UMA metrics - number of accounts per profile and account reconciler.

components/signin/core/browser/signin_metrics.cc

Index: components/signin/core/browser/signin_metrics.cc
diff --git a/components/signin/core/browser/signin_metrics.cc b/components/signin/core/browser/signin_metrics.cc
new file mode 100644
index 0000000000000000000000000000000000000000..4e2c86487d383006a4f6fb1594bb85c6871f4a69
--- /dev/null
+++ b/components/signin/core/browser/signin_metrics.cc
@@ -0,0 +1,33 @@
+// Copyright (c) 2014 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.
+
+#include "components/signin/core/browser/signin_metrics.h"
+
+#include "base/logging.h"
+#include "base/metrics/histogram.h"
+#include "base/metrics/user_metrics.h"
+
+namespace signin_metrics {
+
+void LogSigninAccountReconciliation(int total_number_accounts,
+                                    int count_added_to_cookie_jar,
+                                    int count_added_to_token,
+                                    bool primary_accounts_same,
+                                    bool is_first_reconcile) {
+  UMA_HISTOGRAM_COUNTS_100("Profile.NumberOfAccountsPerProfile",
+                           total_number_accounts);
+  // We want to include zeroes in the counts of added accounts to easily
+  // capture _relatively_ how often we merge accounts.
+  if (is_first_reconcile || count_added_to_cookie_jar ||
+      count_added_to_token) {

[Alexei Svitkine (slow), 2014/05/08 14:39:29]

It seems kind of weird to log a 0 in some cases and not log at all in others.
Can you convince me why this makes sense to do?

[Mike Lerman, 2014/05/08 18:23:14]

My concern is that every execution of the account reconcilor will log data, but
the majority of "routine" executions will pass in zeroes, and I don't want to
over-log them. But I am interested in them on the first pass.

I think I'll split this into two sets of histograms - one that tracks the
"first" execution of the reconcilor, and one that tracks all others. That way I
can interpret the data as necessary.

+    UMA_HISTOGRAM_COUNTS_100("Signin.ReconcilerAddedToCookieJar",
+                             count_added_to_cookie_jar);
+    UMA_HISTOGRAM_COUNTS_100("Signin.ReconcilerAddedToChrome",
+                             count_added_to_token);
+  }
+  if (primary_accounts_same)
+    UMA_HISTOGRAM_BOOLEAN("Signin.ReconcilerDifferentPrimaryAccounts", true);

[Alexei Svitkine (slow), 2014/05/08 14:39:29]

Consider logging both true and false here, so that you have a baseline to
compare against. (i.e. a count just by itself is not super meaningful vs. if you
log both cases, you can see the percentage)

[Mike Lerman, 2014/05/08 18:23:14]

I was going to use the total number of count in
Profile.NumberOfAccountsPerProfile as a baseline, but I can just log both
scenarios here.

+}
+
+}  // namespace signin_metrics