From 608d3ad057cc5acb4ad43e9e75d3c7e877dfd40b Mon Sep 17 00:00:00 2001
From: Mathieu Blondel <mathieu@mblondel.org>
Date: Thu, 14 Nov 2013 18:23:09 +0900
Subject: [PATCH] Add syntactic sugar for pipelines.

---
 sklearn/base.py                | 12 +++++++++++-
 sklearn/tests/test_pipeline.py | 20 ++++++++++++++++++++
 2 files changed, 31 insertions(+), 1 deletion(-)

diff --git a/sklearn/base.py b/sklearn/base.py
index df85044e972b9..4f7a1d749513a 100644
--- a/sklearn/base.py
+++ b/sklearn/base.py
@@ -262,13 +262,23 @@ def __repr__(self):
         class_name = self.__class__.__name__
         return '%s(%s)' % (class_name, _pprint(self.get_params(deep=False),
                                                offset=len(class_name),),)
-
     def __str__(self):
         class_name = self.__class__.__name__
         return '%s(%s)' % (class_name,
                            _pprint(self.get_params(deep=True),
                                    offset=len(class_name), printer=str,),)
 
+    def __or__(self, estimator):
+        from sklearn.pipeline import Pipeline
+        if isinstance(self, Pipeline):
+            name = estimator.__class__.__name__
+            steps = self.steps + [(name, estimator)]
+        else:
+            name1 = self.__class__.__name__
+            name2 = estimator.__class__.__name__
+            steps = [(name1, self), (name2, estimator)]
+        return Pipeline(steps)
+
 
 ###############################################################################
 class ClassifierMixin(object):
diff --git a/sklearn/tests/test_pipeline.py b/sklearn/tests/test_pipeline.py
index c3b2fbc170b70..295e2a01395d0 100644
--- a/sklearn/tests/test_pipeline.py
+++ b/sklearn/tests/test_pipeline.py
@@ -195,6 +195,26 @@ def test_pipeline_methods_preprocessing_svm():
         pipe.score(X, y)
 
 
+def test_pipeline_syntactic_sugar():
+    iris = load_iris()
+    X = iris.data
+    y = iris.target
+    pipeline = StandardScaler() | PCA(n_components=2) | SVC(random_state=0)
+    assert_true(isinstance(pipeline, Pipeline))
+
+    assert_equal(pipeline.steps[0][0], "StandardScaler")
+    assert_true(isinstance(pipeline.steps[0][1], StandardScaler))
+
+    assert_equal(pipeline.steps[1][0], "PCA")
+    assert_true(isinstance(pipeline.steps[1][1], PCA))
+
+    assert_equal(pipeline.steps[2][0], "SVC")
+    assert_true(isinstance(pipeline.steps[2][1], SVC))
+
+    y_pred = pipeline.fit(X, y).predict(X)
+    assert_equal(np.mean(y == y_pred), 0.92)
+
+
 def test_feature_union():
     # basic sanity check for feature union
     iris = load_iris()