While porting my team’s code to Python 3, I came across an interesting PyCharm warning about using set literals instead of the set() function. The code snippet involved converting a list to a set using set(), and PyCharm was not pleased. Upon investigation, I found that set syntax is faster due to lower overhead in Python. Benchmarking confirmed a noticeable performance increase with set literals. I delve into the intricacies of function calls, explaining the additional steps involved.

PyCharm Prefers Set Syntax To A Func Calls

While porting my team’s code over to Python 3, I encountered a PyCharm warning I found interesting: “Function call can be replaced with set literal.” The snippet of code looked a little something like this… We were converting an existing collection into a set using the set() function.

list_to_convert_to_set = [1, 2, 3, 4, 5]  # Imagine this existed somewhere in memory
my_set = set(list_to_convert_to_set)      # PyCharm complained when I did this

So after some digging, I learned a little more about the way Python deals with function calls vs. how they deal with set syntax, in terms of performance. In summary: using set syntax is faster because Python has less overhead involved compared to when it uses a function. That’s why PyCharm recommends set syntax over the set() function. So let’s dive in!

Benchmarking Function Calls Vs. Literals

Try the following code in your terminal. It should print two lines, one for each function call and assignment. The first uses set() to convert an existing collection into a set, and the second uses set syntax (i.e. {*}) to perform the conversion. The * character unpacks the collection. As you can see, there is a non-zero performance increase for the literal.

from timeit import timeit
print(timeit("my_set = set([1, 2, 3, 4, 5, 6, 7, 8, 9, 0])"))
print(timeit("my_set = {*[1, 2, 3, 4, 5, 6, 7, 8, 9, 0]}"))

0.3326582070000086
0.28732166300000017 

You’ll probably see a negligible difference, but there is a difference! That’s because Python is doing a lot more work for the function call then it is when it uses set syntax. Mainly, Python is storing the namespace and stack as it executes the function.

What Python Is Actually Doing

So what is Python doing? Why are funciton calls taking longer? Imagine you’re a Python interpreter that is told to execute a function. You can’t just drop all that you’re doing to execute the newly-provided function! You need to put away whatever you’re doing first before doing something else. Otherwise, you won’t be able to resume your work when you’re done interpreting the function.

Here are a few things that Python needs to do with functions:

1. It needs to look up the function in memory so that it can use it
2. It needs to initialize and manage a frame for the function
3. It needs to store all local variables
4. It needs to store activation records (i.e. call frames) that are kept on the runtime stack
5. It needs to push the frame to the top of the stack when it’s finally called

That’s a lot of stuff Python needs to do in order to handle your set() call! In contrast, here’s what Python needs to do when using set syntax:

1. It needs to load the constant (i.e. constant folding) for the operator

Python can do this pretty quickly, resulting in faster generation of sets when using set syntax than when using the set() function, which itself has to build the set from a generator. The result? Faster set conversion with the set syntax than with the function (for most cases I tested). In reality, you might be able to achieve equivalent performance using set() when your data is already an iterable or a generator object.

Summary Information

Now that we’ve gotten to the bottom of this, we can change that function call to set syntax in our project, thus shutting-up PyCharm! I hope this helps someone who’s wondering why the heck their IDE is angry at them for using a convenient, built-in Python function. The solution: use an even-more convenient Python literal!

FAQ The * Character

The * character unpacks a collection so that it can be converted into a set using set syntax. Try the following in your terminal… A TypeError will be raised when you try to convert my_list to a set. Python complains with the following: TypeError: unhashable type: 'list' To resolve, this, we need to unpack the collection with the * character.

try: 
  my_list = [1, 2, 3, 4, 5]
  my_set = {my_list}
except TypeError:
  print("See! I told you so!")

Now with the * character unpacking our collection, we get a proper set object!

print(type({*[1, 2, 3, 4, 5]}))

<class 'set'>

When you do {[1, 2, 3]}, Python will try to make a set with one element. The * character uses list expansion to construct the set from the containing members, not the list object. Thanks to all the people who emailed me to share their knowledge! You all helped make this blog post more accurate and informative.

Further Reading Materials

Here are some articles and documentation that I found interesting:

1. https://sites.cs.ucsb.edu/~pconrad/cs8/topics.beta/theStack/02/
2. https://www.codementor.io/mjpieters/python-optimization-how-it-can-make-you-a-better-programmer-ajiiftqbo
3. https://docs.python.org/dev/reference/expressions.html#set-displays
4. https://docs.python.org/dev/reference/expressions.html#grammar-token-starred-list