fix(automl): fix TablesClient.predict for list and struct

The Predict request payload is proto. Previously Python dict is
automatically converted to proto. However, the conversion failed for
google.protobuf.ListValue and google.protobuf.Struct. Changing the
structure of the Python dict might fix the problem. However, this PR
fixes the problem by generating the proto message directly. So there
is no auto conversion step.

FIXES #9887

Author: Helin Wang

automl/google/cloud/automl_v1beta1/tables/tables_client.py

@@ -22,8 +22,10 @@
 from google.api_core.gapic_v1 import client_info
 from google.api_core import exceptions
 from google.cloud.automl_v1beta1 import gapic
-from google.cloud.automl_v1beta1.proto import data_types_pb2
+from google.cloud.automl_v1beta1.proto import data_types_pb2, data_items_pb2
 from google.cloud.automl_v1beta1.tables import gcs_client
+from google.protobuf import struct_pb2
+
 
 _GAPIC_LIBRARY_VERSION = pkg_resources.get_distribution("google-cloud-automl").version
 _LOGGER = logging.getLogger(__name__)
@@ -390,21 +392,39 @@ def __column_spec_name_from_args(
 
         return column_spec_name
 
-    def __type_code_to_value_type(self, type_code, value):
+    def __data_type_to_proto_value(self, data_type, value):
+        type_code = data_type.type_code
         if value is None:
-            return {"null_value": 0}
+            return struct_pb2.Value(null_value=struct_pb2.NullValue.NULL_VALUE)
         elif type_code == data_types_pb2.FLOAT64:
-            return {"number_value": value}
-        elif type_code == data_types_pb2.TIMESTAMP:
-            return {"string_value": value}
-        elif type_code == data_types_pb2.STRING:
-            return {"string_value": value}
+            return struct_pb2.Value(number_value=value)
+        elif (
+            type_code == data_types_pb2.TIMESTAMP
+            or type_code == data_types_pb2.STRING
+            or type_code == data_types_pb2.CATEGORY
+        ):
+            return struct_pb2.Value(string_value=value)
         elif type_code == data_types_pb2.ARRAY:
-            return {"list_value": value}
+            if isinstance(value, struct_pb2.ListValue):
+                # in case the user passed in a ListValue.
+                return struct_pb2.Value(list_value=value)
+            array = []
+            for item in value:
+                array.append(
+                    self.__data_type_to_proto_value(data_type.list_element_type, item)
+                )
+            return struct_pb2.Value(list_value=struct_pb2.ListValue(values=array))
         elif type_code == data_types_pb2.STRUCT:
-            return {"struct_value": value}
-        elif type_code == data_types_pb2.CATEGORY:
-            return {"string_value": value}
+            if isinstance(value, struct_pb2.Struct):
+                # in case the user passed in a Struct.
+                return struct_pb2.Value(struct_value=value)
+            struct_value = struct_pb2.Struct()
+            for k, v in value.items():
+                field_value = self.__data_type_to_proto_value(
+                    data_type.struct_type.fields[k], v
+                )
+                struct_value.fields[k].CopyFrom(field_value)
+            return struct_pb2.Value(struct_value=struct_value)
         else:
             raise ValueError("Unknown type_code: {}".format(type_code))
 
@@ -2682,16 +2702,17 @@ def predict(
 
         values = []
         for i, c in zip(inputs, column_specs):
-            value_type = self.__type_code_to_value_type(c.data_type.type_code, i)
+            value_type = self.__data_type_to_proto_value(c.data_type, i)
             values.append(value_type)
 
-        request = {"row": {"values": values}}
+        row = data_items_pb2.Row(values=values)
+        payload = data_items_pb2.ExamplePayload(row=row)
 
         params = None
         if feature_importance:
             params = {"feature_importance": "true"}
 
-        return self.prediction_client.predict(model.name, request, params, **kwargs)
+        return self.prediction_client.predict(model.name, payload, params, **kwargs)
 
     def batch_predict(
         self,

automl/setup.py

@@ -24,6 +24,7 @@
 dependencies = [
     "google-api-core[grpc] >= 1.14.0, < 2.0.0dev",
     'enum34; python_version < "3.4"',
+    "protobuf >= 3.4.0",
 ]
 extras = {
     "pandas": ["pandas>=0.17.1"],

automl/tests/unit/gapic/v1beta1/test_tables_client_v1beta1.py

@@ -23,7 +23,8 @@
 from google.api_core import exceptions
 from google.auth.credentials import AnonymousCredentials
 from google.cloud import automl_v1beta1
-from google.cloud.automl_v1beta1.proto import data_types_pb2
+from google.cloud.automl_v1beta1.proto import data_types_pb2, data_items_pb2
+from google.protobuf import struct_pb2
 
 PROJECT = "project"
 REGION = "region"
@@ -1116,9 +1117,10 @@ def test_predict_from_array(self):
         model.configure_mock(tables_model_metadata=model_metadata, name="my_model")
         client = self.tables_client({"get_model.return_value": model}, {})
         client.predict(["1"], model_name="my_model")
-        client.prediction_client.predict.assert_called_with(
-            "my_model", {"row": {"values": [{"string_value": "1"}]}}, None
+        payload = data_items_pb2.ExamplePayload(
+            row=data_items_pb2.Row(values=[struct_pb2.Value(string_value="1")])
         )
+        client.prediction_client.predict.assert_called_with("my_model", payload, None)
 
