Skip test earlier if metric is not found in compare.py

[fhahn]

Summary

This speeds up processing for inputs with lots of missing metrics, by
avoiding adding them to names.

[serge-sans-paille]

utils/compare.py

37

I don't know if it's relevant, but a missing metrics *and* empty or duplicate name would no longer raise an Exception; it's probably better to continue right after the sanity checks?

[fhahn]

Move metric check after name validation, thanks!

[fhahn]

utils/compare.py

37

Yep, it should be after the name validation, thanks

[serge-sans-paille]

LGTM then, please get another feedback though, I'm not comfortable with that code area

[fhahn]

ping

[fhahn]

Any chance you could have a quick look Jessica?

[fhahn]

ping

[paquette]

LGTM

utils/compare.py

def read_lit_json(filename):
     columnindexes = {}
     names = set()
     info_columns = ['hash']
     # Pass1: Figure out metrics (= the column index)
     if 'tests' not in jsondata:
         print("%s: Could not find toplevel 'tests' key")
         sys.exit(1)
     for test in jsondata['tests']:
+        if "metrics" not in test:
+            print("Warning: '%s' has no metrics, skipping!" % test['name'])
+            continue
         name = test.get("name")
         if name is None:
             sys.stderr.write("Error: Found unnamed test\n" % name)
             sys.exit(1)
         if name in names:
             sys.stderr.write("Error: Multiple tests with name '%s'\n" % name)
             sys.exit(1)
         names.add(name)
-        if "metrics" not in test:
-            print("Warning: '%s' has No metrics!" % test['name'])
-            continue
         for name in test["metrics"].keys():
             if name not in columnindexes:
                 columnindexes[name] = len(columns)
                 columns.append(name)
         for name in test.keys():
             if name not in columnindexes and name in info_columns:
                 columnindexes[name] = len(columns)
                 columns.append(name)
 
     # Pass2 actual data construction
     nan = float('NaN')
     data = []
     testnames = []
     for test in jsondata['tests']:
+        if "metrics" not in test:
+            continue
         name = test['name']
         if 'shortname' in test:
             name = test['shortname']
         testnames.append(name)
 
         datarow = [nan] * len(columns)
-        if "metrics" in test:
-            for (metricname, value) in test['metrics'].items():
-                datarow[columnindexes[metricname]] = value
         for (name, value) in test.items():
             index = columnindexes.get(name)
             if index is not None:
                 datarow[index] = test[name]
         data.append(datarow)
     index = pd.Index(testnames, name='Program')
     return pd.DataFrame(data=data, index=index, columns=columns)