Requests for comment/Accessing page properties from wiki pages

From mediawiki.org
Request for comment (RFC)
Accessing page properties from wiki pages
Component General
Creation date
Author(s) MZMcBride, Legoktm, Jackmcbarn, Matma Rex
Document status in discussion
See Phabricator.

This is a request for comments regarding accessing page properties from wiki pages.

Background[edit]

MediaWiki has a page_props database table where certain attributes and values are stored on a per-page basis. For example:

MariaDB [enwiki_p]> select page_title, pp_propname, pp_value from page join page_props on pp_page = page_id where page_namespace = 0 and page_title = 'Barack_Obama';
+--------------+--------------------------+----------------------------+
| page_title   | pp_propname              | pp_value                   |
+--------------+--------------------------+----------------------------+
| Barack_Obama | defaultsort              | Obama, Barack              |
| Barack_Obama | page_image_free          | President_Barack_Obama.jpg |
| Barack_Obama | wikibase-badge-Q17437796 | 1                          |
| Barack_Obama | wikibase_item            | Q76                        |
+--------------+--------------------------+----------------------------+
4 rows in set (0.00 sec)
MariaDB [enwiki_p]> select page_title, pp_propname, pp_value from page join page_props on pp_page = page_id where page_namespace = 0 and page_title = 'Obama_(disambiguation)';
+------------------------+----------------+----------+
| page_title             | pp_propname    | pp_value |
+------------------------+----------------+----------+
| Obama_(disambiguation) | disambiguation |          |
| Obama_(disambiguation) | wikibase_item  | Q223850  |
+------------------------+----------------+----------+
2 rows in set (0.00 sec)

MediaWiki stores whether a page is a disambiguation page, what its default category sort key is, and what its page image is, each being determined by different means: __DISAMBIG__, {{DEFAULTSORT:Obama, Barack}}, and a heuristic (cf. Extension:PageImages#Image choice).

Problem[edit]

Users cannot access a page's properties from other local wiki pages.

Potential use-cases:

  • phabricator:T71441: Feature request: add detection for disambiguation pages to Scribunto
  • phabricator:T131911: Allow retrieving/getting page image file name from wikitext using Scribunto/Lua or parser function or something
  • phabricator:T154346: Provide "wikitext" means of accessing arbitrary wiki page's default category sort key

Complications[edit]

This makes wikitext more volatile. We already have features such as {{#ifexist:Page name}} and {{CURRENTTIMESTAMP}}, which can be combined in weird and potentially abusive ways. For example: {{DEFAULTSORT:{{CURRENTTIMESTAMP}}}}.

If we added the ability to get page properties, we could end up with cases such as these:

  • {{DEFAULTSORT:{{#getdefaultsort:{{FULLPAGENAME}}}}}}
  • {{#if:{{#invoke:IsThisADabPage|main|{{FULLPAGENAME}}}}||__DISAMBIG__}}

While we could try to restrict page property retrieval to pages other than the current page, you could end up with trivial loops: page A tries to get page B's category sort key, page B tries to get category A's category sort key. These types of problems have been solved by templates.

It's also currently possible to do things like this: w:en:User:Sgeo/null-edit-detector.

You also have a problem that's basically the opposite of volatility: if a page's default category sort key changes, how do know to update the pages, such as its talk page, that are also using this sort key?

Proposal[edit]

Option 1[edit]

Modify Scribunto/Lua to expose these properties as functions or Title object attributes.

We need to track these usages. Would using templatelinks be acceptable?

Option 2[edit]

Add parser functions to MediaWiki core or to extensions such as PageImages. For example:

  • {{#getdefaultsort:Page name}}
  • {{#getpageimage:Page name}}

We need to track these usages. Would using templatelinks be acceptable?

Option 3[edit]

For two of the use-cases, the focus is talk pages. So we could add:

  • {{ISSUBJECTSPACEPAGEADISAMBIGUATIONPAGE}}
  • {{SUBJECTPAGEDEFAULTCATEGORYSORTKEY}}

But this is kind of obnoxious and it boxes us in a bit. We may want to break out of the subject space page–talk page binary at some point (RfC: Associated namespaces (talk)), supporting groupings of three or four or more pages. On a related point, we may want to use this functionality on non-talk pages, which is the case for properties such as page images.

Option 4[edit]

Do nothing. Don't allow access to these page properties. :-(

Option 5[edit]

Here's a dumb idea: use the existing templates/transclusion system, either by creating a template for every page and property or by using Extension:LabeledSectionTransclusion to extract parts of a subject-space page.

This would be pretty terrible, but would save making a new links table and would require almost no development effort.

Option 6[edit]

Don't store these properties in wikitext. A lot of complications described above go away if you don't use wikitext as the storage/input mechanism. But in exchange, you get new complications regarding tracking changes, preventing abuse/misuse, etc. We'd be in violation of everything is a wiki page, but perhaps with something like multi-content revisions, this would be possible. This option would delay any implementation significantly.

A different version of this option would be to store properties in a centralized repository such as Wikidata or in a local repository using the Wikibase Repository extension.