Closed siddhartha-gadgil closed 4 years ago
The above is wrong and unnecesarily complicated. In the case (2) we just take as non-crossing.
Since this is wrong, it is closed. Please see http://math.iisc.ac.in/~gadgil/catg2020/notes/intersections/ for corrected sketch.
The problems
C
andD
with indicesi
andj
so that the vertices of the curves at these indices coincide.c0
andc1
be the edges ofC
before and after the coinciding vertex and letd0
andd1
be similar.D
comes in from aboveC
at (or before) the given intersection. Note that we are not ruling outc0 == d0
.fromAbove(C, D, i, j)
of the curves and indices.The function
fromAbove
c0 == d0
then we replace indices byi - 1
andj - 1
(cyclically) and compute. The hypothesis guarantees that this is not an infinite loop.d0 = inverse(c1)
, then we computefromAbove(C, D, i, j) = !fromAbove(C, inverse(D), i, -j)
. This immediately gives the above case, so we do simplify.As the curves are reduced, we can only get this case if we have a single intersection point, and once we flip this way we will not have to flip again for the same intersection.inverse(c1)
toc0
and returntrue
if and only if we encounterd0
along the way.