Home | History | Annotate | Download | only in src
      1 /*
      2  * Copyright  2009  Red Hat, Inc.
      3  * Copyright  2009  Keith Stribley
      4  * Copyright  2011  Google, Inc.
      5  *
      6  *  This is part of HarfBuzz, a text shaping library.
      7  *
      8  * Permission is hereby granted, without written agreement and without
      9  * license or royalty fees, to use, copy, modify, and distribute this
     10  * software and its documentation for any purpose, provided that the
     11  * above copyright notice and the following two paragraphs appear in
     12  * all copies of this software.
     13  *
     14  * IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE TO ANY PARTY FOR
     15  * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES
     16  * ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN
     17  * IF THE COPYRIGHT HOLDER HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
     18  * DAMAGE.
     19  *
     20  * THE COPYRIGHT HOLDER SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING,
     21  * BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
     22  * FITNESS FOR A PARTICULAR PURPOSE.  THE SOFTWARE PROVIDED HEREUNDER IS
     23  * ON AN "AS IS" BASIS, AND THE COPYRIGHT HOLDER HAS NO OBLIGATION TO
     24  * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
     25  *
     26  * Red Hat Author(s): Behdad Esfahbod
     27  * Google Author(s): Behdad Esfahbod
     28  */
     29 
     30 #include "hb-private.hh"
     31 
     32 #include "hb-icu.h"
     33 
     34 #include "hb-unicode-private.hh"
     35 
     36 #include <unicode/uchar.h>
     37 #include <unicode/unorm2.h>
     38 #include <unicode/ustring.h>
     39 #include <unicode/utf16.h>
     40 #include <unicode/uversion.h>
     41 
     42 
     43 hb_script_t
     44 hb_icu_script_to_script (UScriptCode script)
     45 {
     46   if (unlikely (script == USCRIPT_INVALID_CODE))
     47     return HB_SCRIPT_INVALID;
     48 
     49   return hb_script_from_string (uscript_getShortName (script), -1);
     50 }
     51 
     52 UScriptCode
     53 hb_icu_script_from_script (hb_script_t script)
     54 {
     55   if (unlikely (script == HB_SCRIPT_INVALID))
     56     return USCRIPT_INVALID_CODE;
     57 
     58   for (unsigned int i = 0; i < USCRIPT_CODE_LIMIT; i++)
     59     if (unlikely (hb_icu_script_to_script ((UScriptCode) i) == script))
     60       return (UScriptCode) i;
     61 
     62   return USCRIPT_UNKNOWN;
     63 }
     64 
     65 
     66 static hb_unicode_combining_class_t
     67 hb_icu_unicode_combining_class (hb_unicode_funcs_t *ufuncs HB_UNUSED,
     68 				hb_codepoint_t      unicode,
     69 				void               *user_data HB_UNUSED)
     70 
     71 {
     72   return (hb_unicode_combining_class_t) u_getCombiningClass (unicode);
     73 }
     74 
     75 static unsigned int
     76 hb_icu_unicode_eastasian_width (hb_unicode_funcs_t *ufuncs HB_UNUSED,
     77 				hb_codepoint_t      unicode,
     78 				void               *user_data HB_UNUSED)
     79 {
     80   switch (u_getIntPropertyValue(unicode, UCHAR_EAST_ASIAN_WIDTH))
     81   {
     82   case U_EA_WIDE:
     83   case U_EA_FULLWIDTH:
     84     return 2;
     85   case U_EA_NEUTRAL:
     86   case U_EA_AMBIGUOUS:
     87   case U_EA_HALFWIDTH:
     88   case U_EA_NARROW:
     89     return 1;
     90   }
     91   return 1;
     92 }
     93 
     94 static hb_unicode_general_category_t
     95 hb_icu_unicode_general_category (hb_unicode_funcs_t *ufuncs HB_UNUSED,
     96 				 hb_codepoint_t      unicode,
     97 				 void               *user_data HB_UNUSED)
     98 {
     99   switch (u_getIntPropertyValue(unicode, UCHAR_GENERAL_CATEGORY))
    100   {
    101   case U_UNASSIGNED:			return HB_UNICODE_GENERAL_CATEGORY_UNASSIGNED;
    102 
    103   case U_UPPERCASE_LETTER:		return HB_UNICODE_GENERAL_CATEGORY_UPPERCASE_LETTER;
