third_party/harfbuzz-ng/src/hb-ot-shape-complex-thai.cc

   1 /*
   2  * Copyright © 2010,2012  Google, Inc.
   3  *
   4  *  This is part of HarfBuzz, a text shaping library.
   5  *
   6  * Permission is hereby granted, without written agreement and without
   7  * license or royalty fees, to use, copy, modify, and distribute this
   8  * software and its documentation for any purpose, provided that the
   9  * above copyright notice and the following two paragraphs appear in
  10  * all copies of this software.
  11  *
  12  * IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE TO ANY PARTY FOR
  13  * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES
  14  * ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN
  15  * IF THE COPYRIGHT HOLDER HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
  16  * DAMAGE.
  17  *
  18  * THE COPYRIGHT HOLDER SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING,
  19  * BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
  20  * FITNESS FOR A PARTICULAR PURPOSE.  THE SOFTWARE PROVIDED HEREUNDER IS
  21  * ON AN "AS IS" BASIS, AND THE COPYRIGHT HOLDER HAS NO OBLIGATION TO
  22  * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
  23  *
  24  * Google Author(s): Behdad Esfahbod
  25  */
  26
  27 #include "hb-ot-shape-complex-private.hh"
  28
  29
  30 /* Thai / Lao shaper */
  31
  32
  33 /* PUA shaping */
  34
  35
  36 enum thai_consonant_type_t
  37 {
  38   NC,
  39   AC,
  40   RC,
  41   DC,
  42   NOT_CONSONANT,
  43   NUM_CONSONANT_TYPES = NOT_CONSONANT
  44 };
  45
  46 static thai_consonant_type_t
  47 get_consonant_type (hb_codepoint_t u)
  48 {
  49   if (u == 0x0E1Bu || u == 0x0E1Du || u == 0x0E1Fu/* || u == 0x0E2Cu*/)
  50     return AC;
  51   if (u == 0x0E0Du || u == 0x0E10u)
  52     return RC;
  53   if (u == 0x0E0Eu || u == 0x0E0Fu)
  54     return DC;
  55   if (hb_in_range (u, 0x0E01u, 0x0E2Eu))
  56     return NC;
  57   return NOT_CONSONANT;
  58 }
  59
  60
  61 enum thai_mark_type_t
  62 {
  63   AV,
  64   BV,
  65   T,
  66   NOT_MARK,
  67   NUM_MARK_TYPES = NOT_MARK
  68 };
  69
  70 static thai_mark_type_t
  71 get_mark_type (hb_codepoint_t u)
  72 {
  73   if (u == 0x0E31u || hb_in_range (u, 0x0E34u, 0x0E37u) ||
  74       u == 0x0E47u || hb_in_range (u, 0x0E4Du, 0x0E4Eu))
  75     return AV;
  76   if (hb_in_range (u, 0x0E38u, 0x0E3Au))
  77     return BV;
  78   if (hb_in_range (u, 0x0E48u, 0x0E4Cu))
  79     return T;
  80   return NOT_MARK;
  81 }
  82
  83
  84 enum thai_action_t
  85 {
  86   NOP,
  87   SD,  /* Shift combining-mark down */
  88   SL,  /* Shift combining-mark left */
  89   SDL, /* Shift combining-mark down-left */
  90   RD   /* Remove descender from base */
  91 };
  92
  93 static hb_codepoint_t
  94 thai_pua_shape (hb_codepoint_t u, thai_action_t action, hb_font_t *font)
  95 {
  96   struct thai_pua_mapping_t {
  97     hb_codepoint_t u;
  98     hb_codepoint_t win_pua;
  99     hb_codepoint_t mac_pua;
 100   } const *pua_mappings = NULL;
 101   static const thai_pua_mapping_t SD_mappings[] = {
 102     {0x0E48u, 0xF70Au, 0xF88Bu}, /* MAI EK */
 103     {0x0E49u, 0xF70Bu, 0xF88Eu}, /* MAI THO */
 104     {0x0E4Au, 0xF70Cu, 0xF891u}, /* MAI TRI */
 105     {0x0E4Bu, 0xF70Du, 0xF894u}, /* MAI CHATTAWA */
 106     {0x0E4Cu, 0xF70Eu, 0xF897u}, /* THANTHAKHAT */
 107     {0x0E38u, 0xF718u, 0xF89Bu}, /* SARA U */
 108     {0x0E39u, 0xF719u, 0xF89Cu}, /* SARA UU */
 109     {0x0E3Au, 0xF71Au, 0xF89Du}, /* PHINTHU */
 110     {0x0000u, 0x0000u, 0x0000u}
 111   };
 112   static const thai_pua_mapping_t SDL_mappings[] = {
 113     {0x0E48u, 0xF705u, 0xF88Cu}, /* MAI EK */
 114     {0x0E49u, 0xF706u, 0xF88Fu}, /* MAI THO */
 115     {0x0E4Au, 0xF707u, 0xF892u}, /* MAI TRI */
 116     {0x0E4Bu, 0xF708u, 0xF895u}, /* MAI CHATTAWA */
 117     {0x0E4Cu, 0xF709u, 0xF898u}, /* THANTHAKHAT */
 118     {0x0000u, 0x0000u, 0x0000u}
 119   };
 120   static const thai_pua_mapping_t SL_mappings[] = {
 121     {0x0E48u, 0xF713u, 0xF88Au}, /* MAI EK */
 122     {0x0E49u, 0xF714u, 0xF88Du}, /* MAI THO */
 123     {0x0E4Au, 0xF715u, 0xF890u}, /* MAI TRI */
 124     {0x0E4Bu, 0xF716u, 0xF893u}, /* MAI CHATTAWA */
 125     {0x0E4Cu, 0xF717u, 0xF896u}, /* THANTHAKHAT */
 126     {0x0E31u, 0xF710u, 0xF884u}, /* MAI HAN-AKAT */
 127     {0x0E34u, 0xF701u, 0xF885u}, /* SARA I */
 128     {0x0E35u, 0xF702u, 0xF886u}, /* SARA II */
 129     {0x0E36u, 0xF703u, 0xF887u}, /* SARA UE */
 130     {0x0E37u, 0xF704u, 0xF888u}, /* SARA UEE */
 131     {0x0E47u, 0xF712u, 0xF889u}, /* MAITAIKHU */
 132     {0x0E4Du, 0xF711u, 0xF899u}, /* NIKHAHIT */
 133     {0x0000u, 0x0000u, 0x0000u}
 134   };
 135   static const thai_pua_mapping_t RD_mappings[] = {
 136     {0x0E0Du, 0xF70Fu, 0xF89Au}, /* YO YING */
 137     {0x0E10u, 0xF700u, 0xF89Eu}, /* THO THAN */
 138     {0x0000u, 0x0000u, 0x0000u}
 139   };
 140
 141   switch (action) {
 142     default: assert (false); /* Fallthrough */
 143     case NOP: return u;
 144     case SD:  pua_mappings = SD_mappings; break;
 145     case SDL: pua_mappings = SDL_mappings; break;
 146     case SL:  pua_mappings = SL_mappings; break;
 147     case RD:  pua_mappings = RD_mappings; break;
 148   }
 149   for (; pua_mappings->u; pua_mappings++)
 150     if (pua_mappings->u == u)
 151     {
 152       hb_codepoint_t glyph;
 153       if (hb_font_get_glyph (font, pua_mappings->win_pua, 0, &glyph))
 154         return pua_mappings->win_pua;
 155       if (hb_font_get_glyph (font, pua_mappings->mac_pua, 0, &glyph))
 156         return pua_mappings->mac_pua;
 157       break;
 158     }
 159   return u;
 160 }
 161
 162
 163 static enum thai_above_state_t
 164 {     /* Cluster above looks like: */
 165   T0, /*  ⣤                      */
 166   T1, /*     ⣼                   */
 167   T2, /*        ⣾                */
 168   T3, /*           ⣿             */
 169   NUM_ABOVE_STATES
 170 } thai_above_start_state[NUM_CONSONANT_TYPES + 1/* For NOT_CONSONANT */] =
 171 {
 172   T0, /* NC */
 173   T1, /* AC */
 174   T0, /* RC */
 175   T0, /* DC */
 176   T3, /* NOT_CONSONANT */
 177 };
 178
 179 static const struct thai_above_state_machine_edge_t {
 180   thai_action_t action;
 181   thai_above_state_t next_state;
 182 } thai_above_state_machine[NUM_ABOVE_STATES][NUM_MARK_TYPES] =
 183 {        /*AV*/    /*BV*/    /*T*/
 184 /*T0*/ {{NOP,T3}, {NOP,T0}, {SD, T3}},
 185 /*T1*/ {{SL, T2}, {NOP,T1}, {SDL,T2}},
 186 /*T2*/ {{NOP,T3}, {NOP,T2}, {SL, T3}},
 187 /*T3*/ {{NOP,T3}, {NOP,T3}, {NOP,T3}},
 188 };
 189
 190
 191 static enum thai_below_state_t
 192 {
 193   B0, /* No descender */
 194   B1, /* Removable descender */
 195   B2, /* Strict descender */
 196   NUM_BELOW_STATES
 197 } thai_below_start_state[NUM_CONSONANT_TYPES + 1/* For NOT_CONSONANT */] =
 198 {
 199   B0, /* NC */
