Hi Arnaud > I've tried a completely different approach, that I imagine as 'folding'. I > thought it would improve performance over my previous effort but extremely > limited and crude benchmarking seems to indicate disappointingly comparable > performance...
I wrote a stack-based version yesterday and it's also slow. It keeps track of the stack computation and allows you to backtrack. I'll post it sometime, but it's too slow for my liking and I need to put in some more optimizations. I'm trying not to think about this problem. What was wrong with the very fast(?) code you sent earlier? Terry -- http://mail.python.org/mailman/listinfo/python-list