    104   case U_LOWERCASE_LETTER:		return HB_UNICODE_GENERAL_CATEGORY_LOWERCASE_LETTER;
    105   case U_TITLECASE_LETTER:		return HB_UNICODE_GENERAL_CATEGORY_TITLECASE_LETTER;
    106   case U_MODIFIER_LETTER:		return HB_UNICODE_GENERAL_CATEGORY_MODIFIER_LETTER;
    107   case U_OTHER_LETTER:			return HB_UNICODE_GENERAL_CATEGORY_OTHER_LETTER;
    108 
    109   case U_NON_SPACING_MARK:		return HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK;
    110   case U_ENCLOSING_MARK:		return HB_UNICODE_GENERAL_CATEGORY_ENCLOSING_MARK;
    111   case U_COMBINING_SPACING_MARK:	return HB_UNICODE_GENERAL_CATEGORY_SPACING_MARK;
    112 
    113   case U_DECIMAL_DIGIT_NUMBER:		return HB_UNICODE_GENERAL_CATEGORY_DECIMAL_NUMBER;
    114   case U_LETTER_NUMBER:			return HB_UNICODE_GENERAL_CATEGORY_LETTER_NUMBER;
    115   case U_OTHER_NUMBER:			return HB_UNICODE_GENERAL_CATEGORY_OTHER_NUMBER;
    116 
    117   case U_SPACE_SEPARATOR:		return HB_UNICODE_GENERAL_CATEGORY_SPACE_SEPARATOR;
    118   case U_LINE_SEPARATOR:		return HB_UNICODE_GENERAL_CATEGORY_LINE_SEPARATOR;
    119   case U_PARAGRAPH_SEPARATOR:		return HB_UNICODE_GENERAL_CATEGORY_PARAGRAPH_SEPARATOR;
    120 
    121   case U_CONTROL_CHAR:			return HB_UNICODE_GENERAL_CATEGORY_CONTROL;
    122   case U_FORMAT_CHAR:			return HB_UNICODE_GENERAL_CATEGORY_FORMAT;
    123   case U_PRIVATE_USE_CHAR:		return HB_UNICODE_GENERAL_CATEGORY_PRIVATE_USE;
    124   case U_SURROGATE:			return HB_UNICODE_GENERAL_CATEGORY_SURROGATE;
    125 
    126 
    127   case U_DASH_PUNCTUATION:		return HB_UNICODE_GENERAL_CATEGORY_DASH_PUNCTUATION;
    128   case U_START_PUNCTUATION:		return HB_UNICODE_GENERAL_CATEGORY_OPEN_PUNCTUATION;
    129   case U_END_PUNCTUATION:		return HB_UNICODE_GENERAL_CATEGORY_CLOSE_PUNCTUATION;
    130   case U_CONNECTOR_PUNCTUATION:		return HB_UNICODE_GENERAL_CATEGORY_CONNECT_PUNCTUATION;
    131   case U_OTHER_PUNCTUATION:		return HB_UNICODE_GENERAL_CATEGORY_OTHER_PUNCTUATION;
    132 
    133   case U_MATH_SYMBOL:			return HB_UNICODE_GENERAL_CATEGORY_MATH_SYMBOL;
    134   case U_CURRENCY_SYMBOL:		return HB_UNICODE_GENERAL_CATEGORY_CURRENCY_SYMBOL;
    135   case U_MODIFIER_SYMBOL:		return HB_UNICODE_GENERAL_CATEGORY_MODIFIER_SYMBOL;
    136   case U_OTHER_SYMBOL:			return HB_UNICODE_GENERAL_CATEGORY_OTHER_SYMBOL;
    137 
    138   case U_INITIAL_PUNCTUATION:		return HB_UNICODE_GENERAL_CATEGORY_INITIAL_PUNCTUATION;
    139   case U_FINAL_PUNCTUATION:		return HB_UNICODE_GENERAL_CATEGORY_FINAL_PUNCTUATION;
    140   }
    141 
    142   return HB_UNICODE_GENERAL_CATEGORY_UNASSIGNED;
    143 }
    144 
    145 static hb_codepoint_t
    146 hb_icu_unicode_mirroring (hb_unicode_funcs_t *ufuncs HB_UNUSED,
    147 			  hb_codepoint_t      unicode,
    148 			  void               *user_data HB_UNUSED)
    149 {
    150   return u_charMirror(unicode);
    151 }
    152 
    153 static hb_script_t
    154 hb_icu_unicode_script (hb_unicode_funcs_t *ufuncs HB_UNUSED,
    155 		       hb_codepoint_t      unicode,
    156 		       void               *user_data HB_UNUSED)
    157 {
    158   UErrorCode status = U_ZERO_ERROR;
    159   UScriptCode scriptCode = uscript_getScript(unicode, &status);
    160 
    161   if (unlikely (U_FAILURE (status)))
    162     return HB_SCRIPT_UNKNOWN;
    163 
    164   return hb_icu_script_to_script (scriptCode);
    165 }
    166 
    167 #if U_ICU_VERSION_MAJOR_NUM >= 49
    168 static const UNormalizer2 *normalizer;
    169 #endif
    170 
    171 static hb_bool_t
    172 hb_icu_unicode_compose (hb_unicode_funcs_t *ufuncs HB_UNUSED,
    173 			hb_codepoint_t      a,
    174 			hb_codepoint_t      b,
    175 			hb_codepoint_t     *ab,
    176 			void               *user_data HB_UNUSED)
    177 {
    178 #if U_ICU_VERSION_MAJOR_NUM >= 49
    179   {
    180     UChar32 ret = unorm2_composePair (normalizer, a, b);
    181     if (ret < 0) return false;
