Tamás Nepusz

Hmm, okay, what shall the fix be then? Am I right to think that if there are _u_ sources in $S$, _v_ targets in $T$ and an intersection size of...

As a stopgap measure, maybe we should handle the case of full overlap and no overlap properly and print a warning for partial overlaps until we work the math out?

Maybe [this one](https://snap.stanford.edu/class/cs224w-readings/brandes01centrality.pdf)? Also, two more things: 1. I'm not sure whether the current imlpementation works correctly if a vertex appears more than once in the source selector. This needs...

Case of `igraph_maximum_bipartite_matching` and `igraph_i_maximum_bipartite_matching_weighted`: this is not a true recursion. `igraph_i_maximum_bipartite_matching_weighted()` calculates an initial maximum matching on a set of _tight_ edges; without going into details, this is always...

Case of `igraph_maxflow()` and `igraph_i_maxflow_undirected()`: this is not a true recursion either. `igraph_maxflow()` calls `igraph_i_maxflow_undirected()` for an undirected graph, which converts the graph into directed, and calls back to `igraph_maxflow()`....

Also, passing `duplicate=FALSE` as an argument to `graph_from_adj_list()` also produces a correct graph. So it's really the error check that's missing here.

For the record: the function corresponding to the `graph_from_adj_list()` function in R is `igraph_adjlist()` in the C core. I don't really like the name of this function either because we...

I've been wanting to add this for a long time now so I totally support the idea.

`Graph.automorphism_group()` is now exposed in the Python interface.

Isn't this basically the same as `igraph_automorphism_group()`? What are the key differences?