Def

In Python programs we use the def-keyword to create functions. Statements may be top-level. But it is possible to structure logic into methods.

In a function, we can place a "return" statement to return a value. A tuple can be used to return multiple values at once.

We use def for a simple "dividesafe" method that receives 2 arguments. In its body, it checks the argument "b" against zero. It returns -1 if the division would cause an error.

def dividesafe(a, b):
    # Handle zero denominator.
    if b == 0:
        return -1
    # Divide.
    return a / b

# Use method.
print(dividesafe(10, 5))
print(dividesafe(10, 0))
2.0
-1

Pass

It is possible to create an empty method, one with no return value. In methods that return no value, a return statement is not required. In this case we can use a pass statement.

def waste_of_time():
    # This method does nothing.
    pass

# Call empty method.
waste_of_time()

Default

Method arguments can optionally have default values. These defaults are used when no argument is passed to the method. Here, we show a method that has 2 default-valued arguments.

def computesize(width=10, height=2):
    return width * height

print(computesize()) # 10, 2: defaults used.
print(computesize(5)) # 5, 2: default height.
print(computesize(height=3)) # 10, 3: default width.
20
10
30

One line

We can specify small methods on one line. For short methods, like ones that call other methods (such as print) this syntax is convenient.

# Print uppercased form of the string.
def printupper(s): print(s.upper())

printupper("hello")
printupper("world")
HELLO
WORLD

In recursion, a method calls itself. This helps solve complex problems. This next program solves no problems. Instead it demonstrates recursion. It ceases once its argument exceeds 10.

def recursive(depth):
    # Stop recursion if depth exceeds 10.
    if depth > 10:
        return

    print(depth)

    # Call itself.
    recursive(depth + 1)

# Begin.
recursive(5)
5
6
7
8
9
10

Tuple argument

A method in Python can receive any number of arguments stored in a tuple. This requires the single-star syntax. A formal parameter of name *t can be used as a tuple.

def display(*t):
    # Print tuple.
    print(t)

    # Loop over tuple items.
    for item in t:
        print(item)

# Pass parameters.
display("San Francisco", "Los Angeles", "Sacramento")
display(2, 0, 1, 4)
('San Francisco', 'Los Angeles', 'Sacramento')
San Francisco
Los Angeles
Sacramento
(2, 0, 1, 4)
2
0
1
4

Dictionary argument

Some parameter types have special syntax. A method can receive a dictionary of keys and values. This is specified with two stars before the parameter name.

def display(**values):
    # Loop over dictionary.
    for key in values:
        print(key, "=", values[key])

    # Newline.
    print()

# Pass named parameters.
display(first = "Sigmund", last = "Freud", year = 1899)
display(one = 1, two = 2)
year = 1899
last = Freud
first = Sigmund

two = 2
one = 1

Callable

This built-in is not often useful, but I include it for completeness. We can tell if an object (like an integer or lambda) can be called with it.

method = lambda n: n == 2

# A method is callable.
if callable(method):
    print(True)

number = 8

# An integer is not callable.
if not callable(number):
    print(False)
True
False

Benchmark, inline methods

There is a cost to calling a method. We can reduce this by inlining or combining methods. This can be done by pasting several method bodies together.

import time

# Method that calls other methods.
def method1(v):
    return method2(v) + method3(v) + method4(v)

def method2(v):
    return v + 2

def method3(v):
    return v + 3

def method4(v):
    return v + 4

# Method with all computations inlined.
def method1b(v):
    return v + 2 + v + 3 + v + 4

# Test them.
print(method1(5))
print(method1b(5))

print(time.time())

# Version 1: non-inlined methods.
i = 0
x = 0
while i < 1000000:
    x = method1(5)
    i += 1

print(time.time())

# Version 2: inlined methods.
i = 0
while i < 1000000:
    x = method1b(5)
    i += 1

print(time.time())
24
24
1405711521.510698
1405711522.499755   method1  = 0.989 s
1405711523.037786   method1b = 0.538 s

Notes, return

We can do more than just return 1 number in Python. We can return a tuple, which contains more than 1 value. Or we can return None (like a void method).

Lambda

Often a method has a single statement. The lambda expression syntax, with the keyword "lambda," is helpful here. It makes programs easier to write and shorter.

Timeit

Statements and methods can be benchmarked. You can use simple calls to time(), or the more complex timeit module. Micro-benchmarks tend to waste a lot of time.

With def, a function syntax form, we add further levels of structure to our programs. This yields programs that are clearer, smaller and often faster than unstructured designs.