Hi I typically deal with very large arrays that don't fit in memory. How does Numpy handle this? In Matlab I can use memory mapping but I would prefer caching as is done in The Gimp. Any pointers appreciated. Mathew