fix: Drop timezone in from_arrow (#3392)

* Drop timezone in from_arrow

* Add doc and test

src/awkward/_connect/pyarrow/conversions.py

@@ -471,6 +471,8 @@ def popbuffers(paarray, awkwardarrow_type, storage_type, buffers, generate_bitma
         if to64:
             data = numpy.astype(numpy.frombuffer(data, dtype=np.int32), dtype=np.int64)
         if dt is None:
+            if getattr(storage_type, "tz", None) is not None:
+                storage_type = pyarrow.lib.timestamp(storage_type.unit)
             dt = storage_type.to_pandas_dtype()
 
         out = ak.contents.NumpyArray(
@@ -670,6 +672,8 @@ def form_popbuffers(awkwardarrow_type, storage_type):
     elif isinstance(storage_type, pyarrow.lib.DataType):
         _, dt = _pyarrow_to_numpy_dtype.get(str(storage_type), (False, None))
         if dt is None:
+            if getattr(storage_type, "tz", None) is not None:
+                storage_type = pyarrow.lib.timestamp(storage_type.unit)
             dt = np.dtype(storage_type.to_pandas_dtype())
 
         out = ak.forms.NumpyForm(

src/awkward/operations/ak_from_arrow.py

@@ -40,6 +40,10 @@ def from_arrow(
     low-level #ak.forms.Form), even through Parquet, making Parquet a good way to save
     Awkward Arrays for later use.
 
+    Because awkward uses numpy's dtype system, timestamp types do not have timezones.
+    If encountering timestamp types with timezones in the input arrow data, they
+    will be silently dropped.
+
     See also #ak.to_arrow, #ak.to_arrow_table, #ak.from_parquet, #ak.from_arrow_schema.
     """
     return _impl(array, generate_bitmasks, highlevel, behavior, attrs)

src/awkward/operations/ak_from_arrow_schema.py

@@ -18,6 +18,10 @@ def from_arrow_schema(schema):
 
     Converts an Apache Arrow schema into an Awkward Form.
 
+    Because awkward uses numpy's dtype system, timestamp types do not have timezones.
+    If encountering timestamp types with timezones in the input arrow data, they
+    will be silently dropped.
+
     See also #ak.to_arrow, #ak.to_arrow_table, #ak.from_arrow, #ak.to_parquet, #ak.from_parquet.
     """
     return _impl(schema)

tests/test_3392_from_arrow_tz.py

@@ -0,0 +1,25 @@
+from __future__ import annotations
+
+import pytest
+
+import awkward as ak
+
+pa = pytest.importorskip("pyarrow")
+
+
+def test_tz_is_dropped():
+    data = pa.Table.from_arrays(
+        [
+            pa.array(
+                [
+                    1,
+                    2,
+                    3,
+                ],
+                type=pa.timestamp("ns", tz="UTC"),
+            )
+        ],
+        names=["a"],
+    )
+    ak.from_arrow(data)
+    ak.from_arrow_schema(data.schema)