[NTP Snippets] Report the share of the UI lang vs. the top non-UI lang

components/ntp_snippets/remote/ntp_snippets_fetcher.cc

Index: components/ntp_snippets/remote/ntp_snippets_fetcher.cc
diff --git a/components/ntp_snippets/remote/ntp_snippets_fetcher.cc b/components/ntp_snippets/remote/ntp_snippets_fetcher.cc
index 8e577878ffa620df986580aaed386f777fdc85f4..4055ffcd2756512def0c1bab6f066e31733d804b 100644
--- a/components/ntp_snippets/remote/ntp_snippets_fetcher.cc
+++ b/components/ntp_snippets/remote/ntp_snippets_fetcher.cc
@@ -545,6 +545,15 @@ void NTPSnippetsFetcher::SetUpCommonFetchingParameters(
   for (const LanguageModel::LanguageInfo& info : top_languages) {
     if (info.language_code != params->ui_language.language_code) {
       params->other_top_language = info;
+
+      // Report to UMA how important the UI language is.
+      DCHECK(params->other_top_language.frequency > 0)

[Marc Treib, 2016/10/14 15:32:56]

DCHECK_GT

[jkrcal, 2016/10/14 16:34:15]

Done.

+          << "GetTopLanguages() should not return languages with 0 frequency";
+      float ratio_ui_to_other_top_language =
+          params->ui_language.frequency / params->other_top_language.frequency;

[Marc Treib, 2016/10/14 15:32:56]

Should this be "ui_freq / (ui_freq + other_freq)" ?

[jkrcal, 2016/10/14 16:34:15]

Well, I meant it as it was. Your suggestion actually makes the data more
standard an easier to interpret :) Done.

+      UMA_HISTOGRAM_COUNTS_1000(

[Alexei Svitkine (slow), 2016/10/14 15:30:15]

I think you probably want to use UMA_HISTOGRAM_PERCENT here. Otherwise, you'll
get an exponentially scaled set of buckets.

[jkrcal, 2016/10/14 16:34:15]

This is what I wanted because the result was not supposed to be on the range of
0-100. 
Together with the change of Marc above, I can make it PERCENT. Which makes more
sense.

+          "NewTabPage.Languages.UILanguageToOtherTopLanguageRatio",
+          ratio_ui_to_other_top_language * 100);
       break;
     }
   }