Question

I have a list of 2d numpy arrays. As a test, consider the following list:

lst = [np.arange(10).reshape(5,2)]*10

Now I can get at a particular data element by:

lst[k][j,i]

I would like to convert this to a numpy array so that I can index it:

array[k,j,i]

i.e., the shape should be (10, 5, 2).

This seems to work, but seems completely unnecessary:

z = np.empty((10,5,2))
for i,x in enumerate(z):
    x[:,:] = lst[i]

These don’t work:

np.hstack(lst)
np.vstack(lst)
np.dstack(lst)  #this is closest, but gives wrong shape (5, 2, 10)

I suppose I could pair a np.dstack with a np.rollaxis, but again, that doesn’t seem quite right …

Is there a good way to do this with numpy?

I’ve looked at this very related post, but I can’t quite seem to work it out.

Answer 1

This should work simply by calling the array constructor, i.e. np.array(lst).

>>> l = [np.arange(10).reshape((5,2)) for i in range(10)]
>>> np.array(l).shape
(10, 5, 2)