Use curly braces to initialize a Set in Python
Question:
I’m learning python, and I have a novice question about initializing sets. Through testing, I’ve discovered that a set can be initialized like so:
my_set = {'foo', 'bar', 'baz'}
Are there any disadvantages of doing it this way, as opposed to the standard way of:
my_set = set(['foo', 'bar', 'baz'])
or is it just a question of style?
Answers:
There are two obvious issues with the set literal syntax:
my_set = {'foo', 'bar', 'baz'}
-
It’s not available before Python 2.7
-
There’s no way to express an empty set using that syntax (using {} creates an empty dict)
Those may or may not be important to you.
The section of the docs outlining this syntax is here.
From Python 3 documentation (the same holds for python 2.7):
Curly braces or the set() function can be used to create sets. Note:
to create an empty set you have to use set(), not {}; the latter
creates an empty dictionary, a data structure that we discuss in the
next section.
in python 2.7:
>>> my_set = {'foo', 'bar', 'baz', 'baz', 'foo'}
>>> my_set
set(['bar', 'foo', 'baz'])
Be aware that {} is also used for map/dict:
>>> m = {'a':2,3:'d'}
>>> m[3]
'd'
>>> m={}
>>> type(m)
<type 'dict'>
One can also use comprehensive syntax to initialize sets:
>>> a = {x for x in """didn't know about {} and sets """ if x not in 'set' }
>>> a
set(['a', ' ', 'b', 'd', "'", 'i', 'k', 'o', 'n', 'u', 'w', '{', '}'])
Compare also the difference between {} and set() with a single word argument.
>>> a = set('aardvark')
>>> a
{'d', 'v', 'a', 'r', 'k'}
>>> b = {'aardvark'}
>>> b
{'aardvark'}
but both a and b are sets of course.
You need to do empty_set = set() to initialize an empty set. {} is an empty dict.
I’m surprised nobody has mentioned this, but it appears there is actually a difference between those two syntaxes from what I can tell—and that is performance/optimization.
For most situations the difference should be negligible, but in your example the following is creating a set from items directly:
my_set = {'foo', 'bar', 'baz'}
While the following creates a list and then passes it to the set constructor:
my_set = set(['foo', 'bar', 'baz'])
The end results are equivalent, but we’re getting them two slightly different ways. As you’d expect, the second one is a bit slower:
❯ python -m timeit 'my_set = {"foo", "bar", "baz"}'
10000000 loops, best of 5: 37.3 nsec per loop
❯ python -m timeit 'my_set = set(["foo", "bar", "baz"])'
5000000 loops, best of 5: 92.3 nsec per loop
As far as I’m aware, the {} literal syntax is the only way to skip creating an intermediate iterable when constructing a set. It’s a little odd to me personally that the set constructor wasn’t instead designed to take a variable number of positional arguments like so:
# This usage is invalid in real Python. If it existed,
# I would expect the call to be exactly equivalent
# to the performance of the literal syntax.
>>> set("foo", "bar", "baz")
# You would then construct a `set` from an existing
# `list` by simply unpacking the sequence.
>>> set(*["foo", "bar", "baz"])
Unfortunately, that signature doesn’t exist.
I’m learning python, and I have a novice question about initializing sets. Through testing, I’ve discovered that a set can be initialized like so:
my_set = {'foo', 'bar', 'baz'}
Are there any disadvantages of doing it this way, as opposed to the standard way of:
my_set = set(['foo', 'bar', 'baz'])
or is it just a question of style?
There are two obvious issues with the set literal syntax:
my_set = {'foo', 'bar', 'baz'}
-
It’s not available before Python 2.7
-
There’s no way to express an empty set using that syntax (using
{}creates an empty dict)
Those may or may not be important to you.
The section of the docs outlining this syntax is here.
From Python 3 documentation (the same holds for python 2.7):
Curly braces or the set() function can be used to create sets. Note:
to create an empty set you have to use set(), not {}; the latter
creates an empty dictionary, a data structure that we discuss in the
next section.
in python 2.7:
>>> my_set = {'foo', 'bar', 'baz', 'baz', 'foo'}
>>> my_set
set(['bar', 'foo', 'baz'])
Be aware that {} is also used for map/dict:
>>> m = {'a':2,3:'d'}
>>> m[3]
'd'
>>> m={}
>>> type(m)
<type 'dict'>
One can also use comprehensive syntax to initialize sets:
>>> a = {x for x in """didn't know about {} and sets """ if x not in 'set' }
>>> a
set(['a', ' ', 'b', 'd', "'", 'i', 'k', 'o', 'n', 'u', 'w', '{', '}'])
Compare also the difference between {} and set() with a single word argument.
>>> a = set('aardvark')
>>> a
{'d', 'v', 'a', 'r', 'k'}
>>> b = {'aardvark'}
>>> b
{'aardvark'}
but both a and b are sets of course.
You need to do empty_set = set() to initialize an empty set. {} is an empty dict.
I’m surprised nobody has mentioned this, but it appears there is actually a difference between those two syntaxes from what I can tell—and that is performance/optimization.
For most situations the difference should be negligible, but in your example the following is creating a set from items directly:
my_set = {'foo', 'bar', 'baz'}
While the following creates a list and then passes it to the set constructor:
my_set = set(['foo', 'bar', 'baz'])
The end results are equivalent, but we’re getting them two slightly different ways. As you’d expect, the second one is a bit slower:
❯ python -m timeit 'my_set = {"foo", "bar", "baz"}'
10000000 loops, best of 5: 37.3 nsec per loop
❯ python -m timeit 'my_set = set(["foo", "bar", "baz"])'
5000000 loops, best of 5: 92.3 nsec per loop
As far as I’m aware, the {} literal syntax is the only way to skip creating an intermediate iterable when constructing a set. It’s a little odd to me personally that the set constructor wasn’t instead designed to take a variable number of positional arguments like so:
# This usage is invalid in real Python. If it existed,
# I would expect the call to be exactly equivalent
# to the performance of the literal syntax.
>>> set("foo", "bar", "baz")
# You would then construct a `set` from an existing
# `list` by simply unpacking the sequence.
>>> set(*["foo", "bar", "baz"])
Unfortunately, that signature doesn’t exist.