Simon Riggs <simon@2ndQuadrant.com> wrote:
> Yes, for most people touching != overlap. So it just looks like a
> bug.
A quick search of the web turned up a definition of overlap in
geometry as meaning that two polygons share at least one *internal*
point, which would be consistent with your interpretation; but there
is the issue of breaking existing code. Perhaps people are now
accustomed to following the existing overlaps test with a test that
the area of intersection is non-zero?
Anyway, based on what I found, we should document the current
behavior, as the term in PostgreSQL doesn't seem to match the
conventional definition in geometry.
-Kevin