How to format numpy arrays?

Question

I have data that looks like this

listL= array([array([-12939613.07220617,   3962855.50690994]),
       array([-12939614.67349505,   3962828.80807231]),
       array([-12939484.00289515,   3962828.1637531 ]),
       array([-12939484.98046737,   3962854.91251931]),

print(type(listL)) -> <class 'numpy.ndarray'>
print(type(listL[0])) -> <class 'numpy.ndarray'>
print(type(listL[0][0])) -> <class 'numpy.float64'>

I have a second piece of data that looks like this:

A = array([[ 73.87682896,   8.55827956],
       [ 57.43741519,  10.40224912],
       [ 87.88970753,  75.42971056],

print(type(A)) -> <class 'numpy.ndarray'>
print(type(A[0])) -> <class 'numpy.ndarray'>
print(type(A[0][0])) -> <class 'numpy.float64'>

The types are the same across both sets of data, but I have a function that works with 'A', but not with 'listL' and I cannot figure out why.

A[spatial.KDTree(A).query(coordinate)[1]]

works, but

listL[spatial.KDTree(listL).query(coordinate)[1]]

returns the error:

not enough values to unpack (expected 2, got 1)

The type doesn't help much. Look at shape and dtype instead. listL Iwhich is missing a ])) is a 1d object dtype array (shape (4,). A is a 2d array of numbers, shape (3,2). np.stack(listL) should produce a (4,2) numeric array. — hpaulj
– hpaulj, Commented Jun 8, 2018 at 4:20

Paul Panzer · Accepted Answer · 2018-06-08 04:39:58Z

ListL is in some respects a pathological structure, it is an array of arrays where a 2D array is expected.

numpy/scipy goes to some lengths to graciously accept most things that can be interpreted as 2D array, for example if you repace KDTRee(ListL) with KDTree(list(ListL)) it works.

Why does it work? Because list(listL) behaves like most "essentially 2D" structures (like lists of lists etc.). We can send it through array or asarray or asanyarray etc. to obtain a genuine 2D array.

>>> np.array(list(listL))
array([[-12939613.07220617,   3962855.50690994],
       [-12939614.67349505,   3962828.80807231],
       [-12939484.00289515,   3962828.1637531 ],
       [-12939484.98046737,   3962854.91251931]])
>>> np.array(list(listL)).shape
(4, 2)

An array of arrays is one of the few cases left that still trip up the array conversion machinery:

>>> np.array(listL)
array([array([-12939613.07220617,   3962855.50690994]),
       array([-12939614.67349505,   3962828.80807231]),
       array([-12939484.00289515,   3962828.1637531 ]),
       array([-12939484.98046737,   3962854.91251931])], dtype=object)
>>> np.array(listL).shape
(4,)

We can see despite our attempt to convert to an ordinary array, listL keeps reporting its shape as 1D. Which appears to be what triggers the exception you are observing.

Collectives™ on Stack Overflow

How to format numpy arrays?

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related