There are no plans to include such a function, but pull requests are welcome.

Some notes:

• This is called the joint degree matrix in graph theory (when not normalized), and sometimes studied similarly to degree sequences. This name should be at least mentioned, or perhaps used for the function.
• It is beneficial to return a sparse matrix result for efficiency (think scale-free degree distribution)
• In view of the above (and of the simplicity of implementation), it may be a better idea to implement it directly in the higher level interfaces (R/Python) and use the native or common sparse matrix type of the language. IGraph/M (Mathematica) already has it, and C implementation probably wouldn't be noticeably more performant.

We have a sparse matrix data type in the C layer as well, so we can make use of that if we decide that it should be implemented in the C layer. But, at least in Python, there is no direct dependency between the Python-C glue code and NumPy, and I want to keep it that way, so the best that the Python-C glue code can do (without depending on NumPy) is to return a list of (outdeg, indeg, count) triplets that can then be fed into a sparse matrix constructor. The Python interface can then import numpy dynamically at runtime and convert it into a NumPy sparse matrix if it is installed (and throw an ImportError otherwise).

If this is implemented, it should have an option to (optionally!) set the maximum degree for the result. This is useful when generating joint degree matrices for several graphs, and we want them to be of compatible size so they can operated (e.g. added) together.

If the maximum degree of the graph is smaller than what is requested, the result will simply be padded with zeros.

This advice is based on practical experience from using a joint degree matrix function.

Example: