FIX add sentinel to all subflows when testing

Sentinel was so far only added to the parent flow, thus subflows got
reused during testing. To prevent reusing subflows during testing,
this commits makes sure that each subflow has the same unique
sentinel in it's external version string

openml/flows/flow.py

@@ -238,7 +238,7 @@ def _from_xml(cls, xml_dict):
             OpenMLFlow
 
         """
-        arguments = {}
+        arguments = OrderedDict()
         dic = xml_dict["oml:flow"]
 
         # Mandatory parts in the xml file

openml/flows/sklearn_converter.py

@@ -41,7 +41,7 @@ def sklearn_to_flow(o):
     elif isinstance(o, (bool, int, float)):
         rval = o
     elif isinstance(o, dict):
-        rval = {}
+        rval = OrderedDict()
         for key, value in o.items():
             if not isinstance(key, six.string_types):
                 raise TypeError('Can only use string as keys, you passed '
@@ -120,8 +120,9 @@ def flow_to_sklearn(o, **kwargs):
 
         else:
             # Regular dictionary
-            rval = {flow_to_sklearn(key, **kwargs): flow_to_sklearn(value, **kwargs)
-                    for key, value in o.items()}
+            rval = OrderedDict((flow_to_sklearn(key, **kwargs),
+                                flow_to_sklearn(value, **kwargs))
+                               for key, value in o.items())
     elif isinstance(o, (list, tuple)):
         rval = [flow_to_sklearn(element, **kwargs) for element in o]
         if isinstance(o, tuple):
@@ -271,14 +272,16 @@ def _deserialize_model(flow, **kwargs):
 
     parameters = flow.parameters
     components = flow.components
-    component_dict = defaultdict(dict)
-    parameter_dict = {}
+    component_dict = OrderedDict()
+    parameter_dict = OrderedDict()
 
     for name in components:
         if '__' in name:
             parameter_name, step = name.split('__')
             value = components[name]
             rval = flow_to_sklearn(value)
+            if parameter_name not in component_dict:
+                component_dict[parameter_name] = OrderedDict()
             component_dict[parameter_name][step] = rval
         else:
             value = components[name]
@@ -292,6 +295,8 @@ def _deserialize_model(flow, **kwargs):
         # Replace the component placeholder by the actual flow
         if isinstance(rval, dict) and 'oml:serialized_object' in rval:
             parameter_name, step = rval['value'].split('__')
+            if parameter_name not in component_dict:
+                component_dict[parameter_name] = OrderedDict()
             rval = component_dict[parameter_name][step]
         parameter_dict[name] = rval
 

tests/flows/test_flow.py

@@ -158,7 +158,15 @@ def test_sklearn_to_upload_to_flow(self):
             estimator=model, param_distributions=parameter_grid, cv=cv)
         rs.fit(X, y)
         flow = openml.flows.sklearn_to_flow(rs)
-        flow.external_version = sentinel + flow.external_version
+
+        # Add the sentinel to all external version strings in all subflows
+        to_visit = collections.deque()
+        to_visit.appendleft(flow)
+        while len(to_visit) > 0:
+            current_flow = to_visit.pop()
+            for sub_flow in current_flow.components.values():
+                to_visit.appendleft(sub_flow)
+            current_flow.external_version = sentinel + current_flow.external_version
 
         flow.publish()
         self.assertIsInstance(flow.flow_id, int)