Re: [igraph] graph representation in memory

[Mati Vait]

Hi,
at the moment we are investigating code from version 0.5.1 for no
specific reason. It's probably wise to replace it with latest stable
release.  

I can't tell anything yet about data structures because we haven't
decided which to use. There will probably be two ways for storing data -
in memory and on disk.

As this is a school project we have version control provided by school
(which we have to use) and at the moment we are not planning to start
separate branch of igraph outside of that repository.

I will try to keep you posted on details and developments (or atleast
bored with questions).

Thank you all,
Mati










    


On Mon, 2009-03-23 at 18:35 -0400, Chris Wj wrote:
> I would be interested in following your progress, learning a bit, and
> making the same modifications. Please post to the mailing list as you
> make updates and the such. 
> 
> Question, have you started modifications from the 0.6 main launchpad
> bazaar branch? Or are you modifying 0.5.1/0.5.2?
> 
> Thanks,
> Chris
> 
> On Mon, Mar 23, 2009 at 1:47 PM, Gábor Csárdi <address@hidden>
> wrote:
>         Hi Mati,
>         
>         you are completely right, it is better to use (long) int
>         instead of
>         double, this was a design mistake on my part.
>         
>         As for the work, it should not be very difficult. In theory,
>         all
>         functions that you need to rewrite are in
>         type_indexededgelist.c.
>         Plus, there are three macros in igraph.h that you need to
>         update or
>         rewrite as a function: IGRAPH_FROM, IGRAPH_TO and
>         IGRAPH_OTHER.
>         
>         This is theoretically enough for the C library and probably
>         also for
>         Python. For R you need a bit more, you need to write some
>         conversion
>         functions and rewrite iterators.R as well.
>         
>         Many C functions transform the igraph_t into a plain adjacency
>         list,
>         or one containing edge ids. For better performance you would
>         need to
>         rewrite these too, they are igraph_adjlist_t and
>         igraph_adjedgelist,
>         plus lazy versions of these. You can rewrite these just to use
>         the
>         normal operations of your new data structure.
>         
>         Maybe I forgot some minor things, but I thing roughly this is
>         it.
>         
>         Of course I can help you if you decide to work on this,
>         assuming you
>         need any help. :)
>         
>         May I ask you what kind of data structure you have in mind? Do
>         you
>         intend to put it into igraph, or the goal is just to play with
>         it a
>         bit?
>         
>         Btw. I am not sure what you can actually do with a graph with
>         100
>         million nodes and a general graph library. You cannot
>         calculate too
>         many things and perhaps you're better off coding them without
>         using
>         igraph at all. But I understand that this is a school
>         project. :)
>         
>         Best,
>         Gabor