can regex capture groups be used in GNU find command?

Question

With the GNU find command (GNU findutils 4.4.2), a regular expression can be used to search for files. For example:

$ find pool -regextype posix-extended -regex ".*/mypackage-([a-zA-Z0-9.]+-[0-9]{1,2})-x86_64.pkg.tar.xz+"

Is it possible to extract the capture group defined by that expression and use it in a -printf argument?

So, given a found file called pool/mypackage-1.4.9-1-x86_64.pkg.tar.xz, I would like to include the 1.4.9-1 part in a printf expression.

Is this possible?

No, but you can use -print0 and pipe to GNU sed -Ez (possibly followed by tr '\0' '\n') — Stéphane Chazelas, Jun 26 '14 at 11:38
Note that [a-zA-Z] only makes sense in the C/POSIX locale. — Stéphane Chazelas, Jun 26 '14 at 11:50
I expected the simple answer to be No given copious amounts of searching and reading man pages before asking the question. The answers provide interesting alternaive ways to achieve the desired output. — starfry, Jun 26 '14 at 13:22

score 5 · Answer 1 · edited Jun 26 '14 at 11:48

5

If you use

find pool -regextype posix-extended \
    -regex ".*/mypackage-([a-zA-Z0-9.]+-[0-9]{1,2})-x86_64\.pkg\.tar\.xz" \
    -printf '%f\n' |
  grep -Eo '[a-zA-Z0-9.]+-[0-9]{1,2}'

(assuming GNU grep as well), it should work for any path. The regex doesn't allow for any newlines, so there's no way to make it match for example a directory containing a similar name.

edited Jun 26 '14 at 11:48

Stéphane Chazelas

544,893

answered Jun 26 '14 at 11:39

l0b0

51,350

Note that in this particular case, you could replace the grep with a cut -d- -f2,3 to avoid the overhead of another regexp matching. – Stéphane Chazelas Jun 26 '14 at 12:09
When I used this, I got newlines embedded because the grep expression matches two parts. I guess the expression could be fixed but using cut sorts it. However, it's brittle: say the file was mypackage-other the fied numbers would be out. The other answer may be less efficient (not really an issue here) but it does work. – starfry Jun 26 '14 at 13:19
Yep, just thought it was nice to reuse the regex exactly – l0b0 Jun 26 '14 at 13:19

score 2 · Accepted Answer · answered Jun 26 '14 at 12:06

2

An alternative to l0b0's fine answer (shorter, but potentially slightly less efficient):

Assuming a (recent) GNU sed:

find pool -print0 |
  sed -znE 's|.*/mypackage-([[:alnum:].]+-[0-9]{1,2})-x86_64\.pkg\.tar\.xz$|\1|p'|
  tr '\0' '\n'

Note the expensive part of find is the walking down the tree which it will have to do anyway whether you have -regex or not. So here, we're doing the matching and reporting in sed instead.

answered Jun 26 '14 at 12:06

Stéphane Chazelas

544,893

This works great. I changed the sed field separator to a colon so more complex expressions that I have which include pipes also work. Thanks for taking the time to answer. – starfry Jun 26 '14 at 13:21

can regex capture groups be used in GNU find command?

2 Answers2

Linked