test: improve to_gbq logic unit test coverage (#449)

* 🦉 Updates from OwlBot

See https://github.com/googleapis/repo-automation-bots/blob/main/packages/owl-bot/README.md

Co-authored-by: Owl Bot <gcf-owl-bot[bot]@users.noreply.github.com>

Author: tswast

.coveragerc

@@ -22,7 +22,7 @@ omit =
   google/cloud/__init__.py
 
 [report]
-fail_under = 89
+fail_under = 94
 show_missing = True
 exclude_lines =
     # Re-enable the standard pragma

noxfile.py

@@ -259,7 +259,7 @@ def cover(session):
     test runs (not system test runs), and then erases coverage data.
     """
     session.install("coverage", "pytest-cov")
-    session.run("coverage", "report", "--show-missing", "--fail-under=89")
+    session.run("coverage", "report", "--show-missing", "--fail-under=94")
 
     session.run("coverage", "erase")
 

owlbot.py

@@ -33,7 +33,7 @@
 templated_files = common.py_library(
     unit_test_python_versions=["3.7", "3.8", "3.9", "3.10"],
     system_test_python_versions=["3.7", "3.8", "3.9", "3.10"],
-    cov_level=89,
+    cov_level=94,
     unit_test_extras=extras,
     system_test_extras=extras,
     intersphinx_dependencies={

pandas_gbq/gbq.py

@@ -121,7 +121,20 @@ class InvalidSchema(ValueError):
     table in BigQuery.
     """
 
-    pass
+    def __init__(
+        self, message: str, local_schema: Dict[str, Any], remote_schema: Dict[str, Any]
+    ):
+        super().__init__(message)
+        self._local_schema = local_schema
+        self._remote_schema = remote_schema
+
+    @property
+    def local_schema(self) -> Dict[str, Any]:
+        return self._local_schema
+
+    @property
+    def remote_schema(self) -> Dict[str, Any]:
+        return self._remote_schema
 
 
 class NotFoundException(ValueError):
@@ -354,19 +367,12 @@ def sizeof_fmt(num, suffix="B"):
         return fmt % (num, "Y", suffix)
 
     def get_client(self):
+        import google.api_core.client_info
         import pandas
 
-        try:
-            # This module was added in google-api-core 1.11.0.
-            # We don't have a hard requirement on that version, so only
-            # populate the client_info if available.
-            import google.api_core.client_info
-
-            client_info = google.api_core.client_info.ClientInfo(
-                user_agent="pandas-{}".format(pandas.__version__)
-            )
-        except ImportError:
-            client_info = None
+        client_info = google.api_core.client_info.ClientInfo(
+            user_agent="pandas-{}".format(pandas.__version__)
+        )
 
         # In addition to new enough version of google-api-core, a new enough
         # version of google-cloud-bigquery is required to populate the
@@ -1057,7 +1063,7 @@ def to_gbq(
                 DeprecationWarning,
                 stacklevel=2,
             )
-        elif api_method == "load_csv":
+        else:
             warnings.warn(
                 "chunksize will be ignored when using api_method='load_csv' in a future version of pandas-gbq",
                 PendingDeprecationWarning,
@@ -1122,12 +1128,14 @@ def to_gbq(
             )
         elif if_exists == "replace":
             connector.delete_and_recreate_table(dataset_id, table_id, table_schema)
-        elif if_exists == "append":
+        else:
             if not pandas_gbq.schema.schema_is_subset(original_schema, table_schema):
                 raise InvalidSchema(
                     "Please verify that the structure and "
                     "data types in the DataFrame match the "
-                    "schema of the destination table."
+                    "schema of the destination table.",
+                    table_schema,
+                    original_schema,
                 )
 
             # Update the local `table_schema` so mode (NULLABLE/REQUIRED)
@@ -1283,9 +1291,6 @@ def delete(self, table_id):
         """
         from google.api_core.exceptions import NotFound
 
-        if not self.exists(table_id):
-            raise NotFoundException("Table does not exist")
-
         table_ref = self._table_ref(table_id)
         try:
             self.client.delete_table(table_ref)

pandas_gbq/load.py

@@ -185,6 +185,11 @@ def load_csv_from_file(
     chunksize: Optional[int],
     schema: Optional[Dict[str, Any]],
 ):
+    """Manually encode a DataFrame to CSV and use the buffer in a load job.
+
+    This method is needed for writing with google-cloud-bigquery versions that
+    don't implment load_table_from_dataframe with the CSV serialization format.
+    """
     if schema is None:
         schema = pandas_gbq.schema.generate_bq_schema(dataframe)
 
@@ -203,7 +208,7 @@ def load_chunk(chunk, job_config):
         finally:
             chunk_buffer.close()
 
