Stacking multiple decorators together will result in unexpected behavior. A trivial example:
from functools import wraps
from random import random
import time
from memory_profiler import profile
def timeit(f):
@wraps(wrapped=f)
def wrapper(*args, **kw):
ts = time.time()
result = f(*args, **kw)
te = time.time()
print(f'func: {f.__name__} took: {te-ts:2.4f} sec')
return result
return wrapper
@profile
@timeit
def to_profile(n):
arr = [random() for i in range(n)]
return sum(arr)
to_profile(1_000)What I expect to see:
Line # Mem usage Increment Occurrences Line Contents
=============================================================
20 48.8 MiB 48.8 MiB 1 @timeit
21 @profile
22 def to_profile(n):
23 48.8 MiB 0.0 MiB 1003 arr = [random() for i in range(n)]
24 48.8 MiB 0.0 MiB 1 return sum(arr)
What we actually get:
Line # Mem usage Increment Occurrences Line Contents
=============================================================
9 49.0 MiB 49.0 MiB 1 @wraps(wrapped=f)
10 def wrapper(*args, **kw):
11 49.0 MiB 0.0 MiB 1 ts = time.time()
12 49.0 MiB 0.0 MiB 1 result = f(*args, **kw)
13 49.0 MiB 0.0 MiB 1 te = time.time()
14 49.0 MiB 0.0 MiB 1 print(f'func: {f.__name__} took: {te-ts:2.4f} sec')
15 49.0 MiB 0.0 MiB 1 return result
I realized while checking this that the sample timeit decorator has the same issue and is timing the @profile rather than the underlying to_profile.
Stacking multiple decorators together will result in unexpected behavior. A trivial example:
What I expect to see:
What we actually get:
I realized while checking this that the sample timeit decorator has the same issue and is timing the
@profilerather than the underlyingto_profile.