Improving performance of clipping with rectangles

[postgispaul]

There is some improvement possible when calculating the intersection of a geometry g and a rectangle r that is completely covering the geometry g.

This can be useful f.e. if you want to clip geometries to exclude +90deg or -90deg parts so it doesn't generate errors when converting to Mercator.

If r is indeed fully covering g then the intersection is just g.

This can easily be extended to union (result is r) although usecase is unclear. A bit harder but doable are difference and symdifference.

[mdavis]

Perhaps this need could be met by improving ​ST_ClipByBox2D so that it produces valid output geometry?

[Algunenano]

Perhaps this need could be met by improving ​ST_ClipByBox2D so that it produces valid output geometry?

For this use case, ST_ClipByBox2D won't execute GEOS functions unless it needs to (it has bbox shortcuts).

In any case, I do see the benefit of checking whether one of the input geometries of ST_Interects/ST_Intersection is a box to add some shortcuts (not exactly the same for both functions but still a benefit).