Allow benchmarks to specify running multiple metrics.

tracing/bin/run_metric

Index: tracing/bin/run_metric
diff --git a/tracing/bin/run_metric b/tracing/bin/run_metric
index 91a0eb35fb8cc0bdf1c94b0b24e8072a611e3504..3f7b4d3ed5b29ec0fbc7eddfa87444844665008f 100755
--- a/tracing/bin/run_metric
+++ b/tracing/bin/run_metric
@@ -17,16 +17,17 @@ def Main(argv):
       ['/tracing/metrics/all_metrics.html'])
   parser = argparse.ArgumentParser(
       description='Runs metrics on local traces')
-  parser.add_argument('metric_name',
-                      help=('The function name of a registered metric, NOT '
-                      'filename. Available metrics are: %s' %
+  parser.add_argument('metric_names',

[eakuefner, 2016/06/30 18:45:03]

Instead of doing this, can you make it possible to pass multiple --metric
arguments, like we do with --output-format in Telemetry [0] and implicitly take
the argument before the trace_file_or_dir as a --metric if none is specified,
like how sed will treat the string you pass as the expression to evaluate if you
don't pass -e explicitly?

[0]
https://github.com/catapult-project/catapult/blob/master/telemetry/telemetry/...

[alexandermont, 2016/07/01 21:46:00]

Done

[eakuefner, 2016/07/06 15:52:54]

It doesn't look like this is done in the latest patch set -- did you forget to
upload these changes? To be clear, rather than using commas, I would like to be
able to write:

bin/run_metric --metric=memoryMetric --metric=executionMetric mytrace.json, or
as a shorthand if I only happen to be specifying one metric:

bin/run_metric memoryMetric mytrace.json

[eakuefner, 2016/07/06 17:23:32]

Talked to Alex about this offline and read a bit more about argparse -- it turns
out that normally, arguments with - are "optional", and in this case, we don't
want it to be optional, we want it to be mandatory and variadic. However, the
nargs= functionality that argparse provides assumes such a variadic argument
will be the last one provided. Since trace_file_or_dir is always going to be a
singleton, this works for us, as long as we switch the order of the arguments,
which this CL will now do.

+                      help=('Comma separated list of function names '
+                      'of registered metrics (not filenames.) '
+                      'Available metrics are: %s' %
                       ', '.join(all_metrics)),
                       choices=all_metrics, metavar='metricName')
   parser.add_argument('trace_file_or_dir',
                       help='A trace file, or a dir containing trace files')
 
   args = parser.parse_args(argv[1:])
-  metric = args.metric_name
+  metrics = args.metric_names.split(',')
 
   if os.path.isdir(args.trace_file_or_dir):
     trace_dir = args.trace_file_or_dir
@@ -35,7 +36,7 @@ def Main(argv):
     traces = [args.trace_file_or_dir]
 
   results = {k: v.AsDict()
-      for k, v in metric_runner.RunMetricOnTraces(traces, metric).iteritems()}
+      for k, v in metric_runner.RunMetricOnTraces(traces, metrics).iteritems()}
 
   failures = []
   for trace in traces: