fix: omit NaN values when uploading from insert_rows_from_dataframe (#170)

* fix: omit `NaN` values when uploading from `insert_rows_from_dataframe`

NaN values are most often used to indicate a NULL value in pandas. Also,
even when a column is a floating point column, the BigQuery streaming
API JSON parser doesn't seem to be able to handle NaN literals.

* doc: update docstring to indicate missing NaNs

Author: tswast

google/cloud/bigquery/_pandas_helpers.py

@@ -19,6 +19,7 @@
 import logging
 import warnings
 
+import six
 from six.moves import queue
 
 try:
@@ -780,3 +781,14 @@ def download_dataframe_bqstorage(
         selected_fields=selected_fields,
         page_to_item=page_to_item,
     )
+
+
+def dataframe_to_json_generator(dataframe):
+    for row in dataframe.itertuples(index=False, name=None):
+        output = {}
+        for column, value in six.moves.zip(dataframe.columns, row):
+            # Omit NaN values.
+            if value != value:
+                continue
+            output[column] = value
+        yield output

google/cloud/bigquery/client.py

@@ -2535,7 +2535,9 @@ def insert_rows_from_dataframe(
             ]):
                 The destination table for the row data, or a reference to it.
             dataframe (pandas.DataFrame):
-                A :class:`~pandas.DataFrame` containing the data to load.
+                A :class:`~pandas.DataFrame` containing the data to load. Any
+                ``NaN`` values present in the dataframe are omitted from the
+                streaming API request(s).
             selected_fields (Sequence[google.cloud.bigquery.schema.SchemaField]):
                 The fields to return. Required if ``table`` is a
                 :class:`~google.cloud.bigquery.table.TableReference`.
@@ -2559,10 +2561,7 @@ def insert_rows_from_dataframe(
         insert_results = []
 
         chunk_count = int(math.ceil(len(dataframe) / chunk_size))
-        rows_iter = (
-            dict(six.moves.zip(dataframe.columns, row))
-            for row in dataframe.itertuples(index=False, name=None)
-        )
+        rows_iter = _pandas_helpers.dataframe_to_json_generator(dataframe)
 
         for _ in range(chunk_count):
             rows_chunk = itertools.islice(rows_iter, chunk_size)

tests/system.py

@@ -2335,6 +2335,14 @@ def test_insert_rows_from_dataframe(self):
                     "string_col": "another string",
                     "int_col": 50,
                 },
+                {
+                    "float_col": 6.66,
+                    "bool_col": True,
+                    # Include a NaN value, because pandas often uses NaN as a
+                    # NULL value indicator.
+                    "string_col": float("NaN"),
+                    "int_col": 60,
+                },
             ]
         )
 
@@ -2344,14 +2352,28 @@ def test_insert_rows_from_dataframe(self):
         table = retry_403(Config.CLIENT.create_table)(table_arg)
         self.to_delete.insert(0, table)
 
-        Config.CLIENT.insert_rows_from_dataframe(table, dataframe, chunk_size=3)
+        chunk_errors = Config.CLIENT.insert_rows_from_dataframe(
+            table, dataframe, chunk_size=3
+        )
+        for errors in chunk_errors:
+            assert not errors
 
-        retry = RetryResult(_has_rows, max_tries=8)
-        rows = retry(self._fetch_single_page)(table)
+        # Use query to fetch rows instead of listing directly from the table so
+        # that we get values from the streaming buffer.
+        rows = list(
+            Config.CLIENT.query(
+                "SELECT * FROM `{}.{}.{}`".format(
+                    table.project, table.dataset_id, table.table_id
+                )
+            )
+        )
 
         sorted_rows = sorted(rows, key=operator.attrgetter("int_col"))
         row_tuples = [r.values() for r in sorted_rows]
-        expected = [tuple(data_row) for data_row in dataframe.itertuples(index=False)]
+        expected = [
+            tuple(None if col != col else col for col in data_row)
+            for data_row in dataframe.itertuples(index=False)
+        ]
 
         assert len(row_tuples) == len(expected)
 

tests/unit/test_client.py

@@ -5582,6 +5582,74 @@ def test_insert_rows_from_dataframe(self):
             )
             assert call == expected_call
 
+    @unittest.skipIf(pandas is None, "Requires `pandas`")
+    def test_insert_rows_from_dataframe_nan(self):
+        from google.cloud.bigquery.schema import SchemaField
+        from google.cloud.bigquery.table import Table
+
+        API_PATH = "/projects/{}/datasets/{}/tables/{}/insertAll".format(
+            self.PROJECT, self.DS_ID, self.TABLE_REF.table_id
+        )
+
+        dataframe = pandas.DataFrame(
+            {
+                "str_col": ["abc", "def", float("NaN"), "jkl"],
+                "int_col": [1, float("NaN"), 3, 4],
+                "float_col": [float("NaN"), 0.25, 0.5, 0.125],
+            }
+        )
+
+        # create client
+        creds = _make_credentials()
+        http = object()
+        client = self._make_one(project=self.PROJECT, credentials=creds, _http=http)
+        conn = client._connection = make_connection({}, {})
+
+        # create table
+        schema = [
+            SchemaField("str_col", "STRING"),
+            SchemaField("int_col", "INTEGER"),
+            SchemaField("float_col", "FLOAT"),
+        ]
+        table = Table(self.TABLE_REF, schema=schema)
+
+        with mock.patch("uuid.uuid4", side_effect=map(str, range(len(dataframe)))):
+            error_info = client.insert_rows_from_dataframe(
+                table, dataframe, chunk_size=3, timeout=7.5
+            )
+
+        self.assertEqual(len(error_info), 2)
+        for chunk_errors in error_info:
+            assert chunk_errors == []
+
+        EXPECTED_SENT_DATA = [
+            {
+                "rows": [
+                    {"insertId": "0", "json": {"str_col": "abc", "int_col": 1}},
+                    {"insertId": "1", "json": {"str_col": "def", "float_col": 0.25}},
+                    {"insertId": "2", "json": {"int_col": 3, "float_col": 0.5}},
+                ]
+            },
+            {
+                "rows": [
+                    {
+                        "insertId": "3",
+                        "json": {"str_col": "jkl", "int_col": 4, "float_col": 0.125},
+                    }
+                ]
+            },
+        ]
+
+        actual_calls = conn.api_request.call_args_list
+
+        for call, expected_data in six.moves.zip_longest(
+            actual_calls, EXPECTED_SENT_DATA
+        ):
+            expected_call = mock.call(
+                method="POST", path=API_PATH, data=expected_data, timeout=7.5
+            )
+            assert call == expected_call
+
     @unittest.skipIf(pandas is None, "Requires `pandas`")
     def test_insert_rows_from_dataframe_many_columns(self):
         from google.cloud.bigquery.schema import SchemaField