vector.array method has unexpected behavior with lists of lists vs. tuples

[denehoffman]

Vector Version

Vector 1.0.0

Python Version

Python 3.10.5

OS / Environment

I'm operating on MacOS 13.1, but I can confirm this happens on CentOS7 so it's probably platform independent

Describe the bug

In the tutorial, there is the following example:

# NumPy-like arguments (literally passed through to NumPy)
vector.array(
    [(1.1, 2.1), (1.2, 2.2), (1.3, 2.3), (1.4, 2.4), (1.5, 2.5)],
    dtype=[("x", float), ("y", float)],
)

This seems like it could be a convenient way to convert a numpy array to an array of vector objects. However, this function seems to only play nice if the subitems are tuples and the object is a list. For example, see the following code:

>>> import numpy as np
>>> import vector
>>> v = np.array([[1.1, 2.1], [1.2, 2.2], [1.3, 2.3], [1.4, 2.4], [1.5, 2.5]])
>>> vector.array(v, dtype=[('x', float), ('y', float)])
VectorNumpy2D([[(1.1, 1.1), (2.1, 2.1)],
               [(1.2, 1.2), (2.2, 2.2)],
               [(1.3, 1.3), (2.3, 2.3)],
               [(1.4, 1.4), (2.4, 2.4)],
               [(1.5, 1.5), (2.5, 2.5)]], dtype=[('x', '<f8'), ('y', '<f8')])
>>> v = np.array([(1.1, 2.1), (1.2, 2.2), (1.3, 2.3), (1.4, 2.4), (1.5, 2.5)])
>>> vector.array(v, dtype=[('x', float), ('y', float)])
VectorNumpy2D([[(1.1, 1.1), (2.1, 2.1)],
               [(1.2, 1.2), (2.2, 2.2)],
               [(1.3, 1.3), (2.3, 2.3)],
               [(1.4, 1.4), (2.4, 2.4)],
               [(1.5, 1.5), (2.5, 2.5)]], dtype=[('x', '<f8'), ('y', '<f8')])
>>> vector.array(v.tolist(), dtype=[('x', float), ('y', float)])
VectorNumpy2D([[(1.1, 1.1), (2.1, 2.1)],
               [(1.2, 1.2), (2.2, 2.2)],
               [(1.3, 1.3), (2.3, 2.3)],
               [(1.4, 1.4), (2.4, 2.4)],
               [(1.5, 1.5), (2.5, 2.5)]], dtype=[('x', '<f8'), ('y', '<f8')])

In all of these cases, I expect the result from the tutorial example, and I think a reasonable user would also expect that. Maybe there's a preferred way of doing this, I could transpose the numpy array and call it like

vector.array({"x": v.T[1], "y": v.T[2], "z": v.T[3], "t": v.T[0]})

but my point is that a user reading the docs (me) would try the first example, it wouldn't fail immediately, but it would give a very wrong result/shape.

P.S. I wasn't sure if I should classify this as a bug, since I've strayed from the actual example a bit, so if this is just the way it is and there's not a better way to do this, go ahead and close the issue!

Any additional but relevant log output

No response

[Saransh-cpp]

Thanks for pointing this out! I think this should be the intended behavior given that numpy works with structured arrays in a similar way -

In [1]: v = np.array([[1.1, 2.1], [1.2, 2.2], [1.3, 2.3], [1.4, 2.4], [1.5, 2.5]
   ...: ], dtype=[('x', float), ('y', float)])

In [2]: v
Out[2]: 
array([[(1.1, 1.1), (2.1, 2.1)],
       [(1.2, 1.2), (2.2, 2.2)],
       [(1.3, 1.3), (2.3, 2.3)],
       [(1.4, 1.4), (2.4, 2.4)],
       [(1.5, 1.5), (2.5, 2.5)]], dtype=[('x', '<f8'), ('y', '<f8')])

Moreover, VectorNumpy classes internally call numpy.array and pass a view of it to the users; hence, the behavior stays intact -

vector/src/vector/backends/numpy.py

Lines 940 to 965 in 44620fd

	class VectorNumpy2D(VectorNumpy, Planar, Vector2D, FloatArray): # type: ignore[misc]
	"""
	Two dimensional vector class for the NumPy backend. This class can be directly
	used to construct two dimensional NumPy vectors. For two dimensional Momentum
	NumPy vectors see :class:`vector.backends.numpy.MomentumNumpy2D`.
	Examples:
	>>> import vector
	>>> vec = vector.VectorNumpy2D([(1.1, 2.1), (1.2, 2.2), (1.3, 2.3), (1.4, 2.4), (1.5, 2.5)],
	... dtype=[('x', float), ('y', float)])
	>>> vec
	VectorNumpy2D([(1.1, 2.1), (1.2, 2.2), (1.3, 2.3), (1.4, 2.4), (1.5, 2.5)],
	dtype=[('x', '<f8'), ('y', '<f8')])
	"""
	ObjectClass = vector.backends.object.VectorObject2D
	_IS_MOMENTUM = False
	_azimuthal_type: type[AzimuthalNumpyXY] | type[AzimuthalNumpyRhoPhi]
	def __new__(cls, *args: typing.Any, **kwargs: typing.Any) -> VectorNumpy2D:
	"""Returns the object of ``VectorNumpy2D``. Behaves as ``__init__`` in this case."""
	if len(args) == 1 and len(kwargs) == 0 and isinstance(args[0], dict):
	array = _array_from_columns(args[0])
	else:
	array = numpy.array(*args, **kwargs)
	return array.view(cls)