re: [Help-glpk] implementation of graphs and networks in glpk

[Robbie Morrison]

Hello Andrew

>From my perspective as a network flow user, your graph
and support types look perfectly reasonable.  The
terminology is fine.  The use of an incidence list is
probably a good choice.

Will there be restrictions on the graph objects, for
instance, will parallel arcs be acceptable?  I guess
that depends on the algorithm in use.  In which case,
can you match structural restrictions with graph
algorithms when programming in C?  And, if so, would
you want to?

That then prompts the question of which network
algorithms do you propose to implement?  Least-cost
flow, maximum flow, minimum flow? (I use all three).
And so on!

I also work with nested graphs.  I don't suppose there
will be support for a graph of sub-graphs (as the
Boost.Graph library does) (my client code should deal
with that I guess!).

On a design level, is this new graph type a stand-alone
data structure?  Or is it a convenience data structure
which is really a 'glp_prob' deep down?

Will the user be able to solve their resulting
least-cost flow network problem using both
'glp_simplex' and 'glp_netsolve' (or whatever the call
will be)?  Or can one convert a graph object to a
'glp_prob' object to do so?  And vice-versa?

hope this helps, with best wishes and good luck
Robbie
---

> ------------------------------------------------------------
> To:          address@hidden
> Subject:     [Help-glpk] implementation of graphs and networks in glpk
> Message-ID: <address@hidden>
> From:        Andrew Makhorin <address@hidden>
> Date:        Mon, 22 Dec 2008 19:58:44 +0300
> ------------------------------------------------------------
>
> I plan to include in the next release of glpk some new api data
> structures and routines to deal with graphs and flow networks.

[snip - explaination and later on the C code]

> Any comments and suggestions are appreciated. Thanks.

---
Robbie Morrison
PhD student -- policy-oriented energy system simulation
Technical University of Berlin (TU-Berlin), Germany
University email (redirected) : address@hidden
Webmail (preferred)           : address@hidden
[from IMAP client]