Re: [Boost-users] Boost Graph Library documentation unclear

Subject: Re: [Boost-users] Boost Graph Library documentation unclear
From: Andrew Sutton (andrew.n.sutton_at_[hidden])
Date: 2009-04-26 07:37:30

> I've been trying to use the Boost Graph Library. For the most part I'm
> working it out. However, I'm finding the documentation annoyingly
> decentralized....it isn't interlinked enough, at the very least. Too hard to
> go from "how do I use X" to "oh, here's the documentation on X." Took me
> half an hour just to find *something* on the add_edge function (turns out
> it's on the Concepts page?), and what's there *barely* qualifies as a
> prototype, much less a proper man page.

Concepts are similar "STL Concepts" (
http://www.sgi.com/tech/stl/table_of_contents.html). They classify and
describe the operations available to /all/ graph data structures (e.g.,
adjacency_list or matrix), which is why it might seem to barely qualify as a
prototype.

> My current question is regarding properties. Some of the algorithms specify
> the "edge_weight" property as if it's something all edges have, and the
> vertex_index property as if it's something all vecS-specified graphs have.
> Yet I can't find any documentation explicitly confirming this, and my
> compiler seems to disagree in any case.
>
> http://www.boost.org/doc/libs/1_38_0/libs/graph/doc/kruskal_min_spanning_tree.html

 No edges are weighted by default. The easiest way to weight a graph is to
used bundled properties (
http://www.boost.org/doc/libs/1_38_0/libs/graph/doc/bundles.html) and create
a property_map over a member pointer.

Currently, only adjacency_list<OEL, vecS, D, VP, EP, GP, EL> has a
vertex_index property by default. The OEL, D, etc. are just placeholders for
any template argument.

Can something be done about the state of the documentation, and is there a
> comprehensive list of default properties available somewhere (together with
> how those are replaced or supplemented when user properties are introduced)?

I'm currently rewriting the documentation, but that won't be ready until
1.40 at the earliest.

Andrew Sutton
andrew.n.sutton_at_[hidden]