Add run_time_meta arge to specify values to use in graph

This commit adds a new flag to the run_time_meta command in
subunit2sql-graph. You give a comma seperated list to the cli and
it will only graph the box and whiskers for runs which have metadata
value that is in the list provided. This is useful if you only want
to view a subset of values grouped under a key.

Change-Id: Ic6cadf727ad0421bac93905bc078b2c6a3412b1a

doc/source/graph.rst

@@ -137,6 +137,27 @@ will generate a graph like:
 
 .. image:: graph-dailycount.png
 
+Run Time Grouped By Run Metadata
+--------------------------------
+
+This graph generates a box and whiskers plot to show the distribution of run
+times for runs with each plot grouped by values to a provided run_metadata key
+
+For example, running something like::
+
+  subunit2sql-graph --title "Run times by Job Name" --database-connection mysql://test:test@localhost/subunit2sql -o test.png run_time_meta build_name --filter_list gate-tempest-dsvm-neutron-full-ssh,gate-tempest-dsvm-full-ubuntu-xenial,gate-tempest-dsvm-full-ubuntu-trusty,gate-tempest-dsvm-py35-ubuntu-xenial
+
+will generate a graph like:
+
+.. image:: graph-run_time_meta.png
+   :width: 115%
+
+It's also worth noting the --filter_list argument used in the command above.
+In some cases, especially larger data sets, there are too many distinct values
+for a given run_metadata key to make the graph useful. To workaround this you
+specify a list of values that will be used to filter the output so that only
+matches of that key will be in the output.
+
 subunit2sql-graph plugin interface
 ==================================
 

subunit2sql/analysis/run_time_meta.py

@@ -28,10 +28,16 @@ matplotlib.style.use('ggplot')
 def set_cli_opts(parser):
     parser.add_argument('metadata_key',
                         help="The run_metadata key to group the runs by")
+    parser.add_argument('--filter_list', '-f',
+                        help='A comma seperated list of values to use')
 
 
 def generate_series():
     session = api.get_session()
+    if CONF.command.filter_list:
+        filter_list = CONF.command.filter_list.split(',')
+    else:
+        filter_list = None
     if CONF.start_date:
         start_date = datetime.datetime.strptime(CONF.start_date, '%Y-%m-%d')
     else:
@@ -43,8 +49,13 @@ def generate_series():
     run_times = api.get_run_times_grouped_by_run_metadata_key(
         CONF.command.metadata_key, start_date=start_date,
         stop_date=stop_date, session=session)
-    df = pd.DataFrame(dict(
-        [(k, pd.Series(v)) for k, v in run_times.iteritems()]))
+    if not filter_list:
+        df = pd.DataFrame(dict(
+            [(k, pd.Series(v)) for k, v in run_times.iteritems()]))
+    else:
+        df = pd.DataFrame(dict(
+            [(k, pd.Series(v)) for k, v in run_times.iteritems()
+             if k in filter_list]))
     if not CONF.title:
         title = "Run aggregate run time grouped by metadata"
     else: