Have you tried coding even the brute-force naive search? It is far easier to improve an algorithm when you have someplace relatively concrete to start from. Plus, the naive approach is most likely to return a correct result, so that you can regression test your exotic interval-skipping, second-derivative conjugate interpolation algorithm. :)
In sum, I started to think through your problem this morning, and yes, I can visualize the rectangles floating across the 2D curve, and yes, I can picture that there are likely to be shortcuts and optimizations over the brute-force "check every index" approach. But really, you *should* do some of the work yourself... -- Paul -- http://mail.python.org/mailman/listinfo/python-list
