Dan Schult

Thanks @stefanv I have made your recommended changes and in the process found some more.

`A1d @ A1d` in NumPy returns a scalar. But the array-api says it should return a 0-dim array. Which should sparse arrays return? I think they should match the array-api...

Let's leave off the absolute value -- round-off error is something we have to live with for floating point computations. And who knows if someone might actually need to compute...

I'm working on yet another review... :}

Thanks for these changes -- I will take care of the deprecations. I wasn't expected the name change from root to leaf. I was expecting the name root to stay...

aha! Thanks for pointing out my error... I had the adjacency entry A_ij representing an edge from j to i instead of i to j. This is the reverse of...

The merge conflicts occur when another PR that touches the same code this PR touches is merged. That means `main` changes in a way that git can't figure out whether...

Is there a reason you are using Graphviz 2.43.0? The current version is 9.0.0. I don't know that it makes any difference, but it seems like an obvious place to...

Ahhh... whoops... I got distracted by the version number and lost track of the issue itself. My apologies. To allow multiedges in pygraphviz the AGraph must have its `strict` parameter...

No need to open a new PR. Its a busy time of year, but we'll get to it. Thanks!!