1 /* GLIB - Library of useful routines for C programming
3 * gconvert.c: Convert between character sets using iconv
4 * Copyright Red Hat Inc., 2000
5 * Authors: Havoc Pennington <hp@redhat.com>, Owen Taylor <otaylor@redhat.com
7 * This library is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2 of the License, or (at your option) any later version.
12 * This library is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with this library; if not, write to the
19 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
20 * Boston, MA 02111-1307, USA.
32 #include "gprintfint.h"
33 #include "gthreadinit.h"
35 #ifdef G_PLATFORM_WIN32
43 #if defined(USE_LIBICONV_GNU) && !defined (_LIBICONV_H)
44 #error GNU libiconv in use but included iconv.h not from libiconv
46 #if !defined(USE_LIBICONV_GNU) && defined (_LIBICONV_H)
47 #error GNU libiconv not in use but included iconv.h is from libiconv
53 g_convert_error_quark (void)
57 quark = g_quark_from_static_string ("g_convert_error");
63 try_conversion (const char *to_codeset,
64 const char *from_codeset,
67 *cd = iconv_open (to_codeset, from_codeset);
69 if (*cd == (iconv_t)-1 && errno == EINVAL)
76 try_to_aliases (const char **to_aliases,
77 const char *from_codeset,
82 const char **p = to_aliases;
85 if (try_conversion (*p, from_codeset, cd))
95 extern const char **_g_charset_get_aliases (const char *canonical_name) G_GNUC_INTERNAL;
99 * @to_codeset: destination codeset
100 * @from_codeset: source codeset
102 * Same as the standard UNIX routine iconv_open(), but
103 * may be implemented via libiconv on UNIX flavors that lack
104 * a native implementation.
106 * GLib provides g_convert() and g_locale_to_utf8() which are likely
107 * more convenient than the raw iconv wrappers.
109 * Return value: a "conversion descriptor", or (GIConv)-1 if
110 * opening the converter failed.
113 g_iconv_open (const gchar *to_codeset,
114 const gchar *from_codeset)
118 if (!try_conversion (to_codeset, from_codeset, &cd))
120 const char **to_aliases = _g_charset_get_aliases (to_codeset);
121 const char **from_aliases = _g_charset_get_aliases (from_codeset);
125 const char **p = from_aliases;
128 if (try_conversion (to_codeset, *p, &cd))
131 if (try_to_aliases (to_aliases, *p, &cd))
138 if (try_to_aliases (to_aliases, from_codeset, &cd))
143 return (cd == (iconv_t)-1) ? (GIConv)-1 : (GIConv)cd;
148 * @converter: conversion descriptor from g_iconv_open()
149 * @inbuf: bytes to convert
150 * @inbytes_left: inout parameter, bytes remaining to convert in @inbuf
151 * @outbuf: converted output bytes
152 * @outbytes_left: inout parameter, bytes available to fill in @outbuf
154 * Same as the standard UNIX routine iconv(), but
155 * may be implemented via libiconv on UNIX flavors that lack
156 * a native implementation.
158 * GLib provides g_convert() and g_locale_to_utf8() which are likely
159 * more convenient than the raw iconv wrappers.
161 * Return value: count of non-reversible conversions, or -1 on error
164 g_iconv (GIConv converter,
168 gsize *outbytes_left)
170 iconv_t cd = (iconv_t)converter;
172 return iconv (cd, inbuf, inbytes_left, outbuf, outbytes_left);
177 * @converter: a conversion descriptor from g_iconv_open()
179 * Same as the standard UNIX routine iconv_close(), but
180 * may be implemented via libiconv on UNIX flavors that lack
181 * a native implementation. Should be called to clean up
182 * the conversion descriptor from g_iconv_open() when
183 * you are done converting things.
185 * GLib provides g_convert() and g_locale_to_utf8() which are likely
186 * more convenient than the raw iconv wrappers.
188 * Return value: -1 on error, 0 on success
191 g_iconv_close (GIConv converter)
193 iconv_t cd = (iconv_t)converter;
195 return iconv_close (cd);
199 #define ICONV_CACHE_SIZE (16)
201 struct _iconv_cache_bucket {
208 static GList *iconv_cache_list;
209 static GHashTable *iconv_cache;
210 static GHashTable *iconv_open_hash;
211 static guint iconv_cache_size = 0;
212 G_LOCK_DEFINE_STATIC (iconv_cache_lock);
214 /* caller *must* hold the iconv_cache_lock */
216 iconv_cache_init (void)
218 static gboolean initialized = FALSE;
223 iconv_cache_list = NULL;
224 iconv_cache = g_hash_table_new (g_str_hash, g_str_equal);
225 iconv_open_hash = g_hash_table_new (g_direct_hash, g_direct_equal);
232 * iconv_cache_bucket_new:
234 * @cd: iconv descriptor
236 * Creates a new cache bucket, inserts it into the cache and
237 * increments the cache size.
239 * Returns a pointer to the newly allocated cache bucket.
241 static struct _iconv_cache_bucket *
242 iconv_cache_bucket_new (const gchar *key, GIConv cd)
244 struct _iconv_cache_bucket *bucket;
246 bucket = g_new (struct _iconv_cache_bucket, 1);
247 bucket->key = g_strdup (key);
248 bucket->refcount = 1;
252 g_hash_table_insert (iconv_cache, bucket->key, bucket);
254 /* FIXME: if we sorted the list so items with few refcounts were
255 first, then we could expire them faster in iconv_cache_expire_unused () */
256 iconv_cache_list = g_list_prepend (iconv_cache_list, bucket);
265 * iconv_cache_bucket_expire:
266 * @node: cache bucket's node
267 * @bucket: cache bucket
269 * Expires a single cache bucket @bucket. This should only ever be
270 * called on a bucket that currently has no used iconv descriptors
273 * @node is not a required argument. If @node is not supplied, we
274 * search for it ourselves.
