✔️ Linting Passed

All linting checks passed. Your pull request is in excellent shape! ☀️

_{Generated for commit: dc1dbbe. Link to the linter CI: here}

Thanks for helping clean this up! Looks good to me.

Let's see what the testing robots have to say.

@adrinjalali could you explain in a bit more detail what you meant with "require_fit=False can be tricky if there is validation"? HashingVectorizer does some work in its fit.

So because those estimators do not accept normal NumPy array (they are vectorizer) they are not covered by the common test (tests run in test_common.py and defined in estimator_checks.py).

We therefore need non-regression tests for the FeatureHasher and HashingVectorizer in the file test_feature_hasher.py and test_text.py. The test should check that we can call transform method without calling the fit method and that it does not raise any error.

@adrinjalali could you explain in a bit more detail what you meant with "require_fit=False can be tricky if there is validation"? HashingVectorizer does some work in its fit.

As in, do we expect the transformer to do any kind of feature count / feature name consistency between fit and transform? An estimator might technically not require fit to run transform, but we might expect the transform to do those basic validations. I just meant to make sure that expectation is not there when we set the tags.

As far that I recall, we defined stateless as: if you call fit you get some validation for free but you can call transform and go on but without those validation.

if you call fit you get some validation for free but you can call transform and go on but without those validation.

Time to update the glossary entry?