Example
The core data structure in numpy is the ndarray (short for n-dimensional array). ndarrays are
- homogeneous (i.e. they contain items of the same data-type)
- contain items of fixed sizes (given by a shape, a tuple of n positive integers that specify the sizes of each dimension)
One-dimensional array:
x = np.arange(15)
# array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14])
x.shape
# (15,)
Two-dimensional array:
x = np.asarray([[0, 1, 2, 3, 4], [5, 6, 7, 8, 9], [10, 11, 12, 13, 14]])
x
# array([[ 0, 1, 2, 3, 4],
# [ 5, 6, 7, 8, 9],
# [10, 11, 12, 13, 14]])
x.shape
# (3, 5)
Three-dimensional:
np.arange(12).reshape([2,3,2])
To initialize an array without specifying its contents use:
x = np.empty([2, 2])
# array([[ 0., 0.],
# [ 0., 0.]])
Datatype guessing and automatic casting
The data-type is set to float by default
x = np.empty([2, 2])
# array([[ 0., 0.],
# [ 0., 0.]])
x.dtype
# dtype('float64')
If some data is provided, numpy will guess the data-type:
x = np.asarray([[1, 2], [3, 4]])
x.dtype
# dtype('int32')
Note that when doing assignments numpy will attempt to automatically cast values to suit the ndarray's datatype
x[1, 1] = 1.5 # assign a float value
x[1, 1]
# 1
# value has been casted to int
x[1, 1] = 'z' # value cannot be casted, resulting in a ValueError
Array broadcasting
See also Broadcasting array operations.
x = np.asarray([[1, 2], [3, 4]])
# array([[1, 2],
[3, 4]])
y = np.asarray([[5, 6]])
# array([[5, 6]])
In matrix terminology, we would have a 2x2 matrix and a 1x2 row vector. Still we're able to do a sum
# x + y
array([[ 6, 8],
[ 8, 10]])
This is because the array y is "stretched" to:
array([[5, 6],
[5, 6]])
to suit the shape of x.
Resources:
- Introduction to the ndarray from the official documentation: The N-dimensional array (ndarray)
- Class reference: ndarray.