277 iconv_cache_bucket_expire (GList *node, struct _iconv_cache_bucket *bucket)
279 g_hash_table_remove (iconv_cache, bucket->key);
282 node = g_list_find (iconv_cache_list, bucket);
284 g_assert (node != NULL);
288 node->prev->next = node->next;
290 node->next->prev = node->prev;
294 iconv_cache_list = node->next;
296 node->next->prev = NULL;
299 g_list_free_1 (node);
301 g_free (bucket->key);
302 g_iconv_close (bucket->cd);
310 * iconv_cache_expire_unused:
312 * Expires as many unused cache buckets as it needs to in order to get
313 * the total number of buckets < ICONV_CACHE_SIZE.
316 iconv_cache_expire_unused (void)
318 struct _iconv_cache_bucket *bucket;
321 node = iconv_cache_list;
322 while (node && iconv_cache_size >= ICONV_CACHE_SIZE)
327 if (bucket->refcount == 0)
328 iconv_cache_bucket_expire (node, bucket);
335 open_converter (const gchar *to_codeset,
336 const gchar *from_codeset,
339 struct _iconv_cache_bucket *bucket;
344 key = g_alloca (strlen (from_codeset) + strlen (to_codeset) + 2);
345 _g_sprintf (key, "%s:%s", from_codeset, to_codeset);
347 G_LOCK (iconv_cache_lock);
349 /* make sure the cache has been initialized */
352 bucket = g_hash_table_lookup (iconv_cache, key);
357 cd = g_iconv_open (to_codeset, from_codeset);
358 if (cd == (GIConv) -1)
363 /* Apparently iconv on Solaris <= 7 segfaults if you pass in
364 * NULL for anything but inbuf; work around that. (NULL outbuf
365 * or NULL *outbuf is allowed by Unix98.)
367 gsize inbytes_left = 0;
368 gchar *outbuf = NULL;
369 gsize outbytes_left = 0;
374 /* reset the descriptor */
375 g_iconv (cd, NULL, &inbytes_left, &outbuf, &outbytes_left);
382 cd = g_iconv_open (to_codeset, from_codeset);
383 if (cd == (GIConv) -1)
386 iconv_cache_expire_unused ();
388 bucket = iconv_cache_bucket_new (key, cd);
391 g_hash_table_insert (iconv_open_hash, cd, bucket->key);
393 G_UNLOCK (iconv_cache_lock);
399 G_UNLOCK (iconv_cache_lock);
401 /* Something went wrong. */
405 g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_NO_CONVERSION,
406 _("Conversion from character set '%s' to '%s' is not supported"),
407 from_codeset, to_codeset);
409 g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_FAILED,
410 _("Could not open converter from '%s' to '%s'"),
411 from_codeset, to_codeset);
418 close_converter (GIConv converter)
420 struct _iconv_cache_bucket *bucket;
426 if (cd == (GIConv) -1)
429 G_LOCK (iconv_cache_lock);
431 key = g_hash_table_lookup (iconv_open_hash, cd);
434 g_hash_table_remove (iconv_open_hash, cd);
436 bucket = g_hash_table_lookup (iconv_cache, key);
441 if (cd == bucket->cd)
442 bucket->used = FALSE;
446 if (!bucket->refcount && iconv_cache_size > ICONV_CACHE_SIZE)
448 /* expire this cache bucket */
449 iconv_cache_bucket_expire (NULL, bucket);
454 G_UNLOCK (iconv_cache_lock);
456 g_warning ("This iconv context wasn't opened using open_converter");
458 return g_iconv_close (converter);
461 G_UNLOCK (iconv_cache_lock);
467 * g_convert_with_iconv:
468 * @str: the string to convert
469 * @len: the length of the string, or -1 if the string is
470 * nul-terminated<footnoteref linkend="nul-unsafe"/>.
471 * @converter: conversion descriptor from g_iconv_open()
472 * @bytes_read: location to store the number of bytes in the
473 * input string that were successfully converted, or %NULL.
474 * Even if the conversion was successful, this may be
475 * less than @len if there were partial characters
476 * at the end of the input. If the error
477 * #G_CONVERT_ERROR_ILLEGAL_SEQUENCE occurs, the value
478 * stored will the byte offset after the last valid
480 * @bytes_written: the number of bytes stored in the output buffer (not
481 * including the terminating nul).
482 * @error: location to store the error occuring, or %NULL to ignore
483 * errors. Any of the errors in #GConvertError may occur.
485 * Converts a string from one character set to another.
487 * Note that despite the fact that @byes_read can return information
488 * about partial characters, this function is not generally suitable
489 * for streaming. It may not handle stateful encodings like CP1255
490 * correctly, since it doesn't keep the @converter state across
491 * multiple invocations. If you need to do streaming conversions
492 * which may involve stateful encodings, you have to use g_iconv()
495 * Note that you should use g_iconv() for streaming
496 * conversions<footnote id="streaming-state">
498 * Despite the fact that @byes_read can return information about partial
499 * characters, the <literal>g_convert_...</literal> functions
500 * are not generally suitable for streaming. If the underlying converter
501 * being used maintains internal state, then this won't be preserved
502 * across successive calls to g_convert(), g_convert_with_iconv() or
503 * g_convert_with_fallback(). (An example of this is the GNU C converter
504 * for CP1255 which does not emit a base character until it knows that
505 * the next character is not a mark that could combine with the base
510 * Return value: If the conversion was successful, a newly allocated
511 * nul-terminated string, which must be freed with
512 * g_free(). Otherwise %NULL and @error will be set.
