Search code examples
pythonperformanceif-statementoptimizationreturn

Is using else faster than returning value right away?


Which of the following is faster?

1.

def is_even(num: int):
    if num%2==0:
        return True
    else:
        return False
def is_even(num: int):
    if num%2==0:
        return True
    return False

Regardless of which is faster, is there anything in PEP that states when to use one over the other (I couldn't find anything...)?


Solution

  • Compare the byte codes shown by dis.dis().

    if/else version:

      2           0 LOAD_FAST                0 (num)
                  2 LOAD_CONST               1 (2)
                  4 BINARY_MODULO
                  6 LOAD_CONST               2 (0)
                  8 COMPARE_OP               2 (==)
                 10 POP_JUMP_IF_FALSE       16
    
      3          12 LOAD_CONST               3 (True)
                 14 RETURN_VALUE
    
      5     >>   16 LOAD_CONST               4 (False)
                 18 RETURN_VALUE
                 20 LOAD_CONST               0 (None)
                 22 RETURN_VALUE
    

    if/return version:

      2           0 LOAD_FAST                0 (num)
                  2 LOAD_CONST               1 (2)
                  4 BINARY_MODULO
                  6 LOAD_CONST               2 (0)
                  8 COMPARE_OP               2 (==)
                 10 POP_JUMP_IF_FALSE       16
    
      3          12 LOAD_CONST               3 (True)
                 14 RETURN_VALUE
    
      4     >>   16 LOAD_CONST               4 (False)
                 18 RETURN_VALUE
    

    They're idential up to byte 18. The if/else version has an extra 2 instructions to return None if neither of the blocks return from the function.

    This is wasted memory, but won't affect the running time because both the if and else blocks do return. An optimizing compiler would realize that both if and else end with return so this isn't needed, but the Python compiler doesn't bother (apparently this has been fixed between Python 3.9.2 and 3.11.1).