    182     *ab = ret;
    183     return true;
    184   }
    185 #endif
    186 
    187   /* We don't ifdef-out the fallback code such that compiler always
    188    * sees it and makes sure it's compilable. */
    189 
    190   UChar utf16[4], normalized[5];
    191   unsigned int len;
    192   hb_bool_t ret, err;
    193   UErrorCode icu_err;
    194 
    195   len = 0;
    196   err = false;
    197   U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), a, err);
    198   if (err) return false;
    199   U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), b, err);
    200   if (err) return false;
    201 
    202   icu_err = U_ZERO_ERROR;
    203   len = unorm2_normalize (unorm2_getNFCInstance (&icu_err), utf16, len, normalized, ARRAY_LENGTH (normalized), &icu_err);
    204   if (U_FAILURE (icu_err))
    205     return false;
    206   if (u_countChar32 (normalized, len) == 1) {
    207     U16_GET_UNSAFE (normalized, 0, *ab);
    208     ret = true;
    209   } else {
    210     ret = false;
    211   }
    212 
    213   return ret;
    214 }
    215 
    216 static hb_bool_t
    217 hb_icu_unicode_decompose (hb_unicode_funcs_t *ufuncs HB_UNUSED,
    218 			  hb_codepoint_t      ab,
    219 			  hb_codepoint_t     *a,
    220 			  hb_codepoint_t     *b,
    221 			  void               *user_data HB_UNUSED)
    222 {
    223 #if U_ICU_VERSION_MAJOR_NUM >= 49
    224   {
    225     UChar decomposed[4];
    226     int len;
    227     UErrorCode icu_err = U_ZERO_ERROR;
    228     len = unorm2_getRawDecomposition (normalizer, ab, decomposed,
    229 				      ARRAY_LENGTH (decomposed), &icu_err);
    230     if (U_FAILURE (icu_err) || len < 0) return false;
    231 
    232     len = u_countChar32 (decomposed, len);
    233     if (len == 1) {
    234       U16_GET_UNSAFE (decomposed, 0, *a);
    235       *b = 0;
    236       return *a != ab;
    237     } else if (len == 2) {
    238       len =0;
    239       U16_NEXT_UNSAFE (decomposed, len, *a);
    240       U16_NEXT_UNSAFE (decomposed, len, *b);
    241     }
    242     return true;
    243   }
    244 #endif
    245 
    246   /* We don't ifdef-out the fallback code such that compiler always
    247    * sees it and makes sure it's compilable. */
    248 
    249   UChar utf16[2], normalized[2 * HB_UNICODE_MAX_DECOMPOSITION_LEN + 1];
    250   unsigned int len;
    251   hb_bool_t ret, err;
    252   UErrorCode icu_err;
    253 
    254   /* This function is a monster! Maybe it wasn't a good idea adding a
    255    * pairwise decompose API... */
    256   /* Watchout for the dragons.  Err, watchout for macros changing len. */
    257 
    258   len = 0;
    259   err = false;
    260   U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), ab, err);
    261   if (err) return false;
    262 
    263   icu_err = U_ZERO_ERROR;
    264   len = unorm2_normalize (unorm2_getNFDInstance (&icu_err), utf16, len, normalized, ARRAY_LENGTH (normalized), &icu_err);
    265   if (U_FAILURE (icu_err))
    266     return false;
    267 
    268   len = u_countChar32 (normalized, len);
    269 
    270   if (len == 1) {
    271     U16_GET_UNSAFE (normalized, 0, *a);
    272     *b = 0;
    273     ret = *a != ab;
    274   } else if (len == 2) {
    275     len =0;
    276     U16_NEXT_UNSAFE (normalized, len, *a);
    277     U16_NEXT_UNSAFE (normalized, len, *b);
    278 
    279     /* Here's the ugly part: if ab decomposes to a single character and
    280      * that character decomposes again, we have to detect that and undo
    281      * the second part :-(. */
    282     UChar recomposed[20];
    283     icu_err = U_ZERO_ERROR;
    284     unorm2_normalize (unorm2_getNFCInstance (&icu_err), normalized, len, recomposed, ARRAY_LENGTH (recomposed), &icu_err);
    285     if (U_FAILURE (icu_err))
    286       return false;
    287     hb_codepoint_t c;
