Having statistics on ranges was really missing! The planner was doing
some really, really bad choices on bigger tables regarding seq/random
scans, nested loop/other joins etc.
Is there any chance this makes it into 9.2 final? It would really
round-off the introduction of range types and maybe avoid problems
like "the new range types are slow" (just due to the bad row
estimates).
Thanks for implementing this feature,
-Matthias