Why most of function attributions are writeable?

[Baofeng (Baofeng Wang)]

e.g.

>>> def foo(x, v=3):
...    print(x,v)
...
>>> foo.__defaults__
(3,)
>>> foo()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: foo() missing 1 required positional argument: 'x'
>>> foo(2)
2 3
>>> foo.__defaults__="x",6
>>> foo.__defaults__
('x', 6)
>>> foo()
x 6

It means there is a chance to modify the function contract during runtime. But would that causes surprise when most of users anticipate a function’s behavior per reading its definition(contract)?
Someone may argue that increases the flexibility, though I doubt if increasing flexibility is always a positive thing to take.

And is there any specific use cases demand function attributions being writeable?

[Rosuav (Chris Angelico)]

@Baofeng

It means there is a chance to modify the function contract during runtime. But would that causes surprise when most of users anticipate a function’s behavior per reading its definition(contract)?
Someone may argue that increases the flexibility, though I doubt if increasing flexibility is always a positive thing to take.

Consenting adults. This sort of dynamic programming is incredibly useful, and what would be gained by locking it down?

@Baofeng

And is there any specific use cases demand function attributions being writeable?

Oh yes, lots. Quite a few handy function decorators work, not by constructing an inefficient run-time wrapper, but by modifying the existing function.

[mwichmann (Mats Wichmann)]

To the very specific question about whether it’s modifiable:

3. Data model

says that __defaults__ is writable, so yes - allowed. I’ve never found a need to modify __defaults__ myself, nor have I ever seen it in code, and it feels kind of odd to do so, but other attributes at runtime - absolutely.