DEPR: Categorical.ravel, get_dtype_counts, dtype_str, to_dense

[jbrockmendel]

Starting in on the 0.25.0 section in #6581.

[jreback]

sure

[jbrockmendel]

@datapythonista im having trouble telling what the checks failure is about. one thing suggests it is in asvs, but another suggests annotations

[datapythonista]

I just see this error in the benckmarks: https://github.com/pandas-dev/pandas/pull/29900/checks?check_run_id=323732409#step:13:4913

I guess it means that one of the 3 runs was too slow, but you'll know better.

It took me forever to find the error. I think it shouldn't be difficult to highlight failures in the benchmarks, so it's immediate to know what failed. I'll see if I can open a PR for that.

[jbrockmendel]

Thanks for taking a look. I agree itd be nice to have an easier way to find when there is a broken asv. I'm hopeful that #29906 will fix this particular problem

[WillAyd]

Question / comment but other changes lgtm

[WillAyd]

Not super familiar with this change - was there a consensus on this being the return value or does it need discussion?

[jbrockmendel]

Yes, in #27199

[jreback]

this is fine, though original we were returning self (and not a copy)?

[jbrockmendel]

definitely not a copy, im not sure if it matters if we return self vs self.view

[jreback]

this is a shallow copy (a new object)

   def view(self, dtype=None):
        if dtype is not None:
            raise NotImplementedError(dtype)
        return self._constructor(values=self._codes, dtype=self.dtype, fastpath=True)

but suppose its ok, can you validate that it is a new object in the tests?

[jbrockmendel]

can you validate that it is a new object in the tests?

After adding this to the test_ravel in the extensions tests, it looks like all the other EAs we test return self (because that is was the base class does so ill change this to just use the base class implementation and match the others.

[jreback]

needs a rebase and question.

[WillAyd]

Thanks @jbrockmendel