Ross Barnowski

This is working as expected and should result in a minor improvement in doc build times. Two important lessons are: 1. The parallel execution [doesn't seem to drastically increase the...

> Should I think about adding unit tests for these cases such that changes do not cause breaking changes on previously working code? Yes, additional regression tests are always welcome!...

Adding a `nonedge` parameter to `from_numpy_matrix` seems like a reasonable feature request to me. As you allude to, it's straightforward to imagine use cases where the lack of an edge...

Good point --- thanks Jarrod: all of the above references to `to/from_numpy_matrix` should be replaced by `to/from_numpy_array`. > from_numpy_array has a nonedge parameter. Not that I see: https://github.com/networkx/networkx/blob/c73d4ab4e30c5d0dea0aa58e6d22ccb20061fbe3/networkx/convert_matrix.py#L1255

~- [ ] Request extended `amsmath` support for LaTeX math rendering. See e.g. https://numpy.org/numpy-tutorials/mooreslaw-tutorial/#building-moores-law-as-an-exponential-function~ added this to the list above (bsipocz)

> The latex issue would also solve a bunch of the errors we get for the strict run. And I'm sure there is a solution for it already, I just...

> this is the only change we need to make in order for our code to be accepted? No - generally there is no objective criterion for getting code into...

It's tough to tell what's supposed to happen here - from the traceback, it looks like this is *expected* to "fail" (i.e. raise an exception). In that sense, raising the...

> This behaviour (copying the mask) is counterintuitive I'm sure there are users out there who depend on this behavior though. I don't think it's particularly counterintuitive that masks are...

Also true - my point was more from the perspective of backwards compatibility. I wouldn't necessarily characterize the fact `ma.zeros_like` preserves the input mask (when the input is a masked...