Re: [Boost-users] [boost-users][graph] There is graph::null_vertex, but no "null_edge"!

Subject: Re: [Boost-users] [boost-users][graph] There is graph::null_vertex, but no "null_edge"!
From: Jeremiah Willcock (jewillco_at_[hidden])
Date: 2011-02-17 18:04:35

On Fri, 18 Feb 2011, al.zatv wrote:

> Jeremiah Willcock <jewillco_at_[hidden]> писал(а) в своём письме Fri, 18 Feb 
> 2011 01:47:39 +0300:
>>> There is graph::null_vertex, but no "null_edge".
>
>> It looks like you are correct.  The BGL functions that can return 
>> non-existent edges return iterators instead, in which case edges(g).end() 
>> is a legitimate "null" return value.  If you know the graph is non-empty, 
>> you can also return std::pair<edge_descriptor, bool>, returning 
>> make_pair(*edges(g).first, false) for the "not found" case and 
>> make_pair(..., true) otherwise.
>
> Thank you for answer! I did it this way (for not founded edge)
>  return make_pair( false, Graph::edge_descriptor() );
>
> Am I correct?

No, that won't work in general; I believe there are graph types whose 
edge_descriptors are not default constructible.

> But, I think, it will be much better if null_edge will be added to the 
> library. It can be -1 for vector containers and null for pointer-based 
> descriptors, etc.

It would be, but that would be a change to base concepts that may cause 
old graph types to fail with new algorithms that use null_edge().  What 
does your code that may return a null edge descriptor do?  Would it be 
possible to return an iterator instead?

-- Jeremiah Willcock