On Wed, 2006-01-18 at 20:13 -0500, Tom Lane wrote:
> Come to think of it, the idea also seems to map nicely into bitmap index
> scans: the index will directly hand back a list of potential pages to
> look at, but they are all marked "lossy" because the index doesn't know
> exactly which tuple(s) on the target pages match the query. The
> existing bitmap-heap-scan code can take it from there.
Yes, I've privately suggested this solution in that context.
I think there is enough meat there to make this topic worth discussing
further, but not on list again just yet.
Best Regards, Simon Riggs