http://judy.sourceforge.net is a very fast, very scalable dynamic array facility. I wonder how it compares to Tcl's hashing and whether it is worth looking into as the basis for a more powerful [array] command.

[CMcC] they certainly look interesting, but it seems to me that they are trees which store their keys internally, and they distribute it within the tree (they seem to be based on a trie data structure) so I can't see how you could use them for storing objects with an arbitrary ordering unless you used the address of the object as they key (cast to an integer.) Hmm, I guess that would work after all, as a kind of hash.  It's quite counter-intuitive that something giving a complete (if arbitrary) ordering would be more efficient than something like a hash, but since Judy is designed to play well with CPU caches, I guess such unusual results are to be expected.

http://www.snookles.com/scott/publications/pli2003-slf.pdf  talks about how much faster it is in [Erlang] as compared with Erlang's builtin hashing.

-- [Todd Coram]