In PredictService, allow handling data as a 1d array instead of 2d

When working with text, scikit-learn and others will sometimes expect
the input to be a 1d array of strings rather than a 2d array.  Setting
the new *unwrap_sample* to true will use this representation.

palladium/server.py

@@ -59,9 +59,15 @@ class PredictService:
         }
 
     def __init__(
-            self, mapping, params=(), entry_point='/predict',
-            decorator_list_name='predict_decorators',
-            predict_proba=False, **kwargs):
+        self,
+        mapping,
+        params=(),
+        entry_point='/predict',
+        decorator_list_name='predict_decorators',
+        predict_proba=False,
+        unwrap_sample=False,
+        **kwargs
+    ):
         """
         :param mapping:
           A list of query parameters and their type that should be
@@ -85,12 +91,19 @@ def __init__(
           Instead of returning a single class (the default), when
           *predict_proba* is set to true, the result will instead
           contain a list of class probabilities.
+
+        :param unwrap_sample:
+          When working with text, scikit-learn and others will
+          sometimes expect the input to be a 1d array of strings
+          rather than a 2d array.  Setting *unwrap_sample* to true
+          will use this representation.
         """
         self.mapping = mapping
         self.params = params
         self.entry_point = entry_point
         self.decorator_list_name = decorator_list_name
         self.predict_proba = predict_proba
+        self.unwrap_sample = unwrap_sample
         vars(self).update(kwargs)
 
     def initialize_component(self, config):
@@ -132,7 +145,11 @@ def sample_from_data(self, model, data):
         for key, type_name in self.mapping:
             value_type = self.types[type_name]
             values.append(value_type(data[key]))
-        return np.array(values, dtype=object)
+        if self.unwrap_sample:
+            assert len(values) == 1
+            return np.array(values[0])
+        else:
+            return np.array(values, dtype=object)
 
     def params_from_data(self, model, data):
         """Retrieve additional parameters (keyword arguments) for

palladium/tests/test_server.py

@@ -171,6 +171,67 @@ def test_sample_from_data(self, PredictService):
         assert sample[0] == 'myflower'
         assert sample[1] == 3
 
+    def test_unwrap_sample_get(self, PredictService, flask_app):
+        predict_service = PredictService(
+            mapping=[('text', 'str')],
+            unwrap_sample=True,
+            )
+        model = Mock()
+        model.predict.return_value = np.array([1])
+        with flask_app.test_request_context():
+            request = Mock(
+                args=dict([
+                    ('text', 'Hi this is text'),
+                    ]),
+                method='GET',
+                )
+            resp = predict_service(model, request)
+
+        assert model.predict.call_args[0][0].ndim == 1
+        model.predict.assert_called_with(np.array(['Hi this is text']))
+        resp_data = json.loads(resp.get_data(as_text=True))
+        assert resp.status_code == 200
+        assert resp_data == {
+            "metadata": {
+                "status": "OK",
+                "error_code": 0,
+                },
+            "result": 1,
+            }
+
+    def test_unwrap_sample_post(self, PredictService, flask_app):
+        predict_service = PredictService(
+            mapping=[('text', 'str')],
+            unwrap_sample=True,
+            )
+        model = Mock()
+        model.predict.return_value = np.array([1, 2])
+        with flask_app.test_request_context():
+            request = Mock(
+                json=[
+                    {'text': 'First piece of text'},
+                    {'text': 'Second piece of text'},
+                    ],
+                method='POST',
+                mimetype='application/json',
+                )
+            resp = predict_service(model, request)
+
+        assert model.predict.call_args[0][0].ndim == 1
+        assert (
+            model.predict.call_args[0] ==
+            np.array(['First piece of text', 'Second piece of text'])
+            ).all()
+        resp_data = json.loads(resp.get_data(as_text=True))
+        assert resp.status_code == 200
+        assert resp_data == {
+            "metadata": {
+                "status": "OK",
+                "error_code": 0,
+                },
+            "result": [1, 2],
+            }
+
     def test_probas(self, PredictService, flask_app):
         model = Mock()
         model.predict_proba.return_value = np.array([[0.1, 0.5, math.pi]])