Examples

We discuss some more real-world examples here. Examples may be added or removed from this section later.

range

A generator is the perfect tool to write a range function. range is often used only to index or iterate over a list-like object. Even though range is in-built, we will rewrite a simpler version of range ourselves to see how generators are perfectly suited for this task.

Naive Approach

Let's first consider a naive, non-generator, simple function.

Python

def naive_range(start, stop):
    out = []
    i = start
    while i < stop:
        out.append(i)
        i = i + 1
    return out

print(naive_range(0, 10))
# [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]

# This may take a while and may run out of memory
naive_range(0, 10000000)

for i in naive_range(0, 10000000):
    # use i to index into another list-like object

The simple function naive_range materializes the entire range into a list and returns it. This takes a lot of memory. Imagine you want to iterate over a 10M long list-like object. Using naive_range to iterate over it would first create another 10M long list in memory. This is a waste of memory because you only want one value of the iteration index i in memory at a time.

Generator approach

We already saw a generator-based range-like function but let's see it again.

Python

def generator_range(start, stop):
    i = start
    while i < stop:
        yield i
        i = i + 1

print(list(generator_range(0, 10)))
# [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]

# This does not take a lot of memory and
# returns a generator object immediately.
generator_range(0, 10000000)

for i in generator_range(0, 10000000):
    # use i to index into another list-like object

You should try it out yourself to see the difference. Since generators follow the iterator protocol, we don't need any special syntax when using generator_range in a for loop.

range in Python 2 and 3

The range function in Python 2 used lists much like our naive_range while the range function in Python 3 uses iterators much like our generator_range. See this.

Unbounded range

Generator-based approach allows us to generate integers unboundedly. interestingly, this is even simpler than our generator_range above. This was clearly not possible if we use a naive list-based approach.

Python

def unbounded_range(start):
    i = start
    while True:
        yield i
        i = i + 1

r = unbounded_range(0)
next(r)
# 0

next(r)
# 1

next(r)
# 2

# Use in a for loop with care to avoid
# an infinitely running loop.
for i in unbounded_range(0):
    # do something with i
    print(i)
    if i > 10:
        break
# 0
# ...
# 11

Random number generator

Another surprising perfect fit for generators is a random number generator. Random number generators are typically state machines.

Python

def discrete_uniform_lgc(seed=894965, modulus=2**32,
                         coeff=1664525, constant=1013904223):
    x = seed
    while True:
        x = (coeff * x + constant) % modulus
        yield x

FILL ME

---

Last update: 2022-09-13
Created: 2022-09-13

Comments