Page MenuHomePhabricator

Special:Shortpages needs index fix to run real-time
Closed, ResolvedPublic

Description

Author: allen3-wiki

Description:
The list of pages generated by Special:Shortpages is now displaying articles
that have been deleted or are redirects.

The change in behaviour began sometime after 22:30, November 12, 2005 (UTC).


Version: unspecified
Severity: normal
URL: http://en.wikipedia.org/w/index.php?title=Wikipedia:Village_pump_%28technical%29&oldid=29060221#What_is_up_with_Special:Shortpages.3F

Details

Reference
bz3945

Event Timeline

bzimport raised the priority of this task from to Low.Nov 21 2014, 8:54 PM
bzimport set Reference to bz3945.
bzimport added a subscriber: Unknown Object (MLST).

Shortpages has been switched to cached mode because the indexes turn out to be
badly inefficient for it on the large wikipedias.

Adding an appropriate index for it should be possible, so it can be switched
back.

IIRC it requires something like (page_namespace,page_is_redirect,page_len).

  • Bug 4058 has been marked as a duplicate of this bug. ***

bugzilla_wikipedia_org.to.jamesd wrote:

I tested various indexes on frwiki. Without a use or force,
the optimiser didn't pick (page_namespace,
page_is_redirect, page_len) but did pick (page_namespace,
page_is_redirect, page_len, page_title). So I suggest the
latter. That made it a covering index also, so it should be
a very fast query.

I'll leave the test indexes on frwiki for a few hours so
you can test yourself. Please drop the indexes when done,
or use a private IRC chat to let me know you're done with
them and I'll do it.

I'm not worried about the index being quite big. I don't
expect more than one or two index pages to be loaded and do
expect that it's better to pull in those covering index
pages than the larger number of row pages I expect to be
needed to get the titles for the small pages - expect those
to be unpopular, nearly random and need more like 50 page
loads.

bugs wrote:

Most Specialpages like these are cached, but are shown more user-friendly than they were at the time. WONTFIX'ing.

ayg wrote:

Reopening. There's still clear benefit to having as many special pages as possible real-time, if we're willing to use the indexes. We might judge that it's not worth it to slow table modifications for the sake of two special pages, but if so, that's not your decision to make.

I'm seeing INDEX (page_len), so maybe this is doable?