-    return load_csv(dataframe, chunksize, bq_schema, load_chunk,)
+    return load_csv(dataframe, chunksize, bq_schema, load_chunk)
 
 
 def load_chunks(

pandas_gbq/schema.py

@@ -21,7 +21,19 @@ def to_pandas_gbq(client_schema):
     """Given a sequence of :class:`google.cloud.bigquery.schema.SchemaField`,
     return a schema in pandas-gbq API format.
     """
-    remote_fields = [field_remote.to_api_repr() for field_remote in client_schema]
+    remote_fields = [
+        # Filter out default values. google-cloud-bigquery versions before
+        # 2.31.0 (https://github.com/googleapis/python-bigquery/pull/557)
+        # include a description key, even if not explicitly set. This has the
+        # potential to unset the description unintentionally in cases where
+        # pandas-gbq is updating the schema.
+        {
+            key: value
+            for key, value in field_remote.to_api_repr().items()
+            if value is not None
+        }
+        for field_remote in client_schema
+    ]
     for field in remote_fields:
         field["type"] = field["type"].upper()
         field["mode"] = field["mode"].upper()

setup.py

@@ -28,8 +28,9 @@
     "pandas >=0.24.2",
     "pyarrow >=3.0.0, <7.0dev",
     "pydata-google-auth",
-    "google-auth",
-    "google-auth-oauthlib",
+    "google-api-core >=1.14.0",
+    "google-auth >=1.4.1",
+    "google-auth-oauthlib >=0.0.1",
     # 2.4.* has a bug where waiting for the query can hang indefinitely.
     # https://github.com/pydata/pandas-gbq/issues/343
     "google-cloud-bigquery[bqstorage,pandas] >=1.11.1,<4.0.0dev,!=2.4.*",

testing/constraints-3.7.txt

@@ -6,6 +6,7 @@
 # e.g., if setup.py has "foo >= 1.14.0, < 2.0.0dev",
 # Then this file should have foo==1.14.0
 db-dtypes==0.3.1
+google-api-core==1.14.0
 google-auth==1.4.1
 google-auth-oauthlib==0.0.1
 google-cloud-bigquery==1.11.1

tests/system/test_gbq.py

@@ -1522,11 +1522,6 @@ def test_delete_table(gbq_table):
     assert not gbq_table.exists("test_delete_table")
 
 
-def test_delete_table_not_found(gbq_table):
-    with pytest.raises(gbq.NotFoundException):
-        gbq_table.delete("test_delete_table_not_found")
-
-
 def test_create_table_data_dataset_does_not_exist(
     project, credentials, gbq_dataset, random_dataset_id
 ):

tests/unit/conftest.py

@@ -9,6 +9,36 @@
 import pytest
 
 
+def mock_get_credentials(*args, **kwargs):
+    import google.auth.credentials
+
+    mock_credentials = mock.create_autospec(google.auth.credentials.Credentials)
+    return mock_credentials, "default-project"
+
+
+@pytest.fixture
+def mock_service_account_credentials():
+    import google.oauth2.service_account
+
+    mock_credentials = mock.create_autospec(google.oauth2.service_account.Credentials)
+    return mock_credentials
+
+
+@pytest.fixture
+def mock_compute_engine_credentials():
+    import google.auth.compute_engine
+
+    mock_credentials = mock.create_autospec(google.auth.compute_engine.Credentials)
+    return mock_credentials
+
+
+@pytest.fixture(autouse=True)
+def no_auth(monkeypatch):
+    import pydata_google_auth
+
+    monkeypatch.setattr(pydata_google_auth, "default", mock_get_credentials)
+
+
 @pytest.fixture(autouse=True, scope="function")
 def reset_context():
     import pandas_gbq
@@ -20,41 +50,12 @@ def reset_context():
 @pytest.fixture(autouse=True)
 def mock_bigquery_client(monkeypatch):
     import google.cloud.bigquery
-    import google.cloud.bigquery.table
 
     mock_client = mock.create_autospec(google.cloud.bigquery.Client)
     # Constructor returns the mock itself, so this mock can be treated as the
     # constructor or the instance.
     mock_client.return_value = mock_client
-
-    mock_query = mock.create_autospec(google.cloud.bigquery.QueryJob)
-    mock_query.job_id = "some-random-id"
-    mock_query.state = "DONE"
-    mock_rows = mock.create_autospec(google.cloud.bigquery.table.RowIterator)
-    mock_rows.total_rows = 1
-
-    mock_rows.__iter__.return_value = [(1,)]
-    mock_query.result.return_value = mock_rows
-    mock_client.list_rows.return_value = mock_rows
-    mock_client.query.return_value = mock_query
-    # Mock table creation.
     monkeypatch.setattr(google.cloud.bigquery, "Client", mock_client)
     mock_client.reset_mock()
 
-    # Mock out SELECT 1 query results.
-    def generate_schema():
-        query = mock_client.query.call_args[0][0] if mock_client.query.call_args else ""
-        if query == "SELECT 1 AS int_col":
-            return [google.cloud.bigquery.SchemaField("int_col", "INTEGER")]
-        else:
-            return [google.cloud.bigquery.SchemaField("_f0", "INTEGER")]
-
-    type(mock_rows).schema = mock.PropertyMock(side_effect=generate_schema)
-
-    # Mock out get_table.
-    def get_table(table_ref_or_id, **kwargs):
-        return google.cloud.bigquery.Table(table_ref_or_id)
-
-    mock_client.get_table.side_effect = get_table
-
     return mock_client