Andrey Mokhov

@jmatsushita Yes, I've been thinking about such representations to support hyperedges, however, I used lists in my experiments because with `Set` you essentially force the algebra to be commutative with...

I suggest we keep this PR open, since hypergraphs seem like an interesting direction to explore.

> there could be two graphs with different mappings, right? Yes. > I guess we could combine `index` functions but I don't know what to do with `value` which seems...

> That's right. We know that `AdjacencyIntMap` is fast and we'd like to benefit from its efficiency when working with more complex vertex types `a`. Re-reading what I wrote, I...

@michaelpj Thanks for the ideas! Most likely, we don't really need to make operations like `overlay` and `connect` really fast, because constructing large adjacency maps using these operations is going...

@Avasil Why don't you start with something naive but very simple: just renumber everything (always keeping the indices starting at 0) whenever you do something like `overlay` or `connect`. Then...

@Avasil Thanks for the experiment! I don't have time to look at your implementation in detail at the moment, but it's indeed strange that `vertexCount` is so slow, e.g. there...

@Avasil Is there a chance to bring the overhead in `addVertex` and `addEdge` down to 3-5x?

We expect the type of edge labels to be at least a monoid, and the identity element of this monoid corresponds to the non-edge. You can therefore implement `noEdge`: ```haskell...

> I tried to watch the video but it cannot be played because of "privacy issues". @Josef-Vonasek Strange, I just tried and it worked fine for me. It does require...