PERF: ArrowExtensionArray.from_sequence_of_strings(dtype=pa.bool()) by lukemanley · Pull Request #51730 · pandas-dev/pandas (original) (raw)

ArrowExtensionArray._from_sequence_of_strings currently depends on BooleanArray._from_sequence_of_strings to parse strings for boolean[pyarrow]. This removes the dependency on BooleanArray and uses pyarrow casting to accomplish the same thing. I've adding some logic around pyarrow's casting to match existing behavior of BooleanArray (assuming we want to maintain existing behavior?)

import pyarrow as pa
import numpy as np
from pandas.core.arrays.arrow.array import ArrowExtensionArray

N = 1_000_000
strings_np = np.random.choice(["true", "false", None], N, replace=True)
strings_pa = pa.array(strings_np, type=pa.string())

%timeit ArrowExtensionArray._from_sequence_of_strings(strings_np, dtype=pa.bool_())

# 200 ms ± 4.72 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)    -> main
# 54.5 ms ± 842 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)  -> PR

%timeit ArrowExtensionArray._from_sequence_of_strings(strings_pa, dtype=pa.bool_())

# 298 ms ± 3.45 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)    -> main
# 30.8 ms ± 1.1 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)  -> PR