notes.md

Chapter 16: Operator Overloading — Notes

1. The Immutability Rule of Operators

A fundamental rule of Python operator overloading: Infix operators like + (__add__) and * (__mul__) must never mutate their operands. They must always read from the operands and return a new object.

2. The Operator Dispatch Mechanism

When Python evaluates a + b, it follows this exact sequence:

1. Call a.__add__(b).
2. If __add__ raises an exception, crash.
3. If __add__ returns NotImplemented, or if a has no __add__ method, Python calls b.__radd__(a).
4. If __radd__ returns NotImplemented, or if b has no __radd__ method, Python raises a TypeError: unsupported operand type(s) for +.

This mechanism is what allows (1, 2) + MyVector([3, 4]) to work, even though the built-in tuple class has no idea what MyVector is.

3. NotImplemented vs NotImplementedError

• NotImplemented is a singleton value (like None). You return it from dunder operator methods to tell Python's execution engine to try the reverse fallback.
• NotImplementedError is an exception. You raise it inside abstract base classes to force subclasses to override a method. If you raise it inside an operator method, it breaks the dispatch chain and crashes the program.

4. Augmented Assignment (+=)

Augmented assignment operators like += look for the in-place method __iadd__.

• If __iadd__ exists, Python evaluates a = a.__iadd__(b). The method should mutate a in place, and it must return self.
• If __iadd__ does not exist, Python falls back to a = a.__add__(b). This evaluates the expression and binds the variable a to the brand new object returned by __add__.