Professional Python

Chapter 7Abstract Base Classes

How do you know whether an object you are using conforms to a given specification? The common answer to this in Python is referred to as the “duck typing” model. If it looks like a duck and quacks like a duck, it is probably a duck.

When dealing with programming and objects, this usually translates to verifying that an object implements a given method, or has a given property. If the object has a quack method, then you have decent evidence that it is a Duck. And, furthermore, if all you need is a quack method, it probably does not matter much whether or not it is actually a Duck.

This is often a very useful construct, and it flows naturally from Python's loose typing system. It emphasizes questions of composition over questions of identity, hasattr over isinstance.

Sometimes, however, identity is important. For example, perhaps you are using a library that requires input conforming to a particular identity. Alternatively, sometimes it is too cumbersome to check for a myriad of different properties and methods.

Python 2.6 and Python 3 introduce the concept of abstract base classes. Abstract base classes are a mechanism for assigning identity. They are a way of answering, “Is this class fundamentally a Duck?” Abstract base classes also provide a mechanism for designating abstract methods, requiring other implementers to provide key functionality that is purposefully not provided in a base implementation.

This chapter explores abstract base classes, ...