Indexing

Part of the power of NumPy comes from the versatile ways in which its arrays can be accessed.

In addition to normal list indexing, it allows you to use arrays themselves as indices by performing the following:

>>> a[np.array([2,3,4])] 
array([77, 3, 4])

Coupled with the fact that conditions are also propagated to individual elements, we gain a very convenient way to access our data, using the following:

>>> a>4
array([False, False, True, False, False, True], dtype=bool)
>>> a[a>4]
array([77, 5])

By performing the following command, we can trim outliers:

>>> a[a>4] = 4
>>> a
array([0, 1, 4, 3, 4, 4])

As this is a frequent use case, there is a special clip function for it, clipping the values at both ends of an interval with one function call:

>>> a.clip(0,4)
array([0, 1, 4, 3, 4, 4])