Create a numpy array of arrays of similar shapes (from list)

Asked Mar 04 '20 at 10:57

Active Mar 04 '20 at 14:23

Viewed 19 times

I need to convert a list of numpy arrays into an array of arrays in order to save it using numpy.save (calling save direclty on the list would sort of call numpy.array anyway). When the shapes of the arrays in the list are dissimilar, everything goes fine. For example:

data = np.array([np.zeros((18, 10)), np.zeros((19, 11))])

works, and data has shape (2,), which is what I expect. However, when the shapes of the arrays are the same, for example

data = np.array([np.zeros((18, 10)), np.zeros((18, 10))])

data get the shape (2, 18, 10). How to prevent this behavior?

Even worse, if I only have the same shape in axis 0:

data = np.array([np.zeros((18, 10)), np.zeros((18, 11))])

I get an error:

ValueError: could not broadcast input array from shape (18,10) into shape (18)

Any idea? Thank you for your help!

edited Mar 04 '20 at 14:23

asked Mar 04 '20 at 10:57

Son Pham-Ba

Yes, that's the behavior of `np,array` that we've discussed in many SO. To make an object dtype array in all those cases you have to first make an object dtype array of the right size, and then fill it with the list. – hpaulj Mar 04 '20 at 11:40
Have you considered `pickle` as an alternative to `save`? – hpaulj Mar 04 '20 at 11:41
Please add a python tag to your question – Mad Physicist Mar 04 '20 at 13:09
Also related: https://stackoverflow.com/q/49064548/2988730 – Mad Physicist Mar 04 '20 at 13:10
Thanks for your answers. I opted for `np.array([None, *list_of_arrays])[1:]` – Son Pham-Ba Mar 04 '20 at 14:35

Create a numpy array of arrays of similar shapes (from list)

0 Answers0