glib/glib/gunicollate.c

/* gunicollate.c - Collation
 *
 *  Copyright 2001,2005 Red Hat, Inc.
 *
 * SPDX-License-Identifier: LGPL-2.1-or-later
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public License
 * along with this library; if not, see <http://www.gnu.org/licenses/>.
 */

#include "config.h"

#include <locale.h>
#include <string.h>
#ifdef HAVE_WCHAR_H
#include <wchar.h>
#endif

#ifdef HAVE_CARBON
#include <CoreServices/CoreServices.h>
#endif

#include "gmem.h"
#include "gunicode.h"
#include "gunicodeprivate.h"
#include "gstring.h"
#include "gstrfuncs.h"
#include "gtestutils.h"
#include "gcharset.h"
#include "gconvert.h"

#if SIZEOF_WCHAR_T == 4 && defined(__STDC_ISO_10646__)
#define GUNICHAR_EQUALS_WCHAR_T 1
#endif

#ifdef _MSC_VER
/* Workaround for bug in MSVCR80.DLL */
static gsize
msc_strxfrm_wrapper (char       *string1,
		     const char *string2,
		     gsize       count)
{
  if (!string1 || count <= 0)
    {
      char tmp;

      return strxfrm (&tmp, string2, 1);
    }
  return strxfrm (string1, string2, count);
}
#define strxfrm msc_strxfrm_wrapper
#endif

/**
 * g_utf8_collate:
 * @str1: a UTF-8 encoded string
 * @str2: a UTF-8 encoded string
 * 
 * Compares two strings for ordering using the linguistically
 * correct rules for the [current locale][setlocale].
 * When sorting a large number of strings, it will be significantly 
 * faster to obtain collation keys with g_utf8_collate_key() and 
 * compare the keys with strcmp() when sorting instead of sorting 
 * the original strings.
 * 
 * If the two strings are not comparable due to being in different collation
 * sequences, the result is undefined. This can happen if the strings are in
 * different language scripts, for example.
 *
 * Returns: < 0 if @str1 compares before @str2, 
 *   0 if they compare equal, > 0 if @str1 compares after @str2.
 **/
gint
g_utf8_collate (const gchar *str1,
		const gchar *str2)
{
  gint result;

#ifdef HAVE_CARBON

  UniChar *str1_utf16;
  UniChar *str2_utf16;
  glong len1;
  glong len2;
  SInt32 retval = 0;

  g_return_val_if_fail (str1 != NULL, 0);
  g_return_val_if_fail (str2 != NULL, 0);

  str1_utf16 = g_utf8_to_utf16 (str1, -1, NULL, &len1, NULL);
  str2_utf16 = g_utf8_to_utf16 (str2, -1, NULL, &len2, NULL);

  UCCompareTextDefault (kUCCollateStandardOptions,
                        str1_utf16, len1, str2_utf16, len2,
                        NULL, &retval);
  result = retval;

  g_free (str2_utf16);
  g_free (str1_utf16);

#elif defined(HAVE_WCHAR_H) && defined(GUNICHAR_EQUALS_WCHAR_T)

  gunichar *str1_norm;
  gunichar *str2_norm;

  g_return_val_if_fail (str1 != NULL, 0);
  g_return_val_if_fail (str2 != NULL, 0);

  str1_norm = _g_utf8_normalize_wc (str1, -1, G_NORMALIZE_ALL_COMPOSE);
  str2_norm = _g_utf8_normalize_wc (str2, -1, G_NORMALIZE_ALL_COMPOSE);

  result = wcscoll ((wchar_t *)str1_norm, (wchar_t *)str2_norm);

  g_free (str1_norm);
  g_free (str2_norm);

