Guido van Rossum writes:
That's food for thought.
Thank you. Let me confirm to the proponents that "food for thought" is all I intended. I know a fair amount about statistics, and almost as much about linear algebra, but nothing about physics or engineering.
Certainly I trust [Steven D'Aprano] to come up with a reasonable strawman whose tires we can all kick.
I do, too. It's only fair to give him a preview of (some?) of the arguments against inclusion in the stdlib, that's all. ;-)
[W]hich operations from the OP's list need more than statistics._sum() when limited to NxM matrices for single-digit N and M? (He named "matrix multiplication, transposition, addition, linear problem solving, determinant.")
I believe determinant can be efficiently implemented with statistics._sum. Linear problem solving (to which I would add the closely related operation of square matrix inversion) involves the same kind of principle, but I don't think it can be implemented with statistics._sum. The same methods that one would use for solving/ inversion for small M and N will work efficiently for large M and N. And the algorithms are as obvious as statistics._sum (in hindsight).
I'm not sure whether matrices should try to implement all the different types that statistics._sum does, although I can imagine it might be pedagogically useful to support Fraction and Decimal.
Steve