Your comments seem to be nothing more than "I understand infinite dimension linear algebra better than Haskell", which is fine but not insightful. I would argu

Your comments seem to be nothing more than "I understand infinite dimension linear algebra better than Haskell", which is fine but not insightful. I would argue that having to drag around the notion of everything being indices + projection tuples constructed from a universal, infinite dimension matrix to be needlessly complex compared to the presented version. This only seems to get worse when you generalize to multiple edges between a pair of vertices, with your rank 3 infinite tensor being n x n x m, and requiring a canonical way to index edges.