515 g_convert_with_iconv (const gchar *str,
519 gsize *bytes_written,
525 const gchar *shift_p = NULL;
526 gsize inbytes_remaining;
527 gsize outbytes_remaining;
530 gboolean have_error = FALSE;
531 gboolean done = FALSE;
533 g_return_val_if_fail (converter != (GIConv) -1, NULL);
539 inbytes_remaining = len;
540 outbuf_size = len + 1; /* + 1 for nul in case len == 1 */
542 outbytes_remaining = outbuf_size - 1; /* -1 for nul */
543 outp = dest = g_malloc (outbuf_size);
545 while (!done && !have_error)
547 err = g_iconv (converter, (char **)&p, &inbytes_remaining, &outp, &outbytes_remaining);
549 if (err == (size_t) -1)
554 /* Incomplete text, do not report an error */
558 size_t used = outp - dest;
561 dest = g_realloc (dest, outbuf_size);
564 outbytes_remaining = outbuf_size - used - 1; /* -1 for nul */
569 g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_ILLEGAL_SEQUENCE,
570 _("Invalid byte sequence in conversion input"));
575 g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_FAILED,
576 _("Error during conversion: %s"),
586 /* call g_iconv with NULL inbuf to cleanup shift state */
589 inbytes_remaining = 0;
602 *bytes_read = p - str;
605 if ((p - str) != len)
610 g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_PARTIAL_INPUT,
611 _("Partial character sequence at end of input"));
618 *bytes_written = outp - dest; /* Doesn't include '\0' */
631 * @str: the string to convert
632 * @len: the length of the string, or -1 if the string is
633 * nul-terminated<footnote id="nul-unsafe">
635 Note that some encodings may allow nul bytes to
636 occur inside strings. In that case, using -1 for
637 the @len parameter is unsafe.
640 * @to_codeset: name of character set into which to convert @str
641 * @from_codeset: character set of @str.
642 * @bytes_read: location to store the number of bytes in the
643 * input string that were successfully converted, or %NULL.
644 * Even if the conversion was successful, this may be
645 * less than @len if there were partial characters
646 * at the end of the input. If the error
647 * #G_CONVERT_ERROR_ILLEGAL_SEQUENCE occurs, the value
648 * stored will the byte offset after the last valid
650 * @bytes_written: the number of bytes stored in the output buffer (not
651 * including the terminating nul).
652 * @error: location to store the error occuring, or %NULL to ignore
653 * errors. Any of the errors in #GConvertError may occur.
655 * Converts a string from one character set to another.
657 * Note that despite the fact that @byes_read can return information
658 * about partial characters, this function is not generally suitable
659 * for streaming. It may not handle stateful encodings like CP1255
660 * correctly, since it doesn't keep the @converter state across
661 * multiple invocations. If you need to do streaming conversions
662 * which may involve stateful encodings, you have to use g_iconv()
665 * Note that you should use g_iconv() for streaming
666 * conversions<footnoteref linkend="streaming-state"/>.
668 * Return value: If the conversion was successful, a newly allocated
669 * nul-terminated string, which must be freed with
670 * g_free(). Otherwise %NULL and @error will be set.
673 g_convert (const gchar *str,
675 const gchar *to_codeset,
676 const gchar *from_codeset,
678 gsize *bytes_written,
684 g_return_val_if_fail (str != NULL, NULL);
685 g_return_val_if_fail (to_codeset != NULL, NULL);
686 g_return_val_if_fail (from_codeset != NULL, NULL);
688 cd = open_converter (to_codeset, from_codeset, error);
690 if (cd == (GIConv) -1)
701 res = g_convert_with_iconv (str, len, cd,
702 bytes_read, bytes_written,
705 close_converter (cd);
711 * g_convert_with_fallback:
712 * @str: the string to convert
713 * @len: the length of the string, or -1 if the string is
714 * nul-terminated<footnoteref linkend="nul-unsafe"/>.
715 * @to_codeset: name of character set into which to convert @str
716 * @from_codeset: character set of @str.
717 * @fallback: UTF-8 string to use in place of character not
718 * present in the target encoding. (The string must be
719 * representable in the target encoding).
720 If %NULL, characters not in the target encoding will
721 be represented as Unicode escapes \uxxxx or \Uxxxxyyyy.
722 * @bytes_read: location to store the number of bytes in the
723 * input string that were successfully converted, or %NULL.
724 * Even if the conversion was successful, this may be
725 * less than @len if there were partial characters
726 * at the end of the input.
727 * @bytes_written: the number of bytes stored in the output buffer (not
728 * including the terminating nul).
729 * @error: location to store the error occuring, or %NULL to ignore
730 * errors. Any of the errors in #GConvertError may occur.
732 * Converts a string from one character set to another, possibly
733 * including fallback sequences for characters not representable
734 * in the output. Note that it is not guaranteed that the specification
735 * for the fallback sequences in @fallback will be honored. Some
736 * systems may do a approximate conversion from @from_codeset
737 * to @to_codeset in their iconv() functions,
738 * in which case GLib will simply return that approximate conversion.
740 * Note that you should use g_iconv() for streaming
741 * conversions<footnoteref linkend="streaming-state"/>.
743 * Return value: If the conversion was successful, a newly allocated
744 * nul-terminated string, which must be freed with
745 * g_free(). Otherwise %NULL and @error will be set.
748 g_convert_with_fallback (const gchar *str,
750 const gchar *to_codeset,
751 const gchar *from_codeset,
754 gsize *bytes_written,
760 const gchar *insert_str = NULL;
762 gsize inbytes_remaining;
763 const gchar *save_p = NULL;
764 gsize save_inbytes = 0;
765 gsize outbytes_remaining;
769 gboolean have_error = FALSE;
770 gboolean done = FALSE;
772 GError *local_error = NULL;
774 g_return_val_if_fail (str != NULL, NULL);
775 g_return_val_if_fail (to_codeset != NULL, NULL);
776 g_return_val_if_fail (from_codeset != NULL, NULL);
781 /* Try an exact conversion; we only proceed if this fails
782 * due to an illegal sequence in the input string.
784 dest = g_convert (str, len, to_codeset, from_codeset,
785 bytes_read, bytes_written, &local_error);
789 if (!g_error_matches (local_error, G_CONVERT_ERROR, G_CONVERT_ERROR_ILLEGAL_SEQUENCE))
791 g_propagate_error (error, local_error);
795 g_error_free (local_error);
799 /* No go; to proceed, we need a converter from "UTF-8" to
800 * to_codeset, and the string as UTF-8.
