On Wed, 12 Mar 2003, Tom Lane wrote:
> Stephan Szabo <sszabo@megazone23.bigpanda.com> writes:
> > Of course, I misread what explain did (without trying the
> > enable_seqscan=off case) and this is still not indexable because even
> > after that, you'll not get a clause on the outside that it considers
> > indexable. It is smart enough (7.4 anyway) to make the filter ((t.*).n)=5
> > which I thought it'd index, but doesn't. :(
>
> Note that inline-expansion of SQL functions like this is new for 7.4;
> it's not done in any current release.
I wasn't sure when it was added.
> I think the extra step to make this expression indexable is probably not
> too hard: the constant-expression folder needs to be taught that
> extracting a field from a whole-row Var can be replaced by a Var
> reference to the field, ie, fold "(t.*).n" into "t.n".
That's what I figured, but I wasn't sure if there was some case where it
was unsafe.
