Etienne dg

Your problem is not the same as above, I think you misunderstand the entries of the algorithm. You only have positive demands, so no sinks. As the flow has nowhere...

I don't think the choice of the priority queue is the most important detail of the algorithm, we could use the standard Priority queue of DataStructures, then switch to a...

Ah, there is a [PR for a Fibonacci heap](https://github.com/JuliaCollections/DataStructures.jl/pull/592), but it seems to be[ stuck for the moment.](https://github.com/JuliaCollections/DataStructures.jl/issues/500#issuecomment-726719542)

The bipartite case is the assignment problem, it is solved by the Hungarian algorithm (we have a pure julia implementation). For the general case, it could be nice to rely...

You have `maximum_weight_maximal_matching_hungarian` for bipartite graphs, which should be more efficient. (negate weights to minimize instead of maximize).

I agree that this package is severely undocumented (and in bad shape overall), I will try to fix it. ``` julia help?> MatchingResult search: MatchingResult struct MatchingResult{U} weight::U mate::Vector{Int} end...

In the API, the `weights` function return a weight matrix. This is used in a lot of place in the codebase and probably in a lot of users code. In...