 200   B0, /* AC */
 201   B1, /* RC */
 202   B2, /* DC */
 203   B2, /* NOT_CONSONANT */
 204 };
 205
 206 static const struct thai_below_state_machine_edge_t {
 207   thai_action_t action;
 208   thai_below_state_t next_state;
 209 } thai_below_state_machine[NUM_BELOW_STATES][NUM_MARK_TYPES] =
 210 {        /*AV*/    /*BV*/    /*T*/
 211 /*B0*/ {{NOP,B0}, {NOP,B2}, {NOP, B0}},
 212 /*B1*/ {{NOP,B1}, {RD, B2}, {NOP, B1}},
 213 /*B2*/ {{NOP,B2}, {SD, B2}, {NOP, B2}},
 214 };
 215
 216
 217 static void
 218 do_thai_pua_shaping (const hb_ot_shape_plan_t *plan HB_UNUSED,
 219                      hb_buffer_t              *buffer,
 220                      hb_font_t                *font)
 221 {
 222   thai_above_state_t above_state = thai_above_start_state[NOT_CONSONANT];
 223   thai_below_state_t below_state = thai_below_start_state[NOT_CONSONANT];
 224   unsigned int base = 0;
 225
 226   hb_glyph_info_t *info = buffer->info;
 227   unsigned int count = buffer->len;
 228   for (unsigned int i = 0; i < count; i++)
 229   {
 230     thai_mark_type_t mt = get_mark_type (info[i].codepoint);
 231
 232     if (mt == NOT_MARK) {
 233       thai_consonant_type_t ct = get_consonant_type (info[i].codepoint);
 234       above_state = thai_above_start_state[ct];
 235       below_state = thai_below_start_state[ct];
 236       base = i;
 237       continue;
 238     }
 239
 240     const thai_above_state_machine_edge_t &above_edge = thai_above_state_machine[above_state][mt];
 241     const thai_below_state_machine_edge_t &below_edge = thai_below_state_machine[below_state][mt];
 242     above_state = above_edge.next_state;
 243     below_state = below_edge.next_state;
 244
 245     /* At least one of the above/below actions is NOP. */
 246     thai_action_t action = above_edge.action != NOP ? above_edge.action : below_edge.action;
 247
 248     if (action == RD)
 249       info[base].codepoint = thai_pua_shape (info[base].codepoint, action, font);
 250     else
 251       info[i].codepoint = thai_pua_shape (info[i].codepoint, action, font);
 252   }
 253 }
 254
 255
 256 static void
 257 preprocess_text_thai (const hb_ot_shape_plan_t *plan,
 258                       hb_buffer_t              *buffer,
 259                       hb_font_t                *font)
 260 {
 261   /* This function implements the shaping logic documented here:
 262    *
 263    *   http://linux.thai.net/~thep/th-otf/shaping.html
 264    *
 265    * The first shaping rule listed there is needed even if the font has Thai
 266    * OpenType tables.  The rest do fallback positioning based on PUA codepoints.
 267    * We implement that only if there exist no Thai GSUB in the font.
 268    */
 269
 270   /* The following is NOT specified in the MS OT Thai spec, however, it seems
 271    * to be what Uniscribe and other engines implement.  According to Eric Muller:
 272    *
 273    * When you have a SARA AM, decompose it in NIKHAHIT + SARA AA, *and* move the
 274    * NIKHAHIT backwards over any tone mark (0E48-0E4B).
 275    *
 276    * <0E14, 0E4B, 0E33> -> <0E14, 0E4D, 0E4B, 0E32>
 277    *
 278    * This reordering is legit only when the NIKHAHIT comes from a SARA AM, not
 279    * when it's there to start with. The string <0E14, 0E4B, 0E4D> is probably
 280    * not what a user wanted, but the rendering is nevertheless nikhahit above
 281    * chattawa.
 282    *
 283    * Same for Lao.
