Implementing loglinear classification (without training), for CL classification

appengine/findit/crash/changelist_features/test/min_distance_test.py

Index: appengine/findit/crash/changelist_features/test/min_distance_test.py
diff --git a/appengine/findit/crash/changelist_features/test/min_distance_test.py b/appengine/findit/crash/changelist_features/test/min_distance_test.py
new file mode 100644
index 0000000000000000000000000000000000000000..eb2ae28a6a4940fe73abd7966592d78569ab2d98
--- /dev/null
+++ b/appengine/findit/crash/changelist_features/test/min_distance_test.py
@@ -0,0 +1,67 @@
+# Copyright 2016 The Chromium Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+from crash.changelist_features import min_distance
+from crash.results import AnalysisInfo
+from crash.results import MatchResult
+from crash.results import Result
+from crash.scorers.test.scorer_test_suite import ScorerTestSuite
+from crash.stacktrace import StackFrame
+
+
+class MinDistanceTest(ScorerTestSuite):

[Sharu Jiang, 2016/12/06 20:49:19]

Now that we changed Scorer -> Feature, we should also change the ScorerTestSuite
to FeatureTestSuite.

[wrengr, 2016/12/07 00:55:38]

Will do. I actually just pushed a revised CL that removes the
changelist_features directory for this CL. Reason being that their types weren't
quite right. I will re-add them in the CL that actually uses loglinear models
for the changelist classifier, and will be sure to rename that file in/for that
CL.

+
+  @property
+  def _maximum(self):
+    return float(min_distance.DEFAULT_MAXIMUM)
+
+  def _GetMockResult(self, mock_min_distance):
+    """Returns a ``Result`` with the desired min_distance."""
+    result = Result(self._GetDummyChangeLog(), 'src/')
+    result.file_to_analysis_info = {
+        'file': AnalysisInfo(
+            min_distance=mock_min_distance,
+            min_distance_frame=None)
+    }
+    return result
+
+  def testMinDistanceFeatureNone(self):
+    """Test that the feature returns 0 when there are no frames."""
+    result = Result(self._GetDummyChangeLog(), 'src/')
+    self.assertIsNone(min_distance.MinDistanceFeature()(result))
+
+  def testMinDistanceFeature(self):

[inferno, 2016/12/06 18:07:06]

Some functions missing docstrings! Can you add presubmit rules to findit project
for this. clusterfuzz has some presubmit rules.

[Sharu Jiang, 2016/12/06 20:49:19]

I added a bug to track this.
https://bugs.chromium.org/p/chromium/issues/detail?id=671721

+    result = self._GetMockResult(42.)
+    self.assertEqual(42., min_distance.MinDistanceFeature()(result))
+
+  def testMinDistanceFeatureIsOverMax(self):
+    result = self._GetMockResult(42.)
+    self.assertEqual(10., min_distance.MinDistanceFeature(10.)(result))
+
+  def testLinearMinDistanceFeatureIsZero(self):
+    """Test that the feature returns 1 when the min_distance is 0."""
+    result = self._GetMockResult(0.)
+    self.assertEqual(1., min_distance.LinearMinDistanceFeature()(result))
+
+  def testLinearMinDistanceFeatureMiddling(self):
+    """Test that the feature returns middling scores for middling distances."""
+    result = self._GetMockResult(42.)
+    self.assertEqual(
+        (self._maximum - 42.) / self._maximum,
+        min_distance.LinearMinDistanceFeature()(result))
+
+  def testLinearMinDistanceFeatureIsOverMax(self):
+    """Test that the feature returns 0 when the min_distance is too large."""
+    result = self._GetMockResult(self._maximum + 1)
+    self.assertEqual(0., min_distance.LinearMinDistanceFeature()(result))
+
+    result = self._GetMockResult(42.)
+    self.assertEqual(0., min_distance.LinearMinDistanceFeature(10.)(result))
+
+  def testSquaredMinDistanceFeatureMiddling(self):
+    """Test that the squared feature also returns middling values."""
+    result = self._GetMockResult(42.)
+    self.assertEqual(
+        ((self._maximum - 42.) / self._maximum)**2,
+        min_distance.SquaredMinDistanceFeature()(result))