What does "list comprehension" and similar mean? How does it work and how can I use it?

I have the following code:

[x ** 2 for x in range(10)]

When I run it in the Python shell, it returns:

[0, 1, 4, 9, 16, 25, 36, 49, 64, 81]

I've searched and it seems this is called a list comprehension and similarly there seem to be set/dict comprehensions and generator expressions. But how does it work?

From the documentation:

> List comprehensions provide a concise way to create lists. Common applications are to make new lists where each element is the result of some operations applied to each member of another sequence or iterable, or to create a subsequence of those elements that satisfy a certain condition.

---

About your question, the list comprehension does the same thing as the following "plain" Python code:

>>> l = [] 
>>> for x in range(10):
...     l.append(x**2)
>>> l
[0, 1, 4, 9, 16, 25, 36, 49, 64, 81]

How do you write it in one line? Hmm...we can...probably...use map() with lambda:

>>> list(map(lambda x: x**2, range(10)))
[0, 1, 4, 9, 16, 25, 36, 49, 64, 81]

But isn't it clearer and simpler to just use a list comprehension?

>>> [x**2 for x in range(10)]
[0, 1, 4, 9, 16, 25, 36, 49, 64, 81]

---

Basically, we can do anything with x. Not only x**2. For example, run a method of x:

>>> [x.strip() for x in ('foo\n', 'bar\n', 'baz\n')]
['foo', 'bar', 'baz']

Or use x as another function's argument:

>>> [int(x) for x in ('1', '2', '3')]
[1, 2, 3]

We can also, for example, use x as the key of a dict object. Let's see:

>>> d = {'foo': '10', 'bar': '20', 'baz': '30'}
>>> [d[x] for x in ['foo', 'baz']]
['10', '30']

How about a combination?

>>> d = {'foo': '10', 'bar': '20', 'baz': '30'}
>>> [int(d[x].rstrip('0')) for x in ['foo', 'baz']]
[1, 3]

And so on.

---

You can also use if or if...else in a list comprehension. For example, you only want odd numbers in range(10). You can do:

>>> l = []
>>> for x in range(10):
...     if x%2:
...         l.append(x)
>>> l
[1, 3, 5, 7, 9]

Ah that's too complex. What about the following version?

>>> [x for x in range(10) if x%2]
[1, 3, 5, 7, 9]

To use an if...else ternary expression, you need put the if ... else ... after x, not after range(10):

>>> [i if i%2 != 0 else None for i in range(10)]
[None, 1, None, 3, None, 5, None, 7, None, 9]

---

Have you heard about nested list comprehension? You can put two or more fors in one list comprehension. For example:

>>> [i for x in [[1, 2, 3], [4, 5, 6]] for i in x]
[1, 2, 3, 4, 5, 6]

>>> [j for x in [[[1, 2], [3]], [[4, 5], [6]]] for i in x for j in i]
[1, 2, 3, 4, 5, 6]

Let's talk about the first part, for x in [[1, 2, 3], [4, 5, 6]] which gives [1, 2, 3] and [4, 5, 6]. Then, for i in x gives 1, 2, 3 and 4, 5, 6.

Warning: You always need put for x in [[1, 2, 3], [4, 5, 6]] before for i in x:

>>> [j for j in x for x in [[1, 2, 3], [4, 5, 6]]]
Traceback (most recent call last):
  File "<input>", line 1, in <module>
NameError: name 'x' is not defined

---

We also have set comprehensions, dict comprehensions, and generator expressions.

set comprehensions and list comprehensions are basically the same, but the former returns a set instead of a list:

>>> {x for x in [1, 1, 2, 3, 3, 1]}
{1, 2, 3}

It's the same as:

>>> set([i for i in [1, 1, 2, 3, 3, 1]])
{1, 2, 3}

A dict comprehension looks like a set comprehension, but it uses {key: value for key, value in ...} or {i: i for i in ...} instead of {i for i in ...}.

For example:

>>> {i: i**2 for i in range(5)}
{0: 0, 1: 1, 2: 4, 3: 9, 4: 16}

And it equals:

>>> d = {}
>>> for i in range(5):
...     d[i] = i**2
>>> d
{0: 0, 1: 1, 2: 4, 3: 9, 4: 16}

Does (i for i in range(5)) give a tuple? No!, it's a generator expression. Which returns a generator:

>>> (i for i in range(5))
<generator object <genexpr> at 0x7f52703fbca8>

It's the same as:

>>> def gen():
...     for i in range(5):
...         yield i
>>> gen()
<generator object gen at 0x7f5270380db0>

And you can use it as a generator:

>>> gen = (i for i in range(5))
>>> next(gen)
0
>>> next(gen)
1
>>> list(gen)
[2, 3, 4]
>>> next(gen)
Traceback (most recent call last):
  File "<input>", line 1, in <module>
StopIteration

Note: If you use a list comprehension inside a function, you don't need the [] if that function could loop over a generator. For example, sum():

>>> sum(i**2 for i in range(5))
30

Related (about generators): https://stackoverflow.com/questions/1756096/understanding-generators-in-python.

