如何在 python ctypes 中使用 UTF-16？

Question

我有一个外国 C 库，它在 API 中使用 utf-16：作为函数参数、返回值和结构成员。

在 Windows 上，使用 ctypes.c_wchar_p 没问题，但在 OSX 下，ctypes 在 c_wchar 中使用 UCS-32，我找不到支持 utf-16 的方法。

这是我的研究：

1. 使用 _SimpleCData 子类化来重新定义 _check_retval_。

   • 它允许 utf-16 到 Python 字符串的透明转换。
   • 可以作为C结构成员放置
   • 但它不允许将字符串作为参数处理，它的from_param()方法从未被调用（为什么？）： func('str', b'W\x00B\x00\x00\x00') # passed without conversion

2. 使用自己的类型和from_param()方法。

   • 优点：可以使用构造函数进行初始化，也可以在将字符串传递给函数时即时编码：
   • 缺点：不能用作函数返回类型或结构成员。

这里是：

ustr = myutf16('hello')
func(ustr)
func('hello')   # calls myutf16.from_param('hello')

Answer 1

您可以from_param在c_char_p子类中重写，将字符串编码unicode为 UTF-16。您可以添加一个_check_retval_方法来将 UTF-16 结果解码为unicode字符串。对于结构字段，您可以使用处理设置和获取属性的描述符类。将该字段设置为私有_name类型c_char_p，并将描述符设置为公共name。例如：

import sys
import ctypes

if sys.version_info[0] > 2:
    unicode = str

def decode_utf16_from_address(address, byteorder='little',
                              c_char=ctypes.c_char):
    if not address:
        return None
    if byteorder not in ('little', 'big'):
        raise ValueError("byteorder must be either 'little' or 'big'")
    chars = []
    while True:
        c1 = c_char.from_address(address).value
        c2 = c_char.from_address(address + 1).value
        if c1 == b'\x00' and c2 == b'\x00':
            break
        chars += [c1, c2]
        address += 2
    if byteorder == 'little':
        return b''.join(chars).decode('utf-16le')
    return b''.join(chars).decode('utf-16be')

class c_utf16le_p(ctypes.c_char_p):
    def __init__(self, value=None):
        super(c_utf16le_p, self).__init__()
        if value is not None:
            self.value = value

    @property
    def value(self,
              c_void_p=ctypes.c_void_p):
        addr = c_void_p.from_buffer(self).value
        return decode_utf16_from_address(addr, 'little')

    @value.setter
    def value(self, value,
              c_char_p=ctypes.c_char_p):
        value = value.encode('utf-16le') + b'\x00'
        c_char_p.value.__set__(self, value)

    @classmethod
    def from_param(cls, obj):
        if isinstance(obj, unicode):
            obj = obj.encode('utf-16le') + b'\x00'
        return super(c_utf16le_p, cls).from_param(obj)

    @classmethod
    def _check_retval_(cls, result):
        return result.value

class UTF16LEField(object):
    def __init__(self, name):
        self.name = name

    def __get__(self, obj, cls,
                c_void_p=ctypes.c_void_p,
                addressof=ctypes.addressof):
        field_addr = addressof(obj) + getattr(cls, self.name).offset
        addr = c_void_p.from_address(field_addr).value
        return decode_utf16_from_address(addr, 'little')

    def __set__(self, obj, value):
        value = value.encode('utf-16le') + b'\x00'
        setattr(obj, self.name, value)

例子：

if __name__ == '__main__':
    class Test(ctypes.Structure):
        _fields_ = (('x', ctypes.c_int),
                    ('y', ctypes.c_void_p),
                    ('_string', ctypes.c_char_p))
        string = UTF16LEField('_string')

    print('test 1: structure field')
    t = Test()
    t.string = u'eggs and spam'
    print(t.string)

    print('test 2: parameter and result')
    result = None

    @ctypes.CFUNCTYPE(c_utf16le_p, c_utf16le_p)
    def testfun(string):
        global result
        print('parameter: %s' % string.value)
        # callbacks leak memory except for simple return
        # values such as an integer address, so return the
        # address of a global variable.
        result = c_utf16le_p(string.value + u' and eggs')
        return ctypes.c_void_p.from_buffer(result).value

    print('result: %s' % testfun(u'spam'))

输出：

import sys
import ctypes

if sys.version_info[0] > 2:
    unicode = str

def decode_utf16_from_address(address, byteorder='little',
                              c_char=ctypes.c_char):
    if not address:
        return None
    if byteorder not in ('little', 'big'):
        raise ValueError("byteorder must be either 'little' or 'big'")
    chars = []
    while True:
        c1 = c_char.from_address(address).value
        c2 = c_char.from_address(address + 1).value
        if c1 == b'\x00' and c2 == b'\x00':
            break
        chars += [c1, c2]
        address += 2
    if byteorder == 'little':
        return b''.join(chars).decode('utf-16le')
    return b''.join(chars).decode('utf-16be')

class c_utf16le_p(ctypes.c_char_p):
    def __init__(self, value=None):
        super(c_utf16le_p, self).__init__()
        if value is not None:
            self.value = value

    @property
    def value(self,
              c_void_p=ctypes.c_void_p):
        addr = c_void_p.from_buffer(self).value
        return decode_utf16_from_address(addr, 'little')

    @value.setter
    def value(self, value,
              c_char_p=ctypes.c_char_p):
        value = value.encode('utf-16le') + b'\x00'
        c_char_p.value.__set__(self, value)

    @classmethod
    def from_param(cls, obj):
        if isinstance(obj, unicode):
            obj = obj.encode('utf-16le') + b'\x00'
        return super(c_utf16le_p, cls).from_param(obj)

    @classmethod
    def _check_retval_(cls, result):
        return result.value

class UTF16LEField(object):
    def __init__(self, name):
        self.name = name

    def __get__(self, obj, cls,
                c_void_p=ctypes.c_void_p,
                addressof=ctypes.addressof):
        field_addr = addressof(obj) + getattr(cls, self.name).offset
        addr = c_void_p.from_address(field_addr).value
        return decode_utf16_from_address(addr, 'little')

    def __set__(self, obj, value):
        value = value.encode('utf-16le') + b'\x00'
        setattr(obj, self.name, value)