Zoltan Boszormenyi <[EMAIL PROTECTED]> writes:
> What are the "existing problems"?
Please read the archives; this has been discussed recently.
There's a lot of disagreement about what ALTER should allow
and what pg_dump should do with an altered sequence.
> I am asking because I am experimenting to implement
> the SQL2003 compliant form for the serial type
> to be able specify the underlying sequence
> parameters:
Be aware that the big problem with SQL2003 is that it expects an
"action at a distance" behavior whereby different references to a
generator all return the same result if executed within the same
query cycle. This makes the construct not equivalent to either
nextval() or currval(), but some hybrid with hidden state; and
changing of that state would have to tie into core parts of the
executor. It looks pretty messy :-(
regards, tom lane
---------------------------(end of broadcast)---------------------------
TIP 2: Don't 'kill -9' the postmaster
