Python中, 仿照经典代码实现单例, 却出现了不是单例的的状态, 代码哪里出错了 ?

Question

实际现象

• 期望实现单例, 保持某个属性全局唯一

预期现象

• 仿照经典实现代码, 却出现了不是单例的情况

我要做什么

• 实现scheduler全局唯一, 不派生过多的scheduler

重现步骤

1. 拷贝代码

2. 运行之

3. 查看内存地址

相关代码

• 仿照经典实现

from apscheduler.schedulers.background import BackgroundScheduler

class Borg(object):
    

    __shared_state = {}

    def __init__(self):
        self.__dict__ = Borg.__shared_state

        self.scheduler = BackgroundScheduler()
        
        
s1 = Borg().scheduler
s2 = Borg().scheduler

print s1, s2


# <apscheduler.schedulers.background.BackgroundScheduler object at 0x02623DF0> 
# <apscheduler.schedulers.background.BackgroundScheduler object at 0x02D801D0>

• 我自己想的办法

from apscheduler.schedulers.background import BackgroundScheduler


class Borg(object):
    @classmethod
    def get_scheduler(cls):
        try:
            cls_scheduler = cls.scheduler
        except AttributeError as e:
            cls.scheduler = BackgroundScheduler()

            return cls.scheduler
        else:
            return cls_scheduler
        




bs1 = Borg.get_scheduler()
bs2 = Borg.get_scheduler()
bs3 = Borg.get_scheduler()
print bs1, bs2, bs3


# <apscheduler.schedulers.background.BackgroundScheduler object at 0x02752D70> 
# <apscheduler.schedulers.background.BackgroundScheduler object at 0x02752D70> 
# <apscheduler.schedulers.background.BackgroundScheduler object at 0x02752D70>

上下文环境

• 产品版本: Python 2.7 APScheduler最新

• 操作系统: Linux

• Github链接, 经典实现: https://github.com/faif/pytho...

Answer 1

The classic example is not a pure singleton, but a Brogpattern with singleton characteristics.

The magic lies in using the dictionary of the class attribute __shared_state shared by the class. Subsequent instances created will overwrite the __dict__ of the previous instance. Brog共享的类属性__shared_state的字典，后面创建的实例会覆盖前面实例的__dict__

s1 = Borg().scheduler
s2 = Borg().scheduler

Create an instance

In fact, they are two different instances, but their attributes share the same class attributes, which looks like a singleton. s1，同时初始化属性scheduler，此时的地址是0x02623DF0，再创建一个实例s2，然后修改了属性scheduler，地址为0x02D801D0。s1和s2

If you change the code to this, you should be able to see the process more clearly:

s1 = Borg()
print(s1.scheduler)
s2 = Borg()
print(s1.scheduler)
print(s2.scheduler)
print(s1.scheduler is s2.scheduler)