Dive into NumPy! Start organizing, rearranging, and sampling your data.

Python Data Analysis and Visualization

From Python to Numpy

As explained in the [Subclassing ndarray](https://docs.scipy.org/doc/numpy/user/basics.subclassing.html)
documentation, subclassing `ndarray` is complicated by the fact that new
instances of `ndarray` classes can come about in three different ways:

* Explicit constructor call
* View casting
* New from template

However, our case is simpler because we're only interested in the view
casting. We thus only need to define the `__new__` method that will be called
at each instance creation. As such, the `GPUData` class will be equipped with two
properties:

* `extents`: This represents the full extent of the view relatively to the base
  array. It is stored as a byte offset and a byte size.
* `pending_data`: This represents the contiguous *dirty* area as (byte offset,
  byte size) relatively to the extents property.


In this lesson, we will learn how to subclass ndarray to use it in the GPUData class. 

Introduction

Anatomy of an Array

Code Vectorization

Problem Vectorization

Custom Vectorization

Beyond NumPy

Memory-aware Array: Array Subclass in GPUData class