Hi @eriknw

Oops! I stretched the infix notation a bit too far by testing the following line, which is an expression one would often encounter when transforming matrix-quantities (like graph adjacency-matrices, for example):

AA << op.plus_times(P.T @ A @ P)

where A, AA and P are all matrices. (That is, A is transformed by P into AA. P could be a permutation matrix, for example.)

The expression was executed without throwing an exception, but the result was incorrect. To be precise, the structure of the answer was correct but the values were all 0 (which was incorrect).

Of course, the right way to do it was this:

AA << op.plus_times(P.T @ op.plus_times(A @ P))

which gave the right answer.

But it would be great if the former notation was also possible. What do you think?