tools/metrics/histograms/histograms.xml

Index: tools/metrics/histograms/histograms.xml
diff --git a/tools/metrics/histograms/histograms.xml b/tools/metrics/histograms/histograms.xml
index 57daa7befac55f5d60833ee420c28fd6291f8934..ca26d1b2e18b6aad886e6d735d968e8f59fe8a79 100644
--- a/tools/metrics/histograms/histograms.xml
+++ b/tools/metrics/histograms/histograms.xml
@@ -12918,6 +12918,16 @@ http://cs/file:chrome/histograms.xml - but prefer this file for new entries.
   </summary>
 </histogram>
 
+<histogram name="History.TopHostsVisitsByRank" units="rank">
+  <owner>twifkak@chromium.org</owner>
+  <summary>
+    Page visits to each of a user's top 50 hosts. Visits to all other hosts go
+    into the 51st bucket. Android-only. The list of top hosts is computed
+    approximately nightly. Only count the page visit if it came from user

[Mark P, 2015/06/12 21:09:54]

after "nightly.", you may want to add another sentence saying that this means
that the same bucket can represent multiple hosts if the top hosts section was
recalculated in between emissions to this histogram.  This also means that if
you look at the UMA dashboard (which summarizes counts over a day) that likely a
day's count covers a recalc.  Something about this should be alluded to in the
description.

[twifkak, 2015/06/16 20:25:24]

Done.

+    browsing and only count it once when cycling through a redirect chain.

[Mark P, 2015/06/12 21:09:54]

I suggest removing "came from user browsing".  Look at the alternatives (sync,
extensions, etc.), no one would expect this to count visits caused by those.  In
other words, I think that phrase adds more confusing than it clarifies.

Also, you mention this code fragment:
"visit_source == SOURCE_BROWSED && (transition & ui::PAGE_TRANSITION_CHAIN_END)"
which implies that you only count it _at most_ once when cycling through a
redirect chain,
possibly not at all if it's not at the end of the chain.
Are you sure your statement "count it once" is accurate?

[twifkak, 2015/06/16 20:25:24]

Done.
Are you implying that some redirect chains don't end in a
PAGE_TRANSITION_CHAIN_END? Under what circumstances?

[Mark P, 2015/06/16 20:54:31]

No.  I was thinking from the previous text that you were counting a visit to a
host every time it was visited (even on a redirect chain), just making sure that
if a host was hit multiple times on a chain you'd only count it once.
With the new text, what you're actually doing is clear.  (It wasn't what I
thought before.)

[twifkak, 2015/06/17 00:05:23]

Acknowledged.

+  </summary>
+</histogram>
+
 <histogram name="History.TopSitesRecoveredPercentage" units="%">
   <owner>Please list the metric's owners. Add more owner tags as needed.</owner>
   <summary>