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 9daef964fbafc3ad717b26212140a15dbc37e3fd..392d8cdb33f2a12f380d8553f77c6d5eb109f2fc 100644
--- a/tools/metrics/histograms/histograms.xml
+++ b/tools/metrics/histograms/histograms.xml
@@ -59082,7 +59082,9 @@ http://cs/file:chrome/histograms.xml - but prefer this file for new entries.
   <int value="110800" label="xkb:latam::spa">Latin American keyboard</int>
   <int value="110801" label="xkb:lt::lit">Lithuanian keyboard</int>
   <int value="110802" label="xkb:lv:apostrophe:lav">Latvian keyboard</int>
-  <int value="110900" label="xkb:mn::mon">Mongolian keyboard</int>
+  <int value="110900" label="xkb:mk::mkd">Macedonian keyboard</int>
+  <int value="110901" label="xkb:mn::mon">Mongolian keyboard</int>

[Ilya Sherman, 2015/07/16 04:49:47]

Hmm, why did this value change?

[Shu Chen, 2015/07/16 05:13:35]

The ID sequence matters, because the input method ID for statistics has some
rules:
https://code.google.com/p/chromium/codesearch#chromium/src/chrome/browser/chr...

[Ilya Sherman, 2015/07/16 06:25:31]

You should find a mapping that is more stable than what you're using.  It's not
appropriate for bucket value "110900" to mean "Mongolian Keyboard" for some
versions of Chrome, and "Macedonian keyboard" for others.  Otherwise, you'll end
up with misleading data on the histograms dashboard, which will always use the
latest label definition from histograms.xml.

[Shu Chen, 2015/07/16 06:38:44]

Do you have any idea of making the mapping stable?
I am not sure whether UMA histograms support some kinds of hash values?

The input method ID histograms are used for short-terms usage findings.
So usually we are not interested in long term trend.
We often need to know something like "for now how many 7-day users for this
input method".
Considering we won't add new input methods from day to day, I think we can
endure such kind of instability. 

[Ilya Sherman, 2015/07/16 06:47:46]

The easiest thing to do would be to just compute a hash of the keyboard name --
that would be stable.  Any hashing algorithm that tends to avoid collisions in a
32-bit output space ought to be fine.

Even if you don't anticipate tracking long-term trends, I bet that you often
look at data collected from multiple versions of Chrome at once -- the current
stable, beta, dev, and canary channels, as well as the sizable population of
users who are stuck on older builds.  For this reason, it's really fairly
important to maintain the stability of the histogram bucket value semantics,
even if you only look at short-term data.

+  <int value="110902" label="xkb:mt::mlt">Maltese keyboard</int>
   <int value="111000" label="xkb:no::nob">Norwegian keyboard</int>
   <int value="111200" label="xkb:pl::pol">Polish keyboard</int>
   <int value="111201" label="xkb:pt::por">Portuguese keyboard</int>
@@ -59105,9 +59107,10 @@ http://cs/file:chrome/histograms.xml - but prefer this file for new entries.
   <int value="111705" label="xkb:us:altgr-intl:eng">US Extended keyboard</int>
   <int value="111706" label="xkb:us:colemak:eng">US Colemak keyboard</int>
   <int value="111707" label="xkb:us:dvorak:eng">US Dvorak keyboard</int>
-  <int value="111708" label="xkb:us:intl:eng">US International keyboard</int>
-  <int value="111709" label="xkb:us:intl:nld">US International keyboard</int>
-  <int value="111710" label="xkb:us:intl:por">US International keyboard</int>
+  <int value="111708" label="xkb:us:dvp:eng">US Dvorak Programmer keyboard</int>
+  <int value="111709" label="xkb:us:intl:eng">US International keyboard</int>
+  <int value="111710" label="xkb:us:intl:nld">US International keyboard</int>
+  <int value="111711" label="xkb:us:intl:por">US International keyboard</int>

[Ilya Sherman, 2015/07/16 04:49:47]

Likewise, why did these values change?

[Shu Chen, 2015/07/16 05:13:35]

Ditto. US Dvorak Programmer keyboard was added in cl
https://codereview.chromium.org/1113683003, but forgot to change the
histograms.xml.

   <int value="209700" label="zh-hant-t-i0-array-1992">Array input method</int>
   <int value="209900" label="zh-hant-t-i0-cangjie-1987">
     Cangjie input method