802 cd = open_converter (to_codeset, "UTF-8", error);
803 if (cd == (GIConv) -1)
814 utf8 = g_convert (str, len, "UTF-8", from_codeset,
815 bytes_read, &inbytes_remaining, error);
818 close_converter (cd);
824 /* Now the heart of the code. We loop through the UTF-8 string, and
825 * whenever we hit an offending character, we form fallback, convert
826 * the fallback to the target codeset, and then go back to
827 * converting the original string after finishing with the fallback.
829 * The variables save_p and save_inbytes store the input state
830 * for the original string while we are converting the fallback
834 outbuf_size = len + 1; /* + 1 for nul in case len == 1 */
835 outbytes_remaining = outbuf_size - 1; /* -1 for nul */
836 outp = dest = g_malloc (outbuf_size);
838 while (!done && !have_error)
840 size_t inbytes_tmp = inbytes_remaining;
841 err = g_iconv (cd, (char **)&p, &inbytes_tmp, &outp, &outbytes_remaining);
842 inbytes_remaining = inbytes_tmp;
844 if (err == (size_t) -1)
849 g_assert_not_reached();
853 size_t used = outp - dest;
856 dest = g_realloc (dest, outbuf_size);
859 outbytes_remaining = outbuf_size - used - 1; /* -1 for nul */
866 /* Error converting fallback string - fatal
868 g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_ILLEGAL_SEQUENCE,
869 _("Cannot convert fallback '%s' to codeset '%s'"),
870 insert_str, to_codeset);
878 gunichar ch = g_utf8_get_char (p);
879 insert_str = g_strdup_printf (ch < 0x10000 ? "\\u%04x" : "\\U%08x",
883 insert_str = fallback;
885 save_p = g_utf8_next_char (p);
886 save_inbytes = inbytes_remaining - (save_p - p);
888 inbytes_remaining = strlen (p);
891 /* fall thru if p is NULL */
893 g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_FAILED,
894 _("Error during conversion: %s"),
905 g_free ((gchar *)insert_str);
907 inbytes_remaining = save_inbytes;
912 /* call g_iconv with NULL inbuf to cleanup shift state */
914 inbytes_remaining = 0;
925 close_converter (cd);
928 *bytes_written = outp - dest; /* Doesn't include '\0' */
934 if (save_p && !fallback)
935 g_free ((gchar *)insert_str);
950 strdup_len (const gchar *string,
952 gsize *bytes_written,
959 if (!g_utf8_validate (string, len, NULL))
966 g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_ILLEGAL_SEQUENCE,
967 _("Invalid byte sequence in conversion input"));
972 real_len = strlen (string);
977 while (real_len < len && string[real_len])
982 *bytes_read = real_len;
984 *bytes_written = real_len;
986 return g_strndup (string, real_len);
991 * @opsysstring: a string in the encoding of the current locale. On Windows
992 * this means the system codepage.
993 * @len: the length of the string, or -1 if the string is
994 * nul-terminated<footnoteref linkend="nul-unsafe"/>.
995 * @bytes_read: location to store the number of bytes in the
996 * input string that were successfully converted, or %NULL.
997 * Even if the conversion was successful, this may be
998 * less than @len if there were partial characters
999 * at the end of the input. If the error
1000 * #G_CONVERT_ERROR_ILLEGAL_SEQUENCE occurs, the value
1001 * stored will the byte offset after the last valid
1003 * @bytes_written: the number of bytes stored in the output buffer (not
1004 * including the terminating nul).
1005 * @error: location to store the error occuring, or %NULL to ignore
1006 * errors. Any of the errors in #GConvertError may occur.
1008 * Converts a string which is in the encoding used for strings by
1009 * the C runtime (usually the same as that used by the operating
1010 * system) in the current locale into a UTF-8 string.
1012 * Return value: The converted string, or %NULL on an error.
1015 g_locale_to_utf8 (const gchar *opsysstring,
1018 gsize *bytes_written,
1021 const char *charset;
1023 if (g_get_charset (&charset))
1024 return strdup_len (opsysstring, len, bytes_read, bytes_written, error);
1026 return g_convert (opsysstring, len,
1027 "UTF-8", charset, bytes_read, bytes_written, error);
1031 * g_locale_from_utf8:
1032 * @utf8string: a UTF-8 encoded string
1033 * @len: the length of the string, or -1 if the string is
1034 * nul-terminated<footnoteref linkend="nul-unsafe"/>.
1035 * @bytes_read: location to store the number of bytes in the
1036 * input string that were successfully converted, or %NULL.
1037 * Even if the conversion was successful, this may be
1038 * less than @len if there were partial characters
1039 * at the end of the input. If the error
1040 * #G_CONVERT_ERROR_ILLEGAL_SEQUENCE occurs, the value
1041 * stored will the byte offset after the last valid
1043 * @bytes_written: the number of bytes stored in the output buffer (not
1044 * including the terminating nul).
1045 * @error: location to store the error occuring, or %NULL to ignore
1046 * errors. Any of the errors in #GConvertError may occur.
1048 * Converts a string from UTF-8 to the encoding used for strings by
1049 * the C runtime (usually the same as that used by the operating
1050 * system) in the current locale.
1052 * Return value: The converted string, or %NULL on an error.
1055 g_locale_from_utf8 (const gchar *utf8string,
1058 gsize *bytes_written,
1061 const gchar *charset;
1063 if (g_get_charset (&charset))
1064 return strdup_len (utf8string, len, bytes_read, bytes_written, error);
1066 return g_convert (utf8string, len,
1067 charset, "UTF-8", bytes_read, bytes_written, error);
1070 #ifndef G_PLATFORM_WIN32
1072 typedef struct _GFilenameCharsetCache GFilenameCharsetCache;
1074 struct _GFilenameCharsetCache {
1077 gchar **filename_charsets;
1081 filename_charset_cache_free (gpointer data)
1083 GFilenameCharsetCache *cache = data;
1084 g_free (cache->charset);
1085 g_strfreev (cache->filename_charsets);
1090 * g_get_filename_charsets:
1091 * @charsets: return location for the %NULL-terminated list of encoding names
1093 * Determines the preferred character sets used for filenames.
