Hi Abel,
As long as your x,y,z are next to each other, you can transform from
your structure to an unstructured array via a view, which has very
little cost. Though you need to be a bit careful with offsets, etc., if
there are also other elements in the structured dtype.
Example, with some extra fields:
dtype = np.dtype([("i", np.int64), ("x", np.float64), ("y", np.float64), ("z",
np.float64), ("j", np.int64)])
atoms = np.array(
[
(1, 0.0, 0.0, 0.0, -1),
(2, 1.0, 0.0, 0.0, -2),
(3, 0.0, 1.0, 0.0, -3),
(4, 1.0, 1.0, 1.0, -4),
],
dtype=dtype,
)
dt2 = np.dtype([("i", np.int64), ("xyz", np.float64, (3,)), ("j", np.int64)])
xyz = atoms.view(dt2)["xyz"]
xyz
# array([[0., 0., 0.],
# [1., 0., 0.],
# [0., 1., 0.],
# [1., 1., 1.]])
xyz[:] = 9.
atoms
array([(1, 9., 9., 9., -1), (2, 9., 9., 9., -2), (3, 9., 9., 9., -3),
(4, 9., 9., 9., -4)],
dtype=[('i', '<i8'), ('x', '<f8'), ('y', '<f8'), ('z', '<f8'), ('j',
'<i8')])
All the best,
Marten
[email protected] writes:
> I'm using structured arrays to store atoms data produced by LAMMPS (I'm using
> a structured array that follows its format). I need to rotate the positions:
>
> ```
> import numpy as np
>
> transform = np.array([[1, 0, 0], [0, 1, 0], [0, 0, 1]], dtype=np.float64)
> dtype = np.dtype([("x", np.float64), ("y", np.float64), ("z", np.float64)])
> # real case with more fields, integers, bools, strings
>
> atoms = np.array(
> [
> (0.0, 0.0, 0.0),
> (1.0, 0.0, 0.0),
> (0.0, 1.0, 0.0),
> (1.0, 1.0, 1.0),
> ],
> dtype=dtype,
> )
>
> atoms[["x", "y", "z"]] = atoms[["x", "y", "z"]] @ transform.T
> ```
>
> But this produces:
>
> ```
> Traceback (most recent call last):
> File "c:\Users\acgc99\Desktop\rotation.py", line 16, in <module>
> atoms[["x", "y", "z"]] = atoms[["x", "y", "z"]] @ transform.T
> ~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~
> numpy._core._exceptions._UFuncNoLoopError: ufunc 'matmul' did not contain a
> loop with signature matching types (dtype([('x', '<f8'), ('y', '<f8'), ('z',
> '<f8')]), dtype('float64')) -> None
> ```
>
> I can convert to unstructured arrays, but I guess that doing that change
> multiple times is not efficient when working with tens of millions of atoms.
> _______________________________________________
> NumPy-Discussion mailing list -- [email protected]
> To unsubscribe send an email to [email protected]
> https://mail.python.org/mailman3//lists/numpy-discussion.python.org
> Member address: [email protected]
_______________________________________________
NumPy-Discussion mailing list -- [email protected]
To unsubscribe send an email to [email protected]
https://mail.python.org/mailman3//lists/numpy-discussion.python.org
Member address: [email protected]
