scikit-learn
diff --git a/‎sklearn/utils/testing.py
Lines changed: 78 additions & 5 deletions b/‎sklearn/utils/testing.py
Lines changed: 78 additions & 5 deletions
@@ -18,6 +18,8 @@
 
 import scipy as sp
 import scipy.io
+import numpy as np
+
 from functools import wraps
 try:
     # Python 2
@@ -45,7 +47,6 @@
 from numpy.testing import assert_array_equal
 from numpy.testing import assert_array_almost_equal
 from numpy.testing import assert_array_less
-import numpy as np
 
 from sklearn.base import (ClassifierMixin, RegressorMixin, TransformerMixin,
                           ClusterMixin)
@@ -54,8 +55,9 @@
            "assert_raises_regexp", "raises", "with_setup", "assert_true",
            "assert_false", "assert_almost_equal", "assert_array_equal",
            "assert_array_almost_equal", "assert_array_less",
-           "assert_less", "assert_less_equal",
-           "assert_greater", "assert_greater_equal"]
+           "assert_less", "assert_less_equal", "assert_greater",
+           "assert_greater_equal",
+           "assert_same_model", "assert_fitted_attributes_equal"]
 
 
 try:
@@ -422,6 +424,77 @@ def assert_raise_message(exceptions, message, function, *args, **kwargs):
                              (names, function.__name__))
 
 
+def _assert_same_model_method(method, X, estimator1, estimator2, msg=None):
+    if hasattr(estimator1, method):
+        m = '%r\n\nhas %s, but\n\n%r\n\ndoes not' % (estimator1,
+                                                     method,
+                                                     estimator2)
+
+        if not hasattr(estimator2, method):
+            raise AttributeError(m)
+
+        # Check if the method(X) returns the same for both models.
+        res1 = getattr(estimator1, method)(X)
+        res2 = getattr(estimator2, method)(X)
+        same_model = (res1.shape == res2.shape) and np.allclose(res1, res2)
+
+        if msg is None:
+            msg = ("Models are not equal. \n\n%s method returned"
+                   " different results:\n\n%r\n\n and\n\n%r\n\n for :\n\n%r"
+                   "\n\nand :\n\n%r" % (method, res1, res2,
+                                        estimator1, estimator2))
+
+        assert same_model, msg
+
+
+def assert_same_model(X, estimator1, estimator2, msg=None):
+    """Helper function to check if models are same"""
+    _assert_same_model_method('predict', X, estimator1, estimator2, msg)
+    _assert_same_model_method('transform', X, estimator1, estimator2, msg)
+    _assert_same_model_method('decision_function',
+                              X, estimator1, estimator2, msg)
+    _assert_same_model_method('predict_proba',
+                              X, estimator1, estimator2, msg)
+
+    assert_fitted_attributes_equal(estimator1, estimator2)
+
+
+def assert_not_same_model(X, estimator1, estimator2, msg=None):
+    """Helper function to check if models are different"""
+    try:
+        assert_same_model(X, estimator1, estimator2)
+    except AssertionError:
+        pass
+    else:
+        raise AssertionError(msg)
+
+
+def assert_fitted_attributes_equal(estimator1, estimator2):
+    """Helper function to check if fitted model attributes are equal."""
+    # A list of attributes which are known to be inconsistent.
+    # FIXME embedding_ to be removed after #4299
+    skip_attributes = ('embedding_',)
+
+    est1_dict = estimator1.__dict__.copy()
+    est2_dict = estimator2.__dict__.copy()
+
+    # Remove all keys that are non-attributes, are not comparable types
+    # and those which are found to be inconsistent
+    for attr, value1 in est1_dict.items():
+        if ((not attr.endswith('_')) or attr.endswith('estimator_') or
+                attr.endswith('estimators_') or
+                attr in skip_attributes):
+            est1_dict.pop(attr)
+            try:
+                est2_dict.pop(attr)
+            except KeyError:  # Incase the attribute is not present in est2
+                pass
+
+    # assert_equal is capable of recursively checking for all the items
+    # of the two dicts
+    assert_equal(est1_dict, est2_dict)
+
+
 def fake_mldata(columns_dict, dataname, matfile, ordering=None):
     """Create a fake mldata data set.
 
@@ -628,8 +701,8 @@ def is_abstract(c):
         estimators = filtered_estimators
         if type_filter:
             raise ValueError("Parameter type_filter must be 'classifier', "
-                             "'regressor', 'transformer', 'cluster' or None, got"
-                             " %s." % repr(type_filter))
+                             "'regressor', 'transformer', 'cluster' or None,"
+                             "got %s." % repr(type_filter))
 
     # drop duplicates, sort for reproducibility
     return sorted(set(estimators))