&& vs. &&& what's the difference

[robe]

These seem to be exactly the same in my installs shouldn't one be a 2d interacts check and the other a 3D interacts check?

The definitions look the same and when testing they seem to give the same answers with 3D geometries even in cases where I expect &&& to return false and && to return true

e.g. both have:

CREATE OPERATOR &&&(
  PROCEDURE = geometry_overlaps,
  LEFTARG = geometry,
  RIGHTARG = geometry,
  COMMUTATOR = &&&,
  RESTRICT = contsel,
  JOIN = contjoinsel);

CREATE OPERATOR &&(
  PROCEDURE = geometry_overlaps,
  LEFTARG = geometry,
  RIGHTARG = geometry,
  COMMUTATOR = &&,
  RESTRICT = contsel,
  JOIN = contjoinsel);

[pramsey]

I think it's a bug, try replacing the &&& one with geometry_overlaps_nd. Also, if you force an index scan and build and index you should find you get difference answers.

[robe]

Nope I get a ERROR: function geometry_overlaps_nd(geometry, geometry) does not exist

when I try that

[pramsey]

Might be better now

[robe]

Yap much better