Standard graph API?
wweston at att.net
Mon Aug 23 20:43:35 CEST 2004
Magnus Lie Hetland wrote:
> Is there any interest in a (hypothetical) standard graph API (with
> 'graph' meaning a network, consisting of nodes and edges)? Yes, we
> have the standard ways of implementing graphs through (e.g.) dicts
> mapping nodes to neighbor-sets, but if one wants a graph that's
> implemented in some other way, this may not be the most convenient (or
> abstract) interface to emulate. It might be nice to have the kind of
> polymorphic freedom that one has with, e.g, with the DB-API. One could
> always develop factories or adaptors (such as for PyProtocols) to/from
> the dict-of-sets version...
> So, any interest? Or am I just a lone nut in wanting this?
A know I'd appreciate it. It could be used to configure
neural nets and logic networks; where this api would make
it easy to build an abstraction then "compile" it into a
faster representation for execution - or just run the
tree/graph in "interpreted" mode.
I don't think it would get a lot of use, but the use
would be high end.
More information about the Python-list