Lecture 4

Object-oriented Programming

Special Methods and Overloading

MCS 275 Spring 2021
David Dumas

Lecture 4: Special Methods and Overloading

Course bulletins:

At this point you must have read the syllabus.
Discord open (link in the zoom chat or Blackboard).
Worksheet 2 available.

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.

We DO NOT assume knowledge of: Subclasses, inheritance, special methods.

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

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

Expression		Special method
`A==B`		`A.__eq__(B)`
`A+B`		`A.__add__(B)`
`A-B`		`A.__sub__(B)`
`A*B`		`A.__mul__(B)`
`A/B`		`A.__truediv__(B)`
`A**B`		`A.__pow__(B)`

List of many more in the Python documentation.

More special methods

Expression		Actually calls
`str(A)`		`A.__str__()`
`len(A)`		`A.__len__()`
`abs(A)`		`A.__abs__()`
`bool(A)`		`A.__bool__()`
`A[k]`		`A.__getitem__(k)`
`A[k]=v`		`A.__setitem__(k,v)`

Live coding

Let's build classes:

Point -- point in the plane
Vector -- vector in the plane

Difference of two Points is a Vector.

Can multiply a Vector by a float or add it to a Point.

Language features used

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

add & radd

In evaluating A+B, Python first tries

A.__add__(B)

but if that fails (returns NotImplemented), it will try

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!

Note: This is good advice, but wasn't actually discussed in Lecture 4.

References

I discussed overloading in MCS 260 Fall 2020 Lecture 24, and used this geometric object module as an example. Overloading is often, but not always, covered in MCS 260.
See Lutz, Chapter 30 for more information about overloading.
Lutz, Chapters 26-32 discuss object-oriented programming.

Revision history

2021-01-22 Retrospective editing based on what was covered
2021-01-19 Initial publication