Boost logo

Boost Users :

From: Tony Cook (tony__cook_at_[hidden])
Date: 2005-09-15 09:57:43


Thanks for all the help guys.

>If a graph algorithm does not document its requirements, that's a bug, and
>should be fixed. However, copy_graph does document its requirement for
>vertex_index.

I see that now. However it is written in a difficult form which made little
sense to me without that nudge. Man BGL is very steep!

"Default: get(vertex_index, G)" ??

How about

"Default: requires the presence of the vertex_index property map"

Why else do you think FAQ #5 exists ;)

> A hash map will be constant time, but is still somewhat slow. Boost does
> not yet have one.

:(

> I don't think you understand the nature of generic algorithms.

I'm new and I'm learning. Thanks for the enlightenment :) . OK please reread
this as an issue with the graph_copy algorithm. Surely all graph types
should at least copy - even if *relatively* inefficiently (I can't believe
that its impractical to copy!!).

Worse still you can write

G1 = G2 ;

It compiles but doesn't work! It's shallow.

> That's an exaggeration.

:) :)

What I didn't tell you is that I'm *forced* by VC7.1 to use vecS,listS
graphs as with my particular collection of properties, ALL other graph type
combinations cause VC7.1 to issue an "Internal Compiler Error" - I should
have slammed the Ford attribute onto VC7.1 - it deserves it, not BGL

Anyway I'm happy this issue is being considered. I need no further
assistance on this matter.

Cheers, Tony


Boost-users list run by williamkempf at hotmail.com, kalb at libertysoft.com, bjorn.karlsson at readsoft.com, gregod at cs.rpi.edu, wekempf at cox.net