Only run pre-commit hooks on modified files

[johnnv1]

Motivation

Improve integration with pre-commit

Expected behavior

Can use the nbdev_clear without having a nbdev config at settings.ini or to implicit pass args: ['-fname', 'mydir/'] at pre-commit config.

Explanation

As I explain here, seems to be a workaround need to add the path if you don't have a nbdev config. Mainly, because this will cause the command to be executed for all files at that path, which is unnecessary.

If the command accepts n arguments (multiple filenames), we can use the pre-commit "normally", by letting it command in which files the hook should be executed -- only those that have been changed in the commit.

Consequently, this will improve the hook's execution speed, since it will not be checking for unnecessary files.

---

With this enhancement, we can update the .pre-commit-hooks.yaml to:

--- a/.pre-commit-hooks.yaml
+++ b/.pre-commit-hooks.yaml
@@ -5,7 +5,8 @@
   always_run: true
-  pass_filenames: false
+  pass_filenames: true
+  types: [jupyter]

[seeM]

Good idea! nbdev_clean currently doesn't take any positional arguments so we could make the change so that it optionally takes a list of files

[dijonkitchen]

I think similarly, it'd be good to have other pre-commit behavior like only including certain filetypes or excluding files. Currently, nbdev runs on everything and may be using more time than it needs.

Example .pre-commit-config.yaml:

repos:
  - repo: https://github.com/fastai/nbdev
    rev: 2.3.11
    hooks:
      - id: nbdev_clean
        files: \.ipynb$ # This doesn't work for `nbdev` but does for other pre-commit hook plugins. 

If it works properly, when the pre-commit hook runs, it should say "Skipped" if no .ipynb files were changed.

[jtbr]

This should be an easy change to make to nbdev_clean.

• Eliminate the --fname argument (which currently only accepts a single file and can only be used once). To my knowledge this is not used within nbdev and is non-standard/clunky.
• Allow any number of file paths to be passed as positional arguments at the end. If any are specified, clean works on those and only those files. This allows pre-commit and other posix/unix tools to work as expected. Or the user can use the shell to specify wildcards.
• If none are provided, then nbdev works as now, running on all notebooks in the nbs_path from system.ini (if it exists)
• Return exit code 0 if no changes were made to any files. Return a positive integer if any changes were made.
• Write a simple message for each file changed: "cleaned notebook1.ipynb\n"

(and the last two are not strictly necessary, but do improve usability)