Backport PR #38539: BUG: fix array conversion from Arrow for slided array (#38611)

Co-authored-by: Joris Van den Bossche <[email protected]>

Author: meeseeksmachine

doc/source/whatsnew/v1.2.0.rst

@@ -747,6 +747,7 @@ I/O
 - :meth:`DataFrame.to_html` was ignoring ``formatters`` argument for ``ExtensionDtype`` columns (:issue:`36525`)
 - Bumped minimum xarray version to 0.12.3 to avoid reference to the removed ``Panel`` class (:issue:`27101`)
 - :meth:`DataFrame.to_csv` was re-opening file-like handles that also implement ``os.PathLike`` (:issue:`38125`)
+- Bug in the conversion of a sliced ``pyarrow.Table`` with missing values to a DataFrame (:issue:`38525`)
 
 Period
 ^^^^^^

pandas/core/arrays/_arrow_utils.py

@@ -30,7 +30,7 @@ def pyarrow_array_to_numpy_and_mask(arr, dtype):
     bitmask = buflist[0]
     if bitmask is not None:
         mask = pyarrow.BooleanArray.from_buffers(
-            pyarrow.bool_(), len(arr), [None, bitmask]
+            pyarrow.bool_(), len(arr), [None, bitmask], offset=arr.offset
         )
         mask = np.asarray(mask)
     else:

pandas/tests/arrays/masked/test_arrow_compat.py

@@ -52,3 +52,15 @@ def test_arrow_from_arrow_uint():
     expected = pd.array([1, 2, 3, 4, None], dtype="UInt32")
 
     tm.assert_extension_array_equal(result, expected)
+
+
+@td.skip_if_no("pyarrow", min_version="0.16.0")
+def test_arrow_sliced():
+    # https://github.com/pandas-dev/pandas/issues/38525
+    import pyarrow as pa
+
+    df = pd.DataFrame({"a": pd.array([0, None, 2, 3, None], dtype="Int64")})
+    table = pa.table(df)
+    result = table.slice(2, None).to_pandas()
+    expected = df.iloc[2:].reset_index(drop=True)
+    tm.assert_frame_equal(result, expected)