为什么在没有销毁对象时调用python ctypes类描述符？

Question

>>> from ctypes import *
>>> class A(Structure):
...  _fields_ = [('a', c_int)]
...  def __del__(self):
...   print("destructor called")
... 
>>> a = (A * 10)()
>>> a[0]
<__main__.A object at 0x7f93038cdd08>
>>> a[0]
destructor called
<__main__.A object at 0x7f93038cde18>
>>> a[0]
destructor called
<__main__.A object at 0x7f93038cdd08>
>>> a[0]
destructor called
<__main__.A object at 0x7f93038cde18>

Why is the destructor being called here ? Why is the address of the object different each time ? Why doesn't python crash with a double free error ?

Answer 1

a is a proxy object , representing an array of C structs, in the Python world. Each time you index into this object, ctypes creates a new instance of the A class for you, to proxy the contained C structs.

Now, because you don't store any references to this new object, it is also garbage collected as soon as it's repr() value has been echoed in the interpreter.

You could store the object produced by a[0] in a new variable:

>>> foo = a[0]
>>> foo
<__main__.A object at 0x11061ea60>

This object will always be distinct from any others you create by indexing position 0 of a , but they all represent the same C struct because they reference the same address:

>>> bar = a[0]
>>> foo is bar
False
>>> addressof(foo) == addressof(bar)
True

If you delete this reference the object is destructed again (provided you did not create more references to the Python proxy object of course):

>>> del foo
destructor called