# config.py
default = False

# lib.py
import config

def f(x=config.default):
    print x

# main.py
import config
import lib

config.default = True
lib.f()

The main function sets the configuration, then calls f. One would expect that the program prints True… but it actually prints False.

This behavior can be surprising, but it is perfectly logical once you know the rule:

In Python, default values are evaluated at import time.

This is all there is to know about this problem. Here is what happens at runtime in the above example:

• The main program first imports config. The definition of default is evaluated and its value is False.
• lib is imported, and the definition of lib.f is evaluated. The value of this function includes the default value for x. So, the definition of this default value, config.default, is evaluated and it is False.
• When the value True is assigned to config.default, it is too late: the value False is already part of the function’s value.

That last part is not only a metaphor, the default value is actually part of the function object:

>>> def f(x=3):
...     print x
...
>>> f.func_defaults
(3,)

In order to avoid this caveat, the usual solution is to use None for default values:

def f(x=None):
    if x is None:
        x = config.default
    print x

That way, the evaluation of config.default will happen at runtime, which is what we want here. The above program will indeed print True.