Last modified: 2010-05-15 15:32:53 UTC
The stub detection algorithm that is used on general pages (so that, for logged
in users with stub threshold turned on links to short articles have a
class="stub" added to the link) is not utilised on category pages. Hence all
links are rendered the same way on category pages.
Fixing this bug would be useful because, for instance, there is a huge "stub
sorting" project on the English Wikipedia. That great effort would be rendered
more useful if we could use standard categories rather than special categories
to indicate stub articles.
Clarifying, the category page text itself has stubs. It's the generated list of pages (and probably images, subcategories) in the
category that isn't stubbed.
because i think the plan is actually to remove stub threshold support
rather than make it more widely used.
Thanks for the responses. I guess (and so could be totally wrong) there are three steps to providing stub threshold support
1) At each save, find out the size of the text and store it in the db.
2) At each logged-in page load, look up the size of each linked page.
3) Look-up the preference of the user and for each link add a class=stub if pref > size.
Now I guess the reason that support is slated for removal is that one of these steps is showing up as a resource/speed bottleneck. Now,
because I'm an optimist, I hope that is number 3) - the user-dependent bit. Rather than remove support totally is it possible to somehow
just send the size through in the link and somehow the user js to figures out the rendering?
As you might have guessed, I write as a keen user of the feature :)
In 1.5 we've got a page_size field on page, and have to look up the title from page anyway. Tossing the
third field in should be relatively easy if we're not removing it entirely. Might require some changes to
the link gen functions to explicitly state without copying stuff around.
Done in CVS HEAD for 1.5.
getting the length isn't the issue. the stub threshold will never be
accurate, because it won't be updated in the cached copy until the page