python - Huge time difference when making a set inside a list comprehension compared to outside -

i have test case below, find elements present in 2 strings , return them in order of first string. why second case, set(string2) done inside list comprehension on order of magnitude slower?

import timeit timeit.timeit(     """     import random, string     string1 = ''.join(random.choice(string.ascii_uppercase) _ in range(100000))     string2 = ''.join(random.choice(string.ascii_uppercase) _ in range(100000))     v=set(string2)     [x x in string1 if x in v]     """,     number=10, ) # 2.005337953567505  timeit.timeit(     """     import random, string     string1 = ''.join(random.choice(string.ascii_uppercase) _ in range(100000))     string2 = ''.join(random.choice(string.ascii_uppercase) _ in range(100000))     [x x in string1 if x in set(string2)]     """,     number=10, ) # 22.1062970161438 

python building set(string2) set every iteration of 'loop' in list comprehension. doesn't 'know' it's same set every time.

e.g. every time python executes list comprehension if clause, x in set(string2) evaluated, means set created every time. since x each char of 100000 character string, means (same) set being built 100000 times in second case, once in first case (where define set v).