add support for serializing python dictionaries to json

Author: andremoeller

src/sagemaker/mxnet/model.py

@@ -20,7 +20,8 @@
 class MXNetPredictor(RealTimePredictor):
     """A RealTimePredictor for inference against MXNet Endpoints.
 
-    This is able to serialize Python lists and numpy arrays to multidimensional tensors for MXNet inference."""
+    This is able to serialize Python lists, dictionaries, and numpy arrays to multidimensional tensors for MXNet
+    inference."""
 
     def __init__(self, endpoint_name, sagemaker_session=None):
         """Initialize an ``MXNetPredictor``.

src/sagemaker/predictor.py

@@ -240,7 +240,12 @@ def __call__(self, data):
         if isinstance(data, list):
             if not len(data) > 0:
                 raise ValueError("empty array can't be serialized")
-            return _json_serialize_python_array(data)
+            return _json_serialize_python_object(data)
+
+        if isinstance(data, dict):
+            if not len(data.keys()) > 0:
+                raise ValueError("empty dictionary can't be serialized")
+            return _json_serialize_python_object(data)
 
         # files and buffers
         if hasattr(data, 'read'):
@@ -254,10 +259,10 @@ def __call__(self, data):
 
 def _json_serialize_numpy_array(data):
     # numpy arrays can't be serialized but we know they have uniform type
-    return _json_serialize_python_array(data.tolist())
+    return _json_serialize_python_object(data.tolist())
 
 
-def _json_serialize_python_array(data):
+def _json_serialize_python_object(data):
     return _json_serialize_object(data)
 
 

src/sagemaker/tensorflow/model.py

@@ -19,8 +19,10 @@
 
 
 class TensorFlowPredictor(RealTimePredictor):
-    """A ``RealTimePredictor`` for inference against MXNet ``Endpoint``s."""
+    """A ``RealTimePredictor`` for inference against TensorFlow ``Endpoint``s.
 
+    This is able to serialize Python lists, dictionaries, and numpy arrays to multidimensional tensors for MXNet
+    inference"""
     def __init__(self, endpoint_name, sagemaker_session=None):
         """Initialize an ``TensorFlowPredictor``.
 

src/sagemaker/tensorflow/predictor.py

@@ -32,7 +32,7 @@ def __init__(self):
         self.content_type = CONTENT_TYPE_OCTET_STREAM
 
     def __call__(self, data):
-        # isintance does not work here because a same protobuf message can be imported from a different module.
+        # isinstance does not work here because a same protobuf message can be imported from a different module.
         # for example sagemaker.tensorflow.tensorflow_serving.regression_pb2 and tensorflow_serving.apis.regression_pb2
         predict_type = data.__class__.__name__
 

tests/unit/test_predictor.py

@@ -51,12 +51,26 @@ def test_json_serializer_python_array():
     assert result == '[1, 2, 3]'
 
 
+def test_json_serializer_python_dictionary():
+    d = {"gender": "m", "age": 22, "city": "Paris"}
+
+    result = json_serializer(d)
+
+    assert json.loads(result) == d
+
+
 def test_json_serializer_python_invalid_empty():
     with pytest.raises(ValueError) as error:
         json_serializer([])
     assert "empty array" in str(error)
 
 
+def test_json_serializer_python_dictionary_invalid_empty():
+    with pytest.raises(ValueError) as error:
+        json_serializer({})
+    assert "empty dictionary" in str(error)
+
+
 def test_json_serializer_csv_buffer():
     csv_file_path = os.path.join(DATA_DIR, "with_integers.csv")
     with open(csv_file_path) as csv_file: