Robust Python

Chapter 13. Protocols

I have a confession to make. I’ve been skirting around something in the Python type system that, upon first glance, is contradictory. It has to do with a key difference in philosophy between the Python runtime type system and the static type hints.

Back in Chapter 2, I described how Python supports duck typing. Recall that this means you can use objects in a context as long as that object supports a specific set of behaviors. You don’t need any sort of parent class or predefined inheritance structure to use duck typing.

However, the typechecker has no clue how to handle duck typing without any assistance. The typechecker knows how to handle types known at static analysis time, but how does it handle duck-typed decisions made at runtime?

To remedy this, I’m going to introduce protocols, a feature introduced in Python 3.8. Protocols solve the contradiction listed above; they annotate duck-typed variables during typechecking. I’ll cover why you need protocols, how to define your own, and how to use them in advanced scenarios. But before you start, you need to understand the disconnect between Python’s duck typing and static typecheckers.

Tension Between Typing Systems

In this chapter, you are going to build an automated lunch shop’s digital menu system. This restaurant has a variety of entries that are “splittable,” meaning you can get a half order. Deli sandwiches, wraps, and soups can be split, but entries like drinks and hamburgers cannot be split. In the ...