Change actions.txt to actions.xml

PRESUBMIT.py

Index: PRESUBMIT.py
diff --git a/PRESUBMIT.py b/PRESUBMIT.py
index 1e6526f26ea9c9dca0f46179f948e5a14216afed..0e7e9dc684a2f4588f1830ef4248da64cadaec1c 100644
--- a/PRESUBMIT.py
+++ b/PRESUBMIT.py
@@ -1014,13 +1014,13 @@ def _CheckCygwinShell(input_api, output_api):
 
 def _CheckUserActionUpdate(input_api, output_api):
   """Checks if any new user action has been added."""
-  if any('chromeactions.txt' == input_api.os_path.basename(f) for f in
+  if any('actions.xml' == input_api.os_path.basename(f) for f in
          input_api.LocalPaths()):
-    # If chromeactions.txt is already included in the changelist, the PRESUBMIT
-    # for chromeactions.txt will do a more complete presubmit check.
+    # If actions.xml is already included in the changelist, the PRESUBMIT
+    # for actions.xml will do a more complete presubmit check.
     return []
 
-  with open('tools/metrics/actions/chromeactions.txt') as f:
+  with open('tools/metrics/actions/actions.xml') as f:
     current_actions = f.read()
 
   file_filter = lambda f: f.LocalPath().endswith(('.cc', '.mm'))
@@ -1030,12 +1030,12 @@ def _CheckUserActionUpdate(input_api, output_api):
       match = input_api.re.search(action_re, line)
       if match:
         for action_name in match.groups():
-          name_pattern = r'\t%s\n' % action_name
+          name_pattern = r'\"%s\"' % action_name
           if name_pattern not in current_actions:

[Ilya Sherman, 2014/03/03 23:17:34]

Hmm, it would probably be more efficient to parse the XML file and build a
single set of action names, rather than doing a huge string search each time --
no?

[yiyaoliu, 2014/03/04 15:31:07]

Agree that it would be quicker if a search needed. I put the parsing inside the
loop to prevent an unnecessary parse when the code does not touch
UserMetricsAction function, which should be most of the cases. 

In this way, the for loop becomes bigger, and a new dependency minidom is
introduced. Not sure if the previous way or the current way is better. Redirect
to Alexei to make the decision. 

[Alexei Svitkine (slow), 2014/03/04 15:32:13]

I'm curious as to the answer here as well. Could you try both approaches and
report what the perf looks like? (Parsing the XML file will also be more correct
since it will ensure the string appears as an action entry, rather than e.g. in
a description somewhere.)

[yiyaoliu, 2014/03/04 18:02:54]

I tried 0, 1, 2, 4 invocations of UserMetricsAction. 
There's no very big difference. 
It seems that when there's 1 invocation, parsing xml takes more time (about 2.7s
vs 2.5s), and when there's more invocations, parxing xml takes less time(about
2.8s vs 3.1s). 

(I highly doubt how representative my experiment result is, for the exactly same
path, the presubmit could take 2,4s or 3.2s. I could only say there's no very
big difference when the number of invocation to UserMetricsAction is small.)

Besides, parsing XML is truly more accurate. Let's go with the current parsing
XML version? 

             return [output_api.PresubmitPromptWarning(
               'File %s line %d: %s is missing in '
-              'tools/metrics/actions/chromeactions.txt. Please run '
-              'tools/metrics/actions/extract_actions.py --hash to update.'
+              'tools/metrics/actions/actions.xml. Please run '
+              'tools/metrics/actions/extract_actions.py to update.'
               % (f.LocalPath(), line_num, action_name))]
   return []