Loïc Joly <loic.joly <at> reportive.com> writes:
Daniel Mitchell a écrit :
I'm not sure I'd call that a bug; after all, the graph must contain at
least a
source vertex for DFS to be sensible. I would say that having a non-empty graph is a pre-condition for calling DFS.
It may be so for depth-first-search, I'm not a domain expert at all. In that case, maybe just adding this requirement to the doc would be enough.
However, for topological_sort, this goes against my expectations. std::sort on an empty STL sequence may be useless, it is however defined and does the right thing. And it allows to write generic code easily, without requiring the user to check anything before. I do not see any reason why topological_sort would not behave the same.
I guess adding a simple test such as if( vertices( g ).first == vertices( g ).second ) return; at the beginning of topological_sort wouldn't hurt.