PERF: DataFrame.first_valid_index/last_valid_index for EA dtypes

[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/v2.1.0.rst file if fixing a bug or adding a new feature.

ASV added:

       before                after    ratio
-      21.6±0.1ms          484±9μs     0.02  frame_methods.FindValidIndex.time_last_valid_index('float64[pyarrow]')
-      21.6±0.3ms          427±3μs     0.02  frame_methods.FindValidIndex.time_first_valid_index('float64[pyarrow]')
-      17.0±0.5ms          167±4μs     0.01  frame_methods.FindValidIndex.time_last_valid_index('Float64')
-      16.8±0.2ms          132±2μs     0.01  frame_methods.FindValidIndex.time_first_valid_index('Float64')

[jbrockmendel]

wow thats a huge difference. is it all in the len(values) or something?

[lukemanley]

wow thats a huge difference. is it all in the len(values) or something?

self._values on a multi-column EA-backed frame converts to an object dtype ndarray. This avoids going through the object dtype conversion.

[phofl]

really nice!

This is basically the same as with MultiIndex._values?

[lukemanley]

This is basically the same as with MultiIndex._values?

similar in the sense they're both expensive conversions through object dtype. MultiIndex._values has the additional overhead of wrapping in tuples.

[phofl]

Yep, thx, that's what I was thinking

[phofl]

Thx @lukemanley