Ошибка при профилировании отлично работающего многопроцессорного Python-скрипта с помощью cProfile

Question

Я написал небольшой скрипт на Python, который использует multiprocessing (См. /questions/42934891/mnogoprotsessornaya-obrabotka-na-teed-generatorah/42934897#42934897). Это работает, когда я проверяю это:

$ ./forkiter.py
0
1
2
3
4
sum of x+1: 15
sum of 2*x: 20
sum of x*x: 30

Но когда я пытаюсь профилировать его с cProfileЯ получаю следующее:

$ python3.6 -m cProfile -o forkiter.prof ./forkiter.py
0
1
2
3
4
Traceback (most recent call last):
  File "/home/bli/lib/python3.6/runpy.py", line 193, in _run_module_as_main
    "__main__", mod_spec)
  File "/home/bli/lib/python3.6/runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "/home/bli/lib/python3.6/cProfile.py", line 160, in <module>
    main()
  File "/home/bli/lib/python3.6/cProfile.py", line 153, in main
    runctx(code, globs, None, options.outfile, options.sort)
  File "/home/bli/lib/python3.6/cProfile.py", line 20, in runctx
    filename, sort)
  File "/home/bli/lib/python3.6/profile.py", line 64, in runctx
    prof.runctx(statement, globals, locals)
  File "/home/bli/lib/python3.6/cProfile.py", line 100, in runctx
    exec(cmd, globals, locals)
  File "./forkiter.py", line 71, in <module>
    exit(main())
  File "./forkiter.py", line 67, in main
    sum_tuples, results_generator))
  File "/home/bli/lib/python3.6/multiprocessing/pool.py", line 699, in next
    raise value
  File "/home/bli/lib/python3.6/multiprocessing/pool.py", line 385, in _handle_tasks
    put(task)
  File "/home/bli/lib/python3.6/multiprocessing/connection.py", line 206, in send
    self._send_bytes(_ForkingPickler.dumps(obj))
  File "/home/bli/lib/python3.6/multiprocessing/reduction.py", line 51, in dumps
    cls(buf, protocol).dump(obj)
_pickle.PicklingError: Can't pickle <class '__main__.FuncApplier'>: attribute lookup FuncApplier on __main__ failed

Что просходит?

Вот сценарий:

#!/usr/bin/env python3
"""This script tries to work around some limitations of multiprocessing."""

from itertools import repeat, starmap
from multiprocessing import Pool
from functools import reduce
from operator import add
from time import sleep

# Doesn't work because local functions can't be pickled:
# def make_tuple_func(funcs):
#     def tuple_func(args_list):
#         return tuple(func(args) for func, args in zip(funcs, args_list))
#     return tuple_func
#
# test_tuple_func = make_tuple_func((plus_one, double, square))

class FuncApplier(object):
    """This kind of object can be used to group functions and call them on a
    tuple of arguments."""
    __slots__ = ("funcs", )

    def __init__(self, funcs):
        self.funcs = funcs

    def __len__(self):
        return len(self.funcs)

    def __call__(self, args_list):
        return tuple(func(args) for func, args in zip(self.funcs, args_list))

    def fork_args(self, args_list):
        """Takes an arguments list and repeat them in a n-tuple."""
        return tuple(repeat(args_list, len(self)))


def sum_tuples(*tuples):
    """Element-wise sum of tuple items."""
    return tuple(starmap(add, zip(*tuples)))


# Can't define these functions in main:
# They wouldn't be pickleable.
def plus_one(x):
    return x + 1

def double(x):
    return 2 * x

def square(x):
    return x * x

def main():
    def my_generator():
        for i in range(5):
            print(i)
            yield i


    test_tuple_func = FuncApplier((plus_one, double, square))

    with Pool(processes=5) as pool:
        results_generator = pool.imap_unordered(
            test_tuple_func,
            (test_tuple_func.fork_args(args_list) for args_list in my_generator()))
        print("sum of x+1:\t%s\nsum of 2*x:\t%s\nsum of x*x:\t%s" % reduce(
            sum_tuples, results_generator))
    exit(0)

if __name__ == "__main__":
    exit(main())

---

Некоторые тесты на травление

Некоторые исследования показали, что иногда объектам нужен __setstate__ а также __getstate__ методы, чтобы быть маринованным. Это помогает для некоторых протоколов травления, но, похоже, это не решает проблему в cProfile дело. Смотрите тесты ниже.

