[Findit] Modify build_failure to prepare for new UI.

appengine/findit/handlers/build_failure.py

Index: appengine/findit/handlers/build_failure.py
diff --git a/appengine/findit/handlers/build_failure.py b/appengine/findit/handlers/build_failure.py
index d5097533218ae9d5c6e40c651003ed2b0cf3c288..ab5a3036ec12dd7a3a8d8cb4827f2f71019960d9 100644
--- a/appengine/findit/handlers/build_failure.py
+++ b/appengine/findit/handlers/build_failure.py
@@ -2,6 +2,8 @@
 # Use of this source code is governed by a BSD-style license that can be
 # found in the LICENSE file.
 
+import copy
+from collections import defaultdict

[stgao, 2016/03/25 23:10:51]

nit: order.

[chanli, 2016/03/25 23:44:07]

Done.

 from datetime import datetime
 import os
 
@@ -9,6 +11,9 @@ from google.appengine.api import users
 
 from base_handler import BaseHandler
 from base_handler import Permission
+from handlers import handlers_util
+from handlers.handlers_util import NO_TRY_JOB_REASON_MAP
+from model import wf_analysis_status
 from model.wf_analysis import WfAnalysis
 from model.wf_analysis_result_status import RESULT_STATUS_TO_DESCRIPTION
 from waterfall import build_failure_analysis_pipelines
@@ -46,6 +51,206 @@ def _GetTriageHistory(analysis):
   return triage_history
 
 
+def _GetOrganizedAnalysisResultBySuspectedCL(analysis_result):
+  """Group tests it they have the same suspected CLs."""
+
+  organized_results = defaultdict(list)
+
+  if not analysis_result:
+    return organized_results
+
+  for failure in analysis_result.get('failures', []):
+    step_name = failure['step_name']
+    supported = failure.get('supported', True)
+    step_revisions_index = {}
+    organized_suspected_cls = organized_results[step_name]
+
+    if not failure.get('tests'):
+      # Non swraming.
+      shared_result = {
+          'first_failure': failure['first_failure'],
+          'last_pass': failure.get('last_pass'),
+          'supported': supported,
+          'tests': [],
+          'suspected_cls': failure['suspected_cls']
+      }
+      organized_suspected_cls.append(shared_result)
+      continue
+
+    # Swarming tests.
+    for index, cl in enumerate(failure['suspected_cls']):
+      step_revisions_index[cl['revision']] = index
+
+    # Groups tests by suspected CLs' revision.
+    # Format is as below:
+    # {
+    #     1: {
+    #         'tests': ['test1', 'test2'],
+    #         'revisions': ['rev1'],
+    #         'suspected_cls': [
+    #             # suspected cl info for rev1 at step level.
+    #         ]
+    #     },
+    #     3: {
+    #         'tests': ['test3'],
+    #         'revisions': ['rev3', 'rev2'],
+    #         'suspected_cls': [
+    #             # suspected cl info for rev2, rev3 at step level.
+    #         ]
+    #     }
+    # }
+    tests_group = defaultdict(list)
+    for index, test in enumerate(failure['tests']):
+      test_name = test['test_name']
+      found_group = False
+      revisions = set()
+      for cl in test['suspected_cls']:
+        revisions.add(cl['revision'])
+      for group in tests_group.values():
+        if revisions == set(group['revisions']):
+          group['tests'].append(test_name)
+          found_group = True
+          break
+      if not found_group:
+        group_suspected_cls = []
+        for revision in revisions:
+          group_suspected_cls.append(
+              failure['suspected_cls'][step_revisions_index[revision]])
+        tests_group[index] = {
+            'tests': [test_name],
+            'revisions': list(revisions),
+            'suspected_cls': group_suspected_cls
+        }
+
+    for index, group in tests_group.iteritems():
+      test_result = failure['tests'][index]
+      shared_result = {
+          'first_failure': test_result['first_failure'],
+          'last_pass': test_result.get('last_pass'),
+          'supported': supported,
+          'tests': group['tests'],
+          'suspected_cls': group['suspected_cls']

[stgao, 2016/03/25 23:10:51]

Is it possible that two tests failed due to the same revisions, but start
failing at different build cycles? How should we handle such a case?

[chanli, 2016/03/25 23:44:07]

If they failed due to the same revision, and failed in different cycle, it
sounds just like what I described earlier that the revision broke a former flaky
test. And they will not be grouped together in my opinion.

Is there another possible reason for this scenario?

+      }
+      organized_suspected_cls.append(shared_result)
+
+  return organized_results
+
+
+def _UpdateAnalysisResultWithTryJobInfo(

[stgao, 2016/03/25 23:10:51]

naming nit: Update -> Get?

[chanli, 2016/03/25 23:44:07]

Done.

