Python/fileutils.c

#include "Python.h"
#ifdef MS_WINDOWS
#  include <windows.h>
#endif

#ifdef HAVE_STAT

/* Decode a byte string from the locale encoding with the
   surrogateescape error handler (undecodable bytes are decoded as characters
   in range U+DC80..U+DCFF). If a byte sequence can be decoded as a surrogate
   character, escape the bytes using the surrogateescape error handler instead
   of decoding them.

   Use _Py_wchar2char() to encode the character string back to a byte string.

   Return a pointer to a newly allocated (wide) character string (use
   PyMem_Free() to free the memory), or NULL on error (conversion error or
   memory error). */
wchar_t*
_Py_char2wchar(char* arg)
{
    wchar_t *res;
#ifdef HAVE_BROKEN_MBSTOWCS
    /* Some platforms have a broken implementation of
     * mbstowcs which does not count the characters that
     * would result from conversion.  Use an upper bound.
     */
    size_t argsize = strlen(arg);
#else
    size_t argsize = mbstowcs(NULL, arg, 0);
#endif
    size_t count;
    unsigned char *in;
    wchar_t *out;
#ifdef HAVE_MBRTOWC
    mbstate_t mbs;
#endif
    if (argsize != (size_t)-1) {
        res = (wchar_t *)PyMem_Malloc((argsize+1)*sizeof(wchar_t));
        if (!res)
            goto oom;
        count = mbstowcs(res, arg, argsize+1);
        if (count != (size_t)-1) {
            wchar_t *tmp;
            /* Only use the result if it contains no
               surrogate characters. */
            for (tmp = res; *tmp != 0 &&
                         (*tmp < 0xd800 || *tmp > 0xdfff); tmp++)
                ;
            if (*tmp == 0)
                return res;
        }
        PyMem_Free(res);
    }
    /* Conversion failed. Fall back to escaping with surrogateescape. */
#ifdef HAVE_MBRTOWC
    /* Try conversion with mbrtwoc (C99), and escape non-decodable bytes. */

    /* Overallocate; as multi-byte characters are in the argument, the
       actual output could use less memory. */
    argsize = strlen(arg) + 1;
    res = (wchar_t*)PyMem_Malloc(argsize*sizeof(wchar_t));
    if (!res) goto oom;
    in = (unsigned char*)arg;
    out = res;
    memset(&mbs, 0, sizeof mbs);
    while (argsize) {
        size_t converted = mbrtowc(out, (char*)in, argsize, &mbs);
        if (converted == 0)
            /* Reached end of string; null char stored. */
            break;
        if (converted == (size_t)-2) {
            /* Incomplete character. This should never happen,
               since we provide everything that we have -
               unless there is a bug in the C library, or I
               misunderstood how mbrtowc works. */
            fprintf(stderr, "unexpected mbrtowc result -2\n");
            return NULL;
        }
        if (converted == (size_t)-1) {
            /* Conversion error. Escape as UTF-8b, and start over
               in the initial shift state. */
            *out++ = 0xdc00 + *in++;
            argsize--;
            memset(&mbs, 0, sizeof mbs);
            continue;
        }
        if (*out >= 0xd800 && *out <= 0xdfff) {
            /* Surrogate character.  Escape the original
               byte sequence with surrogateescape. */
            argsize -= converted;
            while (converted--)
                *out++ = 0xdc00 + *in++;
            continue;
        }
        /* successfully converted some bytes */
        in += converted;
        argsize -= converted;
        out++;
    }
#else
    /* Cannot use C locale for escaping; manually escape as if charset
       is ASCII (i.e. escape all bytes > 128. This will still roundtrip
       correctly in the locale's charset, which must be an ASCII superset. */
    res = PyMem_Malloc((strlen(arg)+1)*sizeof(wchar_t));
    if (!res) goto oom;
    in = (unsigned char*)arg;
    out = res;
    while(*in)
        if(*in < 128)
            *out++ = *in++;
        else
            *out++ = 0xdc00 + *in++;
    *out = 0;
#endif
    return res;
oom:
    fprintf(stderr, "out of memory\n");
    return NULL;
}

/* Encode a (wide) character string to the locale encoding with the
   surrogateescape error handler (characters in range U+DC80..U+DCFF are
   converted to bytes 0x80..0xFF).

   This function is the reverse of _Py_char2wchar().

   Return a pointer to a newly allocated byte string (use PyMem_Free() to free
   the memory), or NULL on error (conversion error or memory error). */
char*
_Py_wchar2char(const wchar_t *text)
{
    const size_t len = wcslen(text);
    char *result = NULL, *bytes = NULL;
    size_t i, size, converted;
    wchar_t c, buf[2];

    /* The function works in two steps:
       1. compute the length of the output buffer in bytes (size)
       2. outputs the bytes */
    size = 0;
    buf[1] = 0;
    while (1) {
        for (i=0; i < len; i++) {
            c = text[i];
            if (c >= 0xdc80 && c <= 0xdcff) {
                /* UTF-8b surrogate */
                if (bytes != NULL) {
                    *bytes++ = c - 0xdc00;
                    size--;
                }
                else
                    size++;
                continue;
            }
            else {
                buf[0] = c;
                if (bytes != NULL)
                    converted = wcstombs(bytes, buf, size);
                else
                    converted = wcstombs(NULL, buf, 0);
                if (converted == (size_t)-1) {
                    if (result != NULL)
                        PyMem_Free(result);
                    return NULL;
                }
                if (bytes != NULL) {
                    bytes += converted;
                    size -= converted;
                }
                else
                    size += converted;
            }
        }
        if (result != NULL) {
            *bytes = 0;
            break;
        }

        size += 1; /* nul byte at the end */
        result = PyMem_Malloc(size);
        if (result == NULL)
            return NULL;
        bytes = result;
    }
    return result;
}

