On tor, 2011-10-06 at 20:15 -0400, Alex Goncharov wrote:
> P.S. And on the odd chance that somebody thinks that this
> functionality would be possible and helpful to add to libpq, and
> the problem is in the lack of human resources: I would be more
> then happy to dig into some PostgreSQL (the product) development
> under somebody's coaching, to start with. This topic or other.
> I just wouldn't know where to start myself.
I had some some research on this particular topic/feature recently. My
notes currently say, it's better to not tackle this before the not-null
cataloging patch (see entry in current commitfest) is finished. Because
that patch would presumably already implement much of the logic
necessary to determine whether a give expression implies nullability or
not and catalog this in a simpler fashion. Based on that you will then
have to drag this information around and put it on the wire so that the
client APIs can process it.
