Hello.
I've been working on bitmap index AM for some time.
The situation is --- in general it works, but there're some problems I cannot
fix myself. I've been using Teodor's scripts for concurrency tests (modified a
bit), unfortunately without success.
Anyway, here's the patch. I hope someone will look at it and comment it.
Some notes:
- I haven't tested amrkpos and restrpos functions, don't know how (I've been
trying to make planner choose merge join without luck);
- no space is freed during vacuums, I think I'll elaborate on this in
separate email;
- nothing done around cost estimation.
Waiting for feedback, thanks.
--
Victor Y. Yegorov