It may sound silly but it may be of interest to be able to compute the norm of [] so that one has not to catch this case before (would make my code more readable).

At the moment numpy.linalg.norm(numpy.array([]),ord=):

• works with ord = 0,1,2,... giving the result 0. Although here is a incosistency returning np.int in the case of ord = 0 and np.float in the others.
• works not not with ord = numpy.inf since this computes abs(x).max() without checking if x is nonempty.

It would be very nice, if in the [] case also for ord = inf one would return 0, which would be consistent with matlab behaviour.