fix: load_table_from_dataframe for higher scale decimal

Author: Gaurang033

google/cloud/bigquery/_pandas_helpers.py

@@ -509,6 +509,10 @@ def augment_schema(dataframe, current_bq_schema):
         else:
             detected_mode = field.mode
             detected_type = _pyarrow_helpers.arrow_scalar_ids_to_bq(arrow_table.type.id)
+            if detected_type == "NUMERIC" and (
+                arrow_table.type.precision > 38 or arrow_table.type.scale > 9
+            ):
+                detected_type = "BIGNUMERIC"
 
         if detected_type is None:
             unknown_type_fields.append(field)

tests/unit/test_client.py

@@ -8745,6 +8745,62 @@ def test_load_table_from_dataframe_with_csv_source_format(self):
         sent_config = load_table_from_file.mock_calls[0][2]["job_config"]
         assert sent_config.source_format == job.SourceFormat.CSV
 
+    @unittest.skipIf(
+        pandas is None or PANDAS_INSTALLED_VERSION < PANDAS_MINIUM_VERSION,
+        "Only `pandas version >=1.0.0` supported",
+    )
+    @unittest.skipIf(pyarrow is None, "Requires `pyarrow`")
+    def test_load_table_from_dataframe_w_higher_scale_decimal128_datatype(self):
+        from google.cloud.bigquery.client import _DEFAULT_NUM_RETRIES
+        from google.cloud.bigquery import job
+        from google.cloud.bigquery.schema import SchemaField
+        from decimal import Decimal
+
+        client = self._make_client()
+        dataframe = pandas.DataFrame(
+            {
+                "x": [
+                    Decimal("0.12345678901234560000000000000000000000"),
+                    Decimal("01234567890123456789012345678901234567.1234567891"),
+                ]
+            }
+        )
+        load_patch = mock.patch(
+            "google.cloud.bigquery.client.Client.load_table_from_file", autospec=True
+        )
+
+        get_table_patch = mock.patch(
+            "google.cloud.bigquery.client.Client.get_table",
+            autospec=True,
+            return_value=mock.Mock(schema=[SchemaField("x", "BIGNUMERIC", "NULLABLE")]),
+        )
+
+        with load_patch as load_table_from_file, get_table_patch:
+            client.load_table_from_dataframe(
+                dataframe, self.TABLE_REF, location=self.LOCATION
+            )
+
+        load_table_from_file.assert_called_once_with(
+            client,
+            mock.ANY,
+            self.TABLE_REF,
+            num_retries=_DEFAULT_NUM_RETRIES,
+            rewind=True,
+            size=mock.ANY,
+            job_id=mock.ANY,
+            job_id_prefix=None,
+            location=self.LOCATION,
+            project=None,
+            job_config=mock.ANY,
+            timeout=DEFAULT_TIMEOUT,
+        )
+
+        sent_config = load_table_from_file.mock_calls[0][2]["job_config"]
+        assert sent_config.source_format == job.SourceFormat.PARQUET
+        assert tuple(sent_config.schema) == (
+            SchemaField("x", "BIGNUMERIC", "NULLABLE", None),
+        )
+
     def test_load_table_from_json_basic_use(self):
         from google.cloud.bigquery.client import _DEFAULT_NUM_RETRIES
         from google.cloud.bigquery import job