"Kevin O'Gorman" <kogorman@pacbell.net> writes:
> I'm about to launch into an experiment that will do some new things
> inside the PG server. I'm sure to have a lot of problems, and one
> of them I can already tell is going to be difficult is the business
> of contexts: memory contexts, scan contexts and the like.
There is some doco about memory contexts in
src/backend/utils/mmgr/README. Dunno about anything comparable
for scan handles --- best way to deal with table scans is probably
to find a routine that does something like what you need to do,
and crib the code.
regards, tom lane