Numpy np.array with dtype TypeError

Question

Description

I have spent about an hour trying to hunt down a numpy array construction error. I must not be using numpy dtypes correctly but the error message is not descriptive enough and I am not given a good enough stack trace to find the error.

Simplified Example that creates same error:

 import numpy as np
 names = ['id', 'x']
 formats = [np.int64, np.float64]
 np.array([1, 1.0], dtype={'names': names, 'formats': formats})

The following code results in the error

 ----> 1 np.array([1, 1.0], dtype={'names': names, 'formats': formats})
 TypeError: a bytes-like object is required, not 'int'

So i get that the error occurs from the first element being an int but why is it expecting a bytes like object?

Answer: it has nothing to do with the fact that the first element is in int. The list needs to be a tuple see below.

Trying to construct a record array figured out the solution and it is not at all obvious. Numpy need to fix this. — costrouc
– costrouc, Commented Mar 16, 2017 at 21:16

costrouc · Accepted Answer · 2017-03-16 21:17:38Z

6

The solution is that numpy requires a single tuple or list of tuples. Not a list of lists or list.

The following works

 np.array((1, 1.0), dtype={'names': names, 'formats': formats})

I find it frustrating that numpy gives no indication in the error that this is what is expected.

answered Mar 16, 2017 at 21:17

costrouc

3,1753 gold badges21 silver badges24 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

hpaulj Over a year ago

That requirement isn't as explicit on this docs page as it could be: docs.scipy.org/doc/numpy/user/basics.rec.html. There is also an example without commentary in docs for np.array: x = np.array([(1,2),(3,4)],dtype=[('a','<i4'),('b','<i4')])

Collectives™ on Stack Overflow

Numpy np.array with dtype TypeError

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related