Change ST_Equals to use && instead of ~=

[robe]

This is sort of getting to what Mark was saying. The behavior is borderline wrong, but not for the reason I think Mark was complaining about.

​http://postgis.refractions.net/pipermail/postgis-users/2010-January/025551.html

It is more an issue with soft upgrade.

We can't change the behavor of ~= with soft upgrade because we can't drop operators. That means people upgrading depending on which version they started off with are going to have some ~= with RECHECK and some with not RECHECK. Which will make it either behave like a bounding box equality check or a ST_OrderingEquals check.

~= with RECHECK is the wrong behavior for ST_Equals because a true or false to that answer doesn't tell you if _ST_Equals will be true or false. Because ST_EQUALS follows DE-IM which means geometries with flipped coordinates will be considered _ST_Equals, but fail the RECHECKED ~= and geometries that are invalid but binary equal can pass the ~= test but fail the _ST_Equals.

Until we can figure out how to deal with the

~= inconsistent behavior of installs — make this an && test. Which will slow it down a bit I guess.

[pramsey]

OK, I can buy that. Done at r5118