On Mon, Aug 21, 2000 at 02:32:08PM -0500, Ross J. Reedstrom wrote:
> On Mon, Aug 21, 2000 at 01:39:32PM -0400, Tom Lane wrote:
> >
> > Seems to me the sort order should be
> >
> > -Infinity
> > normal values
> > +Infinity
> > other types of NaN
> > NULL
> >
> > and the reverse in a descending sort.
> >
>
> NULL is taken care of by special casing in the sort code, as I already mentioned,
> and can be fixed immediately.
>
Grr, I take this back. By the time comparetup_* see the tuples, we've no idea
which order we're sorting in, just a pointer to the appropriate sortop.
<whine mode>
Why does every thing I touch in pgsql end up pulling in down into the
guts of the whole system? Even something that looks nicely factored
at first, like the type system? I guess this stuff is _hard_.
</whine mode>
Sigh, back to fixing up referential integrity violations in the DB
I'm finally upgrading from 6.5 to 7.0.X. (DBA life lesson number XX:
implementing RI in the backend _from the very start of a project_ is a
very good thing. Cleansing the data of cruft left from _not_ having RI in
the backend, is a bad thing. And those clever, recursive self-referencing
table structures for representing trees are a pain in the DBA to reload.)
Ross
--
Ross J. Reedstrom, Ph.D., <reedstrm@rice.edu>
NSBRI Research Scientist/Programmer
Computer and Information Technology Institute
Rice University, 6100 S. Main St., Houston, TX 77005