On Fri, 9 Jun 2000, Tom Lane wrote:
> > Any status on this?
>
> Nothing done about it yet.
>
> IIRC, some people were concerned about the fact that freeObject()
> couldn't possibly cope with circular structures, multiply-linked
> subexpressions, etc. I don't think that's a problem for my intended
> use in the relcache --- the only structures I'll be freeing are ones
> previously read in by the node-reading functions, and those aren't
> going to have any surprises like that.
IMHO use separate memory context will better and more fast way than
freeObject(). I use this method in my suggested query cache and in the
SPI (in SPI_freeplan()) and it is very good (without potential leaks).
All in backend/nodes are (IMHO) very dificult keep up and recursion is
and not gratis feature too. (Homework: write PG in Fortran :-)
Karel