#else

  const gchar *charset;
  gchar *str1_norm;
  gchar *str2_norm;

  g_return_val_if_fail (str1 != NULL, 0);
  g_return_val_if_fail (str2 != NULL, 0);

  str1_norm = g_utf8_normalize (str1, -1, G_NORMALIZE_ALL_COMPOSE);
  str2_norm = g_utf8_normalize (str2, -1, G_NORMALIZE_ALL_COMPOSE);

  if (g_get_charset (&charset))
    {
      result = strcoll (str1_norm, str2_norm);
    }
  else
    {
      gchar *str1_locale = g_convert (str1_norm, -1, charset, "UTF-8", NULL, NULL, NULL);
      gchar *str2_locale = g_convert (str2_norm, -1, charset, "UTF-8", NULL, NULL, NULL);

      if (str1_locale && str2_locale)
	result =  strcoll (str1_locale, str2_locale);
      else if (str1_locale)
	result = -1;
      else if (str2_locale)
	result = 1;
      else
	result = strcmp (str1_norm, str2_norm);

      g_free (str1_locale);
      g_free (str2_locale);
    }

  g_free (str1_norm);
  g_free (str2_norm);

#endif

  return result;
}

#if defined(HAVE_WCHAR_H) && defined(GUNICHAR_EQUALS_WCHAR_T)
/* We need UTF-8 encoding of numbers to encode the weights if
 * we are using wcsxfrm. However, we aren't encoding Unicode
 * characters, so we can't simply use g_unichar_to_utf8.
 *
 * The following routine is taken (with modification) from GNU
 * libc's strxfrm routine:
 *
 * Copyright (C) 1995-1999,2000,2001 Free Software Foundation, Inc.
 * Written by Ulrich Drepper <drepper@cygnus.com>, 1995.
 */
static inline int
utf8_encode (char *buf, wchar_t val)
{
  int retval;

  if (val < 0x80)
    {
      if (buf)
	*buf++ = (char) val;
      retval = 1;
    }
  else
    {
      int step;

      for (step = 2; step < 6; ++step)
        if ((val & (~(guint32)0 << (5 * step + 1))) == 0)
          break;
      retval = step;

      if (buf)
	{
	  *buf = (unsigned char) (~0xff >> step);
	  --step;
	  do
	    {
	      buf[step] = 0x80 | (val & 0x3f);
	      val >>= 6;
	    }
	  while (--step > 0);
	  *buf |= val;
	}
    }

  return retval;
}
#endif

#ifdef HAVE_CARBON

static gchar *
collate_key_to_string (UCCollationValue *key,
                       gsize             key_len)
{
  gchar *result;
  gsize result_len;
  long *lkey = (long *) key;

  /* UCCollationValue format:
   *
   * UCCollateOptions (32/64 bits)
   * SizeInBytes      (32/64 bits)
   * Value            (8 bits arrey)
   *
   * UCCollateOptions: ordering option mask of the collator
   * used to create the key. Size changes on 32-bit / 64-bit
   * hosts. On 64-bits also the extra half-word seems to have
   * some extra (unknown) meaning.
   * SizeInBytes: size of the whole structure, in bytes
   * (including UCCollateOptions and SizeInBytes fields). Size
   * changes on 32-bit & 64-bit hosts.
   * Value: array of bytes containing the comparison weights.
   * Seems to have several sub-strings separated by \001 and \002
   * chars. Also, experience shows this is directly strcmp-able.
   */

  result_len = lkey[1];
  result = g_malloc (result_len + 1);
  memcpy (result, &lkey[2], result_len);
  result[result_len] = '\0';

  return result;
}

static gchar *
carbon_collate_key_with_collator (const gchar *str,
                                  gssize       len,
                                  CollatorRef  collator)
{
  UniChar *str_utf16 = NULL;
  glong len_utf16;
  OSStatus ret;
  UCCollationValue staticbuf[512];
  UCCollationValue *freeme = NULL;
  UCCollationValue *buf;
  ItemCount buf_len;
  ItemCount key_len;
  ItemCount try_len;
  gchar *result = NULL;

  str_utf16 = g_utf8_to_utf16 (str, len, NULL, &len_utf16, NULL);
  try_len = len_utf16 * 5 + 2;

