> I'll be doing my Masters project this summer, and am considering
> using Python as my implementation language. The application is
> [...]
> project. My proposal involves manipulating large graphs of objects
> from a database of meta-data about the documents in the library.
> These graphs are undirected and may often by cyclic. Has anyone used
> python to manipulate large cyclic graphs without experiencing memory
> leaks?
> [...]
> I'd like to hear comments from anyone who has done similar work
> previously so that I don't duplicate anything more than necessary.
I may be off-base here, but from his description, aaron watters'
"kjbuckets" python extension sounds extremely well suited to your
purposes. It can be retrieved from the python distribution ftp
archive, at
URL: <ftp://ftp.cwi.nl/pub/python/contrib/kjbuckets.tar.gz>
(There's also a 'kjubuckets.README' in the same dir, if you want a
preview.)
I expect that aaron will respond, himself, when he gets the chance...
> I'd also like to hear from people who can tell me why they chose NOT
> to use Python for such an application. Other languages will
> definately be considered to get around the ref. counting problem.
I, for one, will be interested to hear about which language you decide to
use, btw.
ken
ken.manheimer@nist.gov, 301 975-3539