Master

   ...: %timeit data[list_]
1.44 ms ± 5.3 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

This PR

   ...: %timeit data[list_]
740 µs ± 5.92 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

Will post the pre-indexing with NA timing in a bit.

Edit:

7b35099 (1 commit prior to NA indexing)

728 µs ± 7.57 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

so we're back to where we were.

cool, can you add an asv

can you add an asv

I think Joris caught this because the existing asv CategoricalSlicing.time_getitem_list https://pandas.pydata.org/speed/pandas/#categoricals.CategoricalSlicing.time_getitem_list?commits=6efc2379-b9de33e3

LGTM

So I reported this for Categorical (indeed because of the existing benchmark), but a similar change was made for other dtypes in #30308. Also the boolean and integer (and numpy) arrays have the same issue.