very large graph

bearophileHUGS at lycos.com bearophileHUGS at lycos.com
Tue Jun 24 03:42:34 CEST 2008


chrispoliq... at gmail.com:
> My non-directed graph will have about 63,000 nodes
> and and probably close to 500,000 edges.

That's large, but today not very large anymore. Today very large
graphs probably have more than millions of nodes...
You have to try, but I think any Python graph lib may be fit for your
purpose.
But I think a list of pairs won't do.

For larger graphs you can use Boost graph too, but it may be overkill
for your purposes.
If your graphs are even larger, you can implement a graph with
DataDraw in C and use the compiled module from Python, this is
probably among the faster possible near-prebuilt data structures.
For even larger graphs you may need the "external" STXXL with C++, and
then custom code.

The list of the operations you have to perform on the graph seems very
simple, so any graph lib may be enough, mine too:
http://sourceforge.net/projects/pynetwork/

For 63,000 nodes and 500,000 edges Graphviz may be too much slow, so
you may need to find a faster visualization tool. Two (or more) of
them are free too.

Ask if you need more information.

Bye,
bearophile



More information about the Python-list mailing list