1094 * The first character set from the @charsets is the filename encoding, the
1095 * subsequent character sets are used when trying to generate a displayable
1096 * representation of a filename, see g_filename_display_name().
1098 * On Unix, the character sets are determined by consulting the
1099 * environment variables <envar>G_FILENAME_ENCODING</envar> and
1100 * <envar>G_BROKEN_FILENAMES</envar>. On Windows, the character set
1101 * used in the GLib API is always UTF-8 and said environment variables
1104 * <envar>G_FILENAME_ENCODING</envar> may be set to a comma-separated list
1105 * of character set names. The special token "@locale" is taken to mean the
1106 * character set for the current locale. If <envar>G_FILENAME_ENCODING</envar>
1107 * is not set, but <envar>G_BROKEN_FILENAMES</envar> is, the character set of
1108 * the current locale is taken as the filename encoding. If neither environment
1109 * variable is set, UTF-8 is taken as the filename encoding, but the character
1110 * set of the current locale is also put in the list of encodings.
1112 * The returned @charsets belong to GLib and must not be freed.
1114 * Note that on Unix, regardless of the locale character set or
1115 * <envar>G_FILENAME_ENCODING</envar> value, the actual file names present on a
1116 * system might be in any random encoding or just gibberish.
1118 * Return value: %TRUE if the filename encoding is UTF-8.
1123 g_get_filename_charsets (G_CONST_RETURN gchar ***filename_charsets)
1125 static GStaticPrivate cache_private = G_STATIC_PRIVATE_INIT;
1126 GFilenameCharsetCache *cache = g_static_private_get (&cache_private);
1127 const gchar *charset;
1131 cache = g_new0 (GFilenameCharsetCache, 1);
1132 g_static_private_set (&cache_private, cache, filename_charset_cache_free);
1135 g_get_charset (&charset);
1137 if (!(cache->charset && strcmp (cache->charset, charset) == 0))
1139 const gchar *new_charset;
1143 g_free (cache->charset);
1144 g_strfreev (cache->filename_charsets);
1145 cache->charset = g_strdup (charset);
1147 p = getenv ("G_FILENAME_ENCODING");
1148 if (p != NULL && p[0] != '\0')
1150 cache->filename_charsets = g_strsplit (p, ",", 0);
1151 cache->is_utf8 = (strcmp (cache->filename_charsets[0], "UTF-8") == 0);
1153 for (i = 0; cache->filename_charsets[i]; i++)
1155 if (strcmp ("@locale", cache->filename_charsets[i]) == 0)
1157 g_get_charset (&new_charset);
1158 g_free (cache->filename_charsets[i]);
1159 cache->filename_charsets[i] = g_strdup (new_charset);
1163 else if (getenv ("G_BROKEN_FILENAMES") != NULL)
1165 cache->filename_charsets = g_new0 (gchar *, 2);
1166 cache->is_utf8 = g_get_charset (&new_charset);
1167 cache->filename_charsets[0] = g_strdup (new_charset);
1171 cache->filename_charsets = g_new0 (gchar *, 3);
1172 cache->is_utf8 = TRUE;
1173 cache->filename_charsets[0] = g_strdup ("UTF-8");
1174 if (!g_get_charset (&new_charset))
1175 cache->filename_charsets[1] = g_strdup (new_charset);
1179 if (filename_charsets)
1180 *filename_charsets = (const gchar **)cache->filename_charsets;
1182 return cache->is_utf8;
1185 #else /* G_PLATFORM_WIN32 */
1188 g_get_filename_charsets (G_CONST_RETURN gchar ***filename_charsets)
1190 static const gchar *charsets[] = {
1196 /* On Windows GLib pretends that the filename charset is UTF-8 */
1197 if (filename_charsets)
1198 *filename_charsets = charsets;
1204 /* Cygwin works like before */
1205 result = g_get_charset (&(charsets[0]));
1207 if (filename_charsets)
1208 *filename_charsets = charsets;
1214 #endif /* G_PLATFORM_WIN32 */
1217 get_filename_charset (const gchar **filename_charset)
1219 const gchar **charsets;
1222 is_utf8 = g_get_filename_charsets (&charsets);
1224 if (filename_charset)
1225 *filename_charset = charsets[0];
1230 /* This is called from g_thread_init(). It's used to
1231 * initialize some static data in a threadsafe way.
1234 _g_convert_thread_init (void)
1236 const gchar **dummy;
1237 (void) g_get_filename_charsets (&dummy);
1241 * g_filename_to_utf8:
1242 * @opsysstring: a string in the encoding for filenames
1243 * @len: the length of the string, or -1 if the string is
1244 * nul-terminated<footnoteref linkend="nul-unsafe"/>.
1245 * @bytes_read: location to store the number of bytes in the
1246 * input string that were successfully converted, or %NULL.
1247 * Even if the conversion was successful, this may be
1248 * less than @len if there were partial characters
1249 * at the end of the input. If the error
1250 * #G_CONVERT_ERROR_ILLEGAL_SEQUENCE occurs, the value
1251 * stored will the byte offset after the last valid
1253 * @bytes_written: the number of bytes stored in the output buffer (not
1254 * including the terminating nul).
1255 * @error: location to store the error occuring, or %NULL to ignore
1256 * errors. Any of the errors in #GConvertError may occur.
1258 * Converts a string which is in the encoding used by GLib for
1259 * filenames into a UTF-8 string. Note that on Windows GLib uses UTF-8
1262 * Return value: The converted string, or %NULL on an error.
