我正在使用multiprocessing.Pool来并行化我为仿真而编写的某些Python代码(在D&D 5e中进行战斗)。但是,我收到此错误:can't pickle _thread.RLock objects
。我做了一些研究,看来问题是我的仿真使用了具有记录器的对象(来自logging
模块)。我仍然想支持模拟的单进程运行的日志记录(以便人们可以调试/检查输出),但是对多进程运行的日志记录支持对我而言并不重要,因为一旦用户知道模拟正确运行,他们就可以通常不需要逐个播放。我该如何重构代码,以使日志记录不再成为问题(在有多个进程的情况下禁用日志记录,以多进程接受的方式实现日志记录,或者其他方式)?
我的Simulation类中的一些代码:
def process_run(self, num): # pylint: disable=unused-argument
"""
The method given to an individual process to do work. Run only once per process so we don't need to reset anything
:param num: which iteration of the loop am I
:return:
"""
self._encounter.run()
return self._encounter.get_stats()
def mp_run(self, n=1000, p=None):
"""
Run the Encounter n times (using multiprocessing), then print the results
:param n: number of times to run the encounter
:param p: number of processes to use. If None, use max number of processes
:return: aggregate stats
"""
if p:
pool = Pool(p) # use user-specified number of processes
else:
pool = Pool() # use max number of processes as user didn't specify
for result in pool.imap_unordered(self.process_run, range(n)):
self.update_stats(result)
pool.close()
pool.join()
self.calculate_aggregate_stats(n)
self.print_aggregate_stats()
return self._aggregate_stats
一些使用日志记录的代码(来自Encounter类,每个Simulation都有一个实例):
self.get_logger().info("Round %d", self.get_round())
unconscious = [comb.get_name() for comb in self.get_combatants() if comb.has_condition("unconscious")]
if unconscious:
self.get_logger().info("Unconscious: %s", str(unconscious))
dead = [comb.get_name() for comb in self.get_combatants() if comb.has_condition("dead")]
if dead:
self.get_logger().info("Dead: %s", str(dead))
如果您需要更多代码示例,请告诉我。