Measuring data use of different ModelTypes in Sync Service.

sync/util/data_type_histogram.h

 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #ifndef SYNC_UTIL_DATA_TYPE_HISTOGRAM_H_
 #define SYNC_UTIL_DATA_TYPE_HISTOGRAM_H_
 
 #include "base/metrics/histogram.h"
+#include "base/metrics/sparse_histogram.h"
 #include "base/time/time.h"
 #include "sync/internal_api/public/base/model_type.h"
 
+// This macro adds |value| to |sample| bucket of histogram |name|. |value|
+// should be greater or equal to 1 and |name| can be variable. DataTypes are
+// mapped to proper |sample| bucket by using ModelTypeToHistogramInt() function.
+// So different DataTypes play the role of different buckets in this histogram.
+#define SYNC_RECORD_DATATYPE_BIN(name, sample, value)     \

[sclittle, 2015/08/19 22:41:35]

Does this have to be a macro? Would this be better as a helper function?

[amohammadkhan, 2015/08/20 00:20:13]

Done.

[Nicolas Zea, 2015/08/20 17:59:50]

Actually, if I recall how histograms work you might run into problems having
this be a function. The reason we used macros below is that the normal histogram
macros actually instantiate a static variable when you invoke them. As such, by
having it be a macro, you ensure that the different places you invoke the macro
get different static variables. If you have it be a function, you wind up with a
single static variable, which means your different histograms will
cross-contaminate each other.

Take a look at histogram_macros.h for more detail.

[amohammadkhan, 2015/08/20 18:12:51]

I see your point. But I used it without static pointers, so the histogram name
can vary without making errors. (Or it can be implemented as a function)

+  do {                                                                  \
+    base::HistogramBase* histogram = base::SparseHistogram::FactoryGet( \
+        name, base::HistogramBase::kUmaTargetedHistogramFlag);          \
+    histogram->AddCount(sample, value);                                 \
+  } while (0)
+
 // For now, this just implements UMA_HISTOGRAM_LONG_TIMES. This can be adjusted
 // if we feel the min, max, or bucket count amount are not appropriate.
 #define SYNC_FREQ_HISTOGRAM(name, time) UMA_HISTOGRAM_CUSTOM_TIMES( \
     name, time, base::TimeDelta::FromMilliseconds(1), \
     base::TimeDelta::FromHours(1), 50)
 
 // Helper macro for datatype specific histograms. For each datatype, invokes
 // a pre-defined PER_DATA_TYPE_MACRO(type_str), where |type_str| is the string
 // version of the datatype.
 //
@@ skipping 116 matching lines @@
       case ::syncer::PROXY_TABS: \
         PER_DATA_TYPE_MACRO("Tabs"); \
         break; \
       default: \
         NOTREACHED() << "Unknown datatype " \
                      << ::syncer::ModelTypeToString(datatype);  \
     } \
   } while (0)
 
 #endif  // SYNC_UTIL_DATA_TYPE_HISTOGRAM_H_