I have generated a numpy array of (x, y) values as a N x N grid.
grid = np.meshgrid(np.linspace(0, 1, 50), np.linspace(0, 1, 50))[0]
grid.shape // (50, 50, 1)
I have a function that takes two parameters and returns 3 values.
i.e. (x, y) -> (a, b, c)
How to I apply the function over the 2d numpy array to get a 3d numpy array?
If your function really takes two parameters you probably want to map not 2d to 3d, but rather 2xMxN to 3xMxN. For this change your first line to something like
gridx, gridy = np.meshgrid(np.linspace(0, 1, 50), np.linspace(0, 1, 50))
or even use the more economical ix_ which has the advantage of not swapping axes
gridy, gridx = np.ix_(np.linspace(0, 1, 50), np.linspace(0, 1, 50))
If your function f does not handle array arguments then as @Jacques Gaudin points out np.vectorize is probably what you want. Be warned that vectorize is primarily a convenience function it doesn't make things faster. It does useful things like broadcasting which is why using ix_ actually works
f_wrapped = np.vectorize(f)
result = f_wrapped(gridy, gridx)
Note that result in your case is a 3-tuple of 50 x 50 arrays, i.e. is grouped by output. This is convenient if you want to chain vectorized functions. If you want all in one big array just convert result to array and optionally use transpose to rearrange axes, e.g.
triplets_last = np.array(result).transpose((1, 2, 0))
If I understand correctly, you are after the np.vectorize decorator. By using it you can apply a function over a meshgrid. Your function should take only one parameter though as you do not pass the coordinates but the value at the coordinates (unless the values are tulpes with two elements).
import numpy as np
grid = np.meshgrid(np.linspace(0, 1, 5), np.linspace(0, 1, 5))[0]
@np.vectorize
def func(a):
return (a, a**.5, a**2)
res = np.array(list(func(grid)))
print(res.shape)
print(res)
(x, y)values looks like on anN x Ngrid... Is this to imply that you have a 2D array of tuples?