Allow nbdev_clean to accept multiple filenames or globs (#1480)

[jbwhit]

Overview

This PR addresses #1480 by allowing nbdev_clean to accept multiple filename arguments, rather than just a single file or glob. This makes it easier to clean multiple notebooks in one command.

What Changed

• Multiple Files Support: Updated nbdev_clean so it can handle either a single path/glob string or a list of paths/globs.
• New Tests: Added tests confirming multiple files are handled correctly (they fail on the old code, pass on the new).
• Doc Update: Revised the docstring to reflect this new usage.

Known Similar Issue in nbdev_test

nbdev_test (and possibly other CLI commands) still only accept single-file arguments. This PR focuses on nbdev_clean specifically to keep it scoped. Let me know if you want a follow-up PR for nbdev_test.

Skipped a Failing Test in 10_processors.ipynb

• One cell running _run_procs([add_show_docs, exec_show_docs]) currently fails for unrelated reasons.
• I added #|hide and #|eval: false to skip that cell so nbdev_prepare can complete.
• We can address that failing test in a separate issue or PR.

Please let me know if you have any questions or suggestions—this is my first PR for this project!

[hamelsmu]

currently fails for unrelated reasons.

Please tell us more. What are the reasons?

[jbwhit]

Thank you for asking about this issue. I was able to resolve the failing test by updating my pandas installation with pip install --upgrade pandas.

The test in 10_processors.ipynb with _run_procs([add_show_docs, exec_show_docs]) is now passing, and I've removed the #|hide and #|eval: false markers. It seems it was just an environment-specific issue on my end rather than a problem with the codebase.

I'll update the PR shortly to remove those markers and ensure all tests are running properly.

[jbwhit]

Ok I've updated the PR -- all tests now pass.

[jph00]

Thank you for this PR. IIUC, _run("git config user.email 'nbdev@fast.ai'") will actually modify the git config of a user running the tests. Tests should not modify a user's environment. Also BTW you might find fastcore's listify helpful.

[jbwhit]

I can make a different test without using _run("git config user.email 'nbdev@fast.ai'") -- the reason I included it was that was trying to base my new test off existing tests, and this code already exists in the same notebook (nbs/api/11_clean.ipynb). If this code shouldn't be there at all I can make a new PR that attempts to write the test correctly?

I will pare down my test to the minimum and I will look to use listify as well. Thanks for your attention and for creating this project in the first place!

[jph00]

Oh wow I'd completely missed that! Thanks for noticing. Yes we should get rid of it :D

…

Message ID: ***@***.***>

[jbwhit]

I've updated the PR to enhance nbdev_clean to properly handle multiple notebook paths.

1. The issue(nbdev_clean documentation partly wrong #1480): When passing a list of paths to nbdev_clean, it fails because globtastic expects a string or Path object, not a list. This PR allows nbdev_clean to accept multiple filename arguments, rather than just a single file or glob.

2. The fix: Use fastcore.listify to properly handle both single paths and lists of paths, ensuring that globtastic receives the correct type. I changed the code back to something much closer to the original implementation that my initial commits.

3. Added test: I've added a comprehensive validation test that:

   • Demonstrates that the current implementation fails when given a list of paths
   • Shows that the fixed implementation using listify correctly processes all paths
   • Verifies that proper cleaning occurs on all specified notebooks
   • Does not require the usage of git within the test at all

The test shows that without listify, the function fails when given a list of paths, but works correctly with the fix. This change maintains backward compatibility while adding the ability to pass multiple paths.

I've kept this PR focused solely on fixing the nbdev_clean functionality and will address the git test refactoring in a separate PR as suggested.

[jph00]

Many thanks!

[jph00]

Looks like there's a CI failure.

[jph00]

Closing due to age.