Question

我试图使用ctypes从Python（3）程序中调用iconv(3)。 iconv的C类型签名是

size_t iconv(iconv_t cd,
             char **inptr, size_t *inbytesleft,
             char **outptr, size_t *outbytesleft);

你应该像这样称呼它：

char *inp = "abcdef";
char outbuf[16];
char *outp = outbuf;
size_t ibytes = strlen(inbuf);
size_t obytes = sizeof outbuf;

size_t rv = iconv(cd, &inp, &ibytes, &outp, &obytes);

显然，它会写入outbuf，并且还会修改所有四个变量 inp，outp，ibytes，和obytes表示在遇到问题（如果有的话）之前它与转换的距离。它保证不写入输入字符串，尽管不是const。

现在，天真地，你在ctypes中反映出这样：

iconv = libc.iconv
iconv.restype = ctypes.c_size_t
iconv.argtypes = [ctypes.c_void_p, 
                  ctypes.POINTER(ctypes.c_char_p),
                  ctypes.POINTER(ctypes.c_size_t),
                  ctypes.POINTER(ctypes.c_char_p),
                  ctypes.POINTER(ctypes.c_size_t)]

（iconv_t是C库中void *的typedef我正在测试中）但是当我尝试调用时，我得到错误：< / p>

>>> obuf = ctypes.create_string_buffer(16)
>>> obuflen = ctypes.c_size_t(16)
>>> iconv(utf8_to_utf16,
...       ctypes.byref(ctypes.c_char_p(b"abcdef")),
...       ctypes.byref(ctypes.c_size_t(6)),
...       ctypes.byref(obuf),
...       ctypes.byref(obuflen))
ArgumentError: argument 4: <class 'TypeError'>: expected LP_c_char_p
    instance instead of pointer to c_char_Array_16

尝试将obuf明确转换为c_char_p也不起作用：

>>>  optr = ctypes.c_char_p(obuf)
TypeError: bytes or integer address expected instead of c_char_Array_16 instance

它在错误消息中使用的这些类型名称不会出现在手册中，而且我很难过。什么是正确的方法？

（如果你想知道我为什么要这样做而不是使用Python的内置编码转换器，那么简短的版本是因为Python的转换器不支持相同的一组编码为[GNU] iconv，也没有//TRANSLIT功能。）

Answer 1

下面是我用来查找通过转换转换为'“'的字符的代码。对不起，它不是很优美，但我希望它对某些人仍然可用。:)

我需要在python2.6下运行它，所以我尝试将它写得更加通用。（2.6有一些陷阱，因此在python3下代码可能会更好，更简单）

from __future__ import print_function
import sys
import ctypes
libc = ctypes.cdll.LoadLibrary("libc.so.6")

if sys.version_info[0]>2:
    def unichr(a):
        return chr(a)

LP_c_char2 = ctypes.POINTER( ctypes.c_char_p) 
LP_c_char = ctypes.POINTER(ctypes.create_string_buffer(16).__class__)                                                                                                  

get_errno_loc = libc.__errno_location
get_errno_loc.restype = ctypes.POINTER(ctypes.c_int)

class MyError(OSError):
    def __init__(self, e):
        if sys.version_info[0]<=2:
            super(MyError, self).__init__(e)
        else:
            super().__init__(e)


def errcheck(ret, func, args):
    if ret == -1 or ret == 2**64-1:
        e = get_errno_loc()[0]
        raise MyError(e)
    return ret

iconv_open = libc.iconv_open
iconv_open.restype = ctypes.c_void_p
ret = iconv_open(
    ctypes.c_char_p(b"ISO8859-2//TRANSLIT"), 
    ctypes.c_char_p(b"UTF-8"))

iconv = libc.iconv
iconv.errcheck = errcheck
iconv.restype = ctypes.c_size_t

obuf = ctypes.create_string_buffer(16)
obuflen = ctypes.c_size_t(16)
optr = LP_c_char(obuf)

inp = b'\xe2\x80\x9c'

iconv.argtypes = [ctypes.c_void_p, LP_c_char2, ctypes.POINTER(ctypes.c_size_t),     ctypes.POINTER(LP_c_char), ctypes.POINTER(ctypes.c_size_t)]
r = iconv(ret, LP_c_char2(ctypes.c_char_p(inp)), ctypes.byref(ctypes.c_size_t(len(inp))), ctypes.byref(optr), ctypes.byref(obuflen))

print(obuf.value)

def func(inp = b"bbb"):
    assert len(inp)<16 , "too big input"
    obuf = ctypes.create_string_buffer(16)
    obuflen = ctypes.c_size_t(16)
    optr = LP_c_char(obuf)
    r = iconv(ret, LP_c_char2(ctypes.c_char_p(inp)), ctypes.byref(ctypes.c_size_t(len(inp))), ctypes.byref(optr), ctypes.byref(obuflen))
    return obuf, obuflen, r, obuf.value[:16-obuflen.value]

oo, uu, r, vys = func(b'\xe2\x80\x9c\xe2\x80\x9c')
print(oo.raw, uu, r, vys)

for i in range(sys.maxunicode):
    try:
        oo, uu, r, vys = func(unichr(i).encode('utf-8'))
        if vys==b'"':
            print(i, unichr(i))
    except UnicodeEncodeError:
        pass
    except MyError as E:
        pass
        # print("MyError: {E} , {i}".format(E=E, i=i))

使用ctypes调用一个接受char **的函数，因为它写入了缓冲区和指针

1 个答案: