string representation of a numpy array with commas separating its elements

The function you are looking for is np.set_string_function. source

What this function does is let you override the default __str__ or __repr__ functions for the numpy objects. If you set the repr flag to True, the __repr__ function will be overriden with your custom function. Likewise, if you set repr=False, the __str__ function will be overriden. Since print calls the __str__ function of the object, we need to set repr=False.

For example:

np.set_string_function(lambda x: repr(x), repr=False)
x = np.arange(5)
print(x)

will print the output

array([0, 1, 2, 3, 4])

A more aesthetically pleasing version is

np.set_string_function(lambda x: repr(x).replace('(', '').replace(')', '').replace('array', '').replace("       ", ' ') , repr=False)

print(np.eye(3))

which gives

[[1., 0., 0.],
 [0., 1., 0.],
 [0., 0., 1.]]

Hope this answers your question.

Now, in numpy 1.11, there is numpy.array2string:

In [279]: a = np.reshape(np.arange(25, dtype='int8'), (5, 5))

In [280]: print(np.array2string(a, separator=', '))
[[ 0,  1,  2,  3,  4],
 [ 5,  6,  7,  8,  9],
 [10, 11, 12, 13, 14],
 [15, 16, 17, 18, 19],
 [20, 21, 22, 23, 24]]

Comparing with repr from @mgilson (shows "array()" and dtype):

In [281]: print(repr(a))
array([[ 0,  1,  2,  3,  4],
       [ 5,  6,  7,  8,  9],
       [10, 11, 12, 13, 14],
       [15, 16, 17, 18, 19],
       [20, 21, 22, 23, 24]], dtype=int8)

---

P.S. Still need np.set_printoptions(threshold=np.nan) for large array.

Another way to do it, which is particularly helpful when an object doesn't have a __repr__() method, is to employ Python's pprint module (which has various formatting options). Here is what that looks like, by example:

>>> import numpy as np
>>> import pprint
>>>
>>> A = np.zeros(10, dtype=np.int64)
>>>
>>> print(A)
[0 0 0 0 0 0 0 0 0 0]
>>>
>>> pprint.pprint(A)
array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0])

Try using repr

>>> import numpy as np
>>> points = np.array([[-468.927,  -11.299,   76.271, -536.723],
...                    [-429.379, -694.915, -214.689,  745.763],
...                    [   0.,       0.,       0.,       0.   ]])
>>> print(repr(points))
array([[-468.927,  -11.299,   76.271, -536.723],
       [-429.379, -694.915, -214.689,  745.763],
       [   0.   ,    0.   ,    0.   ,    0.   ]])

If you plan on using large numpy arrays, set np.set_printoptions(threshold=np.nan) first. Without it, the array representation will be truncated after about 1000 entries (by default).

>>> arr = np.arange(1001)
>>> print(repr(arr))
array([   0,    1,    2, ...,  998,  999, 1000])

Of course, if you have arrays that large, this starts to become less useful and you should probably analyze the data some way other than just looking at it and there are better ways of persisting a numpy array than saving it's repr to a file...