Обновленный скрипт:

#!/usr/bin/env python3
"""This script tries to work around some limitations of multiprocessing."""

from itertools import repeat, starmap
from multiprocessing import Pool
from functools import reduce
from operator import add
from time import sleep
import pickle

# Doesn't work because local functions can't be pickled:
# def make_tuple_func(funcs):
#     def tuple_func(args_list):
#         return tuple(func(args) for func, args in zip(funcs, args_list))
#     return tuple_func
#
# test_tuple_func = make_tuple_func((plus_one, double, square))

class FuncApplier(object):
    """This kind of object can be used to group functions and call them on a
    tuple of arguments."""
    __slots__ = ("funcs", )

    def __init__(self, funcs):
        self.funcs = funcs

    def __len__(self):
        return len(self.funcs)

    def __call__(self, args_list):
        return tuple(func(args) for func, args in zip(self.funcs, args_list))

    # Attempt to make it pickleable when under cProfile (doesn't help)
    def __getstate__(self):
        return self.funcs

    def __setstate__(self, state):
        self.funcs = state

    def fork_args(self, args_list):
        """Takes an arguments list and repeat them in a n-tuple."""
        return tuple(repeat(args_list, len(self)))


def sum_tuples(*tuples):
    """Element-wise sum of tuple items."""
    return tuple(starmap(add, zip(*tuples)))


# Can't define these functions in main:
# They wouldn't be pickleable.
def plus_one(x):
    return x + 1

def double(x):
    return 2 * x

def square(x):
    return x * x

def main():
    def my_generator():
        for i in range(5):
            print(i)
            yield i


    test_tuple_func = FuncApplier((plus_one, double, square))

    print("protocol 0")
    try:
        print(pickle.dumps(test_tuple_func, 0))
    except pickle.PicklingError as err:
        print("failed with the following error:\n%s" % err)
    print("protocol 1")
    try:
        print(pickle.dumps(test_tuple_func, 0))
    except pickle.PicklingError as err:
        print("failed with the following error:\n%s" % err)
    print("protocol 2")
    try:
        print(pickle.dumps(test_tuple_func, 0))
    except pickle.PicklingError as err:
        print("failed with the following error:\n%s" % err)
    print("protocol 3")
    try:
        print(pickle.dumps(test_tuple_func, 0))
    except pickle.PicklingError as err:
        print("failed with the following error:\n%s" % err)
    print("protocol 4")
    try:
        print(pickle.dumps(test_tuple_func, 0))
    except pickle.PicklingError as err:
        print("failed with the following error:\n%s" % err)

    with Pool(processes=5) as pool:
        results_generator = pool.imap_unordered(
            test_tuple_func,
            (test_tuple_func.fork_args(args_list) for args_list in my_generator()))
        print("sum of x+1:\t%s\nsum of 2*x:\t%s\nsum of x*x:\t%s" % reduce(
            sum_tuples, results_generator))
    exit(0)

if __name__ == "__main__":
    exit(main())

Тест без cProfile кажется в порядке:

