Page MenuHomePhabricator

Search using insource and regex returns irregular and different set of articles each time
Closed, DuplicatePublic

Description

Author: SpontaneousGrumbler

Description:
On repeat runs of https://en.wikipedia.org/w/index.php?title=Special:Search&redirs=0&limit=500&offset=0&profile=default&srbackend=CirrusSearch&search=insource%3A%22publicly-%22insource%3A/publicly-/ a few minutes apart, different numbers and selections of articles are returned. I had 79, 119, 83, 85, 82 and 84 articles returned over a couple of hours. These are not just the same list truncated at different points; the list of 119 had 45 articles not in the list of 79, but the list of 79 had 5 articles not in the list of 119.


Version: unspecified
Severity: normal

Details

Reference
bz71098

Event Timeline

bzimport raised the priority of this task from to Needs Triage.Nov 22 2014, 3:43 AM
bzimport added a project: CirrusSearch.
bzimport set Reference to bz71098.

SpontaneousGrumbler wrote:

Today I'm getting "An error has occurred while searching: We could not complete your search due to a temporary problem. Please try again later." every time. At least some consistency has been achieved.

Yay? I've started work to speed it up by a pretty wide margin which should improve matters. It isn't going to change overnight though.

The error you are seeing is that the queue for new regex queries is full and it won't let you add another. I have a error messaging change in the pipeline that'll make that make more sense but I can't rush it through because if I did it'd go out untranslated.

The trouble with the regex queries is that there is only one line and some of them are fast because they have other filters but some are slow and the fast ones get behind the slow ones in line. That's why making them faster will help with the error messages.

The inconsistency is likely also caused by partial timeouts. In other words the search backend decides that it has spent long enough on the regex and just gives what it has so far. It won't have to do that if I'm able to speed up regexes though.

Restricted Application added a subscriber: Aklapper. · View Herald Transcript
Deskana moved this task from Needs triage to Search on the Discovery-ARCHIVED board.

Related to, or possibly a duplicate of, T106685.