API/BUG: freq retention in value_counts

[jbrockmendel]

dti = pd.date_range('2016-01-01', periods=5)

dti.value_counts().index.freq    # <-- None
dti.factorize()[1].freq   # <-- None

mi = pd.MultiIndex.from_arrays([dti, dti])

mi.levels[0].freq   # <-- None

There is a comment in tests.indexes.datetimes.test_datetime test_factorize suggesting that freq should be preserved by factorize, but that is not checked and would fail if it were

        # freq must be preserved
        idx3 = date_range("2000-01", periods=4, freq="M", tz="Asia/Tokyo")
        exp_arr = np.array([0, 1, 2, 3], dtype=np.intp)
        arr, idx = idx3.factorize()
        tm.assert_numpy_array_equal(arr, exp_arr)
        tm.assert_index_equal(idx, idx3)

So the question: do we want to try to preserve freq in factorize?

xref #33677 for the MultiIndex case

Update One more: Categorical:

dti = pd.date_range('2016-01-01', periods=5)
cat = pd.Categorical(dti)
cat.categories.freq   # <-- None

[mroeschke]

Looks like all but the value_counts example is preserving freq now. Could use tests for those

In [8]: dti = pd.date_range('2016-01-01', periods=5)
   ...:
   ...: dti.value_counts().index.freq

In [9]: dti.factorize()[1].freq
Out[9]: <Day>

In [10]: mi = pd.MultiIndex.from_arrays([dti, dti])

In [11]: mi.levels[0].freq
Out[11]: <Day>

In [12]: dti = pd.date_range('2016-01-01', periods=5)
    ...: cat = pd.Categorical(dti)
    ...: cat.categories.freq
Out[12]: <Day>

[das-anubhav]

Hi!

I would like to solve this issue.

[bgollop]

take