[illumen]

I guess it would be possible to implement with a with-hack...

with numpy.doing_matrix:

    matrixy_multiply_goodness = m1 * m2

array_multiply_goodness = a1 * a2