ADD propagate external version of subflows to parent flow

Author: mfeurer

openml/flows/sklearn_converter.py

@@ -257,7 +257,6 @@ def _serialize_model(model):
 
     # Create a flow name, which contains all components in brackets, for
     # example RandomizedSearchCV(Pipeline(StandardScaler,AdaBoostClassifier(DecisionTreeClassifier)),StandardScaler,AdaBoostClassifier(DecisionTreeClassifier))
-    # TODO the name above is apparently wrong, I need to test and check this
     class_name = model.__module__ + "." + model.__class__.__name__
 
     # will be part of the name (in brackets)
@@ -274,7 +273,24 @@ def _serialize_model(model):
     else:
         name = class_name
 
-    external_version = _get_external_version_info()
+    # Get the external versions of all sub-components
+    model_package_name = model.__module__.split('.')[0]
+    module = importlib.import_module(model_package_name)
+    model_package_version_number = module.__version__
+    external_version = '%s==%s' % (model_package_name, model_package_version_number)
+
+    external_versions = set()
+    external_versions.add(external_version)
+    to_visit_stack = []
+    to_visit_stack.extend(sub_components.values())
+    while len(to_visit_stack) > 0:
+        visitee = to_visit_stack.pop()
+        for external_version in visitee.external_version.split(','):
+            external_versions.add(external_version)
+        to_visit_stack.extend(visitee.components.values())
+    external_versions = list(sorted(external_versions))
+    external_version = ','.join(external_versions)
+
     flow = OpenMLFlow(name=name,
                       class_name=class_name,
                       description='Automatically created sub-component.',
@@ -470,7 +486,3 @@ def _deserialize_cross_validator(value, **kwargs):
     for parameter in parameters:
         parameters[parameter] = flow_to_sklearn(parameters[parameter])
     return model_class(**parameters)
-
-
-def _get_external_version_info():
-    return 'sklearn_' + sklearn.__version__

tests/__init__.py

@@ -0,0 +1,3 @@
+# Dummy to allow mock classes in the test files to have a version number for
+# their parent module
+__version__ = '0.1'

tests/flows/dummy_learn/__init__.py

@@ -0,0 +1 @@
+__version__ = 1.0

tests/flows/dummy_learn/dummy_forest.py

@@ -0,0 +1,12 @@
+class DummyRegressor(object):
+    def fit(self, X, y):
+        return self
+
+    def predict(self, X):
+        return X[:, 0]
+
+    def get_params(self, deep=False):
+        return {}
+
+    def set_params(self, params):
+        return None

tests/flows/test_sklearn.py

@@ -1,5 +1,7 @@
 from collections import OrderedDict
 import json
+import os
+import sys
 import unittest
 
 import numpy as np
@@ -61,6 +63,7 @@ def test_serialize_model(self):
         serialization = sklearn_to_flow(model)
 
         self.assertEqual(serialization.name, fixture_name)
+        self.assertEqual(serialization.class_name, fixture_name)
         self.assertEqual(serialization.description, fixture_description)
         self.assertEqual(serialization.parameters, fixture_parameters)
 
@@ -78,16 +81,21 @@ def test_serialize_model_with_subcomponent(self):
 
         fixture_name = 'sklearn.ensemble.weight_boosting.AdaBoostClassifier' \
                        '(base_estimator=sklearn.tree.tree.DecisionTreeClassifier)'
+        fixture_class_name = 'sklearn.ensemble.weight_boosting.AdaBoostClassifier'
         fixture_description = 'Automatically created sub-component.'
+        fixture_subcomponent_class_name = 'sklearn.tree.tree.DecisionTreeClassifier'
 
         serialization =  sklearn_to_flow(model)
 
         self.assertEqual(serialization.name, fixture_name)
+        self.assertEqual(serialization.class_name, fixture_class_name)
         self.assertEqual(serialization.description, fixture_description)
         self.assertEqual(serialization.parameters['algorithm'], '"SAMME.R"')
         self.assertIsInstance(serialization.parameters['base_estimator'], str)
         self.assertEqual(serialization.parameters['learning_rate'], '1.0')
         self.assertEqual(serialization.parameters['n_estimators'], '100')
+        self.assertEqual(serialization.components['base_estimator'].class_name,
+                         fixture_subcomponent_class_name)
 
         new_model = flow_to_sklearn(serialization)
 
@@ -403,7 +411,7 @@ def test_error_on_adding_component_multiple_times_to_flow(self):
                   "\('pca2', PCA\(copy=True, iterated_power='auto', " \
                   "n_components=None, random_state=None,\n" \
                   "  svd_solver='auto', tol=0.0, whiten=False\)\)\)\)."
-        #self.assertRaisesRegexp(ValueError, fixture, sklearn_to_flow, pipeline)
+        self.assertRaisesRegexp(ValueError, fixture, sklearn_to_flow, pipeline)
 
         fu = sklearn.pipeline.FeatureUnion((('pca1', pca), ('pca2', pca2)))
         fixture = "Found a second occurence of component sklearn.decomposition.pca.PCA when trying to serialize " \
@@ -416,7 +424,7 @@ def test_error_on_adding_component_multiple_times_to_flow(self):
                   " n_components=None, random_state=None,\n" \
                   "  svd_solver='auto', tol=0.0, whiten=False\)\)\),\n" \
                   "       transformer_weights=None\)."
-        #self.assertRaisesRegexp(ValueError, fixture, sklearn_to_flow, fu)
+        self.assertRaisesRegexp(ValueError, fixture, sklearn_to_flow, fu)
 
         fs = sklearn.feature_selection.SelectKBest()
         fu2 = sklearn.pipeline.FeatureUnion((('pca1', pca), ('fs', fs)))
@@ -435,3 +443,17 @@ def test_error_on_adding_component_multiple_times_to_flow(self):
                   " n_components=None, random_state=None,\n" \
                   "  svd_solver='auto', tol=0.0, whiten=False\)\)\)\)."
         self.assertRaisesRegexp(ValueError, fixture, sklearn_to_flow, pipeline2)
+
+    def test_subflow_version_change(self):
+        this_directory = os.path.dirname(os.path.abspath(__file__))
+        sys.path.append(this_directory)
+        import dummy_learn
+        import dummy_learn.dummy_forest
+        pca = sklearn.decomposition.PCA()
+        dummy = dummy_learn.dummy_forest.DummyRegressor()
+        pipeline = sklearn.pipeline.Pipeline((('pca', pca), ('dummy', dummy)))
+        flow = sklearn_to_flow(pipeline)
+        self.assertEqual(flow.external_version, 'dummy_learn==1.0,sklearn==0.18.1')
+        dummy_learn.__version__ = '1.1.0'
+        flow = sklearn_to_flow(pipeline)
+        self.assertEqual(flow.external_version, 'dummy_learn==1.1.0,sklearn==0.18.1')