+    organized_results, master_name, builder_name, build_number):
+  """Reorganizes analysis result and try job result by step_name and culprit.
+
+  Returns:
+    update_result (dict): A dict of classified results.
+
+    The format for those dicts are as below:
+    {
+        # A dict of results that contains both
+        # heuristic analysis results and try job results.
+        'determined_results': {
+            'step1': {
+                'results': [
+                    {
+                        'try_job':{
+                            'ref_name': 'step1',
+                            'try_job_key': 'm/b/119',
+                            'status': wf_analysis_status.ANALYZED,
+                            'try_job_url': 'url/121',
+                            'try_job_build_number': 121,
+                        },
+                        'heuristic_analysis': {
+                            'suspected_cls': [
+                                {
+                                    'build_number': 98,
+                                    'repo_name': 'chromium',
+                                    'revision': 'r98_1',
+                                    'commit_position': None,
+                                    'url': None,
+                                    'score': 5,
+                                    'hints': {
+                                        'added f98.cc (and it was in log)': 5,
+                                    },
+                                }
+                            ]
+                        }
+                        'tests': ['test1', 'test2'],
+                        'first_failure': 98,
+                        'last_pass': 97,
+                        'supported': True
+                    }
+                ]
+            }
+        },
+        # A dict of result for flaky tests.
+        'flaky_results': {...},
+        # A dict of results for all the other conditions,
+        # such as non-swarming tests or swarming rerun failed.
+        'undetermined_results': {...}
+    }
+  """
+  update_results = defaultdict(lambda: defaultdict(lambda: defaultdict(list)))
+
+  if not organized_results:
+    return update_results
+
+  try_job_info = handlers_util.GetAllTryJobResults(
+      master_name, builder_name, build_number)
+  if not try_job_info:
+    return update_results
+
+  for step_name, try_jobs in try_job_info.iteritems():
+    try_jobs = try_jobs['try_jobs']
+    step_heuristic_results = organized_results[step_name]
+    step_updated_results = update_results[step_name]['results']
+
+    # Finds out try job result index and heuristic result index for each test.
+    test_result_map = defaultdict(lambda: defaultdict(int))
+
+    for index, try_job in enumerate(try_jobs):
+      if not try_job.get('tests'):  # Compile or non-swarming.
+        test_result_map['non-swarming']['try_job_index'] = index
+        continue
+      for test_name in try_job['tests']:
+        test_result_map[test_name]['try_job_index'] = index
+    for index, heuristic_result in enumerate(step_heuristic_results):
+      if not heuristic_result.get('tests'):  # Compile or non-swarming.
+        test_result_map['non-swarming']['heuristic_index'] = index
+        continue
+      for test_name in heuristic_result['tests']:
+        test_result_map[test_name]['heuristic_index'] = index
+
+    # Group tests based on indices.
+    indices_test_map = defaultdict(list)
+    for test_name, indices in test_result_map.iteritems():
+      indices_test_map[
+          (indices['try_job_index'], indices['heuristic_index'])].append(
+              test_name)
+
+    for indices, tests in indices_test_map.iteritems():
+      try_job_result = try_jobs[indices[0]]
+      heuristic_result = step_heuristic_results[indices[1]]
+
+      final_result = {
+          'try_job': try_job_result,
+          'heuristic_analysis': {
+              'suspected_cls': heuristic_result['suspected_cls']
+          },
+          'tests': tests if tests != ['non-swarming'] else [],
+          'first_failure': heuristic_result['first_failure'],
+          'last_pass': heuristic_result['last_pass'],
+          'supported': heuristic_result['supported']
+      }
+
+      if try_job_result['status'] == handlers_util.FLAKY:
+        step_updated_results['flaky_results'].append(final_result)
+      elif try_job_result['status'] in NO_TRY_JOB_REASON_MAP.values():
+        step_updated_results['undetermined_results'].append(final_result)
+      else:
+        step_updated_results['determined_results'].append(final_result)
+
+  return update_results
+
+
 class BuildFailure(BaseHandler):
   PERMISSION_LEVEL = Permission.ANYONE
 
@@ -95,6 +300,11 @@ class BuildFailure(BaseHandler):
           build_completed=build_completed, force=force,
           queue_name=BUILD_FAILURE_ANALYSIS_TASKQUEUE)
 
+    organized_results = _GetOrganizedAnalysisResultBySuspectedCL(
+        analysis.result)
+    analysis_result = _UpdateAnalysisResultWithTryJobInfo(
+        organized_results, *build_info)
+
     data = {
         'master_name': analysis.master_name,
         'builder_name': analysis.builder_name,
@@ -108,7 +318,7 @@ class BuildFailure(BaseHandler):
         'analysis_update_time': _FormatDatetime(analysis.updated_time),
         'analysis_completed': analysis.completed,
         'analysis_failed': analysis.failed,
-        'analysis_result': analysis.result,
+        'analysis_result': analysis_result,
         'analysis_correct': analysis.correct,
         'triage_history': _GetTriageHistory(analysis),
         'show_triage_help_button': self._ShowTriageHelpButton(),