TST: Add xfail tests for Excel header autofilter (openpyxl/xlsxwriter)

[antznette1]

@rhshadrach

• Adds tests specifying expected behavior for header autofilter on both openpyxl and xlsxwriter.
• Tests are marked xfail until the feature is implemented (see ENH: adding autofilter when writing to excel (pandas-dev#61194) #62994).
• No engine or doc changes in this PR; tests-only to lock desired behavior.

[antznette1]

Hi @rhshadrach, I hope you’re doing well.
When you get a chance, could you please review my PR? I’d really appreciate your feedback so we can move it forward.
Thanks a lot!

[rhshadrach]

Thanks for the PR! I think this is looking good. Unfortunately it seems that odfpy does not provide a reasonable way to add an autofilter. If that is the case, I think we should raise when autofilter=True is passed and this engine is used. Can you add this and a test.

Can you also add tests for:

• nonzero startrow / startcol
• A DataFrame with MultiIndex columns (also called a hierarchical index) with both merge_cells=True and merge_cells=False.

[antznette1]

Thanks for the PR! I think this is looking good. Unfortunately it seems that odfpy does not provide a reasonable way to add an autofilter. If that is the case, I think we should raise when autofilter=True is passed and this engine is used. Can you add this and a test.

Can you also add tests for:

• nonzero startrow / startcol
• A DataFrame with MultiIndex columns (also called a hierarchical index) with both merge_cells=True and merge_cells=False.

Okay, Im working on them now

[antznette1]

Hi @rhshadrach .
Could you please review my PR when you are chanced?

[rhshadrach]

@antznette1 - tests are failing, are you able to run them locally?

Also I want to mention #62994 is looking to add the same feature and appears close to being ready.

[antznette1]

@antznette1 - tests are failing, are you able to run them locally?

Also I want to mention #62994 is looking to add the same feature and appears close to being ready.

Yes, I can run locally. I realized my branch currently adds tests that assert behavior for an API not on main yet (and I left some debug prints). I’m cleaning that up now.

[antznette1]

Thanks for the heads-up. I’ve trimmed this PR to tests-only and marked the new autofilter tests xfail until #62994 lands. No engine/docs changes remain here. Once #62994 merges, I’ll flip these to pass and expand coverage as needed.

[rhshadrach]

@antznette1 - each merge of main kicks off a full CI run. This is not necessary to do multiple times a day; once a week is sufficient.

[antznette1]

@antznette1 - each merge of main kicks off a full CI run. This is not necessary to do multiple times a day; once a week is sufficient.

Thanks for the heads-up. I’ll avoid frequent merges of main and will update at most weekly or when necessary (conflicts or upstream-related CI failures). I’ll also prefer rebasing to keep history clean. This PR remains tests-only and xfail until #62994 lands.

[antznette1]

@rhshadrach
Good day, Hope you are well?

I want to ask about the #62994 progress so I can convert from xfail

[rhshadrach]

@antznette1 - #62994 has been merged; I'm not sure if the tests here expand our coverage or are duplicative of the tests there. Can you take a look.

[antznette1]

@antznette1 - #62994 has been merged; I'm not sure if the tests here expand our coverage or are duplicative of the tests there. Can you take a look.

I removed the duplicates so that the tests are in engine-specific files, which helps with:

• Engine-specific debugging
• Clearer organization
• More detailed assertions per engine

These tests expand coverage by adding edge cases not covered in PR #62994.

[rhshadrach]

I removed the duplicates so that the tests are in engine-specific files, which helps with:
...

I think we should prefer parametrized tests. This then will automatically apply to engines that are to be added in the future, and makes removing engines easier as well. In particular, I do not think we should have tests in test_writers.py and tests in each of the enigne-specific test files that have the same coverage.

[antznette1]

I removed the duplicates so that the tests are in engine-specific files, which helps with:
...

I think we should prefer parametrized tests. This then will automatically apply to engines that are to be added in the future, and makes removing engines easier as well. In particular, I do not think we should have tests in test_writers.py and tests in each of the enigne-specific test files that have the same coverage.

Thanks for the suggestion.

I’ve rebased this branch on main so it now includes the parametrized autofilter tests in TestExcelWriter in test_writers.py. To avoid duplicated coverage:

• I kept the engine-specific files (test_openpyxl.py, test_xlsxwriter.py) free of any autofilter tests.
• I moved the additional autofilter edge cases I had (empty DataFrame, single row, single column, header=False) into TestExcelWriter as parametrized tests over engine.
  This way, all shared autofilter behavior is tested once, in the parametrized suite, and any future engines will automatically get this coverage.