Martijn van Oosterhout <kleptog(at)svana(dot)org> writes:
> I think the reason it hasn't been done for general join conditions is
> because we havn't thought of an efficient algorithm.
Right, it's keeping track of the unmatched right-hand rows that's a
problem.
> However, I wonder if youre case couldn't be handled with a
> time-interval datatype such that you condition is merge-joinable on
> that type. I can't quite see it though...
Interesting thought. The use of NULLs in this example is a pretty poor
representational choice --- the queries would get a lot simpler if you
used "exptime = INFINITY" to represent unexpired data. That doesn't get
you all the way to a mergejoinable query though :-(
regards, tom lane