def test_grad(self,field_def,grad_def):
size = np.random.random(3)+1.0
grid = np.random.randint(8,32,(3))
nodes = grid_filters.cell_coord0(grid,size)
my_locals = locals() # needed for list comprehension
field = np.stack([np.broadcast_to(eval(f,globals(),my_locals),grid) for f in field_def],axis=-1)
field = field.reshape(tuple(grid) + ((3,) if len(field_def)==3 else (1,)))
grad = np.stack([np.broadcast_to(eval(c,globals(),my_locals),grid) for c in grad_def], axis=-1)
grad = grad.reshape(tuple(grid) + ((3,3) if len(grad_def)==9 else (3,)))
assert np.allclose(grad,grid_filters.gradient(size,field))
def test_add_gradient(self, default, shape):
if shape == 'pseudo_scalar':
default.add_calculation('#F#[:,0,0:1]', 'x', '1',
'a pseudo scalar')
if shape == 'scalar':
default.add_calculation('#F#[:,0,0]', 'x', '1', 'just a scalar')
if shape == 'vector':
default.add_calculation('#F#[:,:,1]', 'x', '1', 'just a vector')
x = default.place('x').reshape((np.product(default.cells), -1))
default.add_gradient('x')
in_file = default.place('gradient(x)')
in_memory = grid_filters.gradient(
default.size, x.reshape(tuple(default.cells) +
x.shape[1:])).reshape(in_file.shape)
assert (in_file == in_memory).all()