Man <man.tr...@gmail.com> writes: > Additional information. > In 9.6 the second table (lesser tuple) was choosen (the same testdata). > There are something (cost estimation?) different in previous versions.
I'd bet on different statistics in the two installations (either you forgot to ANALYZE, or the random sample came up quite a bit different). And I'm a little suspicious that these tests weren't all done with the same work_mem setting. regards, tom lane -- Sent via pgsql-hackers mailing list (firstname.lastname@example.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers