Almost there...

I have something that compiles... but doesn't give the correct result. I suspect I've blown the interface to van Laarhoven's SemilatticeSearchTree module in a way that, while it passes syntactic muster, is keeping the search from working right. The presence of a bunch of zeroes in the output might indicate that I'm consistently getting back Nothing.

What does the time output (0.3240.347s real time) tell us about what we might see when it's correct? Well, it's certainly a lower bound--but how much longer will the correct version run? The buggy one has presumably built the trees, and is doing all the work, but if the maxima are taking it immediately to Nothing, we're only skipping a couple of O(log(n)) operations times a thousand input lines, where n is no bigger than a bit over 4,000 (the number of 49-digit Ys). I'm hoping the corrected version won't take all that much longer... but we can't know until we have it right. More news as it happens.