This is my simple of piece of code
Everything is a numpy array. I welcome manipulation using lists too.
a = [1,2,3,4,5]
b = [3,2,2,2,8]
c = ['test1', 'test2', 'test3','test4','test5']
expected Outcome:
d = [ 1, 2, 3, 4, 5;
3, 2, 2, 2, 8;
'test1','test2', 'test3', 'test4','test5' ]
OR
d = [ 1 3 'test1';
2 2 'test2';
3 2 'test3';
4 2 'test4';
5 8 'test5']
Adam's answer using numpy.concat is also correct, but in terms of specifying the exact shape you are expecting — rows stacked vertically — you'll want to look at numpy.vstack:
>>> import numpy as np
>>> np.vstack([a, b, c])
array([['1', '2', '3', '4', '5'],
['3', '2', '2', '2', '8'],
['test1', 'test2', 'test3', 'test4', 'test5']],
dtype='<U21')
There's a catch here either way you do it: since your separate arrays (int64, int64, <U5) are all being put together, the new array will automatically use the least restrictive type, which in this case is the unicode type.
See also: numpy.hstack.
Check out the concat method.
>>> a = np.array([[1, 2], [3, 4]])
>>> b = np.array([[5, 6]])
>>> np.concatenate((a, b), axis=0)
array([[1, 2],
[3, 4],
[5, 6]])
numpy does not use ; to divide or display its dimensions. It's normal 2d display looks a lot like a list of lists.Your a,b,c are lists; you'd have to use np.array([1,2,3]) to get an array, and it will display as [1 2 3] (without the commas).
Simply creating a new list from those lists produces a list of lists
In [565]: d=[a,b,c]
In [566]: d
Out[566]:
[[1, 2, 3, 4, 5],
[3, 2, 2, 2, 8],
['test1', 'test2', 'test3', 'test4', 'test5']]
and simply concatenating the lists produces one longer one
In [567]: a+b+c
Out[567]: [1, 2, 3, 4, 5, 3, 2, 2, 2, 8, 'test1', 'test2', 'test3', 'test4', 'test5']
numpy arrays have problems containing both numbers and strings. You have to make a 'structured array'.
The easiest way to combine these into one array is with a fromarrays utility function:
In [561]: x=np.rec.fromarrays(([1,2,3],[3,2,2],['test1','test2','test3']))
In [562]: x['f0']
Out[562]: array([1, 2, 3])
In [563]: x['f2']
Out[563]:
array(['test1', 'test2', 'test3'],
dtype='<U5')
In [568]: x
Out[568]:
rec.array([(1, 3, 'test1'), (2, 2, 'test2'), (3, 2, 'test3')],
dtype=[('f0', '<i4'), ('f1', '<i4'), ('f2', '<U5')])
or with a little editing of the display:
In [569]: print(x)
[(1, 3, 'test1')
(2, 2, 'test2')
(3, 2, 'test3')]
This is not a 2d array; it's 1d (here 3 elements) with 3 fields.
Perhaps the easiest way to format this array in a way that looks like your spec would be with a csv writer:
In [570]: np.savetxt('x.txt',x,fmt='%d %d %s;')
In [571]: cat x.txt # shell command to display the file
1 3 test1;
2 2 test2;
3 2 test3;
;is part of a display, not a data structure.