Python全局变量在模块之间引用的问题

昨天在Python群里有人提到一个问题，在模块a中引入模块b的全局变量x，并在a中调用b的函数来修改x的值，结果出现很难理解的现象。经过反思发现自己对这个概念也是一知半解，好像在《Python源码剖析》那本书中看到过相关的知识但是又有些记不太清，故编写程序做一些验证以扫清编码障碍，至于实现层面的理解等晚一点再翻翻书（PS: 这件事情显示读书笔记还是很重要的，不然每次都得去书中重新找相关信息）。

这其实是两个问题：1.from moduleA import * 和import moduleA在使用全局变量上的区别； 2.__main__模块横插一脚会怎样？

1.程序

[dongsong@localhost python_study]$ cat testc.py
import pprint, sys, pdb
a = list()
b = list()
def A():
        global a
        a = dict()
        b.append(100)
        print "A() a = %r, b = %r" % (a,b)
[dongsong@localhost python_study]$ cat testd.py 
#encoding=utf8
import pprint, sys, pdb

if __name__ == "__main__":
        if len(sys.argv) == 1:
                print "Usage:\n\tpython %s 1/2" % __file__
                sys.exit()

        if int(sys.argv[1]) == 1:
                print "import testc ..."
                import testc
                print "testc.a = %r, testc.b = %r" % (testc.a, testc.b)
                testc.A()
                print "testc.a = %r, testc.b = %r" % (testc.a, testc.b)

        elif int(sys.argv[1]) == 2:
                print "from testc import *"
                print "before import: globals().keys() = %r" % globals().keys()
                from testc import *
                print "after import: globals().keys() = %r" % globals().keys() #可以看到testc模块中的符号都被导入__main__模块的全局名字空间里了
                print "a = %r, b = %r, sys.modules['testc'].a = %r, sys.modules['testc'].b = %r" % (a, b, sys.modules['testc'].a, sys.modules['testc'].b)
                A() #WARNING:A()函数让testc模块中的符号a指向了新的对象，故testc和__main__中的"a"符号指向了不同对象；A()函数没有改变"b"符号指向的对象，只是修改了该对象的元素，故testc和__main__中的"b"符号对应的数值都跟着变化（还是指向同一个对象）
                print "a = %r, b = %r, sys.modules['testc'].a = %r, sys.modules['testc'].b = %r" % (a, b, sys.modules['testc'].a, sys.modules['testc'].b)

        else:
                print "error argv.."

结果

[dongsong@localhost python_study]$ vpython testd.py 1
import testc ...
testc.a = [], testc.b = []
A() a = {}, b = [100]
testc.a = {}, testc.b = [100]
[dongsong@localhost python_study]$ vpython testd.py 2
from testc import *
before import: globals().keys() = ['__builtins__', '__file__', 'pprint', '__package__', 'sys', 'pdb', '__name__', '__doc__']
after import: globals().keys() = ['a', 'A', 'b', '__builtins__', '__file__', 'pprint', '__package__', 'sys', 'pdb', '__name__', '__doc__']
a = [], b = [], sys.modules['testc'].a = [], sys.modules['testc'].b = []
A() a = {}, b = [100]
a = [], b = [100], sys.modules['testc'].a = {}, sys.modules['testc'].b = [100]

一目了然，不做赘述！

2.程序

[dongsong@localhost python_study]$ cat testa.py
import pprint, sys, pdb
a = list()
b = list()
def A():
        global a
        a = dict() 
        b.append(100)
        print "A() a = %r, b = %r" % (a,b)

if __name__ == '__main__':
        import testa
        testaInModules = sys.modules['testa']
        testaAsMain = sys.modules['__main__']
        print "testa: id = %r, file = %r" % (id(testa), testa.__file__)
        print "testaInModules: id = %r, file = %r" % (id(testaInModules), testaInModules.__file__)
        print "testaAsMain: id = %r, file = %r" % (id(testaAsMain), testaAsMain.__file__)
        print ""

        A()

        print ""
        print "testa: a = %r, b = %r" % (testa.a,testa.b)
        print "testaInModules: a = %r, b = %r" % (testaInModules.a,testaInModules.b)
        print "testaAsMain: a = %r, b = %r" % (testaAsMain.a,testaAsMain.b)
        print ""

        #import testb
        #testb.B()

结果

[dongsong@localhost python_study]$ vpython testa.py
testa: id = 140681127912120, file = '/home/dongsong/python_study/testa.py'
testaInModules: id = 140681127912120, file = '/home/dongsong/python_study/testa.py'
testaAsMain: id = 140681230481008, file = 'testa.py'

A() a = {}, b = [100]

testa: a = [], b = []
testaInModules: a = [], b = []
testaAsMain: a = {}, b = [100]

把testa模块作为__main__模块时，在内存中__main__模块跟testa模块虽然是由相同的源码(相同的字节码文件testa.pyc)加载的，但是它们是两个独立的module对象