PERF: DataFrame.transpose for pyarrow-backed

[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.

Similar to #52836, but for pyarrow-backed frames.

import pandas as pd
import pyarrow as pa
import numpy as np

values = np.random.randn(1000, 1000)
df = pd.DataFrame(values, dtype=pd.ArrowDtype(pa.float64()))

%timeit df.T

# 346 ms ± 67.5 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)    -> main
# 42.5 ms ± 2.12 ms per loop (mean ± std. dev. of 7 runs, 10 loops each) -> PR

Renamed the masked implementation from homogenous -> homogeneous. Both may be valid(?), but homogeneous is used elsewhere.

[jbrockmendel]

cant do this unconditionally?

[lukemanley]

Do you mean use _iter_column_arrays in both cases? The masked array implementation just above this uses the condition as well, presumably for performance:

import pandas as pd

pd.options.mode.data_manager = "array"

df = pd.DataFrame(columns=range(100000), dtype="int64[pyarrow]")

%timeit list(df._iter_column_arrays())
# 57.9 ms ± 2.7 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)

%timeit df._mgr.arrays
# 110 ns ± 8.79 ns per loop (mean ± std. dev. of 7 runs, 10,000,000 loops each)

[jbrockmendel]

Fair enough, thanks for taking a look

[lukemanley]

on second thought, I moved the condition into DataFrame._iter_column_arrays to avoid repeating it elsewhere

[mroeschke]

Thanks @lukemanley