Hello all,
numpy.vectorize is a very convenient way to build universal function from
scalar functions, and handle broadcasting... A use case that is not handled
is the generalization to functions already returning nd array:

For example: If foo(x,y) is a scalar function returning a array of dim k,
it would be convenient to create a vectorized version of it, such that if
the broadcasted shape of x and y has dimension d, then foo(x,y) returns a
(d+k)-dimensional array.

For example:
# Example 1
x = array([1,2,3])
y = array([4,5,6])
# foo(x,y)[i] is equal to foo(x[i], y[i]), a vector of dim k,
# Example 2
x = array([1,2,3])
y = array([4,5,6])
# foo(x,y[:, newaxis])[i,j] is equal to foo(x[i], y[j]), a vector of dim k,

Does any one know if there is already a way to do this?

Best,
Sylvain
_______________________________________________
NumPy-Discussion mailing list
[email protected]
http://mail.scipy.org/mailman/listinfo/numpy-discussion

Reply via email to