Chase Riley Roberts

`tn.conj`?

I don't understand what you mean.

> You could not live with your own failed OO design. And where did that bring you? Back to me. -- The TensorNetwork class

And I like all these ideas. My only request is that the NconBuilder remains functional. (Immutable, functions have no side effects)

This seams very much in the spirit of the "functional node" design we discussed months prior. The idea was very similar to what you are describing, only instead of wrapping...

Ultimately, it seems like it would essentially be a shim to just calling ncon outright with a different user facing API. I welcome PRs for new desired interfaces! No reason...

> While I'm all in favor of developing ideas for new interfaces, I think we should avoid having too many different APIs available to the user in parallel. I would...

Yeah Glen's answer might be the correct one. We currently disallow 0, so reallowing it and always making it a dangling leg might be ok.

@alewis @mganahl Didn't 0 have a special meaning in the original matlab ncon?

@repst Here is a short example of applying an H gate twice. ```python a = np.array(...) H = np.array(...) result = tn.ncon([a, H, H], [[1], [1, 2], [2, -1]]) ```