  if (try_len <= sizeof staticbuf)
    {
      buf = staticbuf;
      buf_len = sizeof staticbuf;
    }
  else
    {
      freeme = g_new (UCCollationValue, try_len);
      buf = freeme;
      buf_len = try_len;
    }

  ret = UCGetCollationKey (collator, str_utf16, len_utf16,
                           buf_len, &key_len, buf);

  if (ret == kCollateBufferTooSmall)
    {
      freeme = g_renew (UCCollationValue, freeme, try_len * 2);
      buf = freeme;
      buf_len = try_len * 2;
      ret = UCGetCollationKey (collator, str_utf16, len_utf16,
                               buf_len, &key_len, buf);
    }

  if (ret == 0)
    result = collate_key_to_string (buf, key_len);
  else
    result = g_strdup ("");

  g_free (freeme);
  g_free (str_utf16);
  return result;
}

static gchar *
carbon_collate_key (const gchar *str,
                    gssize       len)
{
  static CollatorRef collator;

  if (G_UNLIKELY (!collator))
    {
      UCCreateCollator (NULL, 0, kUCCollateStandardOptions, &collator);

      if (!collator)
        {
          static gboolean been_here;
          if (!been_here)
            g_warning ("%s: UCCreateCollator failed", G_STRLOC);
          been_here = TRUE;
          return g_strdup ("");
        }
    }

  return carbon_collate_key_with_collator (str, len, collator);
}

static gchar *
carbon_collate_key_for_filename (const gchar *str,
                                 gssize       len)
{
  static CollatorRef collator;

  if (G_UNLIKELY (!collator))
    {
      /* http://developer.apple.com/qa/qa2004/qa1159.html */
      UCCreateCollator (NULL, 0,
                        kUCCollateComposeInsensitiveMask
                         | kUCCollateWidthInsensitiveMask
                         | kUCCollateCaseInsensitiveMask
                         | kUCCollateDigitsOverrideMask
                         | kUCCollateDigitsAsNumberMask
                         | kUCCollatePunctuationSignificantMask, 
                        &collator);

      if (!collator)
        {
          static gboolean been_here;
          if (!been_here)
            g_warning ("%s: UCCreateCollator failed", G_STRLOC);
          been_here = TRUE;
          return g_strdup ("");
        }
    }

  return carbon_collate_key_with_collator (str, len, collator);
}

#endif /* HAVE_CARBON */

/**
 * g_utf8_collate_key:
 * @str: a UTF-8 encoded string.
 * @len: length of @str, in bytes, or -1 if @str is nul-terminated.
 *
 * Converts a string into a collation key that can be compared
 * with other collation keys produced by the same function using 
 * strcmp(). 
 *
 * The results of comparing the collation keys of two strings 
 * with strcmp() will always be the same as comparing the two 
 * original keys with g_utf8_collate().
 * 
 * Note that this function depends on the [current locale][setlocale].
 * 
 * Returns: a newly allocated string. This string should
 *   be freed with g_free() when you are done with it.
 **/
gchar *
g_utf8_collate_key (const gchar *str,
		    gssize       len)
{
  gchar *result;

#ifdef HAVE_CARBON

  g_return_val_if_fail (str != NULL, NULL);
  result = carbon_collate_key (str, len);

#elif defined(HAVE_WCHAR_H) && defined(GUNICHAR_EQUALS_WCHAR_T)

  gsize xfrm_len;
  gunichar *str_norm;
  wchar_t *result_wc;
  gsize i;
  gsize result_len = 0;

  g_return_val_if_fail (str != NULL, NULL);

  str_norm = _g_utf8_normalize_wc (str, len, G_NORMALIZE_ALL_COMPOSE);

  g_return_val_if_fail (str_norm != NULL, NULL);

  xfrm_len = wcsxfrm (NULL, (wchar_t *)str_norm, 0);
  result_wc = g_new (wchar_t, xfrm_len + 1);
  wcsxfrm (result_wc, (wchar_t *)str_norm, xfrm_len + 1);

