CODEWITHSUNDEEP
pythonnumpy
elbarto
elbarto

Reputation: 211

How to concatenate columns on top of each other in an array

Say I have the following array:

a = np.array([[1,3,5,2,3],[3,2,5,6,7],[1,7,3,6,5]]);

How can I stack the columns on top of each other to form a single column vector to produce the following?

b =np.array([[1], [3], [1], [3], [2], [7],[5],[5],[3],[2],[6],[6],[3],[7],[5]]);

Upvotes: 2

Views: 609

Answers (5)

B. M.
B. M.

Reputation: 18638

The shortest is a.T.reshape(-1,1) :

  • a.T for the good layout,
  • reshape for the good shape, with one column.

or equivalent : a.reshape(-1,1,order='F').

Upvotes: 0

hpaulj
hpaulj

Reputation: 231395

To stick with the stacking idea, a list of a.T is:

In [87]: list(a.T)
Out[87]: 
[array([1, 3, 1]),
 array([3, 2, 7]),
 array([5, 5, 3]),
 array([2, 6, 6]),
 array([3, 7, 5])]

which can then be concatenated on the one axis

In [90]: np.concatenate(a.T)
Out[90]: array([1, 3, 1, 3, 2, 7, 5, 5, 3, 2, 6, 6, 3, 7, 5])

And turn it into a column vector by adding a dimension:

In [91]: _[:,None]
Out[91]: 
array([[1],
       [3],
       [1],
       [3],
       [2],
       [7],
       [5],
       [5],
       [3],
       [2],
       [6],
       [6],
       [3],
       [7],
       [5]])

It may be worth noting that a.T, the transpose, is produced by changing the order to F. So this is a variation on the a.ravel(order='F') approach. In order to stack the columns it has to reorder the elements of the array (default is 'c' row oriented).

Upvotes: 0

Paul Panzer
Paul Panzer

Reputation: 53029

You can use 'F' for Fortran order together with ravel or reshape:

a.ravel('F')[:, None]
# array([[1],
#        [3],
#        [1],
#        [3],
#        [2],
# ...

Upvotes: 1

timgeb
timgeb

Reputation: 78690

You can flatten the transposed array, create a new axis, and transpose again.

>>> np.ravel(a.T)[None].T
array([[1],
       [3],
       [1],
       [3],
       [2],
       [7],
       [5],
       [5],
       [3],
       [2],
       [6],
       [6],
       [3],
       [7],
       [5]])

Upvotes: 1

Guillaume Jacquenot
Guillaume Jacquenot

Reputation: 11717

You can use reshape function and the transpose .T operator

np.reshape(a.T, (a.size,1))

Upvotes: 1

Related Questions