roadmap.md

Dimensionality Broadcasting

Automatically broadcast scalar functions to arbitrary dimensions, for example:

def my_demo(a, b, c):
    return a*a + b*b + c*c

Can be applied arbitrarily to:

arr = np.zeros((1_000_000, 3))
...

but also can be applied to

arr = np.zeros((1_000_000, 3, 500, 1000))

without changing the code.

Inline function definitions

Currently defining a function inline such as

@fs
def parent(a):
    return child(a) * 2

def my_massive_function():
    x = 4

    def cube(a):
        return a * a * a

    return parent(x, child=cube)

will fail because the cube inline function is not within the parent function's scope when the AST is re-built within fastats.