  for (i = 0; i < xfrm_len; i++)
    result_len += utf8_encode (NULL, result_wc[i]);

  result = g_malloc (result_len + 1);
  result_len = 0;
  for (i = 0; i < xfrm_len; i++)
    result_len += utf8_encode (result + result_len, result_wc[i]);

  result[result_len] = '\0';

  g_free (result_wc);
  g_free (str_norm);

  return result;
#else

  gsize xfrm_len = 0;
  const gchar *charset;
  gchar *str_norm;

  g_return_val_if_fail (str != NULL, NULL);

  str_norm = g_utf8_normalize (str, len, G_NORMALIZE_ALL_COMPOSE);

  result = NULL;

  if (g_get_charset (&charset))
    {
      xfrm_len = strxfrm (NULL, str_norm, 0);
      if (xfrm_len < G_MAXINT - 2)
        {
          result = g_malloc (xfrm_len + 1);
          strxfrm (result, str_norm, xfrm_len + 1);
        }
    }
  else
    {
      gchar *str_locale = g_convert (str_norm, -1, charset, "UTF-8", NULL, NULL, NULL);

      if (str_locale)
	{
	  xfrm_len = strxfrm (NULL, str_locale, 0);
	  if (xfrm_len >= G_MAXINT - 2)
	    {
	      g_free (str_locale);
	      str_locale = NULL;
	    }
	}
      if (str_locale)
	{
	  result = g_malloc (xfrm_len + 2);
	  result[0] = 'A';
	  strxfrm (result + 1, str_locale, xfrm_len + 1);
	  
	  g_free (str_locale);
	}
    }
    
  if (!result) 
    {
      xfrm_len = strlen (str_norm);
      result = g_malloc (xfrm_len + 2);
      result[0] = 'B';
      memcpy (result + 1, str_norm, xfrm_len);
      result[xfrm_len+1] = '\0';
    }

  g_free (str_norm);
#endif

  return result;
}

/* This is a collation key that is very very likely to sort before any
 * collation key that libc strxfrm generates. We use this before any
 * special case (dot or number) to make sure that its sorted before
 * anything else.
 */
#define COLLATION_SENTINEL "\1\1\1"

/**
 * g_utf8_collate_key_for_filename:
 * @str: a UTF-8 encoded string.
 * @len: length of @str, in bytes, or -1 if @str is nul-terminated.
 *
 * Converts a string into a collation key that can be compared
 * with other collation keys produced by the same function using strcmp(). 
 * 
 * In order to sort filenames correctly, this function treats the dot '.' 
 * as a special case. Most dictionary orderings seem to consider it
 * insignificant, thus producing the ordering "event.c" "eventgenerator.c"
 * "event.h" instead of "event.c" "event.h" "eventgenerator.c". Also, we
 * would like to treat numbers intelligently so that "file1" "file10" "file5"
 * is sorted as "file1" "file5" "file10".
 * 
 * Note that this function depends on the [current locale][setlocale].
 *
 * Returns: a newly allocated string. This string should
 *   be freed with g_free() when you are done with it.
 *
 * Since: 2.8
 */
gchar *
g_utf8_collate_key_for_filename (const gchar *str,
				 gssize       len)
{
#ifndef HAVE_CARBON
  GString *result;
  GString *append;
  const gchar *p;
  const gchar *prev;
  const gchar *end;
  gchar *collate_key;
  gint digits;
  gint leading_zeros;

