Add arrays to the numpy array

Question

I try to add newly created arrays to other numpy array, but I'm doing something wrong. What I want, is to add multiple arrays like numpy.array([0, 1, 2, 3]), to already created array, so I could get something like this:

x = numpy.array([])

for i in np.arange(5):
    y = numpy.array([0, 1, 2, 3])
    x = np.append(x, y)

result:
x = [0, 1, 2, 3],
    [0, 1, 2, 3],
    [0, 1, 2, 3],
    [0, 1, 2, 3],
    [0, 1, 2, 3]

However, with the loop shown above I get this:

x = [0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3]

Bi Rico · Accepted Answer · 2015-02-11 19:05:27Z

4

Try this:

x = []
for i in range(5):
    y = numpy.array([0, 1, 2, 3])
    x.append(y)

x = numpy.array(x)

or:

N = 5
x = numpy.zeros((N, 4))

for i in range(N):
     x[i] = numpy.array([0, 1, 2, 3])

Here I avoid numpy.append and numpy.vstack inside the loop because it can be quite slow. Every call to numpy.append or numpy.vstack creates an empty array and copies both x and y into the new empty array. If you use a list to hold the rows of array until the loop is over, the array just gets copied once at the end.

If neither of the above work for you, you could do something like this (but it'll be slower):

x = numpy.zeros((0, 4))

for i in range(5):
     y = numpy.array([0, 1, 2, 3])
     x = numpy.vstack(x, y)

edited Feb 11, 2015 at 19:05

answered Feb 11, 2015 at 18:30

Bi Rico

25.9k3 gold badges57 silver badges75 bronze badges

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

1 Comment

TransferOrbit Over a year ago

For what it’s worth, I believe the last, suboptimal example should read: for i in range(5): y = numpy.array([0, 1, 2, 3]) x = numpy.vstack((x, y))

Alex Riley · Accepted Answer · 2015-02-11 18:49:33Z

append adds to the end of the array. Since x only has one dimension (it has shape (0,) to begin with) it can grow only in the way you observe.

It's not generally the right tool to use to build multi-dimensional arrays incrementally as you're doing - you can add append to a specific access (and so stack arrays) but you need to ensure that both arrays are the same shape, and same size along that axis. On top of this the array you're appending to must be copied each time.

A more succinct way to build your required array could be to use np.tile instead:

>>> np.tile([1, 2, 3, 4, 5], (5, 1)) # (5,1) means 5/1 copies along axis 0/1
array([[1, 2, 3, 4, 5],
       [1, 2, 3, 4, 5],
       [1, 2, 3, 4, 5],
       [1, 2, 3, 4, 5],
       [1, 2, 3, 4, 5]])

Collectives™ on Stack Overflow

Add arrays to the numpy array

2 Answers 2

1 Comment

Comments

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

1 Comment

Comments

Related