You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
@jgallowa07, PR #103 makes a ton of very minor changes, mostly to docs, but some restructuring of GH workflows to simplify and also increase what is tested. A major problem I'm having is that almost none of the doctest code throughout the package (the many examples in docstrings) runs after switching on testing of the docstring examples (running pytest --doctest-modules). My impression is that, like the notebooks, it turns out this code is not tested at all. To avoid breaking this code, each time a developer changes the API, they would have to stare at all these docstring code examples in every file to reason about what needs changing. I don't think that's how we want to work, so I recommend fixing the doctests so that they actually run, and checking them in CI. While one of the Jupyter notebooks has a long runtime that arguably warrants not checking it in tests, we should design docstring examples that are actually runnable and that we can check.
I made some progress toward the doctests, but it's going to need another chunk of work, and I won't be available for that in the near term. I think what I'll do is retreat from doctest in #103, leaving the GH action line commented out, and we can work the PR in without that, and later (hopefully soon) someone can fix the docstrings.
The text was updated successfully, but these errors were encountered:
@jgallowa07, PR #103 makes a ton of very minor changes, mostly to docs, but some restructuring of GH workflows to simplify and also increase what is tested. A major problem I'm having is that almost none of the doctest code throughout the package (the many examples in docstrings) runs after switching on testing of the docstring examples (running
pytest --doctest-modules
). My impression is that, like the notebooks, it turns out this code is not tested at all. To avoid breaking this code, each time a developer changes the API, they would have to stare at all these docstring code examples in every file to reason about what needs changing. I don't think that's how we want to work, so I recommend fixing the doctests so that they actually run, and checking them in CI. While one of the Jupyter notebooks has a long runtime that arguably warrants not checking it in tests, we should design docstring examples that are actually runnable and that we can check.I made some progress toward the doctests, but it's going to need another chunk of work, and I won't be available for that in the near term. I think what I'll do is retreat from doctest in #103, leaving the GH action line commented out, and we can work the PR in without that, and later (hopefully soon) someone can fix the docstrings.
The text was updated successfully, but these errors were encountered: