Telemetry media_measurement plus play action and tests.

tools/perf/perf_tools/media_metrics.py

+# Copyright (c) 2013 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.
+
+"""Media Metrics class injects and calls JS responsible for recording metrics.
+
+Default media metrics are collected for every media element in the page, such as
+decoded_frame_count, dropped_frame_count, decoded_video_bytes, and
+decoded_audio_bytes.
+"""
+
+import logging
+import os
+
+
+class MediaMetrics(object):
+  def __init__(self, tab):
+    # Measurement class and media actions class can both load media_metrics.js.

[nduca, 2013/07/09 23:07:19]

is this comment still true?

[shadi, 2013/07/11 00:49:13]

Oops, removed.

+    deps_found = tab.EvaluateJavaScript('window.__mediaRecorders != undefined')
+    if not deps_found:
+      with open(
+          os.path.join(os.path.dirname(__file__), 'media_metrics.js')) as f:
+        js = f.read()
+        tab.ExecuteJavaScript(js)
+    self.tab = tab
+

[nduca, 2013/07/09 23:07:19]

you should have a Start/Stop/Add in style of memory metrics, per the comments
before about the distinction between injection vs recording

+  def ReportCollectedMetrics(self, results):

[nduca, 2013/07/09 23:07:19]

your naming shoudl conform to memory_metrics.py

[shadi, 2013/07/11 00:49:13]

Done.

+    """Reports all recorded metrics as Telemetry perf results.
+
+    Metrics are recorded for each media element in the page. Each metric result
+    has media info + metric values as a map, for example:
+    Media metric = {
+      'info': {
+          'src': 'file.webm',
+          'id': 'video1',
+          ...
+      },
+      'metrics': {
+          'ttp': [120, 'ms'],
+          'decoded_bytes': [13233, 'bytes'],
+          ...
+      }
+    }
+    """
+    media_metrics = self.tab.EvaluateJavaScript('window.__getAllMetrics()')
+    for media_metric in media_metrics:
+      AddResult(media_metric, results)
+
+
+def AddResult(media_metric, results):
+  """Adds the media metric result to global results, needed for graphing."""
+  info = media_metric['info']
+  metrics = media_metric['metrics']
+  for metric in metrics:

[nduca, 2013/07/09 23:07:19]

i'd love this to be a bit more strongly typed. Where you explictly add the
results one by one that you have, rather than just blindly converting the stuff
from js to the right hting. It may be more typing but its going to be a LOT
easier to read.

[shadi, 2013/07/11 00:49:13]

The idea is that we don't know how many metrics have been recorded at this time.
That's why I look through all metrics recorded and report them.

Are you suggesting keeping a hard coded list of all possible metrics that can be
collected here and searching for each if recorded or not?  Then whenever we want
to add a new metric, we need to update both the media_metrics.js and
media_metrics.py vs now you can simply add a new metric in the media_metrics.js
following the same metric format 'name'" [value, unit].

+    value, units = GetValueUnit(metrics[metric])
+    if not units:
+      logging.error('Missing units value for media metric %s.', metric)
+      continue
+    trace = GetTraceName(info)
+    results.Add(trace, units, value, chart_name=metric, data_type='default')
+
+
+def GetValueUnit(metric):
+  """Returns a (value, units) pair stored in a metric if available."""
+  if isinstance(metric, (list, tuple)):

[nduca, 2013/07/09 23:07:19]

per note above, rather this wasn't here and you hand typed out a lot of
addResult lines

[shadi, 2013/07/11 00:49:13]

See earlier comment.

+    if len(metric) > 1:
+      return (metric[0], metric[1])
+    else:
+      return (metric[0], None)
+  return (metric, None)
+
+
+def GetTraceName(info):

[nduca, 2013/07/09 23:07:19]

per note above, rather this wasn't here and you hand typed out a lot of
addResult lines

[shadi, 2013/07/11 00:49:13]

See earlier comment.

+  """Returns a media trace name based on the metric info map.
+
+  In order, returns the first available:
+    - 'id' key value.
+    - file name based on 'src' key value.
+    - the first key value available.
+    - 'no_trace'.
+  """
+  if 'id' in info:
+    return info['id']
+  elif 'src' in info:
+    # Return only file name in src.
+    src = info['src']
+    return src[src.rfind('/') + 1:]
+  else:
+    for key in info:
+      return info[key]
+    return 'no_trace'