1265 g_filename_to_utf8 (const gchar *opsysstring,
1268 gsize *bytes_written,
1271 const gchar *charset;
1273 if (get_filename_charset (&charset))
1274 return strdup_len (opsysstring, len, bytes_read, bytes_written, error);
1276 return g_convert (opsysstring, len,
1277 "UTF-8", charset, bytes_read, bytes_written, error);
1282 #undef g_filename_to_utf8
1284 /* Binary compatibility version. Not for newly compiled code. */
1287 g_filename_to_utf8 (const gchar *opsysstring,
1290 gsize *bytes_written,
1293 const gchar *charset;
1295 if (g_get_charset (&charset))
1296 return strdup_len (opsysstring, len, bytes_read, bytes_written, error);
1298 return g_convert (opsysstring, len,
1299 "UTF-8", charset, bytes_read, bytes_written, error);
1305 * g_filename_from_utf8:
1306 * @utf8string: a UTF-8 encoded string.
1307 * @len: the length of the string, or -1 if the string is
1309 * @bytes_read: location to store the number of bytes in the
1310 * input string that were successfully converted, or %NULL.
1311 * Even if the conversion was successful, this may be
1312 * less than @len if there were partial characters
1313 * at the end of the input. If the error
1314 * #G_CONVERT_ERROR_ILLEGAL_SEQUENCE occurs, the value
1315 * stored will the byte offset after the last valid
1317 * @bytes_written: the number of bytes stored in the output buffer (not
1318 * including the terminating nul).
1319 * @error: location to store the error occuring, or %NULL to ignore
1320 * errors. Any of the errors in #GConvertError may occur.
1322 * Converts a string from UTF-8 to the encoding GLib uses for
1323 * filenames. Note that on Windows GLib uses UTF-8 for filenames.
1325 * Return value: The converted string, or %NULL on an error.
1328 g_filename_from_utf8 (const gchar *utf8string,
1331 gsize *bytes_written,
1334 const gchar *charset;
1336 if (get_filename_charset (&charset))
1337 return strdup_len (utf8string, len, bytes_read, bytes_written, error);
1339 return g_convert (utf8string, len,
1340 charset, "UTF-8", bytes_read, bytes_written, error);
1345 #undef g_filename_from_utf8
1347 /* Binary compatibility version. Not for newly compiled code. */
1350 g_filename_from_utf8 (const gchar *utf8string,
1353 gsize *bytes_written,
1356 const gchar *charset;
1358 if (g_get_charset (&charset))
1359 return strdup_len (utf8string, len, bytes_read, bytes_written, error);
1361 return g_convert (utf8string, len,
1362 charset, "UTF-8", bytes_read, bytes_written, error);
1367 /* Test of haystack has the needle prefix, comparing case
1368 * insensitive. haystack may be UTF-8, but needle must
1369 * contain only ascii. */
1371 has_case_prefix (const gchar *haystack, const gchar *needle)
1375 /* Eat one character at a time. */
1380 g_ascii_tolower (*n) == g_ascii_tolower (*h))
1390 UNSAFE_ALL = 0x1, /* Escape all unsafe characters */
1391 UNSAFE_ALLOW_PLUS = 0x2, /* Allows '+' */
1392 UNSAFE_PATH = 0x8, /* Allows '/', '&', '=', ':', '@', '+', '$' and ',' */
1393 UNSAFE_HOST = 0x10, /* Allows '/' and ':' and '@' */
1394 UNSAFE_SLASHES = 0x20 /* Allows all characters except for '/' and '%' */
1395 } UnsafeCharacterSet;
1397 static const guchar acceptable[96] = {
1398 /* A table of the ASCII chars from space (32) to DEL (127) */
1399 /* ! " # $ % & ' ( ) * + , - . / */
1400 0x00,0x3F,0x20,0x20,0x28,0x00,0x2C,0x3F,0x3F,0x3F,0x3F,0x2A,0x28,0x3F,0x3F,0x1C,
1401 /* 0 1 2 3 4 5 6 7 8 9 : ; < = > ? */
1402 0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x38,0x20,0x20,0x2C,0x20,0x20,
1403 /* @ A B C D E F G H I J K L M N O */
1404 0x38,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,
1405 /* P Q R S T U V W X Y Z [ \ ] ^ _ */
1406 0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x20,0x20,0x20,0x20,0x3F,
1407 /* ` a b c d e f g h i j k l m n o */
1408 0x20,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,
1409 /* p q r s t u v w x y z { | } ~ DEL */
1410 0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x20,0x20,0x20,0x3F,0x20
1413 static const gchar hex[16] = "0123456789ABCDEF";
1415 /* Note: This escape function works on file: URIs, but if you want to
1416 * escape something else, please read RFC-2396 */
1418 g_escape_uri_string (const gchar *string,
1419 UnsafeCharacterSet mask)
1421 #define ACCEPTABLE(a) ((a)>=32 && (a)<128 && (acceptable[(a)-32] & use_mask))
1428 UnsafeCharacterSet use_mask;
1430 g_return_val_if_fail (mask == UNSAFE_ALL
1431 || mask == UNSAFE_ALLOW_PLUS
1432 || mask == UNSAFE_PATH
1433 || mask == UNSAFE_HOST
1434 || mask == UNSAFE_SLASHES, NULL);
1438 for (p = string; *p != '\0'; p++)
1441 if (!ACCEPTABLE (c))
1445 result = g_malloc (p - string + unacceptable * 2 + 1);
1448 for (q = result, p = string; *p != '\0'; p++)
1452 if (!ACCEPTABLE (c))
1454 *q++ = '%'; /* means hex coming */
1469 g_escape_file_uri (const gchar *hostname,
1470 const gchar *pathname)
1472 char *escaped_hostname = NULL;
1477 char *p, *backslash;
1479 /* Turn backslashes into forward slashes. That's what Netscape
1480 * does, and they are actually more or less equivalent in Windows.
