X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=src%2Ftext%2Funicode.c;h=bed5f1f96d7cbd91693690e78297357c1bea7a62;hb=9ce1a13fb8fa77a38e0dfdfe3387829e1df3f085;hp=e1c32612dee4c5952561a521b42cc8c2c608cd96;hpb=45915c74f0baaaed6527dc888f3e4263fb50e7d7;p=vlc diff --git a/src/text/unicode.c b/src/text/unicode.c index e1c32612de..bed5f1f96d 100644 --- a/src/text/unicode.c +++ b/src/text/unicode.c @@ -2,8 +2,7 @@ * unicode.c: Unicode <-> locale functions ***************************************************************************** * Copyright (C) 2005-2006 the VideoLAN team - * Copyright © 2005-2006 Rémi Denis-Courmont - * $Id$ + * Copyright © 2005-2010 Rémi Denis-Courmont * * Authors: Rémi Denis-Courmont * @@ -25,7 +24,11 @@ /***************************************************************************** * Preamble *****************************************************************************/ -#include +#ifdef HAVE_CONFIG_H +# include "config.h" +#endif + +#include #include #include @@ -33,95 +36,63 @@ #include #include #include -#include #include -#ifdef HAVE_DIRENT_H -# include -#endif #ifdef UNDER_CE # include #endif -#ifdef HAVE_SYS_STAT_H -# include -#endif -#ifdef HAVE_FCNTL_H -# include -#endif -#ifdef WIN32 -# include -#else -# include -#endif - -#ifndef HAVE_LSTAT -# define lstat( a, b ) stat(a, b) -#endif - -#ifdef __APPLE__ -/* Define this if the OS always use UTF-8 internally */ -# define ASSUME_UTF8 1 -#endif +#include #if defined (ASSUME_UTF8) /* Cool */ + #elif defined (WIN32) || defined (UNDER_CE) # define USE_MB2MB 1 -#elif defined (HAVE_ICONV) -# define USE_ICONV 1 -#else -# error No UTF8 charset conversion implemented on this platform! -#endif - -#if defined (USE_ICONV) -static char charset[sizeof ("CSISO11SWEDISHFORNAMES//translit")] = ""; +# include -static void find_charset_once (void) +static char *locale_dup (const char *string, bool from) { - char *psz_charset; - if (vlc_current_charset (&psz_charset) - || (psz_charset == NULL) - || ((size_t)snprintf (charset, sizeof (charset), "%s//translit", - psz_charset) >= sizeof (charset))) - strcpy (charset, "UTF-8"); - - free (psz_charset); -} + char *out; + int len; -static int find_charset (void) -{ - static pthread_once_t once = PTHREAD_ONCE_INIT; - pthread_once (&once, find_charset_once); - return !strcmp (charset, "UTF-8"); + len = 1 + MultiByteToWideChar (from ? CP_ACP : CP_UTF8, + 0, string, -1, NULL, 0); + wchar_t *wide = malloc (len * sizeof (wchar_t)); + if (wide == NULL) + return NULL; + + MultiByteToWideChar (from ? CP_ACP : CP_UTF8, 0, string, -1, wide, len); + len = 1 + WideCharToMultiByte (from ? CP_UTF8 : CP_ACP, 0, wide, -1, + NULL, 0, NULL, NULL); + out = malloc (len); + if (out != NULL) + WideCharToMultiByte (from ? CP_UTF8 : CP_ACP, 0, wide, -1, out, len, + NULL, NULL); + free (wide); + return out; } -#endif +#elif defined (HAVE_ICONV) +# define USE_ICONV 1 -static char *locale_fast (const char *string, vlc_bool_t from) +static char *locale_dup (const char *string, bool from) { -#if defined (USE_ICONV) - if (find_charset ()) - return (char *)string; - - vlc_iconv_t hd = vlc_iconv_open (from ? "UTF-8" : charset, - from ? charset : "UTF-8"); + vlc_iconv_t hd = vlc_iconv_open (from ? "UTF-8" : "", + from ? "" : "UTF-8"); if (hd == (vlc_iconv_t)(-1)) - return strdup (string); /* Uho! */ + return NULL; /* Uho! */ const char *iptr = string; size_t inb = strlen (string); size_t outb = inb * 6 + 1; char output[outb], *optr = output; - if (string == NULL) - return NULL; - while (vlc_iconv (hd, &iptr, &inb, &optr, &outb) == (size_t)(-1)) { *optr++ = '?'; outb--; iptr++; inb--; - vlc_iconv (hd, NULL, NULL, NULL, NULL); + vlc_iconv (hd, NULL, NULL, NULL, NULL); /* reset */ } *optr = '\0'; vlc_iconv_close (hd); @@ -131,60 +102,31 @@ static char *locale_fast (const char *string, vlc_bool_t from) assert (*optr == '\0'); assert (strlen (output) == (size_t)(optr - output)); return strdup (output); -#elif defined (USE_MB2MB) - char *out; - int len; - - if (string == NULL) - return NULL; - - len = 1 + MultiByteToWideChar (from ? CP_ACP : CP_UTF8, - 0, string, -1, NULL, 0); - wchar_t wide[len]; - - MultiByteToWideChar (from ? CP_ACP : CP_UTF8, 0, string, -1, wide, len); - len = 1 + WideCharToMultiByte (from ? CP_UTF8 : CP_ACP, 0, wide, -1, NULL, 0, NULL, NULL); - out = malloc (len); - if (out == NULL) - return NULL; - - WideCharToMultiByte (from ? CP_UTF8 : CP_ACP, 0, wide, -1, out, len, NULL, NULL); - return out; -#else - return (char *)string; -#endif } - -static inline char *locale_dup (const char *string, vlc_bool_t from) -{ -#if defined (USE_ICONV) - if (find_charset ()) - return strdup (string); - return locale_fast (string, from); -#elif defined (USE_MB2MB) - return locale_fast (string, from); #else - return strdup (string); +# error No UTF8 charset conversion implemented on this platform! #endif -} +/** + * Releases (if needed) a localized or uniformized string. + * @param str non-NULL return value from FromLocale() or ToLocale(). + */ void LocaleFree (const char *str) { -#if defined (USE_ICONV) - if (!find_charset ()) - free ((char *)str); -#elif defined (USE_MB2MB) +#ifdef ASSUME_UTF8 + (void) str; +#else free ((char *)str); #endif } /** - * FromLocale: converts a locale string to UTF-8 + * Converts a string from the system locale character encoding to UTF-8. * - * @param locale nul-terminated string to be converted + * @param locale nul-terminated string to convert * * @return a nul-terminated UTF-8 string, or NULL in case of error. * To avoid memory leak, you have to pass the result to LocaleFree() @@ -192,17 +134,34 @@ void LocaleFree (const char *str) */ char *FromLocale (const char *locale) { - return locale_fast (locale, VLC_TRUE); +#ifdef ASSUME_UTF8 + return (char *)locale; +#else + return locale ? locale_dup (locale, true) : NULL; +#endif } +/** + * converts a string from the system locale character encoding to utf-8, + * the result is always allocated on the heap. + * + * @param locale nul-terminated string to convert + * + * @return a nul-terminated utf-8 string, or null in case of error. + * The result must be freed using free() - as with the strdup() function. + */ char *FromLocaleDup (const char *locale) { - return locale_dup (locale, VLC_TRUE); +#ifdef ASSUME_UTF8 + return strdup (locale); +#else + return locale_dup (locale, true); +#endif } /** - * ToLocale: converts a UTF-8 string to local system encoding. + * ToLocale: converts an UTF-8 string to local system encoding. * * @param utf8 nul-terminated string to be converted * @@ -212,381 +171,96 @@ char *FromLocaleDup (const char *locale) */ char *ToLocale (const char *utf8) { - return