Telemetry: Sanitize a key and add description to speedindex metrics

tools/perf/metrics/speedindex.py

 # Copyright 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.
 
 import collections
 
 from metrics import Metric
 from telemetry.core import bitmap
 from telemetry.value import scalar
 
@@ skipping 35 matching lines @@
     self._impl.Stop(tab)
 
   # Optional argument chart_name is not in base class Metric.
   # pylint: disable=W0221
   def AddResults(self, tab, results, chart_name=None):
     """Calculate the speed index and add it to the results."""
     index = self._impl.CalculateSpeedIndex(tab)
     # Release the tab so that it can be disconnected.
     self._impl = None
     results.AddValue(scalar.ScalarValue(
-        results.current_page, '%s.speed_index' % chart_name, 'ms', index))
+        results.current_page, '%s_speed_index' % chart_name, 'ms', index,
+        description='Speed Index. This focuses on time when visible parts of '
+                    'page are displayed and shows the time when the '
+                    'first look is "almost" composed. If the contents of the '
+                    'testing page are composed by only static resources, load '
+                    'time can measure more accurately and speed index will be '
+                    'smaller than the load time. On the other hand, If the '
+                    'contents are composed by many XHR requests with small '
+                    'main resource and javascript, speed index will be able to '
+                    'get the features of performance more accurately than load '
+                    'time because the load time will measure the time when '
+                    'static resources are loaded. If you want to get more '
+                    'detail, please refer to http://goo.gl/Rw3d5d. Currently '
+                    'there are two implementations: for Android and for '
+                    'Desktop. The Android version uses video capture; the '
+                    'Desktop one uses paint events and has extra overhead to '
+                    'catch paint events.'))
 
   def IsFinished(self, tab):
     """Decide whether the timeline recording should be stopped.
 
     When the timeline recording is stopped determines which paint events
     are used in the speed index metric calculation. In general, the recording
     should continue if there has just been some data received, because
     this suggests that painting may continue.
 
     A page may repeatedly request resources in an infinite loop; a timeout
@@ skipping 231 matching lines @@
     frame = paint_event.args['frameId']
     return (frame,) + GetBox(paint_event.args['data']['clip'])
 
   def _GroupEventByRectangle(self, paint_events):
     """Group all paint events according to the rectangle that they update."""
     result = collections.defaultdict(list)
     for event in paint_events:
       assert event.name == 'Paint'
       result[self._GetRectangle(event)].append(event)
     return result