1483 pathname = g_strdup (pathname);
1484 p = (char *) pathname;
1486 while ((backslash = strchr (p, '\\')) != NULL)
1493 if (hostname && *hostname != '\0')
1495 escaped_hostname = g_escape_uri_string (hostname, UNSAFE_HOST);
1498 escaped_path = g_escape_uri_string (pathname, UNSAFE_PATH);
1500 res = g_strconcat ("file://",
1501 (escaped_hostname) ? escaped_hostname : "",
1502 (*escaped_path != '/') ? "/" : "",
1507 g_free ((char *) pathname);
1510 g_free (escaped_hostname);
1511 g_free (escaped_path);
1517 unescape_character (const char *scanner)
1522 first_digit = g_ascii_xdigit_value (scanner[0]);
1523 if (first_digit < 0)
1526 second_digit = g_ascii_xdigit_value (scanner[1]);
1527 if (second_digit < 0)
1530 return (first_digit << 4) | second_digit;
1534 g_unescape_uri_string (const char *escaped,
1536 const char *illegal_escaped_characters,
1537 gboolean ascii_must_not_be_escaped)
1539 const gchar *in, *in_end;
1540 gchar *out, *result;
1543 if (escaped == NULL)
1547 len = strlen (escaped);
1549 result = g_malloc (len + 1);
1552 for (in = escaped, in_end = escaped + len; in < in_end; in++)
1558 /* catch partial escape sequences past the end of the substring */
1559 if (in + 3 > in_end)
1562 c = unescape_character (in + 1);
1564 /* catch bad escape sequences and NUL characters */
1568 /* catch escaped ASCII */
1569 if (ascii_must_not_be_escaped && c <= 0x7F)
1572 /* catch other illegal escaped characters */
1573 if (strchr (illegal_escaped_characters, c) != NULL)
1582 g_assert (out - result <= len);
1595 is_asciialphanum (gunichar c)
1597 return c <= 0x7F && g_ascii_isalnum (c);
1601 is_asciialpha (gunichar c)
1603 return c <= 0x7F && g_ascii_isalpha (c);
1606 /* allows an empty string */
1608 hostname_validate (const char *hostname)
1611 gunichar c, first_char, last_char;
1618 /* read in a label */
1619 c = g_utf8_get_char (p);
1620 p = g_utf8_next_char (p);
1621 if (!is_asciialphanum (c))
1627 c = g_utf8_get_char (p);
1628 p = g_utf8_next_char (p);
1630 while (is_asciialphanum (c) || c == '-');
1631 if (last_char == '-')
1634 /* if that was the last label, check that it was a toplabel */
1635 if (c == '\0' || (c == '.' && *p == '\0'))
1636 return is_asciialpha (first_char);
1643 * g_filename_from_uri:
1644 * @uri: a uri describing a filename (escaped, encoded in ASCII).
1645 * @hostname: Location to store hostname for the URI, or %NULL.
1646 * If there is no hostname in the URI, %NULL will be
1647 * stored in this location.
1648 * @error: location to store the error occuring, or %NULL to ignore
1649 * errors. Any of the errors in #GConvertError may occur.
1651 * Converts an escaped ASCII-encoded URI to a local filename in the
1652 * encoding used for filenames.
1654 * Return value: a newly-allocated string holding the resulting
1655 * filename, or %NULL on an error.
1658 g_filename_from_uri (const gchar *uri,
1662 const char *path_part;
1663 const char *host_part;
1664 char *unescaped_hostname;
1675 if (!has_case_prefix (uri, "file:/"))
1677 g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_BAD_URI,
1678 _("The URI '%s' is not an absolute URI using the \"file\" scheme"),
1683 path_part = uri + strlen ("file:");
1685 if (strchr (path_part, '#') != NULL)
1687 g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_BAD_URI,
1688 _("The local file URI '%s' may not include a '#'"),
1693 if (has_case_prefix (path_part, "///"))
1695 else if (has_case_prefix (path_part, "//"))
1698 host_part = path_part;
1700 path_part = strchr (path_part, '/');
1702 if (path_part == NULL)
1704 g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_BAD_URI,
1705 _("The URI '%s' is invalid"),
1710 unescaped_hostname = g_unescape_uri_string (host_part, path_part - host_part, "", TRUE);
1712 if (unescaped_hostname == NULL ||
1713 !hostname_validate (unescaped_hostname))
1715 g_free (unescaped_hostname);
1716 g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_BAD_URI,
1717 _("The hostname of the URI '%s' is invalid"),
1723 *hostname = unescaped_hostname;
1725 g_free (unescaped_hostname);
1728 filename = g_unescape_uri_string (path_part, -1, "/", FALSE);
1730 if (filename == NULL)
1732 g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_BAD_URI,
1733 _("The URI '%s' contains invalidly escaped characters"),
1740 /* Drop localhost */
1741 if (hostname && *hostname != NULL &&
1742 g_ascii_strcasecmp (*hostname, "localhost") == 0)
1748 /* Turn slashes into backslashes, because that's the canonical spelling */
1750 while ((slash = strchr (p, '/')) != NULL)
1756 /* Windows URIs with a drive letter can be like "file://host/c:/foo"
1757 * or "file://host/c|/foo" (some Netscape versions). In those cases, start
1758 * the filename from the drive letter.
1760 if (g_ascii_isalpha (filename[1]))
1762 if (filename[2] == ':')
1764 else if (filename[2] == '|')
1772 result = g_strdup (filename + offs);
1780 #undef g_filename_from_uri
1783 g_filename_from_uri (const gchar *uri,
1787 gchar *utf8_filename;
1788 gchar *retval = NULL;
1790 utf8_filename = g_filename_from_uri_utf8 (uri, hostname, error);
1793 retval = g_locale_from_utf8 (utf8_filename, -1, NULL, NULL, error);
1794 g_free (utf8_filename);
1802 * g_filename_to_uri:
1803 * @filename: an absolute filename specified in the GLib file name encoding,
1804 * which is the on-disk file name bytes on Unix, and UTF-8 on
1806 * @hostname: A UTF-8 encoded hostname, or %NULL for none.
