I have defined a custom dtype. For example:
vec = np.dtype([('x', float), ('y', float), ('z', float)])
quat = np.dtype([('w', float), ('v', vec)])
Now I want to make a scalar quaternion:
quat((1.0, (0.0, 0.0, 0.0)))
I would expect that if anything, my tuple syntax is unacceptable. However, instead, I get the following error:
TypeError: 'numpy.dtype' object is not callable
The relevant portion of the documentation on scalars implies that it is possible to have a scalar of a structured type built like this in numpy.
How do instantiate a quat scalar? Is it even possible?
By the way, I've played with the following workaround:
np.array([(1.0, (0.0, 0.0, 0.0))], dtype=quat)
This does not produce an actual scalar (although it honestly works well enough for my purposes, making the question mostly theoretical). Calling item on the result returns a tuple, not a scalar quat object.
Calling item on your array produced a tuple because item is specifically designed to convert NumPy types to Python types. Indexing the array produces a NumPy scalar of type numpy.void:
scalar = np.array([(1.0, (0.0, 0.0, 0.0))], dtype=quat)[0]