  /*
   * How it works:
   *
   * Split the filename into collatable substrings which do
   * not contain [.0-9] and special-cased substrings. The collatable 
   * substrings are run through the normal g_utf8_collate_key() and the 
   * resulting keys are concatenated with keys generated from the 
   * special-cased substrings.
   *
   * Special cases: Dots are handled by replacing them with '\1' which 
   * implies that short dot-delimited substrings are before long ones, 
   * e.g.
   * 
   *   a\1a   (a.a)
   *   a-\1a  (a-.a)
   *   aa\1a  (aa.a)
   * 
   * Numbers are handled by prepending to each number d-1 superdigits 
   * where d = number of digits in the number and SUPERDIGIT is a 
   * character with an integer value higher than any digit (for instance 
   * ':'). This ensures that single-digit numbers are sorted before 
   * double-digit numbers which in turn are sorted separately from 
   * triple-digit numbers, etc. To avoid strange side-effects when 
   * sorting strings that already contain SUPERDIGITs, a '\2'
   * is also prepended, like this
   *
   *   file\21      (file1)
   *   file\25      (file5)
   *   file\2:10    (file10)
   *   file\2:26    (file26)
   *   file\2::100  (file100)
   *   file:foo     (file:foo)
   * 
   * This has the side-effect of sorting numbers before everything else (except
   * dots), but this is probably OK.
   *
   * Leading digits are ignored when doing the above. To discriminate
   * numbers which differ only in the number of leading digits, we append
   * the number of leading digits as a byte at the very end of the collation
   * key.
   *
   * To try avoid conflict with any collation key sequence generated by libc we
   * start each switch to a special cased part with a sentinel that hopefully
   * will sort before anything libc will generate.
   */

  if (len < 0)
    len = strlen (str);

  result = g_string_sized_new (len * 2);
  append = g_string_sized_new (0);

  end = str + len;

  /* No need to use utf8 functions, since we're only looking for ascii chars */
  for (prev = p = str; p < end; p++)
    {
      switch (*p)
	{
	case '.':
	  if (prev != p) 
	    {
	      collate_key = g_utf8_collate_key (prev, p - prev);
	      g_string_append (result, collate_key);
	      g_free (collate_key);
	    }
	  
	  g_string_append (result, COLLATION_SENTINEL "\1");
	  
	  /* skip the dot */
	  prev = p + 1;
	  break;
	  
	case '0':
	case '1':
	case '2':
	case '3':
	case '4':
	case '5':
	case '6':
	case '7':
	case '8':
	case '9':
	  if (prev != p) 
	    {
	      collate_key = g_utf8_collate_key (prev, p - prev);
	      g_string_append (result, collate_key);
	      g_free (collate_key);
	    }
	  
	  g_string_append (result, COLLATION_SENTINEL "\2");
	  
	  prev = p;
	  
	  /* write d-1 colons */
	  if (*p == '0')
	    {
	      leading_zeros = 1;
	      digits = 0;
	    }
	  else
	    {
	      leading_zeros = 0;
	      digits = 1;
	    }
	  
	  while (++p < end)
	    {
	      if (*p == '0' && !digits)
		++leading_zeros;
	      else if (g_ascii_isdigit(*p))
		++digits;
	      else
                {
 		  /* count an all-zero sequence as
                   * one digit plus leading zeros
                   */
          	  if (!digits)
                    {
                      ++digits;
                      --leading_zeros;
                    }        
		  break;
                }
	    }

	  while (digits > 1)
	    {
	      g_string_append_c (result, ':');
	      --digits;
	    }

	  if (leading_zeros > 0)
	    {
	      g_string_append_c (append, (char)leading_zeros);
	      prev += leading_zeros;
	    }
	  
	  /* write the number itself */
	  g_string_append_len (result, prev, p - prev);
	  
	  prev = p;
	  --p;	  /* go one step back to avoid disturbing outer loop */
	  break;
	  
	default:
	  /* other characters just accumulate */
	  break;
	}
    }
  
  if (prev != p) 
    {
      collate_key = g_utf8_collate_key (prev, p - prev);
      g_string_append (result, collate_key);
      g_free (collate_key);
    }
  
  g_string_append (result, append->str);
  g_string_free (append, TRUE);

  return g_string_free (result, FALSE);
#else /* HAVE_CARBON */
  return carbon_collate_key_for_filename (str, len);
#endif
}