1807 * @error: location to store the error occuring, or %NULL to ignore
1808 * errors. Any of the errors in #GConvertError may occur.
1810 * Converts an absolute filename to an escaped ASCII-encoded URI.
1812 * Return value: a newly-allocated string holding the resulting
1813 * URI, or %NULL on an error.
1816 g_filename_to_uri (const gchar *filename,
1817 const gchar *hostname,
1822 g_return_val_if_fail (filename != NULL, NULL);
1824 if (!g_path_is_absolute (filename))
1826 g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_NOT_ABSOLUTE_PATH,
1827 _("The pathname '%s' is not an absolute path"),
1833 !(g_utf8_validate (hostname, -1, NULL)
1834 && hostname_validate (hostname)))
1836 g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_ILLEGAL_SEQUENCE,
1837 _("Invalid hostname"));
1842 /* Don't use localhost unnecessarily */
1843 if (hostname && g_ascii_strcasecmp (hostname, "localhost") == 0)
1847 escaped_uri = g_escape_file_uri (hostname, filename);
1854 #undef g_filename_to_uri
1857 g_filename_to_uri (const gchar *filename,
1858 const gchar *hostname,
1861 gchar *utf8_filename;
1862 gchar *retval = NULL;
1864 utf8_filename = g_locale_to_utf8 (filename, -1, NULL, NULL, error);
1868 retval = g_filename_to_uri_utf8 (utf8_filename, hostname, error);
1869 g_free (utf8_filename);
1878 * g_uri_list_extract_uris:
1879 * @uri_list: an URI list
1881 * Splits an URI list conforming to the text/uri-list
1882 * mime type defined in RFC 2483 into individual URIs,
1883 * discarding any comments. The URIs are not validated.
1885 * Returns: a newly allocated %NULL-terminated list of
1886 * strings holding the individual URIs. The array should
1887 * be freed with g_strfreev().
1892 g_uri_list_extract_uris (const gchar *uri_list)
1903 /* We don't actually try to validate the URI according to RFC
1904 * 2396, or even check for allowed characters - we just ignore
1905 * comments and trim whitespace off the ends. We also
1906 * allow LF delimination as well as the specified CRLF.
1908 * We do allow comments like specified in RFC 2483.
1914 while (g_ascii_isspace (*p))
1918 while (*q && (*q != '\n') && (*q != '\r'))
1924 while (q > p && g_ascii_isspace (*q))
1929 uris = g_slist_prepend (uris, g_strndup (p, q - p + 1));
1934 p = strchr (p, '\n');
1939 result = g_new (gchar *, n_uris + 1);
1941 result[n_uris--] = NULL;
1942 for (u = uris; u; u = u->next)
1943 result[n_uris--] = u->data;
1945 g_slist_free (uris);
1951 make_valid_utf8 (const gchar *name)
1954 const gchar *remainder, *invalid;
1955 gint remaining_bytes, valid_bytes;
1959 remaining_bytes = strlen (name);
1961 while (remaining_bytes != 0)
1963 if (g_utf8_validate (remainder, remaining_bytes, &invalid))
1965 valid_bytes = invalid - remainder;
1968 string = g_string_sized_new (remaining_bytes);
1970 g_string_append_len (string, remainder, valid_bytes);
1971 g_string_append_c (string, '?');
1973 remaining_bytes -= valid_bytes + 1;
1974 remainder = invalid + 1;
1978 return g_strdup (name);
1980 g_string_append (string, remainder);
1981 g_string_append (string, " (invalid encoding)");
1983 g_assert (g_utf8_validate (string->str, -1, NULL));
1985 return g_string_free (string, FALSE);
1989 * g_filename_display_basename:
1990 * @filename: an absolute pathname in the GLib file name encoding
1992 * Returns the display basename for the particular filename, guaranteed
1993 * to be valid UTF-8. The display name might not be identical to the filename,
1994 * for instance there might be problems converting it to UTF-8, and some files
1995 * can be translated in the display
1997 * You must pass the whole absolute pathname to this functions so that
1998 * translation of well known locations can be done.
2000 * This function is preferred over g_filename_display_name() if you know the
2001 * whole path, as it allows translation.
2003 * Return value: a newly allocated string containing
2004 * a rendition of the basename of the filename in valid UTF-8
2009 g_filename_display_basename (const gchar *filename)
2014 g_return_val_if_fail (filename != NULL, NULL);
2016 basename = g_path_get_basename (filename);
2017 display_name = g_filename_display_name (basename);
2019 return display_name;
2023 * g_filename_display_name:
2024 * @filename: a pathname hopefully in the GLib file name encoding
2026 * Converts a filename into a valid UTF-8 string. The
2027 * conversion is not necessarily reversible, so you
2028 * should keep the original around and use the return
2029 * value of this function only for display purposes.
2030 * Unlike g_filename_to_utf8(), the result is guaranteed
2031 * to be non-NULL even if the filename actually isn't in the GLib
2032 * file name encoding.
2034 * If you know the whole pathname of the file you should use
2035 * g_filename_display_basename(), since that allows location-based
2036 * translation of filenames.
2038 * Return value: a newly allocated string containing
2039 * a rendition of the filename in valid UTF-8
2044 g_filename_display_name (const gchar *filename)
2047 const gchar **charsets;
2048 gchar *display_name = NULL;
2051 is_utf8 = g_get_filename_charsets (&charsets);
2055 if (g_utf8_validate (filename, -1, NULL))
2056 display_name = g_strdup (filename);
2061 /* Try to convert from the filename charsets to UTF-8.
2062 * Skip the first charset if it is UTF-8.
2064 for (i = is_utf8 ? 1 : 0; charsets[i]; i++)
2066 display_name = g_convert (filename, -1, "UTF-8", charsets[i],
2074 /* if all conversions failed, we replace invalid UTF-8
2075 * by a question mark
2078 display_name = make_valid_utf8 (filename);
2080 return display_name;
2083 #define __G_CONVERT_C__
2084 #include "galiasdef.c"