tests: add zsh completion test for Path arguments

[konuktor1-jpg]

Adds a small regression test for zsh path completion in tests/test_completion/test_completion_path.py.

The file already had a Bash completion test, but no zsh-specific assertion. This test checks that Path completion returns the _files signal for zsh.

Tested locally with:

• uv run pytest tests/test_completion/test_completion_path.py -q

[konuktor1-jpg]

Hi! This PR is currently blocked because it has no required label, and the workflows are awaiting maintainer approval. I think bug would fit this change since it adds a regression test for existing zsh path completion behavior. Could someone please add the appropriate label and approve the workflows?

[ADiTyaRaj8969]

Hi @konuktor1-jpg, took a look at this on Windows with the project venv and ran the full file:

tests/test_completion/test_completion_path.py
  test_script              PASSED
  test_completion_path_bash PASSED
  test_completion_path_zsh  PASSED   (new)
3 passed in 2.75s

Why the "_files" in result.stdout assertion is the right check

For a Path argument with no autocompletion callback there are no string completions to feed back, so ZshComplete.format_completion (in typer/_completion_classes.py) returns the literal string "_files" — that's the signal that tells the zsh completion script to fall back to the shell's own file-completion. Asserting on "_files" therefore verifies the path-completion code path actually reaches the fallback, which the existing bash test (which only asserts returncode == 0) does not. Useful regression coverage.

Scope

Keeping this PR to zsh-only feels right — complete_bash, complete_fish, and complete_powershell/complete_pwsh all use different fallback mechanisms (or no fallback at all, depending on the shell), so a Path-completion test for each would warrant separate, shell-specific assertions and isn't a natural fit for this PR.

Bumping for visibility — looks ready as-is.

[svlandeg]

Thank you for your contribution @konuktor1-jpg!

The original test arbitrarily picked bash just to be able to test the code path for a Path argument. If we're going to extend this test suite, we should do it for all shells and we should come up with a way to properly test the behaviour instead of just ensuring coverage. See also my comment here.

So TLDR, I think we shouldn't merge this as-is. Let me know if you'd like to work on extending this PR in line with my comment 👆

[svlandeg]

@ADiTyaRaj8969: your review reads like it's generated by AI - please refrain from leaving AI-generated reviews, especially without disclaimer.

[ADiTyaRaj8969]

Sorry, ma'am. I will avoid AI styled writing in future reviews and make them more personal.