Pipeline checks (#166)

* All intermediate estimators should not implement both sample and transform methods
* All intermediate estimators should not be or inherit from the Pipeline class

Author: chkoar

imblearn/pipeline.py

@@ -24,6 +24,32 @@
 __all__ = ['Pipeline']
 
 
+def _validate_step_methods(step):
+
+    if (not (hasattr(step, "fit") or hasattr(step, "fit_transform") or hasattr(
+            step, "fit_sample")) or
+            not (hasattr(step, "transform") or hasattr(step, "sample"))):
+        raise TypeError(
+            "All intermediate steps of the chain should "
+            "be estimators that implement fit and transform or sample (but not both)"
+            " '%s' (type %s) doesn't)" % (step, type(t)))
+
+
+def _validate_step_behaviour(step):
+    if (hasattr(step, "fit_sample") and hasattr(step, "fit_transform")) or (
+            hasattr(step, "sample") and hasattr(step, "transform")):
+        raise TypeError(
+            "All intermediate steps of the chain should "
+            "be estimators that implement fit and transform or sample."
+            " '%s' implements both)" % (step))
+
+
+def _validate_step_class(step):
+    if isinstance(step, pipeline.Pipeline):
+        raise TypeError(
+            "All intermediate steps of the chain should not be Pipelines")
+
+
 class Pipeline(pipeline.Pipeline):
 
     """Pipeline of transforms and resamples with a final estimator.
@@ -104,12 +130,9 @@ def __init__(self, steps):
         estimator = estimators[-1]
 
         for t in transforms:
-            if (not (hasattr(t, "fit") or hasattr(t, "fit_transform") or
-                     hasattr(t, "fit_sample")) or
-                    not (hasattr(t, "transform") or hasattr(t, "sample"))):
-                raise TypeError("All intermediate steps of the chain should "
-                                "be transforms and implement fit and transform"
-                                " '%s' (type %s) doesn't)" % (t, type(t)))
+            _validate_step_methods(t)
+            _validate_step_behaviour(t)
+            _validate_step_class(t)
 
         if not hasattr(estimator, "fit"):
             raise TypeError("Last step of chain should implement fit "

imblearn/tests/test_pipeline.py

@@ -74,6 +74,20 @@ def predict(self, X):
         return self.successful
 
 
+
+class FitTransformSample(T):
+    """Mock classifier
+    """  
+
+    def fit(self, X, y, should_succeed=False):
+        pass
+
+    def sample(X, y=None):
+        return X, y
+
+    def transform(X, y=None):
+        return  X
+
 def test_pipeline_init():
     # Test the various init parameters of the pipeline.
     assert_raises(TypeError, Pipeline)
@@ -431,3 +445,32 @@ def test_pipeline_methods_anova_rus():
     pipe.predict_proba(X)
     pipe.predict_log_proba(X)
     pipe.score(X, y)
+
+
+
+
+def test_pipeline_with_step_that_implements_both_sample_and_transform():
+    # Test the various methods of the pipeline (anova).
+    X, y = make_classification(n_classes=2, class_sep=2, weights=[0.1, 0.9],
+                               n_informative=3, n_redundant=1, flip_y=0,
+                               n_features=20, n_clusters_per_class=1,
+                               n_samples=5000, random_state=0)
+
+    clf = LogisticRegression()
+    assert_raises(TypeError, Pipeline, [('step', FitTransformSample()), ('logistic', clf)])
+    #assert_raises(TypeError, lambda x: [][0])
+
+
+def test_pipeline_with_step_that_it_is_pipeline():
+    # Test the various methods of the pipeline (anova).
+    X, y = make_classification(n_classes=2, class_sep=2, weights=[0.1, 0.9],
+                               n_informative=3, n_redundant=1, flip_y=0,
+                               n_features=20, n_clusters_per_class=1,
+                               n_samples=5000, random_state=0)
+    # Test with RandomUnderSampling + Anova + LogisticRegression
+    clf = LogisticRegression()
+    rus = RandomUnderSampler(random_state=0)
+    filter1 = SelectKBest(f_classif, k=2)
+    pipe1 = Pipeline([('rus', rus), ('anova', filter1)])
+    assert_raises(TypeError, Pipeline, [('pipe1', pipe1), ('logistic', clf)])
+