numpy array equivalent for += operator

Question

I often do the following:

import numpy as np

def my_generator_fun():
    yield x # some magically generated x

A = []
for x in my_generator_fun():
    A += [x]
A = np.array(A)

Is there a better solution to this which operates on a numpy array from the start and avoids the creation of a standard python list?

Note that the += operator allows to extend an empty and dimensionless array with an arbitrarily dimensioned array whereas np.append and np.concatenate demand for equally dimensioned arrays.

possible duplicate of http://stackoverflow.com/questions/4535374/initialize-a-numpy-array — John Ruddell, Jul 15 '14 at 15:06
no, my point is slightly different: image the case where you successively built up the array or you want to initialize each element differently. However, I really need the exactly same functionality, not only the same result. — morph, Jul 15 '14 at 15:13
why are you even doing a loop with this? why not just make `A = [[0,1],[1,2],[3,4]]` ? — John Ruddell, Jul 15 '14 at 15:19
another possible duplicate.. you really need to research what you are looking for first... http://stackoverflow.com/questions/10346336/list-of-lists-into-array-in-numpy — John Ruddell, Jul 15 '14 at 15:20
i just edited my query again.. maybe something like that? not sure what you are getting returned from your function — John Ruddell, Jul 15 '14 at 15:44

score 4 · Accepted Answer · answered Jul 15 '14 at 17:27

Use np.fromiter:

def f(n):
    for j in range(n):
        yield j

>>> np.fromiter(f(5), dtype=np.intp)
array([0, 1, 2, 3, 4])

If you know beforehand the number of items the iterator is going to return, you can speed things up using the count keyword argument:

>>> np.fromiter(f(5), dtype=np.intp, count=5)
array([0, 1, 2, 3, 4])

score 0 · Answer 2 · answered Jul 15 '14 at 15:10

0

To get the same array A, do:

A = numpy.arange(5)

Arrays are not in general meant to be dynamically sized, but you could use numpy.concatenate.

answered Jul 15 '14 at 15:10

mdurant

27,272
5
45
74

I see your requirement has changed - what in general are you trying to achieve? – mdurant Jul 15 '14 at 15:12
basically said, i will achieve the same functionality like I know it from the += operator in basic python lists for numpy arrays – morph Jul 15 '14 at 15:36
@morph look into `np.hstack` and `np.vstack` – Paul H Jul 15 '14 at 15:37

numpy array equivalent for += operator

2 Answers2