Merge pull request #989 from lennon310/status_code

Set Http Status Code in REST Predict

python/seldon_core/wrapper.py

@@ -48,8 +48,12 @@ def Predict():
         requestJson = get_request()
         logger.debug("REST Request: %s", request)
         response = seldon_core.seldon_methods.predict(user_model, requestJson)
+        json_response = jsonify(response)
+        if 'status' in response and 'code' in response['status']:
+            json_response.status_code = response['status']['code']
+
         logger.debug("REST Response: %s", response)
-        return jsonify(response)
+        return json_response
 
     @app.route("/send-feedback", methods=["GET", "POST"])
     @app.route("/api/v0.1/feedback", methods=["POST"])

python/tests/test_model_microservice.py

@@ -103,6 +103,57 @@ def send_feedback_rest(self, request):
     def send_feedback_grpc(self, request):
         print("Feedback called")
 
+class UserObjectLowLevelWithStatusInResponse(SeldonComponent):
+    def __init__(self, metrics_ok=True, ret_nparray=False):
+        self.metrics_ok = metrics_ok
+        self.ret_nparray = ret_nparray
+        self.nparray = np.array([1, 2, 3])
+
+    def predict_rest(self, request):
+        return {"data": {"ndarray": [9, 9]}, "status": {"code": 400, "status": "FAILURE"}}
+
+    def predict_grpc(self, request):
+        arr = np.array([9, 9])
+        datadef = prediction_pb2.DefaultData(
+            tensor=prediction_pb2.Tensor(
+                shape=(2, 1),
+                values=arr
+            )
+        )
+        request = prediction_pb2.SeldonMessage(data=datadef)
+        return request
+
+    def send_feedback_rest(self, request):
+        print("Feedback called")
+
+    def send_feedback_grpc(self, request):
+        print("Feedback called")
+
+
+class UserObjectLowLevelWithStatusInResponseWithPredictRaw(SeldonComponent):
+    def __init__(self, check_name):
+        self.check_name=check_name
+
+    def predict_raw(self, msg):
+        msg=json_to_seldon_message(msg)
+        if self.check_name == 'img':
+            file_data=msg.binData
+            img = Image.open(io.BytesIO (file_data))
+            img.verify()
+            return {"meta": seldon_message_to_json(msg.meta),
+                    "data": {"ndarray": [rs232_checksum(file_data).decode('utf-8')]},
+                    "status": {"code": 400, "status": "FAILURE"}}
+        elif self.check_name == 'txt':
+            file_data=msg.binData
+            return {"meta": seldon_message_to_json(msg.meta),
+                    "data": {"ndarray": [file_data.decode('utf-8')]},
+                    "status": {"code": 400, "status": "FAILURE"}}
+        elif self.check_name == 'strData':
+            file_data=msg.strData
+            return {"meta": seldon_message_to_json(msg.meta),
+                    "data": {"ndarray": [file_data]},
+                    "status": {"code": 400, "status": "FAILURE"}}
+
 
 class UserObjectLowLevelWithPredictRaw(SeldonComponent):
     def __init__(self, check_name):
@@ -259,6 +310,18 @@ def test_model_lowlevel_multi_form_data_strData_ok():
     )
 
 
+
+def test_model_lowlevel_multi_form_data_strData_non200status():
+    user_object = UserObjectLowLevelWithStatusInResponseWithPredictRaw('strData')
+    app = get_rest_microservice(user_object)
+    client = app.test_client()
+    rv = client.post('/predict',data={"meta":'{"puid":"1234"}',"strData":(f'./tests/resources/test.txt','test.txt')},content_type='multipart/form-data')
+    j = json.loads(rv.data)
+    assert rv.status_code == 400
+    assert j["meta"]["puid"] == "1234"
+    assert j["data"]["ndarray"][0] == "this is test file for testing multipart/form-data input\n"
+
+
 def test_model_multi_form_data_ok():
     user_object = UserObject()
     app = get_rest_microservice(user_object)
@@ -302,6 +365,16 @@ def test_model_feedback_lowlevel_ok():
     assert rv.status_code == 200
 
 
+def test_model_non200status_lowlevel():
+    user_object = UserObjectLowLevelWithStatusInResponse()
+    app = get_rest_microservice(user_object)
+    client = app.test_client()
+    rv = client.get('/predict?json={"request":{"data":{"ndarray":[]}},"reward":1.0}')
+    j = json.loads(rv.data)
+    print(j)
+    assert rv.status_code == 400
+
+
 @skipif_tf_missing
 def test_model_tftensor_ok():
     user_object = UserObject()