 284    *
 285    * Note:
 286    *
 287    * Uniscribe also does some below-marks reordering.  Namely, it positions U+0E3A
 288    * after U+0E38 and U+0E39.  We do that by modifying the ccc for U+0E3A.
 289    * See unicode->modified_combining_class ().  Lao does NOT have a U+0E3A
 290    * equivalent.
 291    */
 292
 293
 294   /*
 295    * Here are the characters of significance:
 296    *
 297    *                    Thai    Lao
 298    * SARA AM:           U+0E33  U+0EB3
 299    * SARA AA:           U+0E32  U+0EB2
 300    * Nikhahit:          U+0E4D  U+0ECD
 301    *
 302    * Testing shows that Uniscribe reorder the following marks:
 303    * Thai:      <0E31,0E34..0E37,0E47..0E4E>
 304    * Lao:       <0EB1,0EB4..0EB7,0EC7..0ECE>
 305    *
 306    * Note how the Lao versions are the same as Thai + 0x80.
 307    */
 308
 309   /* We only get one script at a time, so a script-agnostic implementation
 310    * is adequate here. */
 311 #define IS_SARA_AM(x) (((x) & ~0x0080u) == 0x0E33u)
 312 #define NIKHAHIT_FROM_SARA_AM(x) ((x) - 0x0E33u + 0x0E4Du)
 313 #define SARA_AA_FROM_SARA_AM(x) ((x) - 1)
 314 #define IS_TONE_MARK(x) (hb_in_ranges ((x) & ~0x0080u, 0x0E34u, 0x0E37u, 0x0E47u, 0x0E4Eu, 0x0E31u, 0x0E31u))
 315
 316   buffer->clear_output ();
 317   unsigned int count = buffer->len;
 318   for (buffer->idx = 0; buffer->idx < count;)
 319   {
 320     hb_codepoint_t u = buffer->cur().codepoint;
 321     if (likely (!IS_SARA_AM (u))) {
 322       buffer->next_glyph ();
 323       continue;
 324     }
 325
 326     /* Is SARA AM. Decompose and reorder. */
 327     hb_codepoint_t decomposed[2] = {hb_codepoint_t (NIKHAHIT_FROM_SARA_AM (u)),
 328                                     hb_codepoint_t (SARA_AA_FROM_SARA_AM (u))};
 329     buffer->replace_glyphs (1, 2, decomposed);
 330     if (unlikely (buffer->in_error))
 331       return;
 332
 333     /* Make Nikhahit be recognized as a mark when zeroing widths. */
 334     unsigned int end = buffer->out_len;
 335     _hb_glyph_info_set_general_category (&buffer->out_info[end - 2], HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK);
 336
 337     /* Ok, let's see... */
 338     unsigned int start = end - 2;
 339     while (start > 0 && IS_TONE_MARK (buffer->out_info[start - 1].codepoint))
 340       start--;
 341
 342     if (start + 2 < end)
 343     {
 344       /* Move Nikhahit (end-2) to the beginning */
 345       buffer->merge_out_clusters (start, end);
 346       hb_glyph_info_t t = buffer->out_info[end - 2];
 347       memmove (buffer->out_info + start + 1,
 348                buffer->out_info + start,
 349                sizeof (buffer->out_info[0]) * (end - start - 2));
 350       buffer->out_info[start] = t;
 351     }
 352     else
 353     {
 354       /* Since we decomposed, and NIKHAHIT is combining, merge clusters with the
 355        * previous cluster. */
 356       if (start)
 357         buffer->merge_out_clusters (start - 1, end);
 358     }
 359   }
 360   buffer->swap_buffers ();
 361
 362   /* If font has Thai GSUB, we are done. */
 363   if (plan->props.script == HB_SCRIPT_THAI && !plan->map.found_script[0])
 364     do_thai_pua_shaping (plan, buffer, font);
 365 }
 366
 367 const hb_ot_complex_shaper_t _hb_ot_complex_shaper_thai =
 368 {
 369   "thai",
 370   NULL, /* collect_features */
 371   NULL, /* override_features */
 372   NULL, /* data_create */
 373   NULL, /* data_destroy */
 374   preprocess_text_thai,
 375   HB_OT_SHAPE_NORMALIZATION_MODE_DEFAULT,
 376   NULL, /* decompose */
 377   NULL, /* compose */
 378   NULL, /* setup_masks */
 379   HB_OT_SHAPE_ZERO_WIDTH_MARKS_DEFAULT,
 380   false,/* fallback_position */
 381 };