Wikipedia field should switch to an equivalent article in a local language

[1ec5]

Since #3265, the Wikipedia field has automatically defaulted to the user’s preferred language. This makes sense, since the user is only able to reliably choose an article in a language that they speak; we don’t want them guessing about article titles in languages they don’t speak. However, the documented expectation is that the wikipedia=* tag should link to the article in the local language’s Wikipedia edition, if available, not necessarily the mapper’s own language. If the user happens to be mapping in a region that speaks a different language, a wikipedia=* tag isn’t very useful to other local mappers as a human-readable companion to wikidata=*. To accommodate both the mapper and the local community, the Wikipedia field should automatically replace the selected value with the equivalent value in the local language.

For example, if an English speaker maps in a Spanish-speaking region and select a feature that doesn’t have wikipedia=* tagged yet, they should continue to see English as the default language. As soon as they enter an article title or select one from the dropdown menu and we fetch the corresponding Wikidata item (to fill in wikidata=*), we should replace the wikipedia=en:* tag with a wikipedia=es:* tag that’s linked to the same Wikidata item. On the other hand, if they select a feature that’s already tagged with wikipedia=*, regardless of the language, we should show what’s tagged verbatim. If the user needs to know what it refers to in their own language, they can look at the adjacent Wikidata field, which shows the Wikidata item’s label and description in their language.

[tyrasd]

Thinking about multilingual regions, it might not always be super trivial to automatically find which the "local language" is. A start could be to consider everything cldr as in the territory languages for a country as "potentially local" and keep untouched. Other languages could be mapped to the most common language for that particular county (although that does not seem ideal for about half of Belgium, some regions of Switzerland, etc.).

[matkoniecz]

it might not always be super trivial to automatically find which the "local language"

and in some regions it will have serious political implications to say which language is local/dominant/preferred

[1ec5]

Let’s not overthink this: some local communities already have strict expectations about which Wikipedia languages get linked, while others have an expectation of “not English”. We already have a rough mapping from territories to languages for things like prioritizing certain languages in the Multilingual Names field. If this isn’t reliable enough, then we can maintain a separate mapping from territories to their preferred Wikipedia languages, starting with the ones we know about and allowing communities to request additions to the table.

Communities that need a more language-agnostic solution can always deemphasize wikipedia=* tags in favor of wikidata=* tags, as was done in Israel.

[tordans]

I wonder if https://github.com/rapideditor/country-coder would be a good place to manage the available and "official" and maybe "primary" (as in the one we fall back to when multiple official) languages per geographic area. I assume we use that for things like the traffic speed unit, already

[1ec5]

🤷‍♂ rapideditor/country-coder#131