>> I thought that the bad case for a tuplestore was if the set returning
>> function was expensive and the user used it with a LIMIT clause. In the
>> tuplestore case you evaluate everything then throw it away.
>
> I'm not terribly excited by that example --- but in any case, the real
> solution to any problem that involves communication between function and
> calling query is to make sure that the function can get inlined into the
> query. That was an option we didn't have back in 8.2; but it's there
> now. My test case deliberately disables that optimization ...
I'm pretty excited by that example. LIMIT/OFFSET is really useful as
a way of paginating query results for display on a web page (show
results 1-100, 101-200, etc), and I use it on potentially expensive
SRFs just as I do on tables and views. I agree that inlining is a
better solution when it's possible, but who is to say that's always
the case? Of course if it's PL/pgsql with RETURN QUERY the way
forward is fairly obvious, but what if it isn't that simple?
...Robert