Wikidata fields showing non-local language data

[boothym]

This city node is showing non-English information in the wikidata fields for my en-GB browsers (Firefox/Chrome):

Seems to be coming from the Wu Chinese name and French description here: https://www.wikidata.org/wiki/Q182923

[1ec5]

Here’s the request I’d expect iD to make in this case; it shows “Stirling” and “city in Scotland” as expected.

I think the problem is that we’re passing the browser’s language directly into the MediaWiki API:

iD/modules/services/wikidata.js

Line 61 in f8f5b69

languages: lang,

The API expects lowercase language codes (en-gb), whereas browsers typically supply the region in uppercase (en-GB). An unrecognized language code results in all the languages being returned, and iD picks the first item in each object: Simplified Chinese for the label, French for the description.

Case aside, it’s possible for the browser to supply a language that Wikidata doesn’t support, or for the browser’s language code to contain some additional detail that doesn’t match Wikidata. (For example, WebKit on Apple platforms would probably use zh-Hant-TW instead of zh-tw.) We should supply all the browser’s languages via navigator.languages (plural) and also add a fallback for English, which is more likely to be understood than a combination of Chinese and French (or whatever happens to be present on the item).

[bhousel]

Ah yeah @nyurik showed me how to fix this the other day.. The MediaWiki API lets us request a list of preferred locales and fallbacks.

[1ec5]

#5543 would be another reason to supply fallback language codes in the API request.