* unicode.c: Unicode <-> locale functions
*****************************************************************************
* Copyright (C) 2005-2006 the VideoLAN team
- * Copyright © 2005-2008 Rémi Denis-Courmont
+ * Copyright © 2005-2010 Rémi Denis-Courmont
*
* Authors: Rémi Denis-Courmont <rem # videolan.org>
*
#include <sys/types.h>
#ifdef UNDER_CE
# include <tchar.h>
+#elif defined(WIN32)
+# include <io.h>
#endif
#include <errno.h>
-
-#if defined (__APPLE__) || defined (HAVE_MAEMO)
-/* Define this if the OS always use UTF-8 internally */
-# define ASSUME_UTF8 1
-#endif
-
-#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)
-# include <langinfo.h>
-static char charset[sizeof ("CSISO11SWEDISHFORNAMES")] = "";
-
-static void find_charset_once (void)
-{
- strlcpy (charset, nl_langinfo (CODESET), sizeof (charset));
- if (!strcasecmp (charset, "ASCII")
- || !strcasecmp (charset, "ANSI_X3.4-1968"))
- strcpy (charset, "UTF-8"); /* superset... */
-}
-
-static int find_charset (void)
-{
- static pthread_once_t once = PTHREAD_ONCE_INIT;
- pthread_once (&once, find_charset_once);
- return !strcasecmp (charset, "UTF-8");
-}
-#endif
-
-
-static char *locale_fast (const char *string, bool from)
-{
- if( string == NULL )
- return NULL;
-
-#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");
- if (hd == (vlc_iconv_t)(-1))
- return NULL; /* Uho! */
-
- const char *iptr = string;
- size_t inb = strlen (string);
- size_t outb = inb * 6 + 1;
- char output[outb], *optr = output;
-
- while (vlc_iconv (hd, &iptr, &inb, &optr, &outb) == (size_t)(-1))
- {
- *optr++ = '?';
- outb--;
- iptr++;
- inb--;
- vlc_iconv (hd, NULL, NULL, NULL, NULL); /* reset */
- }
- *optr = '\0';
- vlc_iconv_close (hd);
-
- assert (inb == 0);
- assert (*iptr == '\0');
- assert (*optr == '\0');
- assert (strlen (output) == (size_t)(optr - output));
- return strdup (output);
-#elif defined (USE_MB2MB)
- char *out;
- int len;
-
- 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;
-#else
- (void)from;
- return (char *)string;
-#endif
-}
-
-
-static inline char *locale_dup (const char *string, bool from)
-{
- assert( string );
-
-#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
- (void)from;
- return strdup (string);
-#endif
-}
+#include <wctype.h>
/**
* Releases (if needed) a localized or uniformized string.
*/
void LocaleFree (const char *str)
{
-#if defined (USE_ICONV)
- if (!find_charset ())
- free ((char *)str);
-#elif defined (USE_MB2MB)
- free ((char *)str);
+#ifdef ASSUME_UTF8
+ (void) str;
#else
- (void)str;
+ free ((char *)str);
#endif
}
*/
char *FromLocale (const char *locale)
{
- return locale_fast (locale, true);
+#ifdef ASSUME_UTF8
+ return (char *)locale;
+#else
+ return locale ? FromCharset ("", locale, strlen(locale)) : NULL;
+#endif
}
/**
*/
char *FromLocaleDup (const char *locale)
{
- return locale_dup (locale, true);
+#ifdef ASSUME_UTF8
+ return strdup (locale);
+#else
+ return FromCharset ("", locale, strlen(locale));
+#endif
}
*/
char *ToLocale (const char *utf8)
{
- return locale_fast (utf8, false);
+#ifdef ASSUME_UTF8
+ return (char *)utf8;
+#else
+ size_t outsize;
+ return utf8 ? ToCharset ("", utf8, &outsize) : NULL;
+#endif
}
*/
char *ToLocaleDup (const char *utf8)
{
- return locale_dup (utf8, false);
-}
-
-/**
- * Formats an UTF-8 string as vasprintf(), then print it to stdout, with
- * appropriate conversion 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 )
- return -1;
-
- *str = ToLocaleDup( utf8 );
- free( utf8 );
- return res;
+#ifdef ASSUME_UTF8
+ return strdup (utf8);
+#else
+ size_t outsize;
+ return ToCharset ("", utf8, &outsize);
+#endif
}
/**
*/
int utf8_vfprintf( FILE *stream, const char *fmt, va_list ap )
{
+#ifdef ASSUME_UTF8
+ return vfprintf (stream, fmt, ap);
+#else
char *str;
- int res = utf8_vasprintf( &str, fmt, ap );
- if( res == -1 )
+ int res;
+
+# 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))
+ {
+ res = vasprintf (&str, fmt, ap);
+ if (unlikely(res == -1))
+ return -1;
+
+ size_t wlen = 2 * (res + 1);
+ wchar_t *wide = malloc (wlen);
+ if (likely(wide != NULL))
+ {
+ wlen = MultiByteToWideChar (CP_UTF8, 0, str, res + 1, wide, wlen);
+ if (wlen > 0)
+ {
+ HANDLE h = (HANDLE)(intptr_t)_get_osfhandle (fd);
+ DWORD out;
+
+ WriteConsoleW (h, wide, wlen - 1, &out, NULL);
+ }
+ else
+ res = -1;
+ free (wide);
+ }
+ else
+ res = -1;
+ free (str);
+ return res;
+ }
+# endif
+
+ res = vasprintf (&str, fmt, ap);
+ if (unlikely(res == -1))
return -1;
- fputs( str, stream );
- free( str );
+ char *ansi = ToLocaleDup (str);
