Lecture 4

Object-oriented Programming

Operator overloading

MCS 275 Spring 2022
David Dumas

Lecture 4: Operator overloading

• At this point you must have read the syllabus.
• Discord open (link in the zoom chat or Blackboard).

Object-Oriented Programming

Today we're starting our unit on object-oriented programming (OOP).

We assume knowledge of: Class definitions, creating instances, accessing attributes, calling methods.

Need to review these? See:

• MCS 260 Lecture 25
• MCS 260 Fall 2021 OOP sample code

We DO NOT assume knowledge of: Subclasses, inheritance, operator overloading.

Review of some key concepts

• class -- A type in that combines attributes (data) and methods (behavior).
• instance or object -- A value whose type is a certain class (e.g. "hello" is an instance of str)
• attribute -- A variable local to an object, accessed as objname.attrname.
• constructor -- The method named __init__ that is called when a new object is created.

Special methods / overloading

In Python, built-in operations are often silently translated into method calls.

e.g.   A+B turns into A.__add__(B)

These special method names begin and end with two underscores (__). They are used to customize the way your classes work with built-in language features.

Using these to add special behavior for operators like +,-,* is called operator overloading.

Operator examples

List of many more in the Python documentation.

More special methods

Live coding

Let's build classes:

• Point2 — point in the plane (a location in 2D)
• Vector2 — vector in the plane (e.g. the displacement between two points)

Difference of two Point2s is a Vector2.

Can multiply a Vector2 by a float or add it to a Point2.

Point2 plus Vector2 is a Point2.

Language features used

• isinstance(obj,classname) -- returns bool indicating whether obj is an instance of the named class (or subclass thereof)
• NotImplemented -- Special value that operators should return if the operation is not supported

__add__ & __radd__

In evaluating A+B, Python first tries

A.__add__(B)

B.__radd__(A)

There are reflected versions of all the binary operations (e.g. __rmul__).

Overloading danger

Given the very flexible overloading system in Python, it's easy to be too clever.

Overloading is best used when a function or operator has a clear meaning for a class, and when the operation is so frequently used that direct method calls would be cumbersome.

Avoid overloading when it makes code harder to understand!

Singletons

When a class is designed so that it only ever has one instance, the class (or the only instance of it) is called a singleton.

We've seen two of these so far:

• None, the only instance of NoneType
• NotImplemented, the only instance of NotImplementedType