Bruce Momjian wrote:
> > I already have a patch for this one. The change is a few pretty simple
> > changes
> > to gram.y.
>
> OK, you will have to canvas the general list to make sure this does not
> break things for people, though our inheritance system needs an overhaul
> badly.
This is already CCed to the general list.
> I fear it is totally against the way our API works. How does someone
> see how many columns in the returned row?
A new API PQnfieldsv(PQresult, tupnum) or some such.
> Yikes. Strange.
Strange for C code perhaps. Very useful for constructing real objects in
OO application code framework.
> Can we just return nulls for the empty fields?
Well, I think we should probably distinguish between a field that is
null,
and a field that simply doesn't exist.
> How many new API calls are required?
Perhaps just the one. (above).
