For following:
d = np.array([[0,1,4,3,2],[10,18,4,7,5]])
print(d.shape)
Output is:
(2, 5)
It is expected.
But, for this(difference in number of elements in individual rows):
d = np.array([[0,1,4,3,2],[10,18,4,7]])
print(d.shape)
Output is:
(2,)
How to explain this behaviour?
Short answer: It parses it as an array of two objects: two lists.
Numpy is used to process "rectangular" data. In case you pass it non-rectangular data, the np.array(..) function will fallback on considering it a list of objects.
Indeed, take a look at the dtype of the array here:
>>> d
array([list([0, 1, 4, 3, 2]), list([10, 18, 4, 7])], dtype=object)It is an one-dimensional array that contains two items two lists. These lists are simply objects.
numpy documentation. But the problem with your example is that it's a kind of error, trying to make a 2d array from lists that differ in size. np.array normally makes a multidimensional numeric array. Making the 1d object dtype array is a fall-back alternative. It get's a lot of attention in SO questions, not so much in the formal documentation or tutorials.