[Findit] Flake Checker: Get swarming task triggered by flake try jobs.

appengine/findit/waterfall/flake/process_flake_try_job_result_pipeline.py

 # Copyright 2017 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
 
 from common.pipeline_wrapper import BasePipeline
-from model.flake.flake_try_job import FlakeTryJob
+from gae_libs.http.http_client_appengine import HttpClientAppengine
 from model.flake.master_flake_analysis import DataPoint
+from waterfall import swarming_util
+
+
+def _GetSwarmingTaskIdForTryJob(report, revision, step_name):
+  """Check json output for each task and return id of the one with test result.
+  """
+  if not report:
+    return None
+
+  http_client = HttpClientAppengine()
+  task_ids = report.get('result', {}).get(revision, {}).get(

[stgao, 2017/01/12 07:43:27]

If there is only one task id, do we still have to query Swarming or Isolated
server?

If the test doesn't exist in the tested revision, should we just pick any task
id or return None?

[chanli, 2017/01/12 20:28:29]

Good point. Just updated.
If test doesn't exist, would there be swarming tasks?

[stgao, 2017/01/13 02:18:18]

Example: a test is newly added in the suspected build number. Bisecting at
commits before the test being added will still run Swarming tasks, but all
Swarming tasks won't have any iteration data for the test.

+      step_name, {}).get('step_metadata', {}).get('swarm_task_ids', [])
+
+  for task_id in task_ids:
+    output_json = swarming_util.GetIsolatedOutputForTask(task_id, http_client)
+    if output_json:
+      for data in output_json.get('per_iteration_data', []):
+        # If this task doesn't have result, per_iteration_data will look like
+        # [{}, {}, ...]
+        if data:
+          return task_id
+
+  return None
 
 
 class ProcessFlakeTryJobResultPipeline(BasePipeline):
   """A pipeline for processing a flake try job result."""
 
   # Arguments number differs from overridden method - pylint: disable=W0221
   def run(self, revision, commit_position, try_job_result, urlsafe_try_job_key,
           urlsafe_flake_analysis_key):
     """Extracts pass rate information and updates flake analysis.
 
@@ skipping 38 matching lines @@
       tries = pass_count + fail_count
       pass_rate = float(pass_count) / tries
     else:  # Test does not exist.
       pass_rate = -1
 
     data_point = DataPoint()
     data_point.commit_position = commit_position
     data_point.pass_rate = pass_rate
     data_point.try_job_id = try_job.try_job_ids[-1]
     data_point.try_job_url = try_job.flake_results[-1].get('url')
-    # TODO: Add swarming task data.
+    data_point.task_id = _GetSwarmingTaskIdForTryJob(
+        try_job.flake_results[-1].get('report'), revision, step_name)
     flake_analysis.data_points.append(data_point)