On Aug 30, 2008, at 10:33 AM, Albretch Mueller wrote:
> well, yeah! I would totally agree with you, but since I doubt very
> much "COPY FROM CSV" is part of the SQL standard to beging with, why
> not spice it up a little more?
I'd guess that coming up with a general algorithm to guess the type
from a column of CSV text would satisfy no one, since we'd always
miss a particular case that is important to someone (are times local
or UTC? Should we assume integer or float? Varchar or text?), and the
option is a forest of switches that would be extremely complex and
error prone.
This sounds very much like an application-domain problem, best solved
in the application domain.