[Findit] Group failures by culprit and send notification to codereview.

appengine/findit/model/wf_culprit.py

Index: appengine/findit/model/wf_culprit.py
diff --git a/appengine/findit/model/wf_culprit.py b/appengine/findit/model/wf_culprit.py
new file mode 100644
index 0000000000000000000000000000000000000000..f328a50a70d6ac08b816b8277eb495feb4e5eabe
--- /dev/null
+++ b/appengine/findit/model/wf_culprit.py
@@ -0,0 +1,43 @@
+# 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 google.appengine.ext import ndb
+
+
+class WfCulprit(ndb.Model):
+  """Represents a culprit that causes a group of failures on Chromium waterfall.
+
+  'Wf' is short for waterfall.
+  """
+
+  # Repo or project name of the culprit, eg: chromium, etc.
+  repo_name = ndb.StringProperty()
+
+  # The Git hash revision of the culprit.
+  revision = ndb.StringProperty(indexed=False)
+
+  # When this culprit was first found.
+  found_time = ndb.DateTimeProperty()

[lijeffrey, 2016/06/21 00:07:34]

do we want to index found_time should we ever want to query for historical
WfCulprit entities? I imagine this could eventually be used in conjunction with
auto revert and we may want metrics for that

It also doesn't look like found_time is being set anywhere

[stgao, 2016/06/21 15:14:29]

We do want to query for historical data to know which and how many culprits we
have sent notification to code-review, and later trigger auto-revert when
supported.
Ooops, good catch. Changed to notification_time instead.

+
+  # The status of notification delivery.

[chanli, 2016/06/21 17:50:31]

Do you want to list the statuses here? For reference?

[stgao, 2016/06/24 16:10:28]

Done.

+  notification_status = ndb.IntegerProperty()
+
+  # The list of builds in which the culprit caused some breakage.
+  failed_builds = ndb.JsonProperty(indexed=False, default=[])
+
+  @property
+  def project_name(self):  # pragma: no cover

[chanli, 2016/06/21 17:50:31]

Why don't just use repo_name? 

[stgao, 2016/06/24 16:10:28]

This is for UI.

I'm more leaning to make repo_name the full path of repo checkout dir (i.e.
third_party/pdfium), while project name as the short name (i.e. pdfium).

+    return self.repo_name
+
+  @classmethod
+  def _CreateKey(cls, repo_name, revision):  # pragma: no cover
+    return ndb.Key(cls.__name__, '%s/%s' % (repo_name, revision))
+
+  @classmethod
+  def Create(cls, repo_name, revision):  # pragma: no cover
+    return cls(key=cls._CreateKey(repo_name, revision))
+
+  @classmethod
+  def Get(cls, repo_name, revision):  # pragma: no cover
+    return cls._CreateKey(repo_name, revision).get()