[11:58] wgrant: you said "the usual 3-character threshold" the other day - do you know where that's implemented? it is, perhaps obviously, near-impossible to search for [11:59] cjwatson: That's a good question. It could be in the JS picker frontend. [12:14] thanks, will hunt [12:47] min_search_chars, indeed. [12:54] Hm. This is all very reminiscent of the DSP picker work from 2011 that I tried to resurrect once ... maybe it makes more sense to try to finish that than to continue with the BinaryAndSourcePackageName vocabulary. [12:55] SourcePackage.setBranch ensures a DistributionSourcePackage row, so that would be charms-safe. [13:05] * cjwatson digs up year-old branch for that and is pleasantly surprised to find it not entirely exploding [13:05] cjwatson: Still using the updated cache tables, though? [13:05] DistributionSourcePackageInDatabase doesn't know about binary names AFAICR [13:06] The stuff I dug up uses DistributionSourcePackageCache [13:06] (and the table isn't necessary for DistributionSourcePackage objects) [13:06] But it goes through that for binary names rather than using DistroSeriesPackageCache [13:06] Which seems potentially likely to perform better; will have to see [13:08] I didn't at all understand the constraints at the time I was last working on this (I was just resurrecting some old work by Purple in an attempt to deal with https://bugs.launchpad.net/launchpad/+bug/42298), but there's a query here that's ranking exact source name, exact binary name, partial source name, partial binary name [13:08] Bug #42298: package picker lists unpublished (invalid) packages

[13:09] cache-aware and archive-aware === frankban is now known as frankban|afk