This article shows 5 ways of flattening a list of lists, ranked from worst to best.

5 ways to flatten a list of lists

This short article will show 5 ways of flattening a list of lists, ranked from worst to best.

This is the list we'll be using:

list_of_lists = [
    [1, 2, 3],
    [4, 5],
    [6],
    [7, 8, 9],
]

Let's start.

5th – using functools.reduce

>>> from functools import reduce
>>> from operator import add
>>> flat_list = reduce(add, list_of_lists, [])
>>> flat_list
[1, 2, 3, 4, 5, 6, 7, 8, 9]

This is not a great idea because using reduce(add, ...) is almost always worse than using the equivalent built-in sum.

In other words, sum is the specialised version of reduce that adds things, so you'll almost never need reduce(add, ...).

(reduce isn't useless, though. You can and should read about reduce and its use-cases.)

So, this leads clearly to the next version.

4th – using sum

>>> flat_list = sum(list_of_lists, [])
>>> flat_list
[1, 2, 3, 4, 5, 6, 7, 8, 9]

This is better than option 5.

But it's still not great.

The main reason this isn't great is because sum will do too much work. (Can you see why?)

Every time we add two lists (the accumulated one and one from list_of_lists), we need to create a third list with all of the elements of the two, so we'll waste a lot of time recomputing lists.

So, using sum is a neat party trick, and it shows you understand the underlying way in which sum works, but it isn't practical.

3rd – two nested loops

This is the KISS solution. It's pretty straightforward and it's a great solution! Two for loops and an append:

>>> flat_list = []
>>> for sublist in list_of_lists:
...     for element in sublist:
...         flat_list.append(element)
...
>>> flat_list
[1, 2, 3, 4, 5, 6, 7, 8, 9]

This is a brilliant solution! The reason it's placed in 3rd is not because it's bad, but because you can do even better!

2nd – using a list comprehension

>>> flat_list = [
...     element
...     for sublist in list_of_lists
...     for element in sublist
... ]
>>> flat_list
[1, 2, 3, 4, 5, 6, 7, 8, 9]

This solution is better than the two loops because there are many advantages to using list comprehensions when you have simple for loops whose only job is to append items to a new list.

I've written extensively about list comprehensions before! If you're interested (and you should be!) you can learn more about list comprehensions and their advantages.

The list comprehension is always better than the two loops. So, that's why it comes after the two loops. But the next solution is only better than the list comprehension in some cases.

1st – using itertools.chain:

>>> from itertools import chain
>>> flat_list = list(chain.from_iterable(list_of_lists))
>>> flat_list
[1, 2, 3, 4, 5, 6, 7, 8, 9]

Now, when can this be better than the list comprehension?

Using chain is better than the list comprehension when you don't actually need to wrap the call to chain in a call to list! In the code above, I used list(chain...) because I wanted you to see the resulting list. But chain is a generator, which means it's lazy.

chain being lazy means it's very useful when you want to iterate through the flat list, but don't necessarily need a list per se.

For example, suppose you're traversing the flat list in a for loop and might have a condition that breaks out of the loop if you find something. In that case, the chain might be better.

So, this solution is only better than the list comprehension in some cases, but what's good about me showing it to you is that it teaches you about chain, which is quite useful.

Bonus – flatten a list with heterogeneous depth

The five options above flatten a list of lists, assuming the depth is constant and is always 2. What if you get something that is heterogeneous in depth?

For example, how would you flatten something like the list shown below?

crazy_list = [[[1, 2], [3], [[4, [5, [6, 7, 8]]]], 9], [10, 11], 12]

I propose the following recursive generator function:

def flatten(obj):
    if isinstance(obj, list):
        for item in obj:
            yield from flatten(item)
    else:
        yield obj

print(list(flatten(crazy_list)))
# [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]

Bonus alternatives

If you think I should include another way of flattening a list of lists, even if it's not better than all of the five I already shared, feel free to comment below! Some of you already suggested some other alternatives:

• using unpacking instead of from_iterable and list:

>>> from itertools import chain
>>> flat_list = [*chain(*list_of_lists)]
>>> flat_list
[1, 2, 3, 4, 5, 6, 7, 8, 9]

• using yield from on each sublist:

>>> def flatten(list_of_lists):
...     for sublist in list_of_lists:
...         yield from sublist
... 
>>> flat_list = list(flatten(list_of_lists))
>>> flat_list
[1, 2, 3, 4, 5, 6, 7, 8, 9]

• using more_itertools.flatten (this requires installing the package more_itertools!). If you look closely, though, you'll see that more_itertools.flatten uses itertools.chain:

>>> from more_itertools import flatten
>>> list_of_lists = [
...     [1, 2, 3],
...     [4, 5],
...     [6],
...     [7, 8, 9],
... ]
>>> list(flatten(list_of_lists))
[1, 2, 3, 4, 5, 6, 7, 8, 9]

# more_itertools.flatten uses itertools.chain:
>>> flatten(list_of_lists)
<itertools.chain object at 0x100238fd0>

Become a better Python 🐍 developer 🚀

+35 chapters. +400 pages. Hundreds of examples. Over 30,000 readers!

My book “Pydon'ts” teaches you how to write elegant, expressive, and Pythonic code, to help you become a better developer. >>> Download it here 🐍🚀.