Load data with rows of different sizes into Numpy array

Padding a list of lists can be done in various ways, but since you are already reading this from a file, I think the itertools.zip_longest will be a good start.

In [201]: txt = """1  2  3  4  5
     ...: 6  7  8 
     ...: 9  10 11 12 13 14
     ...: 15 16 17 18 19"""

read and parse the text lines:

In [202]: alist = []
In [203]: for line in txt.splitlines():
     ...:     alist.append([int(i) for i in line.split()])
     ...:     
In [204]: alist
Out[204]: [[1, 2, 3, 4, 5], [6, 7, 8], [9, 10, 11, 12, 13, 14], [15, 16, 17, 18, 19]]

zip_longest (here in PY3 form) takes a fillvalue:

In [205]: from itertools import zip_longest
In [206]: list(zip_longest(*alist, fillvalue=0))
Out[206]: 
[(1, 6, 9, 15),
 (2, 7, 10, 16),
 (3, 8, 11, 17),
 (4, 0, 12, 18),
 (5, 0, 13, 19),
 (0, 0, 14, 0)]
In [207]: np.array(_).T
Out[207]: 
array([[ 1,  2,  3,  4,  5,  0],
       [ 6,  7,  8,  0,  0,  0],
       [ 9, 10, 11, 12, 13, 14],
       [15, 16, 17, 18, 19,  0]])

zip(*) can also be used to 'transpose' the list of lists:

In [209]: list(zip(*alist1))
Out[209]: 
[(1, 2, 3, 4, 5, 0),
 (6, 7, 8, 0, 0, 0),
 (9, 10, 11, 12, 13, 14),
 (15, 16, 17, 18, 19, 0)]

I'm guessing you are doing something more like:

In [211]: maxlen = max([len(i) for i in alist])
In [212]: maxlen
Out[212]: 6
In [213]: arr = np.zeros((len(alist), maxlen),int)
In [214]: for row, line in zip(arr, alist):
     ...:     row[:len(line)] = line
     ...:     
In [215]: arr
Out[215]: 
array([[ 1,  2,  3,  4,  5,  0],
       [ 6,  7,  8,  0,  0,  0],
       [ 9, 10, 11, 12, 13, 14],
       [15, 16, 17, 18, 19,  0]])

Which looks pretty good to me.

A regular poster, Divakar, likes to post a solution that uses cumsum. Let's see if I can reproduce it. It involves constructing a 1d mask where the nonzero values are supposed to go. Working backwards we need a mask like:

In [240]: mask=arr.ravel()>0
In [241]: mask
Out[241]: 
array([ True,  True,  True,  True,  True, False,  True,  True,  True,
       False, False, False,  True,  True,  True,  True,  True,  True,
        True,  True,  True,  True,  True, False], dtype=bool)

So that:

In [242]: arr.flat[mask] = np.hstack(alist)

There's a trick to this mapping that I haven't quite internalized!

The trick is the broadcast the lengths against [0,1,2,3,4,5]:

In [276]: lens=[len(i) for i in alist]
In [277]: maxlen=max(lens)
In [278]: mask=np.array(lens)[:,None]>np.arange(maxlen)
In [279]: mask
Out[279]: 
array([[ True,  True,  True,  True,  True, False],
       [ True,  True,  True, False, False, False],
       [ True,  True,  True,  True,  True,  True],
       [ True,  True,  True,  True,  True, False]], dtype=bool)
In [280]: arr = np.zeros((len(alist), maxlen),int)
In [281]: arr[mask] = np.hstack(alist)
In [282]: arr
Out[282]: 
array([[ 1,  2,  3,  4,  5,  0],
       [ 6,  7,  8,  0,  0,  0],
       [ 9, 10, 11, 12, 13, 14],
       [15, 16, 17, 18, 19,  0]])