/* In principle, this should use HAVE__WSTAT, and _wstat
   should be detected by autoconf. However, no current
   POSIX system provides that function, so testing for
   it is pointless.
   Not sure whether the MS_WINDOWS guards are necessary:
   perhaps for cygwin/mingw builds?
*/
#if defined(HAVE_STAT) && !defined(MS_WINDOWS)

/* Get file status. Encode the path to the locale encoding. */

int
_Py_wstat(const wchar_t* path, struct stat *buf)
{
    int err;
    char *fname;
    fname = _Py_wchar2char(path);
    if (fname == NULL) {
        errno = EINVAL;
        return -1;
    }
    err = stat(fname, buf);
    PyMem_Free(fname);
    return err;
}
#endif

/* Call _wstat() on Windows, or encode the path to the filesystem encoding and
   call stat() otherwise. Only fill st_mode attribute on Windows.

   Return 0 on success, -1 on _wstat() / stat() error or (if PyErr_Occurred())
   unicode error. */

int
_Py_stat(PyObject *path, struct stat *statbuf)
{
#ifdef MS_WINDOWS
    int err;
    struct _stat wstatbuf;

    err = _wstat(PyUnicode_AS_UNICODE(path), &wstatbuf);
    if (!err)
        statbuf->st_mode = wstatbuf.st_mode;
    return err;
#else
    int ret;
    PyObject *bytes = PyUnicode_EncodeFSDefault(path);
    if (bytes == NULL)
        return -1;
    ret = stat(PyBytes_AS_STRING(bytes), statbuf);
    Py_DECREF(bytes);
    return ret;
#endif
}

/* Open a file. Use _wfopen() on Windows, encode the path to the locale
   encoding and use fopen() otherwise. */

FILE *
_Py_wfopen(const wchar_t *path, const wchar_t *mode)
{
#ifndef MS_WINDOWS
    FILE *f;
    char *cpath;
    char cmode[10];
    size_t r;
    r = wcstombs(cmode, mode, 10);
    if (r == (size_t)-1 || r >= 10) {
        errno = EINVAL;
        return NULL;
    }
    cpath = _Py_wchar2char(path);
    if (cpath == NULL)
        return NULL;
    f = fopen(cpath, cmode);
    PyMem_Free(cpath);
    return f;
#else
    return _wfopen(path, mode);
#endif
}

/* Call _wfopen() on Windows, or encode the path to the filesystem encoding and
   call fopen() otherwise.

   Return the new file object on success, or NULL if the file cannot be open or
   (if PyErr_Occurred()) on unicode error */

FILE*
_Py_fopen(PyObject *path, const char *mode)
{
#ifdef MS_WINDOWS
    wchar_t wmode[10];
    int usize;

    usize = MultiByteToWideChar(CP_ACP, 0, mode, -1, wmode, sizeof(wmode));
    if (usize == 0)
        return NULL;

    return _wfopen(PyUnicode_AS_UNICODE(path), wmode);
#else
    FILE *f;
    PyObject *bytes = PyUnicode_EncodeFSDefault(path);
    if (bytes == NULL)
        return NULL;
    f = fopen(PyBytes_AS_STRING(bytes), mode);
    Py_DECREF(bytes);
    return f;
#endif
}

#ifdef HAVE_READLINK

/* Read value of symbolic link. Encode the path to the locale encoding, decode
   the result from the locale encoding. */

int
_Py_wreadlink(const wchar_t *path, wchar_t *buf, size_t bufsiz)
{
    char *cpath;
    char cbuf[PATH_MAX];
    int res;
    size_t r1;

    cpath = _Py_wchar2char(path);
    if (cpath == NULL) {
        errno = EINVAL;
        return -1;
    }
    res = (int)readlink(cpath, cbuf, PATH_MAX);
    PyMem_Free(cpath);
    if (res == -1)
        return -1;
    if (res == PATH_MAX) {
        errno = EINVAL;
        return -1;
    }
    cbuf[res] = '\0'; /* buf will be null terminated */
    r1 = mbstowcs(buf, cbuf, bufsiz);
    if (r1 == -1) {
        errno = EINVAL;
        return -1;
    }
    return (int)r1;
}
#endif

#ifdef HAVE_REALPATH

/* Return the canonicalized absolute pathname. Encode path to the locale
   encoding, decode the result from the locale encoding. */

wchar_t*
_Py_wrealpath(const wchar_t *path,
              wchar_t *resolved_path, size_t resolved_path_size)
{
    char *cpath;
    char cresolved_path[PATH_MAX];
    char *res;
    size_t r;
    cpath = _Py_wchar2char(path);
    if (cpath == NULL) {
        errno = EINVAL;
        return NULL;
    }
    res = realpath(cpath, cresolved_path);
    PyMem_Free(cpath);
    if (res == NULL)
        return NULL;
    r = mbstowcs(resolved_path, cresolved_path, resolved_path_size);
    if (r == (size_t)-1 || r >= PATH_MAX) {
        errno = EINVAL;
        return NULL;
    }
    return resolved_path;
}
#endif

/* Get the current directory. Decode the path from the locale encoding. */

wchar_t*
_Py_wgetcwd(wchar_t *buf, size_t size)
{
#ifdef MS_WINDOWS
    return _wgetcwd(buf, size);
#else
    char fname[PATH_MAX];
    if (getcwd(fname, PATH_MAX) == NULL)
        return NULL;
    if (mbstowcs(buf, fname, size) >= size) {
        errno = ERANGE;
        return NULL;
    }
    return buf;
#endif
}

#endif