There are list, dictionary, and set comprehensions, but no tuple comprehensions (though do explore "generator expressions").

They address the problem that traditional loops in Python are statements (don't return anything) not expressions which return a value.

They are not the solution to every problem and can be rewritten as traditional loops. They become awkward when state needs to be maintained & updated between iterations.

They typically consist of:

[<output expr> <loop expr <input expr>> <optional predicate expr>]

but can be twisted in lots of interesting and bizarre ways.

They can be analogous to the traditional map() and filter() operations which still exist in Python and continue to be used.

When done well, they have a high satisfaction quotient.

I've seen a lot of confusion lately (on other SO questions and from coworkers) about how list comprehensions work. A wee bit of math education can help with why the syntax is like this, and what list comprehensions really mean.

The syntax

It's best to think of list comprehensions as predicates over a set/collection, like we would in mathematics by using set builder notation. The notation actually feels pretty natural to me, because I hold an undergrad degree in Mathematics. But forget about me, Guido van Rossum (inventor of Python) holds a masters in Mathematics and has a math background.

Set builder notation crash course

Here's the (very basics) of how set builder notation works:

So, this set builder notation represents the set of numbers that are strictly positive (i.e. [1,2,3,4,...]).

Points of confusion

1. The predicate filter in set builder notation only specifies which items we want to keep, and list comprehension predicates do the same thing. You don't have to include special logic for omitting items, they are omitted unless included by the predicate. The empty predicate (i.e. no conditional at the end) includes all items in the given collection.

2. The predicate filter in set builder notation goes at the end, and similarly in list comprehensions. (some) Beginners think something like [x < 5 for x in range(10)] will give them the list [0,1,2,3,4], when in fact it outputs [True, True, True, True, True, False, False, False, False, False]. We get the output [True, True, True, True, True, False, False, False, False, False] because we asked Python to evaluate x < 5 for all items in range(10). No predicate implies that we get everything from the set (just like in set builder notation).

If you keep set builder notation in the back of your mind while using list comprehensions, they're a bit easier to swallow.

HTH!

If you prefer a more visual way of figuring out what's going on then maybe this will help:

# for the example in the question...

y = []
for x in range(10):
    y += [x**2]

# is equivalent to...

y = [x**2 for x in range(10)]

# for a slightly more complex example, it is useful
# to visualize  where the various x's end up...

a = [1,2,3,4]
b = [3,4,5,6]
c = []

for x in a:
          if x in b:
                  c += [x]
#   \         \        /
#    \    _____\______/
#     \  /      \
#      \/        \
#      /\         \
#     /  \         \
#    /    \         \
c = [x for x in a if x in b]

print(c)

...produces the output [3, 4]

Introduction

> A list comprehension is a high level, declarative way to create a list in Python. The main benefits of comprehensions are readability and maintainability. A lot of people find them very readable, and even developers who have never seen them before can usually guess correctly what it means.

# Snippet 1
squares = [n ** 2 for n in range(5)]

# Snippet 2
squares = []
for n in range(5):
    squares.append(n ** 2)

Both snippets of code will produce squares to be equal to [0, 1, 4, 9, 16].

Notice that in the first snippet, what you type is declaring what kind of list you want, while the second is specifying how to create it. This is why a comprehension is a high-level and declarative.

Syntax

[EXPRESSION for VARIABLE in SEQUENCE]

EXPRESSION is any Python expression, but it is typical to have some variable in it. This variable is stated in VARIABLE field. SEQUENCE defines the source of values the variable enumerates through.

Considering Snippet 1, [n ** 2 for n in range(5)]:

• EXPRESSION is n ** 2
• VARIABLE is n
• SEQUENCE is range(5)

Notice that if you check the type of squares you will get that the list comprehension is just a regular list:

>>> type(squares)
<class 'list'>

More about EXPRESSION

The expression can be anything that reduces to a value:

• Arithmetic expressions such as n ** 2 + 3 * n + 1
• A function call like f(n) using n as variable
• A slice operation like s[::-1]
• Method calls bar.foo()
• ...

Some examples:

>>> [2 * x + 3 for x in range(5)]
[3, 5, 7, 9, 11]
>>> [abs(num) for num in range(-5, 5)]
[5, 4, 3, 2, 1, 0, 1, 2, 3, 4]
>>> animals = ['dog', 'cat', 'lion', 'tiger']
>>> [animal.upper() for animal in animals]
['DOG', 'CAT', 'LION', 'TIGER']

Filtering:

The order of elements in the final list is determined by the order of SEQUENCE. However, you can filter out elements adding an if clause:

[EXPRESSION for VARIABLE in SEQUENCE if CONDITION]

CONDITION is an expression that evaluates to True or False. Technically, the condition doesn't have to depend upon VARIABLE, but it typically uses it.

Examples:

>>> [n ** 2 for n in range(5) if n % 2 == 0]
[0, 4, 16]
>>> animals = ['dog', 'cat', 'lion', 'tiger']
>>> [animal for animal in animals if len(animal) == 3]
['dog', 'cat']

Also, remember that Python allows you to write other kinds of comprehensions other than lists:

• dictionary comprehensions
• set comprehensions