Re: Using .ignore file in find

[Bernhard Voelker]

On 2/13/25 09:16, Grigorii Sokolik wrote:
> Dear findutils maintainers team!
>
> In the software development world there is a widespread practice of using
> ignore files. Examples are:
>    - .gitignore
>    - .dockerignore
>    - .clang-format-ignore
>    - ...
> All of them have the same syntax and the same logic. It would be super
> great if it would be possible to use the same syntax file in find for
> listing ignored files or unignore files
>
> Use cases:
>
> # list all ignored files
> find . -type f -name .gitignore | sed
> 's/\(.*\.gitignore\)/--match-patterns \1/' | xargs find . -type f
>
> # list all files not violating the rules
> find . -type f -name .my-fancy-system-ignore | sed
> 's/\(.*\.my-fancy-system-ignore\)/--ignore-patterns \1/' | xargs find . |
> xargs add_file_to_my_fancy_system_context
>
> What should I do to make it possible? I could try to implement it myself,
> but I would like to discuss this feature first.

It's tempting to read in white- or black-lists via a file or file descriptor.
E.g. rsync(1) has --exclude-from=FILE and --include-from=FILE (and FWIW the
--cvs-exclude option).
Also grep(1) has --exclude-from=FILE.

What I'm worried about is being hooked on certain input file formats of
various seemingly similar systems:

While - as far as I know - CVS only allowed patterns for regular files,
.gitignore files allow:
- to distinguish in .gitignore files between files and directories
  ('configure' vs. '.deps/'),
- to bind the pattern to the current directory with a leading slash
  ('/ABOUT-NLS'),
- to specify to ignore files in certain subdirectories
  ('/xargs/test-sigusr'),
- and even to negate with a leading '!'.
Every system or tool will have their own subtleties.

Therefore, I'd rather not add support for specific ignore formats to find(1).
Eventually a --exclude-from=FILE in the style of grep(1) to filter globbing
patterns may make sense.

Re. this one:

> # list all ignored files

This seems to be feasible with:
  $ find | git check-ignore --stdin

One can feed that into another find(1) with the rather new -files0-from option
for further tests; e.g. in my findutils directory, find all non-versioned files
which are older than 100 days ('check-ignore' seems to have problems with the
'gnulib' submodule, hence the prune-ing):

  $ find -name gnulib -prune -o -print0 \
      git check-ignore -z --stdin \
    | find -files0-from - -mtime +100 -ls
  7995557     48 -rw-r--r--   1 berny    users       49140 Nov 11  2022 
./po/.reference/id.po
  7995578     84 -rw-r--r--   1 berny    users       82251 Nov 28  2022 
./po/.reference/ru.po
  ...

Have a nice day,
Berny