     def test_predict_from_dict(self):
         data_type = mock.Mock(type_code=data_types_pb2.CATEGORY)
@@ -1131,11 +1133,15 @@ def test_predict_from_dict(self):
         model.configure_mock(tables_model_metadata=model_metadata, name="my_model")
         client = self.tables_client({"get_model.return_value": model}, {})
         client.predict({"a": "1", "b": "2"}, model_name="my_model")
-        client.prediction_client.predict.assert_called_with(
-            "my_model",
-            {"row": {"values": [{"string_value": "1"}, {"string_value": "2"}]}},
-            None,
+        payload = data_items_pb2.ExamplePayload(
+            row=data_items_pb2.Row(
+                values=[
+                    struct_pb2.Value(string_value="1"),
+                    struct_pb2.Value(string_value="2"),
+                ]
+            )
         )
+        client.prediction_client.predict.assert_called_with("my_model", payload, None)
 
     def test_predict_from_dict_with_feature_importance(self):
         data_type = mock.Mock(type_code=data_types_pb2.CATEGORY)
@@ -1150,10 +1156,16 @@ def test_predict_from_dict_with_feature_importance(self):
         client.predict(
             {"a": "1", "b": "2"}, model_name="my_model", feature_importance=True
         )
+        payload = data_items_pb2.ExamplePayload(
+            row=data_items_pb2.Row(
+                values=[
+                    struct_pb2.Value(string_value="1"),
+                    struct_pb2.Value(string_value="2"),
+                ]
+            )
+        )
         client.prediction_client.predict.assert_called_with(
-            "my_model",
-            {"row": {"values": [{"string_value": "1"}, {"string_value": "2"}]}},
-            {"feature_importance": "true"},
+            "my_model", payload, {"feature_importance": "true"}
         )
 
     def test_predict_from_dict_missing(self):
@@ -1167,18 +1179,32 @@ def test_predict_from_dict_missing(self):
         model.configure_mock(tables_model_metadata=model_metadata, name="my_model")
         client = self.tables_client({"get_model.return_value": model}, {})
         client.predict({"a": "1"}, model_name="my_model")
-        client.prediction_client.predict.assert_called_with(
-            "my_model",
-            {"row": {"values": [{"string_value": "1"}, {"null_value": 0}]}},
-            None,
+        payload = data_items_pb2.ExamplePayload(
+            row=data_items_pb2.Row(
+                values=[
+                    struct_pb2.Value(string_value="1"),
+                    struct_pb2.Value(null_value=struct_pb2.NullValue.NULL_VALUE),
+                ]
+            )
         )
+        client.prediction_client.predict.assert_called_with("my_model", payload, None)
 
     def test_predict_all_types(self):
         float_type = mock.Mock(type_code=data_types_pb2.FLOAT64)
         timestamp_type = mock.Mock(type_code=data_types_pb2.TIMESTAMP)
         string_type = mock.Mock(type_code=data_types_pb2.STRING)
-        array_type = mock.Mock(type_code=data_types_pb2.ARRAY)
-        struct_type = mock.Mock(type_code=data_types_pb2.STRUCT)
+        array_type = mock.Mock(
+            type_code=data_types_pb2.ARRAY,
+            list_element_type=mock.Mock(type_code=data_types_pb2.FLOAT64),
+        )
+        struct = data_types_pb2.StructType()
+        struct.fields["a"].CopyFrom(
+            data_types_pb2.DataType(type_code=data_types_pb2.CATEGORY)
+        )
+        struct.fields["b"].CopyFrom(
+            data_types_pb2.DataType(type_code=data_types_pb2.CATEGORY)
+        )
+        struct_type = mock.Mock(type_code=data_types_pb2.STRUCT, struct_type=struct)
         category_type = mock.Mock(type_code=data_types_pb2.CATEGORY)
         column_spec_float = mock.Mock(display_name="float", data_type=float_type)
         column_spec_timestamp = mock.Mock(
@@ -1211,29 +1237,33 @@ def test_predict_all_types(self):
                 "timestamp": "EST",
                 "string": "text",
                 "array": [1],
-                "struct": {"a": "b"},
+                "struct": {"a": "label_a", "b": "label_b"},
                 "category": "a",
                 "null": None,
             },
             model_name="my_model",
         )
-        client.prediction_client.predict.assert_called_with(
-            "my_model",
-            {
-                "row": {
-                    "values": [
-                        {"number_value": 1.0},
-                        {"string_value": "EST"},
-                        {"string_value": "text"},
-                        {"list_value": [1]},
-                        {"struct_value": {"a": "b"}},
-                        {"string_value": "a"},
-                        {"null_value": 0},
-                    ]
-                }
-            },
-            None,
+        struct = struct_pb2.Struct()
+        struct.fields["a"].CopyFrom(struct_pb2.Value(string_value="label_a"))
+        struct.fields["b"].CopyFrom(struct_pb2.Value(string_value="label_b"))
+        payload = data_items_pb2.ExamplePayload(
+            row=data_items_pb2.Row(
+                values=[
+                    struct_pb2.Value(number_value=1.0),
+                    struct_pb2.Value(string_value="EST"),
+                    struct_pb2.Value(string_value="text"),
+                    struct_pb2.Value(
+                        list_value=struct_pb2.ListValue(
+                            values=[struct_pb2.Value(number_value=1.0)]
+                        )
+                    ),
+                    struct_pb2.Value(struct_value=struct),
+                    struct_pb2.Value(string_value="a"),
+                    struct_pb2.Value(null_value=struct_pb2.NullValue.NULL_VALUE),
+                ]
+            )
         )
+        client.prediction_client.predict.assert_called_with("my_model", payload, None)
 
     def test_predict_from_array_missing(self):
         data_type = mock.Mock(type_code=data_types_pb2.CATEGORY)