Add a caveat pointing out that constraint exclusion doesn't work with
authorTom Lane <tgl@sss.pgh.pa.us>
Wed, 20 Jun 2007 23:11:38 +0000 (23:11 +0000)
committerTom Lane <tgl@sss.pgh.pa.us>
Wed, 20 Jun 2007 23:11:38 +0000 (23:11 +0000)
commita94669fdf6978d7c1d2ed7933bff749c0ba01f97
tree8a040e79c31ce3c9da34474e2ab4833191ed76be
parentab7ccb82fe3903efadde88670058dd59a37f13d2
Add a caveat pointing out that constraint exclusion doesn't work with
constraints the planner is unable to disprove, hence simple btree-compatible
conditions should be used.  We've seen people try to get cute with stuff
like date_part(something) = something at least twice now.  Even if we
wanted to try to teach predtest.c about the properties of date_part,
most of the useful variants aren't immutable so nothing could be proved.
doc/src/sgml/ddl.sgml