Skip to content

PERF: avoid NumPy fallback when casting Arrow int/string to string#63401

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
mroeschke merged 6 commits into from
Dec 19, 2025
+25 −5
Merged

PERF: avoid NumPy fallback when casting Arrow int/string to string #63401

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
65482ac
PERF: avoid NumPy fallback in ArrowStringArray._from_sequence for int…
fangchenli Dec 17, 2025
ae3f464
Merge remote-tracking branch 'upstream' into perf/astype-string-arrow…
fangchenli Dec 18, 2025
d368576
update test
fangchenli Dec 18, 2025
36097d7
Merge remote-tracking branch 'upstream' into perf/astype-string-arrow…
fangchenli Dec 19, 2025
c3284ba
TST: fix test_convert_dtypes for PANDAS_FUTURE_INFER_STRING=0
fangchenli Dec 19, 2025
c19fb29
use arrow for bool array
fangchenli Dec 19, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 20 additions & 0 deletions pandas/core/arrays/string_arrow.py
View file
Open in desktop
Original file line numberDiff line numberDiff line change
Expand Up@@ -210,6 +210,26 @@ def _from_sequence(
result=scalars._data
result=lib.ensure_string_array(result, copy=copy, convert_na_value=False)
pa_arr=pa.array(result, mask=na_values, type=pa.large_string())
elifisinstance(scalars, ArrowExtensionArray):
pa_type=scalars._pa_array.type
# Use PyArrow's native cast for integer, string, and boolean types.
# Float has different representation in PyArrow: 1.0 -> "1" instead
# of "1.0", and uses different scientific notation (1e+10 vs 1e10).
# Boolean needs capitalize (true -> True, false -> False).
if (
pa.types.is_integer(pa_type)
orpa.types.is_large_string(pa_type)
orpa.types.is_string(pa_type)
orpa.types.is_boolean(pa_type)
):
pa_arr=pc.cast(scalars._pa_array, pa.large_string())
ifpa.types.is_boolean(pa_type):
pa_arr=pc.utf8_capitalize(pa_arr)
else:
# Fall back for types where PyArrow's string representation
# differs from Python's str()
result=lib.ensure_string_array(scalars, copy=copy)
pa_arr=pa.array(result, type=pa.large_string(), from_pandas=True)
elifisinstance(scalars, (pa.Array, pa.ChunkedArray)):
pa_arr=pc.cast(scalars, pa.large_string())
else:
Expand Down
10 changes: 5 additions & 5 deletions pandas/tests/copy_view/test_astype.py
View file
Open in desktop
Original file line numberDiff line numberDiff line change
Expand Up@@ -3,8 +3,6 @@
importnumpyasnp
importpytest

frompandas.compatimportHAS_PYARROW

frompandasimport (
DataFrame,
Series,
Expand DownExpand Up@@ -218,10 +216,12 @@ def test_convert_dtypes(using_infer_string):
df_orig=df.copy()
df2=df.convert_dtypes()

ifHAS_PYARROW:
assertnottm.shares_memory(get_array(df2, "a"), get_array(df, "a"))
else:
ifusing_infer_string:
# String column is already Arrow-backed, so memory is shared
asserttm.shares_memory(get_array(df2, "a"), get_array(df, "a"))
else:
# String column converts from object to Arrow, no memory sharing
assertnottm.shares_memory(get_array(df2, "a"), get_array(df, "a"))
asserttm.shares_memory(get_array(df2, "d"), get_array(df, "d"))
asserttm.shares_memory(get_array(df2, "b"), get_array(df, "b"))
asserttm.shares_memory(get_array(df2, "c"), get_array(df, "c"))
Expand Down
Loading