Pandas to_csv always substitute long numpy.ndarray with ellipsis

In some sense this is a duplicate of printing the entire numpy array, since to_csv simply asks each item in your DataFrame for it's __str__, so you need to see how that prints:

In [11]: np.arange(10000)
Out[11]: array([   0,    1,    2, ..., 9997, 9998, 9999])

In [12]: np.arange(10000).__str__()
Out[12]: '[   0    1    2 ..., 9997 9998 9999]'

as you can see when it's over a certain threshold it prints with ellipsis, set it to NaN:

np.set_printoptions(threshold='nan')

To give an example:

In [21]: df = pd.DataFrame([[np.arange(10000)]])

In [22]: df  # Note: pandas printing is different!!
Out[22]:
                                                   0
0  [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13,...

In [23]: s = StringIO()

In [24]: df.to_csv(s)

In [25]: s.getvalue()  # ellipsis
Out[25]: ',0\n0,"[   0    1    2 ..., 9997 9998 9999]"\n'

Once changed to_csv records the entire array:

In [26]: np.set_printoptions(threshold='nan')

In [27]: s = StringIO()

In [28]: df.to_csv(s)

In [29]: s.getvalue()  # no ellipsis (it's all there)
Out[29]: ',0\n0,"[   0    1    2    3    4    5    6    7    8    9   10   11   12   13   14\n   15   16   17   18   19   20   21   22   23   24   25   26   27   28   29\n   30   31   32   33   34   35   36   37   38   39   40   41   42   43   44\n   45   46   47   48   49   50   51   52   53   54   55   56   57   58   59\n   60   61  # the whole thing is here...

As mentioned this is not usually a good choice of structure for a DataFrame (numpy arrays in object columns) as you lose much of the pandas speed/efficiency/magic sauce.