Teodor Sigaev <teodor@sigaev.ru> writes:
> 4) OR clauses optimizations
> http://www.sigaev.ru/misc/OR_82-0.6.gz
This seems kinda ugly --- it looks very expensive and unlikely to find
useful optimizations most of the time. Have you done any benchmarking
to find out what the cost is when the optimizations don't succeed?
I guess my big problem with it is that it's a huge expansion of the
single ugliest, most ad-hoc part of the planner, ie, orindxqual.c.
And the new code is just as ad-hoc as what was there before ...
Also, what's with the pull_tlist bit?
regards, tom lane