locale_fast (utf8, VLC_FALSE); -} - - -static char *ToLocaleDup (const char *utf8) -{ - return locale_dup (utf8, VLC_FALSE); -} - - -/** - * utf8_open: open() wrapper for UTF-8 filenames - */ -int utf8_open (const char *filename, int flags, mode_t mode) -{ -#if defined (WIN32) || defined (UNDER_CE) - if (GetVersion() < 0x80000000) - { - /* for Windows NT and above */ - wchar_t wpath[MAX_PATH + 1]; - - if (!MultiByteToWideChar (CP_UTF8, 0, filename, -1, wpath, MAX_PATH)) - { - errno = ENOENT; - return -1; - } - wpath[MAX_PATH] = L'\0'; - - /* - * open() cannot open files with non-“ANSI” characters on Windows. - * We use _wopen() instead. Same thing for mkdir() and stat(). - */ - return _wopen (wpath, flags, mode); - } -#endif - const char *local_name = ToLocale (filename); - - if (local_name == NULL) - { - errno = ENOENT; - return -1; - } - - int fd = open (local_name, flags, mode); - LocaleFree (local_name); - return fd; -} - -/** - * utf8_fopen: fopen() wrapper for UTF-8 filenames - */ -FILE *utf8_fopen (const char *filename, const char *mode) -{ - int rwflags = 0, oflags = 0; - vlc_bool_t append = VLC_FALSE; - - for (const char *ptr = mode; *ptr; ptr++) - { - switch (*ptr) - { - case 'r': - rwflags = O_RDONLY; - break; - - case 'a': - rwflags = O_WRONLY; - oflags |= O_CREAT; - append = VLC_TRUE; - break; - - case 'w': - rwflags = O_WRONLY; - oflags |= O_CREAT | O_TRUNC; - break; - - case '+': - rwflags = O_RDWR; - break; - -#ifdef O_TEXT - case 't': - oflags |= O_TEXT; - break; -#endif - } - } - - int fd = utf8_open (filename, rwflags | oflags, 0666); - if (fd == -1) - return NULL; - - if (append && (lseek (fd, 0, SEEK_END) == -1)) - { - close (fd); - return NULL; - } - - FILE *stream = fdopen (fd, mode); - if (stream == NULL) - close (fd); - - return stream; -} - -/** - * utf8_mkdir: Calls mkdir() after conversion of file name to OS locale - * - * @param dirname a UTF-8 string with the name of the directory that you - * want to create. - * @return A 0 return value indicates success. A -1 return value indicates an - * error, and an error code is stored in errno - */ -int utf8_mkdir( const char *dirname ) -{ -#if defined (UNDER_CE) || defined (WIN32) - wchar_t wname[MAX_PATH + 1]; - char mod[MAX_PATH + 1]; - int i; - - /* Convert '/' into '\' */ - for( i = 0; *dirname; i++ ) - { - if( i == MAX_PATH ) - return -1; /* overflow */ - - if( *dirname == '/' ) - mod[i] = '\\'; - else - mod[i] = *dirname; - dirname++; - - } - mod[i] = 0; - - if( MultiByteToWideChar( CP_UTF8, 0, mod, -1, wname, MAX_PATH ) == 0 ) - { - errno = ENOENT; - return -1; - } - wname[MAX_PATH] = L'\0'; - - if( CreateDirectoryW( wname, NULL ) == 0 ) - { - if( GetLastError( ) == ERROR_ALREADY_EXISTS ) - errno = EEXIST; - else - errno = ENOENT; - return -1; - } - return 0; +#ifdef ASSUME_UTF8 + return (char *)utf8; #else - char *locname = ToLocale( dirname ); - int res; - - if( locname == NULL ) - { - errno = ENOENT; - return -1; - } - res = mkdir( locname, 0755 ); - - LocaleFree( locname ); - return res; + return utf8 ? locale_dup (utf8, false) : NULL; #endif } + /** - * utf8_opendir: wrapper that converts dirname to the locale in use by the OS + * converts a string from UTF-8 to the system locale character encoding, + * the result is always allocated on the heap. * - * @param dirname UTF-8 representation