On Fri, 4 Aug 2000, Thomas Lockhart wrote:
> > > btw, it appears that SQL99 (haven't checked SQL92) specifies that
> > > test=# select (1,2,3) = (1,2,null);
> > > ?column?
> > > ----------
> > > (1 row)
> > > should return FALSE, not NULL.
> > What? If so, they broke it pretty badly. This should be equivalent to
> > 1 = 1 AND 2 = 2 AND 3 = NULL, which should reduce to TRUE AND TRUE AND NULL,
> > which should reduce to NULL. Anything else is not self-consistent.
> Hmm. I could have sworn I looked this up (and was suprised at the
> result). But I'm not finding the example anywhere, and Section 8.2
> General Rule 1 seems to indicate that we do the right thing here
> Also, I *think* we have settled on the following facts:
> 1) "3 = NULL" is typical of an expression generated by M$ Access.
> 2) "3 = NULL" is *not* legal SQL9x syntax, which specifies "3 IS NULL"
> for the comparison "does three have a value of NULL?".
> 3) New versions of M$ Access continue to generate bogus queries
> containing these comparisons.
> 4) Postgres will continue to understand (at least) the special case of
> "column/value = NULL" to retain compatibility with M$.
Stupid question here ... but ... can't this kludge be "faked" in the ODBC
driver itself, vs in the server? *raised eyebrow*
In response to
pgsql-hackers by date
|Next:||From: Ross J. Reedstrom||Date: 2000-08-04 14:47:45|
|Subject: Re: [HACKERS] pg_dump/restore to convert BLOBs to LZTEXT (optiona l!)|
|Previous:||From: Mike Mascari||Date: 2000-08-04 09:15:11|
|Subject: Re: comparing rows|