Python type annotations style for functions with long arguments

Question

What is the canon (based on some PEP) way to style a function definition with long type annotations?

Some alternatives I can conceive:

• Option 1:

def my_long_function_name_super_important(
   first_long_argument_to_my_function: Union[int, str],
   second: Dict[int, str],
   third_arg: int = 0 ) -> (bool, int):
   
   return (True, 1)

• Option 2:

def my_long_function_name_super_important(
   first_long_argument_to_my_function: Union[int, str],
   second: Dict[int, str],
   third_arg: int = 0
) -> (bool, int):
   
   return (True, 1)

• Option 3:

def my_long_function_name_super_important(
   first_long_argument_to_my_function: Union[int, str],
   second: Dict[int, str],  third_arg: int = 0
) -> (bool, int):
   
   return (True, 1)

• Option 4:

def my_long_function_name_super_important(
   first_long_argument_to_my_function: Union[int, str],
   second: Dict[int, str],
   third_arg: int = 0,
   ) -> (bool, int):
   
   return (True, 1)

Note: targeting Python 3.8+ to get an up-to-date answer.

Answer 1

As hinted by @MisterMiyagi, there is no official coding style based on a PEP for these type annotations in functions.

That being said, after playing a bit with the different options, I realized options 2 and 3 generate some minor problems in some python code formatters (e.g.: code folding in functions in jupyter notebook) due to the lack of indentation in the closing ). As such, and in search of the highest readability, I ended up using something similar to option 2, but with a small indentation at the last line of the function definition (easier to spot), and with one argument per line (easier to read):

def my_long_function_name_super_important(
   first_long_argument_to_my_function: Union[int, str],
   second: Dict[int, str],
   third_arg: int = 0
  ) -> (bool, int):
   
   return (True, 1)

F.Y.I., the black formatter of @MisterMiyagi seems to be exactly Option 2 (can be tried out here)

edit: from @pabouk 's comment, Google style seems to lean towards options 2-3.