ctypes：c_ulong类型神奇地更改为long

Question

I am writing the deepcopy function using python2.7 for my class. I encountered a weird problem My code is the following

import copy
from ctypes import *
class Graph (Structure):
    _fields_ = [("numVertices", c_ulong),
                    ("numEdges", c_ulong)]
    def __init__(self):
        self.numVertices = c_ulong(0)
        self.numEdges = c_ulong(0)

    def __deepcopy__(self,memo={}):
        newInstance = Graph()
        newInstance.numVertices = c_ulong(self.numVertices.value)
        newInstance.numEdges = c_ulong(self.numEdges.value)
        return newInstance

graph = Graph()
anotherGraph = copy.deepcopy(graph)

I get the following error:

<ipython-input-46-a0cdaa4ef3f7> in __deepcopy__(self, memo)
      9     def __deepcopy__(self,memo={}):
     10         newInstance = Graph()
---> 11         newInstance.numVertices = c_ulong(self.numVertices.value)
     12         newInstance.numEdges = c_ulong(self.numEdges.value)
     13         return newInstance

AttributeError: 'long' object has no attribute 'value'

If you try:

type(graph.numVertices)

The result is long

I declared the numVertices as c_ulong(). Why does it become long?

Answer 1

The type of the fields in the structure is still maintained, but ctypes has some "helpful" conversions when reading the values:

from ctypes import *

class Test(Structure):
    _fields_ = [('a',c_ulong),
                ('b',c_char_p)]

t = Test(1,b'hello')
print(type(t.a),type(t.b))
print(t._fields_)

Output:

<class 'int'> <class 'bytes'>
[('a', <class 'ctypes.c_ulong'>), ('b', <class 'ctypes.c_char_p'>)]

So you can write your code as the following and it will work correctly:

import copy
from ctypes import *
class Graph (Structure):
    _fields_ = [("numVertices", c_ulong),
                    ("numEdges", c_ulong)]
    def __init__(self):
        self.numVertices = 0
        self.numEdges = 0

    def __deepcopy__(self,memo={}):
        newInstance = Graph()
        newInstance.numVertices = self.numVertices
        newInstance.numEdges = self.numEdges
        return newInstance

graph = Graph()
anotherGraph = copy.deepcopy(graph)

You can suppress the conversions by deriving from the classes, but it is usually unnecessary. One use case is when using ctypes to call a function that returns an allocated string. You need to suppress the c_char_p to Python byte string conversion so you can later free the c_char_p.

from ctypes import *

class ulong(c_ulong): pass
class char_p(c_char_p): pass

class Test(Structure):
    _fields_ = [('a',ulong),
                ('b',char_p)]

t = Test(1,b'hello')
print(type(t.a),type(t.b))
print(t.a,t.b)

Output:

<class '__main__.ulong'> <class '__main__.char_p'>
<ulong object at 0x0000000006263748> char_p(b'hello')