On Tue, Jan 14, 2014 at 9:28 PM, Alexander Korotkov <aekorot...@gmail.com>wrote:
> On Tue, Jan 14, 2014 at 11:16 PM, Marti Raudsepp <ma...@juffo.org> wrote: > >> Oh, this actually highlights a performance regression with the partial >> sort patch. >> > > Interesting. Could you share the dataset? > It occurs with many datasets if work_mem is sufficiently low (10MB in my case). Here's a quicker way to reproduce a similar issue: create table foo as select i, i as j from generate_series(1,10000000) i; create index on foo(i); explain analyze select * from foo a join foo b using (i, j); The real data is from the "release" table from MusicBrainz database dump: https://musicbrainz.org/doc/MusicBrainz_Database/Download . It's nontrivial to set up though, so if you still need the real data, I can upload a pgdump for you. Regards, Marti