On Wed, Apr 4, 2012 at 6:21 PM, Mark H Weaver <m...@netris.org> wrote:
> Hi Stefan, > > Did you intend to send this as private email? Wouldn't it be better to > discuss this on the mailing list? > > Stefan Israelsson Tampe <stefan.ita...@gmail.com> writes: > > Because what would be cool though is to use an > > assoc-list as in kanren and when it grows to large issue a batch > > synchronisation and bring out that whole assoc into modifying the > > thread0 datastructure in one go. > > Another possibility is to simply replace the assoc-list with a more > efficient purely-functional map structure. Guile already has vhashes, > but even better would be persistent Hash Array Mapped Tries (HAMT) as > used in Clojure. See Phil Bagwell's paper "Ideal Hash Trees" for the > basic idea, and the persistent purely-functional variant used in Clojure > simply copies the path from the new leaf to the root in the obvious way > to add or delete without mutation. I'd like to add a highly-optimized > implementation of this to Guile soon. > I do think that this can be interesting to try out. Also the lookup could be coded in C in order to speed up kanren for guile. /Stefan > > Regards, > Mark >