    288     U16_GET_UNSAFE (recomposed, 0, c);
    289     if (c != *a && c != ab) {
    290       *a = c;
    291       *b = 0;
    292     }
    293     ret = true;
    294   } else {
    295     /* If decomposed to more than two characters, take the last one,
    296      * and recompose the rest to get the first component. */
    297     U16_PREV_UNSAFE (normalized, len, *b); /* Changes len in-place. */
    298     UChar recomposed[18 * 2];
    299     icu_err = U_ZERO_ERROR;
    300     len = unorm2_normalize (unorm2_getNFCInstance (&icu_err), normalized, len, recomposed, ARRAY_LENGTH (recomposed), &icu_err);
    301     if (U_FAILURE (icu_err))
    302       return false;
    303     /* We expect that recomposed has exactly one character now. */
    304     if (unlikely (u_countChar32 (recomposed, len) != 1))
    305       return false;
    306     U16_GET_UNSAFE (recomposed, 0, *a);
    307     ret = true;
    308   }
    309 
    310   return ret;
    311 }
    312 
    313 static unsigned int
    314 hb_icu_unicode_decompose_compatibility (hb_unicode_funcs_t *ufuncs HB_UNUSED,
    315 					hb_codepoint_t      u,
    316 					hb_codepoint_t     *decomposed,
    317 					void               *user_data HB_UNUSED)
    318 {
    319   UChar utf16[2], normalized[2 * HB_UNICODE_MAX_DECOMPOSITION_LEN + 1];
    320   unsigned int len;
    321   int32_t utf32_len;
    322   hb_bool_t err;
    323   UErrorCode icu_err;
    324 
    325   /* Copy @u into a UTF-16 array to be passed to ICU. */
    326   len = 0;
    327   err = false;
    328   U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), u, err);
    329   if (err)
    330     return 0;
    331 
    332   /* Normalise the codepoint using NFKD mode. */
    333   icu_err = U_ZERO_ERROR;
    334   len = unorm2_normalize (unorm2_getNFKDInstance (&icu_err), utf16, len, normalized, ARRAY_LENGTH (normalized), &icu_err);
    335   if (U_FAILURE (icu_err))
    336     return 0;
    337 
    338   /* Convert the decomposed form from UTF-16 to UTF-32. */
    339   icu_err = U_ZERO_ERROR;
    340   u_strToUTF32 ((UChar32*) decomposed, HB_UNICODE_MAX_DECOMPOSITION_LEN, &utf32_len, normalized, len, &icu_err);
    341   if (U_FAILURE (icu_err))
    342     return 0;
    343 
    344   return utf32_len;
    345 }
    346 
    347 
    348 static hb_unicode_funcs_t *static_icu_funcs = nullptr;
    349 
    350 #ifdef HB_USE_ATEXIT
    351 static
    352 void free_static_icu_funcs (void)
    353 {
    354   hb_unicode_funcs_destroy (static_icu_funcs);
    355 }
    356 #endif
    357 
    358 hb_unicode_funcs_t *
    359 hb_icu_get_unicode_funcs (void)
    360 {
    361 retry:
    362   hb_unicode_funcs_t *funcs = (hb_unicode_funcs_t *) hb_atomic_ptr_get (&static_icu_funcs);
    363 
    364   if (unlikely (!funcs))
    365   {
    366 #if U_ICU_VERSION_MAJOR_NUM >= 49
    367     if (!hb_atomic_ptr_get (&normalizer)) {
    368       UErrorCode icu_err = U_ZERO_ERROR;
    369       /* We ignore failure in getNFCInstace(). */
    370       (void) hb_atomic_ptr_cmpexch (&normalizer, nullptr, unorm2_getNFCInstance (&icu_err));
    371     }
    372 #endif
    373 
    374     funcs = hb_unicode_funcs_create (nullptr);
    375 
    376 #define HB_UNICODE_FUNC_IMPLEMENT(name) \
    377     hb_unicode_funcs_set_##name##_func (funcs, hb_icu_unicode_##name, nullptr, nullptr);
    378       HB_UNICODE_FUNCS_IMPLEMENT_CALLBACKS
    379 #undef HB_UNICODE_FUNC_IMPLEMENT
    380 
    381     hb_unicode_funcs_make_immutable (funcs);
    382 
    383     if (!hb_atomic_ptr_cmpexch (&static_icu_funcs, nullptr, funcs)) {
    384       hb_unicode_funcs_destroy (funcs);
    385       goto retry;
    386     }
    387 
    388 #ifdef HB_USE_ATEXIT
    389     atexit (free_static_icu_funcs); /* First person registers atexit() callback. */
    390 #endif
    391   };
    392 
    393   return hb_unicode_funcs_reference (funcs);
    394 }
    395