On Monday 19 July 2004 12:14 pm, Guillaume Melquiond wrote:
I think I understand the representation. But I don't even understand how the previous code could pass the regression test. I think it was only pure luck.
Seems like pure luck to me :)
I'm currently looking at correctly defining the iterator (2) so that it goes as far as necessary. But I wonder if anybody ever used the BGL with undirected graphs. I just spotted another obvious mistake: if you use a constant graph, get_edge(u,v) is given by "m_matrix[u * (u - 1)/2 + v]", but if you use a non-constant graph the result is "m_matrix[u * (u + 1)/2 + v]". So you will not get the same result depending on the const-ness of your adjacency matrix... I will also submit a patch for this.
Thanks! Doug