>>> On Mon, Dec 17, 2007 at 1:58 PM, in message
<200712171958.lBHJwOBb037317@wwwmaster.postgresql.org>, "Pedro Gimeno"
<pgsql-001@personal.formauri.es> wrote:=20
=20
> Description: Nonstandard precedence for comparison operators
=20
> The operators <>, <=3D and >=3D are expected to have the same precedence =
as =3D, <
> and >, but according to the docs (and matching actual behaviour) they have
> the same precedence as e.g. ||.
>=20
> This leads to surprises
=20
> Of course the same applies to !=3D since it is converted to <>.
>=20
> Now being picky, the precedence of < and > should be the same as that of =
=3D
> (for comparison, not for assignment), which makes a difference in rare ca=
ses
=20
Everything Pedro said in this post matched my understanding of the
SQL specification. I went back and read those sections closely,
and I can't find any basis for interpreting the spec any other way.
This seems like a pretty serious point of non-compliance to me.
=20
That said, bringing PostgreSQL into compliance with the standard
would undoubtedly break some people's existing applications. It
seems to call for the same phased approach as the standard
conforming string literals, with GUCs to control warnings for
problem constructs and legacy versus standard runtime behavior.
=20
-Kevin
=20
