From: alexanderdz_at_[hidden]
Date: 2001-07-25 03:19:32
Hi Jeremy,
--- In boost_at_y..., Jeremy Siek <jsiek_at_r...> wrote:
> On Tue, 24 Jul 2001 alexanderdz_at_y... wrote:
> >
> > I wrote test program which you can find below and found following
> > problems:
> >
> > A. The depth_first_search algorithm always calls back_edge method
...
>
> If you record predecessors you can check if the edge passed to
> back_edge() is really a tree edge. Perhaps this check should be
> encorporated into depth_first_search().
I would like to. Otherwise it require me to dig to the roots of
graph theory and implementation of concrete algorithms. Which I was
lazy to do.
> Anyways, here's how you
> could change your cycle_detector visitor to do this:
>
> template <class ParentMap>
> struct cycle_detector : public dfs_visitor<>
...
Thanks! It works.
> > B. The library quite often brings compiler to its knees. For
> > example, if you comment the line following "// Internal compiler
> > error if commented out", VC 6 sp5 fails with internal compiler >
> > error. ...
>
> I've tried to avoid as many VC++ internal compiler errors as
> possible, and have spent a lot of time doing this, but I'm afraid I
> can't dedicate more time to this. Please report internal compiler
> errors to the VC++ team.
I'll do that. Unfortunately, it can't help me. Hope to find the
workaround next time.
> > C. To got compile error in line B if I pass visitor the way
> > analogous to line A. To be more precise, I can't write line B
> > following way, as I did for line A
> >
> > vertex_processor proc;
> > depth_first_visit(g, vertices(g).first[1], visitor(proc), ...
>
> The reason is that the depth_first_visit() function does not use
> named template parameters, but just normal template parameters. So
> it is incorrect to pass visitor(proc) to depth_first_visit().
>
> http://www.boost.org/libs/graph/doc/depth_first_visit.html
What is the reason having different interfaces for two closely
related functions?
Thank you for your help!
With best wishes,
Alexander