On Wed, Jun 26, 2013 at 8:23 PM, Elan Feingold <e...@plexapp.com> wrote:
> Hi, > > We've observed a severe performance regression in a query. > > We had a query Q1, which was running fast, on 3.7.15.2. We made a minor > change to the query, which we'll call Q2. > > Q2 runs excruciatingly slowly in 3.7.15.2 (1m 28sec). We then observed > that Q2 was fast in 3.7.14 (300ms), but slow in 3.7.15.2 and later (tested > up until 3.7.17). > Have you tried it with the NGQP snapshot? The NGQP was announced on this mailing list yesterday, but I see that the OP only subscribed as of today. Please see the original posting at http://www.mail-archive.com/sqlite-users%40sqlite.org/msg77855.html > > The slow query Q2 can be found in a test program here ~ > http://www.plexapp.com/elan/test.c > > The database which shows the slow query in action is here ~ > http://www.plexapp.com/elan/com.plexapp.plugins.library.db-slow.zip > > The crux of the query which makes it pathologically slow in 3.7.15+ > appears to be this bit: "join metadata_item_settings on > metadata_item_settings.guid=grandchild.guid ****and > metadata_item_settings.account_id=1****". If I take out the account_id bit, > it's then fast on every version. > > Where is it spending all this time? I'm glad you asked: > https://dl.dropbox.com/s/vnfv5n7m2vgn1p4/2013-06-26%20at%202.21.58%20PM.png > > Please let me know if you have any other questions! > > -elan > _______________________________________________ > sqlite-users mailing list > sqlite-users@sqlite.org > http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users > -- D. Richard Hipp d...@sqlite.org _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
