Python, hstack column numpy arrays (column vectors) of different types

Question

I currently have a numpy multi-dimensional array (of type float) and a numpy column array (of type int). I want to combine the two into a mutli-dimensional numpy array.

import numpy

>> dates.shape
(1251,)
>> data.shape
(1251,10)
>> test = numpy.hstack((dates, data))
ValueError: all the input arrays must have same number of dimensions

To show that the types of the arrays are different:

>> type(dates[0])
<type 'numpy.int64'>
>> type(data[0,0])
<type 'numpy.float64'>

I'm showing numpy.dstack as stacking along the 3rd axis. I want to take a 1251, 10 (in this case) and turn it into a 1251, 11, are you suggesting that i would use dstack for that? — benjaminmgross
– benjaminmgross, Commented Dec 31, 2011 at 3:09
Short of using an object array, you can't do this. Numpy arrays can, by definition, only contain a single type. So the only alternative is to cast the integer array to floating point, then stack them. — talonmies
– talonmies, Commented Dec 31, 2011 at 3:17
I believe a 1-D numpy array is treated as a row vector, not a column vector. I recall tripping over this recently as my intuition is that 1-D arrays should be column vectors :) Have you tried dates.shape = (1251,1)? — jrennie
– jrennie, Commented Dec 31, 2011 at 3:40
Thanks, that actually allows me to changes things more cleanly when I pre-allocate. I had: m,n = data.shape dates = numpy.empty(shape = temp.shape) now I can just make it dates = numpy.empty(shape = (m,1)) (because the number of dates must equal the number of rows of the date vector) and everything works fine. — benjaminmgross
– benjaminmgross, Commented Dec 31, 2011 at 3:49

Community · Accepted Answer · 2012-02-15 16:49:32Z

11

import numpy as np

np.column_stack((dates, data))

The types are cast automatically to the most precise, so your int array will be converted to float.

edited Feb 15, 2012 at 16:49

CommunityBot

11 silver badge

answered Dec 31, 2011 at 3:23

Griffith Rees

1,3942 gold badges16 silver badges25 bronze badges

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

2 Comments

benjaminmgross Over a year ago

Thanks so much Griffith, this certainly did the trick, much appreciated, I was hoping to not have to cast the floats to get back to ordinals, but it'll work just fine, thanks again!

Griffith Rees Over a year ago

If you like my answer give me a vote and check it as the correct one ;)

Bi Rico · Accepted Answer · 2011-12-31 03:17:59Z

1

The types don't matter, you should reshape dates to be (1251, 1) before using hstack.

Ps. The ints will be cast to float.

answered Dec 31, 2011 at 3:17

Bi Rico

25.9k3 gold badges57 silver badges75 bronze badges

Comments

Ricardo Cárdenes · Accepted Answer · 2011-12-31 03:29:14Z

0

test = numpy.hstack((dates[:,numpy.newaxis], data))

answered Dec 31, 2011 at 3:29

Ricardo Cárdenes

9,1941 gold badge23 silver badges35 bronze badges

Collectives™ on Stack Overflow

Python, hstack column numpy arrays (column vectors) of different types

3 Answers 3

2 Comments

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

2 Comments

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related