of the directory name + * @param utf8 nul-terminated string to convert * - * @return a pointer to the DIR struct. Release with closedir(). + * @return a nul-terminated string, or null in case of error. + * The result must be freed using free() - as with the strdup() function. */ -DIR *utf8_opendir( const char *dirname ) +char *ToLocaleDup (const char *utf8) { -#ifdef WIN32 - wchar_t wname[MAX_PATH + 1]; - - if (MultiByteToWideChar (CP_UTF8, 0, dirname, -1, wname, MAX_PATH)) - { - wname[MAX_PATH] = L'\0'; - return (DIR *)vlc_wopendir (wname); - } +#ifdef ASSUME_UTF8 + return strdup (utf8); #else - const char *local_name = ToLocale( dirname ); - - if( local_name != NULL ) - { - DIR *dir = opendir( local_name ); - LocaleFree( local_name ); - return dir; - } + return locale_dup (utf8, false); #endif - - errno = ENOENT; - return NULL; } /** - * utf8_readdir: a readdir wrapper that returns the name of the next entry - * in the directory as a UTF-8 string. - * - * @param dir The directory that is being read - * - * @return a UTF-8 string of the directory entry. Use free() to free this memory. + * Formats an UTF-8 string as vfprintf(), then print it, with + * appropriate conversion to local encoding. */ -char *utf8_readdir( DIR *dir ) +int utf8_vfprintf( FILE *stream, const char *fmt, va_list ap ) { -#ifdef WIN32 - struct _wdirent *ent = vlc_wreaddir (dir); - if (ent == NULL) - return NULL; - - return FromWide (ent->d_name); +#ifdef ASSUME_UTF8 + return vfprintf (stream, fmt, ap); #else - struct dirent *ent; - - ent = readdir( (DIR *)dir ); - if( ent == NULL ) - return NULL; - - return vlc_fix_readdir( ent->d_name ); -#endif -} - -static int dummy_select( const char *str ) -{ - (void)str; - return 1; -} - -int utf8_loaddir( DIR *dir, char ***namelist, - int (*select)( const char * ), - int (*compar)( const char **, const char ** ) ) -{ - if( select == NULL ) - select = dummy_select; + char *str; + int res; - if( dir == NULL ) - return -1; - else +# if defined( WIN32 ) && !defined( UNDER_CE ) + /* Writing to the console is a lot of fun on Microsoft Windows. + * If you use the standard I/O functions, you must use the OEM code page, + * which is different from the usual ANSI code page. Or maybe not, if the + * user called "chcp". Anyway, we prefer Unicode. */ + int fd = _fileno (stream); + if (likely(fd != -1) && _isatty (fd)) { - char **tab = NULL; - char *entry; - unsigned num = 0; - - rewinddir( dir ); + res = vasprintf (&str, fmt, ap); + if (unlikely(res == -1)) + return -1; - while( ( entry = utf8_readdir( dir ) ) != NULL ) + size_t wlen = 2 * (res + 1); + wchar_t *wide = malloc (wlen); + if (likely(wide != NULL)) { - char **newtab; - - if( !select( entry ) ) + wlen = MultiByteToWideChar (CP_UTF8, 0, str, res + 1, wide, wlen); + if (wlen > 0) { - free( entry ); - continue; - } + HANDLE h = (HANDLE)(intptr_t)_get_osfhandle (fd); + DWORD out; - newtab = realloc( tab, sizeof( char * ) * (num + 1) ); - if( newtab == NULL ) - { - free( entry ); - goto error; + WriteConsoleW (h, wide, wlen - 1, &out, NULL); } - tab = newtab; - tab[num++] = entry; + else + res = -1; + free (wide); } - - if( compar != NULL ) - qsort( tab, num, sizeof( tab[0] ), - (int (*)( const void *, const void *))compar ); - - *namelist = tab; - return num; - - error:{ - unsigned i; - - for( i = 0; i < num; i++ ) - free( tab[i] ); - if( tab != NULL ) - free( tab ); - } - } - return -1; -} - -int utf8_scandir( const char *dirname, char ***namelist, - int (*select)( const char * ), - int (*compar)( const char **, const char ** ) ) -{ - DIR *dir = utf8_opendir (dirname); - int val = -1; - - if (dir != NULL) - { - val = utf8_loaddir (dir, namelist, select, compar); - closedir (dir); - } - return val; -} - -static int utf8_statEx( const char *filename, struct stat *buf, - vlc_bool_t deref ) -{ -#if defined (WIN32) || defined (UNDER_CE) - /* retrieve Windows OS version */ - if( GetVersion() < 0x80000000 ) - { - /* for Windows NT and above */ - wchar_t wpath[MAX_PATH + 1]; - - if( !MultiByteToWideChar( CP_UTF8, 0, filename, -1, wpath, MAX_PATH ) ) - { - errno = ENOENT; - return -1; - } - wpath[MAX_PATH] = L'\0'; - - return _wstati64( wpath, buf ); - } -#endif -#ifdef HAVE_SYS_STAT_H - const char *local_name = ToLocale( filename ); - - if( local_name != NULL ) - { - int res = deref ? stat( local_name, buf ) - : lstat( local_name, buf ); - LocaleFree( local_name ); + else + res = -1; + free (str); return res; } - errno = ENOENT; -#endif - return -1; -} - - -int utf8_stat( const char *filename, struct stat *buf) -{ - return utf8_statEx( filename, buf, VLC_TRUE ); -} +# endif -int utf8_lstat( const char *filename, struct stat *buf) -{ - return utf8_statEx( filename, buf, VLC_FALSE ); -} - -/** - * utf8_*printf: *printf with conversion from UTF-8 to local encoding - */ -static int utf8_vasprintf( char **str, const char *fmt, va_list ap ) -{ - char *utf8; - int res = vasprintf( &utf8, fmt, ap ); - if( res == -1 ) + res = vasprintf (&str, fmt, ap); + if (unlikely(res == -1)) return -1; - *str = ToLocaleDup( utf8 ); - free( utf8 ); - return res; -} + char *ansi = ToLocaleDup (str); + free (str); -int utf8_vfprintf( FILE *stream, const char *fmt, va_list ap ) -{ - char *str; - int res = utf8_vasprintf( &str, fmt, ap ); - if( res == -1 ) - return -1; - - fputs( str, stream ); - free( str ); + fputs (ansi, stream); + free (ansi); return res; +#endif } +/** + * Formats an UTF-8 string as fprintf(), then print it, with + * appropriate conversion to local encoding. + */ int utf8_fprintf( FILE *stream, const char *fmt, ... ) { va_list ap; @@ -599,72 +273,76 @@ int utf8_fprintf( FILE *stream, const char *fmt, ... ) } -static char *CheckUTF8( char *str, char rep ) +/** + * Converts the first character from a UTF-8 sequence into a code point. + * + * @param str an UTF-8 bytes sequence + * @return 0 if str points to an empty string, i.e. the first character is NUL; + * number of bytes that the first character occupies (from 1 to 4) otherwise; + * -1 if the byte sequence was not a valid UTF-8 sequence. + */ +static size_t vlc_towc (const char *str, uint32_t *restrict pwc) { uint8_t *ptr = (uint8_t *)str; assert (str != NULL); - for (;;) - { - uint8_t c = ptr[0]; - int charlen = -1; - - if (c == '\0') - break; + uint8_t c = ptr[0]; - for (int i = 0; i < 7; i++) - if ((c >> (7 - i)) == ((0xff >> (7 - i)) ^ 1)) - { - charlen = i; - break; - } - - switch (charlen) - { - case 0: // 7-bit ASCII character -> OK - ptr++; - continue; - - case -1: // 1111111x -> error - case 1: // continuation byte -> error - goto error; - } + if (unlikely(c == '\0')) + { + *pwc = 0; + return 0; + } - assert (charlen >= 2); + if (unlikely(c > 0xF4)) + return -1; - uint32_t cp = c & ~((0xff >> (7 - charlen)) << (7 - charlen)); - for (int i = 1; i < charlen; i++) - { - assert (cp < (1 << 26)); - c = ptr[i]; + int charlen = clz8 (c ^ 0xFF); + switch (charlen) + { + case 0: // 7-bit ASCII character -> OK + *pwc = c; + return 1; - if ((c == '\0') // unexpected end of string - || ((c >> 6) != 2)) // not a continuation byte - goto error; + case 1: // continuation byte -> error + return -1; + } - cp = (cp << 6) | (ptr[i] & 0x3f); - } + assert (charlen >= 2 && charlen <= 4); - if (cp < 128) // overlong (special case for ASCII) - goto error; - if (cp < (1u << (5 * charlen - 3))) // overlong - goto error; + uint32_t cp = c & ~((0xff >> (7 - charlen)) << (7 - charlen)); + for (int i = 1; i < charlen; i++) + { + assert (cp < (1 << 26)); + c = ptr[i]; - ptr += charlen; - continue; + if (unlikely((c >> 6) != 2)) // not a continuation byte + return -1; - error: - if (rep == 0) - return NULL; - *ptr++ = rep; - str = NULL; + cp = (cp << 6) | (ptr[i] & 0x3f); } - return str; + switch (charlen) + { + case 4: + if (unlikely(cp > 0x10FFFF)) // beyond Unicode + return -1; + case 3: + if (unlikely(cp >= 0xD800 && cp < 0xC000)) // UTF-16 surrogate + return -1; + case 2: + if (unlikely(cp < 128)) // ASCII overlong + return -1; + if (unlikely(cp < (1u << (5 * charlen - 3)))) // overlong + return -1; + } + *pwc = cp; + return charlen; } + /** - * EnsureUTF8: replaces invalid/overlong UTF-8 sequences with question marks + * Replaces invalid/overlong UTF-8 sequences with question marks. * Note that it is not possible to convert from Latin-1 to UTF-8 on the fly, * so we don't try that, even though it would be less disruptive. * @@ -672,12 +350,24 @@ static char *CheckUTF8( char *str, char rep ) */ char *EnsureUTF8( char *str ) { - return CheckUTF8( str, '?' ); + char *ret = str; + size_t n; + uint32_t cp; + + while ((n = vlc_towc (str, &cp)) != 0) + if (likely(n != (size_t)-1)) + str += n; + else + { + *str++ = '?'; + ret = NULL; + } + return ret; } /** - * IsUTF8: checks whether a string is a valid UTF-8 byte sequence. + * Checks whether a string is a valid UTF-8 byte sequence. * * @param str nul-terminated string to be checked * @@ -685,5 +375,50 @@ char *EnsureUTF8( char *str ) */ const char *IsUTF8( const char *str ) { - return CheckUTF8( (char *)str, 0 ); + size_t n; + uint32_t cp; + + while ((n = vlc_towc (str, &cp)) != 0) + if (likely(n != (size_t)-1)) + str += n; + else + return NULL; + return str; } + +/** + * Converts a string from the given character encoding to utf-8. + * + * @return a nul-terminated utf-8 string, or null in case of error. + * The result must be freed using free(). + */ +char *FromCharset(const char *charset, const void *data, size_t data_size) +{ + vlc_iconv_t handle = vlc_iconv_open ("UTF-8", charset); + if (handle == (vlc_iconv_t)(-1)) + return NULL; + + char *out = NULL; + for(unsigned mul = 4; mul < 8; mul++ ) + { + size_t in_size = data_size; + const char *in = data; + size_t out_max = mul * data_size; + char *tmp = out = malloc (1 + out_max); + if (!out) + break; + + if (vlc_iconv (handle, &in, &in_size, &tmp, &out_max) != (size_t)(-1)) { + *tmp = '\0'; + break; + } + free(out); + out = NULL; + + if (errno != E2BIG) + break; + } + vlc_iconv_close(handle); + return out; +} +