$ ./forkiter.py
protocol 0
b'ccopy_reg\n_reconstructor\np0\n(c__main__\nFuncApplier\np1\nc__builtin__\nobject\np2\nNtp3\nRp4\n(c__main__\nplus_one\np5\nc__main__\ndouble\np6\nc__main__\nsquare\np7\ntp8\nb.'
protocol 1
b'ccopy_reg\n_reconstructor\np0\n(c__main__\nFuncApplier\np1\nc__builtin__\nobject\np2\nNtp3\nRp4\n(c__main__\nplus_one\np5\nc__main__\ndouble\np6\nc__main__\nsquare\np7\ntp8\nb.'
protocol 2
b'ccopy_reg\n_reconstructor\np0\n(c__main__\nFuncApplier\np1\nc__builtin__\nobject\np2\nNtp3\nRp4\n(c__main__\nplus_one\np5\nc__main__\ndouble\np6\nc__main__\nsquare\np7\ntp8\nb.'
protocol 3
b'ccopy_reg\n_reconstructor\np0\n(c__main__\nFuncApplier\np1\nc__builtin__\nobject\np2\nNtp3\nRp4\n(c__main__\nplus_one\np5\nc__main__\ndouble\np6\nc__main__\nsquare\np7\ntp8\nb.'
protocol 4
b'ccopy_reg\n_reconstructor\np0\n(c__main__\nFuncApplier\np1\nc__builtin__\nobject\np2\nNtp3\nRp4\n(c__main__\nplus_one\np5\nc__main__\ndouble\np6\nc__main__\nsquare\np7\ntp8\nb.'
0
1
2
3
4
sum of x+1: 15
sum of 2*x: 20
sum of x*x: 30

Тест в cProfile не проходит при каждом протоколе протравливания (и, следовательно, в многопроцессорной обработке):

$ python3.6 -m cProfile -o forkiter.prof ./forkiter.py
protocol 0
failed with the following error:
Can't pickle <class '__main__.FuncApplier'>: attribute lookup FuncApplier on __main__ failed
protocol 1
failed with the following error:
Can't pickle <class '__main__.FuncApplier'>: attribute lookup FuncApplier on __main__ failed
protocol 2
failed with the following error:
Can't pickle <class '__main__.FuncApplier'>: attribute lookup FuncApplier on __main__ failed
protocol 3
failed with the following error:
Can't pickle <class '__main__.FuncApplier'>: attribute lookup FuncApplier on __main__ failed
protocol 4
failed with the following error:
Can't pickle <class '__main__.FuncApplier'>: attribute lookup FuncApplier on __main__ failed
0
1
2
3
4
Traceback (most recent call last):
  File "/home/bli/lib/python3.6/runpy.py", line 193, in _run_module_as_main
    "__main__", mod_spec)
  File "/home/bli/lib/python3.6/runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "/home/bli/lib/python3.6/cProfile.py", line 160, in <module>
    main()
  File "/home/bli/lib/python3.6/cProfile.py", line 153, in main
    runctx(code, globs, None, options.outfile, options.sort)
  File "/home/bli/lib/python3.6/cProfile.py", line 20, in runctx
    filename, sort)
  File "/home/bli/lib/python3.6/profile.py", line 64, in runctx
    prof.runctx(statement, globals, locals)
  File "/home/bli/lib/python3.6/cProfile.py", line 100, in runctx
    exec(cmd, globals, locals)
  File "./forkiter.py", line 105, in <module>
    exit(main())
  File "./forkiter.py", line 101, in main
    sum_tuples, results_generator))
  File "/home/bli/lib/python3.6/multiprocessing/pool.py", line 699, in next
    raise value
  File "/home/bli/lib/python3.6/multiprocessing/pool.py", line 385, in _handle_tasks
    put(task)
  File "/home/bli/lib/python3.6/multiprocessing/connection.py", line 206, in send
    self._send_bytes(_ForkingPickler.dumps(obj))
  File "/home/bli/lib/python3.6/multiprocessing/reduction.py", line 51, in dumps
    cls(buf, protocol).dump(obj)
_pickle.PicklingError: Can't pickle <class '__main__.FuncApplier'>: attribute lookup FuncApplier on __main__ failed

Answer 1

Похоже на то cProfile просто не работает с многопроцессорностью.

Если вы готовы изменить код только для профилирования основного процесса (или добавить специальное профилирование для подпроцессов), cProfile.run() кажется, работает в определенной степени.

В вашем примере замените

exit(main())

с

exit(cProfile.run('main()')

По крайней мере, это работает, если параллельная функция является функцией глобальной области видимости, но не уверен, что это также верно для класса, как в вашем случае.