How concatenate 2 Numpy array efficiently?

Question

I have 2 Numpy array <type 'numpy.ndarray'> with shape of (10,) (10, 6) and I would like to concat the first one with the second. The numpy array provided below,

r1 
['467c8100-7f13-4244-81ee-5e2a0f8218a8',
 '71a4b5b2-80d6-4c12-912f-fc71be8d923e',
 '7a3e0168-e47d-4203-98f2-a54a46c62ae0',
 '7dfd43e7-ced1-435f-a0f9-80cfd00ae246',
 '85dbc70e-c773-43ee-b434-8f458d295d10',
 'a56b2bc3-4a81-469e-bc5f-b3aaa520db05',
 'a9e8996f-ff35-4bfb-bbd9-ede5ffecd4d8',
 'c3037410-0c2e-40f8-a844-ac0664a05783',
 'c5618563-10c0-425b-a11b-2fcf931f0ff7',
 'f65e6cea-892e-4335-8e86-bf7f083b5f53'] 

r2 
[[1.55000000e+02, 5.74151515e-01, 1.55000000e+02, 5.74151515e-01, 3.49000000e+02, 1.88383585e+00],
 [5.00000000e+00, 1.91871554e-01, 1.03000000e+02, 1.22893828e+00, 2.95000000e+02, 3.21148368e+00],
 [7.10000000e+01, 1.15231270e-01, 2.42000000e+02, 5.78527276e-01, 4.09000000e+02, 2.67915246e+00],
 [3.60000000e+01, 7.10066720e-01, 2.42000000e+02, 1.80213634e+00, 4.12000000e+02, 4.16314391e+00],
 [1.15000000e+02, 1.05120284e+00, 1.30000000e+02, 1.71697773e+00, 2.53000000e+02, 2.73640301e+00],
 [4.70000000e+01, 2.19434656e-01, 3.23000000e+02, 4.84093786e+00, 5.75000000e+02, 7.00530186e+00],
 [5.50000000e+01, 1.22614463e+00, 1.04000000e+02, 1.55392099e+00, 4.34000000e+02, 4.13661261e+00],
 [3.90000000e+01, 3.34816889e-02, 1.10000000e+02, 2.54431753e-01, 2.76000000e+02, 1.52322736e+00],
 [3.43000000e+02, 2.93550948e+00, 5.84000000e+02, 5.27968165e+00, 7.45000000e+02, 7.57657633e+00],
 [1.66000000e+02, 1.01436635e+00, 2.63000000e+02, 2.69197514e+00, 8.13000000e+02, 7.96477735e+00]]

I tried to concatenate with the command np.concatenate((r1, r2)), it returns with the message of ValueError: all the input arrays must have same number of dimensions which I don't understand. Because, the r1 can possibly concat with the r2 and can form a whole new array and make a new array of 10 x 7 as result.

How to solve this problem ?

Answer 1

These 2 array have a dtype and shape mismatch:

In [174]: r1.shape
Out[174]: (10,)
In [175]: r1.dtype
Out[175]: dtype('<U36')

In [177]: r2.shape
Out[177]: (10, 6)
In [178]: r2.dtype
Out[178]: dtype('float64')

If you add a dimension to r1, so it is now (10,1), you can concatenate on axis=1. But note the dtype - the floats have been turned into strings:

In [181]: r12 =np.concatenate((r1[:,None], r2), axis=1)
In [182]: r12.shape
Out[182]: (10, 7)
In [183]: r12.dtype
Out[183]: dtype('<U36')
In [184]: r12[0,:]
Out[184]: 
array(['467c8100-7f13-4244-81ee-5e2a0f8218a8', '155.0', '0.574151515',
       '155.0', '0.574151515', '349.0', '1.88383585'], 
      dtype='<U36')

A way to mix string and floats is with structured array, for example:

In [185]: res=np.zeros((10,),dtype='U36,(6)f')
In [186]: res.dtype
Out[186]: dtype([('f0', '<U36'), ('f1', '<f4', (6,))])
In [187]: res['f0']=r1
In [188]: res['f1']=r2
In [192]: res.shape
Out[192]: (10,)
In [193]: res[0]
Out[193]: ('467c8100-7f13-4244-81ee-5e2a0f8218a8', [ 155.        ,    0.57415152,  155.        ,    0.57415152,  349.        ,    1.88383579])

We could also make a (10,7) array with dtype=object. But most array operations won't work with such a mix of strings and floats. And the ones that work are slower.

Why do you want to concatenate these arrays? What do you intend to do with the result? That dtype mismatch is more serious than the shape mismatch.

Answer 2

Numpy offers an easy way to concatenate along the second axis.

np.c_[r2,r1]

Answer 3

You can reshape r1 to make it two-dimensional and specify the axis along which the arrays should be joined:

import numpy as np

r1 = np.ones((10,))
r2 = np.zeros((10, 6))
np.concatenate((r1.reshape(10, 1), r2), axis=1)