[postgis-devel] Exclusion Constraints in PostgreSQL 8.5

[Paul Ramsey]

On Tue, Dec 8, 2009 at 9:54 AM, Jeff Davis <pgsql at j-davis.com> wrote:

> Is there a reason the operators don't use RECHECK? If they did use
> RECHECK, PostGIS would work very nicely with Exclusion Constraints right
> now. Is PostGIS interested in supporting a set of operators that can be
> used with Exclusion Constraints?

Because the semantics of the operators are all documented as being
100% bounding box based. So no re-check is needed.

We could define "additional" operators that have strict,
geometric-based, semantics and use re-checks to complete their tests.
But the standards we follow are based on function signatures, so we
already have ST_Contains(A, B) which includes both a bbox index test
and a geometry exact test. So we would end up having an operator
analogue with some special character set (A ## B) that does a re-check
and has the same semantics as ST_Contains().

Is there any big advantage to exact-operator-with-recheck over doing
inexact-operator-combined-with-exact-function-in-sql? Being able to do
Exclusion Constraints would be one :) Are there others?

P.