Interpolant returns inconsistent results

[cdhainaut]

Interpolant returns inconsistent results

[b-mare]

hi @cdhainaut

The way you provided the flattened arrays to underlying_function_3D to compute y_3d does not match how the interpolant function creates the grid.

This should work for your 3D example:

import numpy as np
from casadi import interpolant

x1 = np.linspace(0, 100, 2)
x2 = np.linspace(0, 5, 3)
x3 = np.linspace(0, 9, 4)

X1, X2, X3 = np.meshgrid(x1, x2, x3, indexing="ij")
y_3d = X1**2 + (X2 + 1) ** 2 + X3**2
y_3d_flat = y_3d.ravel(order='F')
X_3d = np.stack((X1.ravel("F"), X2.ravel("F"), X3.ravel("F")))

f_3d = interpolant(f"three_d", "linear", [x1, x2, x3], y_3d_flat)
y_i_3d = f_3d(X_3d)
np.testing.assert_allclose(y_3d_flat, np.array(y_i_3d)[0])