Do geomean computation on filtered common keys

When we have 2 sets of statistics that are not equal

Change-Id: I3811b216f6dc8503df0ba22bb8cdf2bacfbe9009

2 files changed, 20 insertions, 9 deletions

diff --git a/compare.py b/compare.py
index a7f27a7..3d1a780 100755
--- a/compare.py
+++ b/compare.py
@@ -194,11 +194,12 @@ if __name__ == "__main__":
               print_extended=args.print_extended,
               order_by_diff=args.order_by_diff,
               filter_stats_warnings=args.output_for_linaro_automation)
-    if utils.HaveSameKeys(res_1, res_2):
-        utils_stats.ComputeAndPrintRelationGeomean(
-            utils.Unflatten(res_1),
-            utils.Unflatten(res_2),
-            args.print_extended >= print_extended_raw_data)
-    else:
-        utils.Info("Not comparing the geomeans because the two result sets "
-                   "have different keys.")
+
+    if not utils.HaveSameKeys(res_1, res_2):
+        utils.Warning("Computing geomean on a subset of statistics which only " \
+                      "includes keys common to both datasets.")
+        res_1, res_2 = utils.KeepSameKeys(res_1, res_2)
+    utils_stats.ComputeAndPrintRelationGeomean(
+        utils.Unflatten(res_1),
+        utils.Unflatten(res_2),
+        args.print_extended >= print_extended_raw_data)
diff --git a/tools/utils.py b/tools/utils.py
index 77d5965..409ee47 100644
--- a/tools/utils.py
+++ b/tools/utils.py
@@ -416,7 +416,17 @@ def Flatten(data, key=''):
 
     return res
 
-
+def KeepSameKeys(data_1, data_2):
+    for stats_type in data_1:
+        keys_1, keys_2 = data_1[stats_type].keys(), data_2[stats_type].keys()
+        diff_keys = keys_1 ^ keys_2
+        for k in list(keys_1):
+            if k in diff_keys:
+                del data_1[stats_type][k]
+        for k in list(keys_2):
+            if k in diff_keys:
+                del data_2[stats_type][k]
+    return data_1, data_2
 
 def HaveSameKeys(data_1, data_2):
     if IsDictionary(data_1) and IsDictionary(data_2):