+ free (str);
+
+ if (ansi == NULL)
+ return -1;
+ fputs (ansi, stream);
+ free (ansi);
return res;
+#endif
}
/**
}
-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];
+
+ if (unlikely(c == '\0'))
{
- uint8_t c = ptr[0];
- int charlen = -1;
+ *pwc = 0;
+ return 0;
+ }
- if (c == '\0')
- break;
+ if (unlikely(c > 0xF4))
+ return -1;
- for (int i = 0; i < 7; i++)
- if ((c >> (7 - i)) == ((0xff >> (7 - i)) ^ 1))
- {
- charlen = i;
- break;
- }
+ int charlen = clz8 (c ^ 0xFF);
+ switch (charlen)
+ {
+ case 0: // 7-bit ASCII character -> OK
+ *pwc = c;
+ return 1;
- switch (charlen)
- {
- case 0: // 7-bit ASCII character -> OK
- ptr++;
- continue;
+ case 1: // continuation byte -> error
+ return -1;
+ }
- case -1: // 1111111x -> error
- case 1: // continuation byte -> error
- goto error;
- }
+ assert (charlen >= 2 && charlen <= 4);
- assert (charlen >= 2);
+ uint32_t cp = c & ~((0xff >> (7 - charlen)) << (7 - charlen));
+ for (int i = 1; i < charlen; i++)
+ {
+ assert (cp < (1 << 26));
+ c = ptr[i];
- uint32_t cp = c & ~((0xff >> (7 - charlen)) << (7 - charlen));
- for (int i = 1; i < charlen; i++)
- {
- assert (cp < (1 << 26));
- c = ptr[i];
+ if (unlikely((c >> 6) != 2)) // not a continuation byte
+ return -1;
- if ((c == '\0') // unexpected end of string
- || ((c >> 6) != 2)) // not a continuation byte
- goto error;
+ cp = (cp << 6) | (ptr[i] & 0x3f);
+ }
- cp = (cp << 6) | (ptr[i] & 0x3f);
- }
+ 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;
+}
+
+/**
+ * Look for an UTF-8 string within another one in a case-insensitive fashion.
+ * Beware that this is quite slow. Contrary to strcasestr(), this function
+ * works regardless of the system character encoding, and handles multibyte
+ * code points correctly.
+
+ * @param haystack string to look into
+ * @param needle string to look for
+ * @return a pointer to the first occurence of the needle within the haystack,
+ * or NULL if no occurence were found.
+ */
+char *vlc_strcasestr (const char *haystack, const char *needle)
+{
+ ssize_t s;
- if (cp < 128) // overlong (special case for ASCII)
- goto error;
- if (cp < (1u << (5 * charlen - 3))) // overlong
- goto error;
+ do
+ {
+ const char *h = haystack, *n = needle;
- ptr += charlen;
- continue;
+ for (;;)
+ {
+ uint32_t cph, cpn;
- error:
- if (rep == 0)
- return NULL;
- *ptr++ = rep;
- str = NULL;
+ s = vlc_towc (n, &cpn);
+ if (s == 0)
+ return (char *)haystack;
+ if (unlikely(s < 0))
+ return NULL;
+ n += s;
+
+ s = vlc_towc (h, &cph);
+ if (s <= 0 || towlower (cph) != towlower (cpn))
+ break;
+ h += s;
+ }
+
+ s = vlc_towc (haystack, &(uint32_t) { 0 });
+ haystack += s;
}
+ while (s != 0);
- return str;
+ return NULL;
}
/**
*/
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;
}
*/
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;
}
/**
* @return a nul-terminated utf-8 string, or null in case of error.
* The result must be freed using free().
*/
-static char *FromCharset(const char *charset, const void *data, size_t data_size)
+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 out;
}
+/**
+ * Converts a nul-terminated UTF-8 string to a given character encoding.
+ * @param charset iconv name of the character set
+ * @param in nul-terminated UTF-8 string
+ * @param outsize pointer to hold the byte size of result
+ *
+ * @return A pointer to the result, which must be released using free().
+ * The UTF-8 nul terminator is included in the conversion if the target
+ * character encoding supports it. However it is not included in the returned
+ * byte size.
+ * In case of error, NULL is returned and the byte size is undefined.
+ */
+void *ToCharset(const char *charset, const char *in, size_t *outsize)
+{
+ vlc_iconv_t hd = vlc_iconv_open (charset, "UTF-8");
+ if (hd == (vlc_iconv_t)(-1))
+ return NULL;
+
+ const size_t inlen = strlen (in);
+ void *res;
+
+ for (unsigned mul = 4; mul < 16; mul++)
+ {
+ size_t outlen = mul * (inlen + 1);
+ res = malloc (outlen);
+ if (unlikely(res == NULL))
+ break;
+
+ const char *inp = in;
+ char *outp = res;
+ size_t inb = inlen;
+ size_t outb = outlen;
+
+ if (vlc_iconv (hd, &inp, &inb, &outp, &outb) != (size_t)(-1))
+ {
+ *outsize = outlen - outb;
+ inb = 1; /* append nul terminator if possible */
+ if (vlc_iconv (hd, &inp, &inb, &outp, &outb) != (size_t)(-1))
+ break;
+ if (errno == EILSEQ) /* cannot translate nul terminator!? */
+ break;
+ }
+
+ free (res);
+ res = NULL;
+ if (errno != E2BIG) /* conversion failure */
+ break;
+ }
+ vlc_iconv_close (hd);
+ return res;
+}
+
projects / vlc / blobdiff