Inconsistent RBNF Data?

Cameron Dutro cameron at
Wed Nov 9 11:11:27 CST 2016

Hey Steven et al.,

This turned out to be my fault. I had two different versions of ICU on my
classpath, one recent and one quite old. I *thought* the newer one was
loaded but the older one took precedence because it occurred earlier in the
classpath and caused my script to generate invalid test cases.

As you suggested I wrote a small bit of Java code to try and reproduce the
problem, which to my surprise produced the correct result.

Apologies for dragging everyone into this! Thank you all for your help :)


On Tue, Nov 8, 2016 at 3:16 PM, Steven R. Loomis <srl at>

> So- can you reproduce the issue with ICU4C or ICU4J of a certain version?
> There’s an API to request the CLDR version.  In ICU4C you can use the
> ‘icuinfo’ app  or ulocdata_getCLDRVersion(), in J you can do ‘java –jar
> icu4j.jar’ or LocaleData.getCLDRVersion()
> El 11/8/16 12:20 PM, "CLDR-Users en nombre de Cameron Dutro" <
> cldr-users-bounces at en nombre de cameron at>
> escribió:
> Ah right, I forgot to mention the version of ICU. I'm using v57.1 which I
> thought was the version that corresponds to CLDR v29.
> The source code is actually Ruby code (running on JRuby). You can see the
> code in question here
> <>
> .
> Steven, it looks like that changeset was submitted 3 years ago, but isn't
> reflected in v29 or v30 of CLDR (but appears to have made it into ICU
> somehow).
> Thanks for your help!
> -Cameron
> On Tue, Nov 8, 2016 at 11:27 AM, Steven R. Loomis <srl at>
> wrote:
>> It can be helpful give some ICU source code, and which version is being
>> used.
>> But probably relevant is>> perhaps you are comparing an ICU older than this commit?
>> -s
>> El 11/8/16 10:43 AM, "CLDR-Users en nombre de Cameron Dutro" <
>> cldr-users-bounces at en nombre de cameron at>
>> escribió:
>> Hey everyone,
>> I'm running into a strange inconsistency between ICU's output and the
>> data available in CLDR when formatting numbers using RBNF rules.
>> One specific example is the spellout-cardinal-feminine rule set in
>> Spanish. In CLDR v30
>> <>
>> and v29
>> <>,
>> the rule for 101 is "ciento" which is incorrect for the feminine case. ICU
>> however formats feminine spellouts correctly by using "cienta."
>> Where in the world is ICU getting its data? Why does it appear as if ICU
>> isn't actually using the currently available CLDR data?
>> Thanks for your help,
>> -Cameron
>> _______________________________________________ CLDR-Users mailing list
>> CLDR-Users at unicode.org
> _______________________________________________ CLDR-Users mailing list
> CLDR-Users at
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the CLDR-Users mailing list