BUG/PERF: Series.replace with dtype="category"

[lukemanley]

• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed.
• Added type annotations to new arguments/methods/functions.
• Added an entry in the latest doc/source/whatsnew/v2.0.0.rst file if fixing a bug or adding a new feature.

Refactor of Categorical._replace to fix a few bugs with Series(..., dtype="category").replace and improve performance.

BUG 1: overlap between to_replace and value:

Series([1, 2, 3], dtype="category").replace({1:2, 2:3, 3:4})


# main:

0    4
1    4
2    4
dtype: category
Categories (1, int64): [4]


# PR:

0    2
1    3
2    4
dtype: category
Categories (3, int64): [2, 3, 4]

BUG 2: losing nullable dtypes of underlying categories:

Series(["a", "b"], dtype="string").astype("category").replace("b", "c")


# main:

0    a
1    c
dtype: category
Categories (2, object): ['a', 'c']


# PR:

0    a
1    c
dtype: category
Categories (2, string): [a, c]

Perf improvements:

import pandas as pd
import numpy as np

arr = np.repeat(np.arange(1000), 1000)
ser = pd.Series(arr, dtype="category")

%timeit ser.replace(np.arange(200), 5)

681 ms ± 9.65 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)   <- main
11 ms ± 690 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)  <- PR

"""

[github-actions]

This pull request is stale because it has been open for thirty days with no activity. Please update and respond to this comment if you're still interested in working on this.

[mroeschke]

LGTM merge when ready @jbrockmendel

[jbrockmendel]

hopefully ill get time to take a good look this week but dont wait for me if this is a blocker.

[mroeschke]

Thanks @lukemanley

[phofl]

This broke the ci, opened #50848 to revert if there is no quick-win here