28 '20 at 22:58

Globbing pattern: exclude a directory

2 Answers2