googleapis · tswast · Mar 28, 2024 · Mar 28, 2024 · Mar 28, 2024 · Mar 28, 2024
@@ -49,8 +49,7 @@
  db_dtypes_import_exception = exc
  date_dtype_name = time_dtype_name = "" # Use '' rather than None because pytype
 
-pyarrow = _versions_helpers.PYARROW_VERSIONS.try_import(raise_if_error=True)
-from pyarrow import ArrowTypeError # type: ignore # noqa: E402
+pyarrow = _versions_helpers.PYARROW_VERSIONS.try_import()
 
 _BIGNUMERIC_SUPPORT = False
 if pyarrow is not None: # pragma: NO COVER
@@ -309,10 +308,10 @@ def bq_to_arrow_array(series, bq_field):
  if field_type_upper in schema._STRUCT_TYPES:
  return pyarrow.StructArray.from_pandas(series, type=arrow_type)
  return pyarrow.Array.from_pandas(series, type=arrow_type)
- except ArrowTypeError: # pragma: NO COVER
+ except pyarrow.ArrowTypeError: # pragma: NO COVER
  msg = f"""Error converting Pandas column with name: "{series.name}" and datatype: "{series.dtype}" to an appropriate pyarrow datatype: Array, ListArray, or StructArray"""
  _LOGGER.error(msg)
- raise ArrowTypeError(msg)
+ raise pyarrow.ArrowTypeError(msg)
 
 
 def get_column_or_index(dataframe, name):

@@ -49,7 +49,7 @@ def pyarrow_timestamp():
 _BQ_TO_ARROW_SCALARS = {}
 _ARROW_SCALAR_IDS_TO_BQ = {}
 
-if pyarrow: # pragma: NO COVER
+if pyarrow:
  # This dictionary is duplicated in bigquery_storage/test/unite/test_reader.py
  # When modifying it be sure to update it there as well.
  # Note(todo!!): type "BIGNUMERIC"'s matching pyarrow type is added in _pandas_helpers.py

@@ -86,7 +86,7 @@ def default(session, install_extras=True):
  install_target = ".[all]"
  else:
  install_target = "."
- session.install("-e", install_target)
+ session.install("-e", install_target, "-c", constraints_path)
  session.run("python", "-m", "pip", "freeze")
 
  # Run py.test against the unit tests.
@@ -115,14 +115,15 @@ def unit(session):
 def unit_noextras(session):
  """Run the unit test suite."""
 
- # Install optional dependencies that are out-of-date.
+ # Install optional dependencies that are out-of-date to see that
+ # we fail gracefully.
  # https://github.com/googleapis/python-bigquery/issues/933
- # There is no pyarrow 1.0.0 package for Python 3.9.
-
+ #
+ # We only install this extra package on one of the two Python versions
+ # so that it continues to be an optional dependency.
+ # https://github.com/googleapis/python-bigquery/issues/1877
  if session.python == UNIT_TEST_PYTHON_VERSIONS[0]:
- session.install("pyarrow>=3.0.0")
- elif session.python == UNIT_TEST_PYTHON_VERSIONS[-1]:
- session.install("pyarrow")
+ session.install("pyarrow==1.0.0")
 
  default(session, install_extras=False)
 

@@ -2,4 +2,3 @@ google-cloud-testutils==1.4.0
 pytest===7.4.4; python_version == '3.7'
 pytest==8.1.1; python_version >= '3.8'
 mock==5.1.0
-pyarrow>=3.0.0
@@ -1,5 +1,5 @@
+# samples/snippets should be runnable with no "extras"
 google-cloud-testutils==1.4.0
 pytest===7.4.4; python_version == '3.7'
 pytest==8.1.1; python_version >= '3.8'
 mock==5.1.0
-pyarrow>=3.0.0
@@ -1 +1,2 @@
-google-cloud-bigquery==3.19.0
+# samples/snippets should be runnable with no "extras"
+google-cloud-bigquery==3.19.0
@@ -1 +0,0 @@
-pyarrow>=3.0.0

@@ -1 +0,0 @@
-pyarrow>=3.0.0

@@ -27,9 +27,9 @@ packaging==20.0.0
 pandas==1.1.0
 proto-plus==1.22.0
 protobuf==3.19.5
-pyarrow>=3.0.0
+pyarrow==3.0.0
 python-dateutil==2.7.3
 requests==2.21.0
 Shapely==1.8.4
 six==1.13.0
-tqdm==4.7.4
+tqdm==4.7.4
@@ -53,8 +53,7 @@
 if pyarrow:
  import pyarrow.parquet
  import pyarrow.types
- from pyarrow import ArrowTypeError # type: ignore # noqa: E402
-else: # pragma: NO COVER
+else:
  # Mock out pyarrow when missing, because methods from pyarrow.types are
  # used in test parameterization.
  pyarrow = mock.Mock()
@@ -572,9 +571,9 @@ def test_bq_to_arrow_array_w_conversion_fail(module_under_test): # pragma: NO C
  series = pandas.Series(rows, name="test_col", dtype="object")
  bq_field = schema.SchemaField("field_name", "STRING", mode="REPEATED")
  exc_msg = f"""Error converting Pandas column with name: "{series.name}" and datatype: "{series.dtype}" to an appropriate pyarrow datatype: Array, ListArray, or StructArray"""
- with pytest.raises(ArrowTypeError, match=exc_msg):
+ with pytest.raises(pyarrow.ArrowTypeError, match=exc_msg):
  module_under_test.bq_to_arrow_array(series, bq_field)
- raise ArrowTypeError(exc_msg)
+ raise pyarrow.ArrowTypeError(exc_msg)
 
 
 @pytest.mark.parametrize("bq_type", ["RECORD", "record", "STRUCT", "struct"])

@@ -3408,6 +3408,7 @@ def test_to_dataframe_datetime_out_of_pyarrow_bounds(self):
 
  def test_to_dataframe_progress_bar(self):
  pytest.importorskip("pandas")
+ pytest.importorskip("pyarrow")
  pytest.importorskip("tqdm")
 
  from google.cloud.bigquery.schema import SchemaField
@@ -3447,6 +3448,7 @@ def test_to_dataframe_progress_bar(self):
  @mock.patch("google.cloud.bigquery._tqdm_helpers.tqdm", new=None)
  def test_to_dataframe_no_tqdm_no_progress_bar(self):
  pytest.importorskip("pandas")
+ pytest.importorskip("pyarrow")
  from google.cloud.bigquery.schema import SchemaField
 
  schema = [
@@ -3711,7 +3713,7 @@ def test_to_dataframe_w_dtypes_mapper(self):
  if hasattr(pandas, "Float64Dtype"):
  self.assertEqual(list(df.miles), [1.77, 6.66, 2.0])
  self.assertEqual(df.miles.dtype.name, "Float64")
- else: # pragma: NO COVER
+ else:
  self.assertEqual(list(df.miles), ["1.77", "6.66", "2.0"])
  self.assertEqual(df.miles.dtype.name, "string")