Page MenuHomePhabricator

Store structured data needed for MediaWiki documentation
Open, Needs TriagePublic

Description

There are some infobox-like templates in MediaWiki documentation that should be auto-generated but are currently hand-created (e.g. hook pages, configuration variable pages, extension infoboxes, categories about extractable features such as which extension uses which hook). That results in wasted time (these pages are created by hand, even though their content is mostly mechanical) and spotty coverage (many hooks and most of the variables have no page at all; many existing pages are outdated; documentation for older versions of extensions is only available via the page history).

The pages for these templates should be bot-created, and the templates should fetch their data from some sort of structured data repository which (unlike wikitext) is an easy target for scripts that extract the information from git.

Some candidates for the data storage:

  • Wikidata (the vocabulary and some items exist already; OTOH this seems like a stretch of the notability guidelines)
  • a local Wikibase installation on mediawiki.org (this would probably be blocked on the CommonsData work that allows Wikibase to reuse vocabulary from a remote Wikibase site)
  • Page Forms / Cargo installed on mediawiki.org (see T149612#2833923)
  • tabular data on Commons?
  • something horrible in wikitext (e.g. mw.loadData in Scribunto)
  • have some extension (or the combination of an extension + Scribunto) extract the data directly from git (see T91626)

It would be nice to have rough agreement on which method to pursue.

Event Timeline

Tgr created this task.Jan 10 2017, 6:33 PM
Restricted Application added a subscriber: Aklapper. · View Herald TranscriptJan 10 2017, 6:33 PM

@Tgr - thank you for creating this task, and for including me. I am the author of Page Forms and Cargo, and I think they would be good solutions for this problem - the data would still be created by hand (no bots), but editing would become a lot easier, and the data could be displayed in a variety of useful ways. Extracting data and text from Git would also be useful, but not all data is stored there (and not all extensions are on Git), so I would say that these are two complementary things - ideally mediawiki.org would have Page Forms, Cargo and Git extraction.

Tgr renamed this task from Store structured needed for MediaWiki documentation to Store structured data needed for MediaWiki documentation.May 5 2017, 2:14 PM
Tgr added a subscriber: Bawolff.May 7 2018, 5:39 AM

something horrible in wikitext (e.g. mw.loadData in Scribunto)

That has been implemented a while ago via extjsonuploader (thanks Bawolff!).