gfx/harfbuzz/src/hb-ot-shaper-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.hh"
  28
  29 #ifndef HB_NO_OT_SHAPE
  30
  31 #include "hb-ot-shaper.hh"
  32
  33
  34 /* Thai / Lao shaper */
  35
  36
  37 /* PUA shaping */
  38
  39
  40 enum thai_consonant_type_t
  41 {
  42   NC,
  43   AC,
  44   RC,
  45   DC,
  46   NOT_CONSONANT,
  47   NUM_CONSONANT_TYPES = NOT_CONSONANT
  48 };
  49
  50 static thai_consonant_type_t
  51 get_consonant_type (hb_codepoint_t u)
  52 {
  53   if (u == 0x0E1Bu || u == 0x0E1Du || u == 0x0E1Fu/* || u == 0x0E2Cu*/)
  54     return AC;
  55   if (u == 0x0E0Du || u == 0x0E10u)
  56     return RC;
  57   if (u == 0x0E0Eu || u == 0x0E0Fu)
  58     return DC;
  59   if (hb_in_range<hb_codepoint_t> (u, 0x0E01u, 0x0E2Eu))
  60     return NC;
  61   return NOT_CONSONANT;
  62 }
  63
  64
  65 enum thai_mark_type_t
  66 {
  67   AV,
  68   BV,
  69   T,
  70   NOT_MARK,
  71   NUM_MARK_TYPES = NOT_MARK
  72 };
  73
  74 static thai_mark_type_t
  75 get_mark_type (hb_codepoint_t u)
  76 {
  77   if (u == 0x0E31u || hb_in_range<hb_codepoint_t> (u, 0x0E34u, 0x0E37u) ||
  78       u == 0x0E47u || hb_in_range<hb_codepoint_t> (u, 0x0E4Du, 0x0E4Eu))
  79     return AV;
  80   if (hb_in_range<hb_codepoint_t> (u, 0x0E38u, 0x0E3Au))
  81     return BV;
  82   if (hb_in_range<hb_codepoint_t> (u, 0x0E48u, 0x0E4Cu))
  83     return T;
  84   return NOT_MARK;
  85 }
  86
  87
  88 enum thai_action_t
  89 {
  90   NOP,
  91   SD,  /* Shift combining-mark down */
  92   SL,  /* Shift combining-mark left */
  93   SDL, /* Shift combining-mark down-left */
  94   RD   /* Remove descender from base */
  95 };
  96
  97 static hb_codepoint_t
  98 thai_pua_shape (hb_codepoint_t u, thai_action_t action, hb_font_t *font)
  99 {
 100   struct thai_pua_mapping_t {
 101     uint16_t u;
 102     uint16_t win_pua;
 103     uint16_t mac_pua;
 104   } const *pua_mappings = nullptr;
 105   static const thai_pua_mapping_t SD_mappings[] = {
 106     {0x0E48u, 0xF70Au, 0xF88Bu}, /* MAI EK */
 107     {0x0E49u, 0xF70Bu, 0xF88Eu}, /* MAI THO */
 108     {0x0E4Au, 0xF70Cu, 0xF891u}, /* MAI TRI */
 109     {0x0E4Bu, 0xF70Du, 0xF894u}, /* MAI CHATTAWA */
 110     {0x0E4Cu, 0xF70Eu, 0xF897u}, /* THANTHAKHAT */
 111     {0x0E38u, 0xF718u, 0xF89Bu}, /* SARA U */
 112     {0x0E39u, 0xF719u, 0xF89Cu}, /* SARA UU */
 113     {0x0E3Au, 0xF71Au, 0xF89Du}, /* PHINTHU */
 114     {0x0000u, 0x0000u, 0x0000u}
 115   };
 116   static const thai_pua_mapping_t SDL_mappings[] = {
 117     {0x0E48u, 0xF705u, 0xF88Cu}, /* MAI EK */
 118     {0x0E49u, 0xF706u, 0xF88Fu}, /* MAI THO */
 119     {0x0E4Au, 0xF707u, 0xF892u}, /* MAI TRI */
 120     {0x0E4Bu, 0xF708u, 0xF895u}, /* MAI CHATTAWA */
 121     {0x0E4Cu, 0xF709u, 0xF898u}, /* THANTHAKHAT */
 122     {0x0000u, 0x0000u, 0x0000u}
 123   };
 124   static const thai_pua_mapping_t SL_mappings[] = {
 125     {0x0E48u, 0xF713u, 0xF88Au}, /* MAI EK */
 126     {0x0E49u, 0xF714u, 0xF88Du}, /* MAI THO */
 127     {0x0E4Au, 0xF715u, 0xF890u}, /* MAI TRI */
 128     {0x0E4Bu, 0xF716u, 0xF893u}, /* MAI CHATTAWA */
 129     {0x0E4Cu, 0xF717u, 0xF896u}, /* THANTHAKHAT */
 130     {0x0E31u, 0xF710u, 0xF884u}, /* MAI HAN-AKAT */
 131     {0x0E34u, 0xF701u, 0xF885u}, /* SARA I */
 132     {0x0E35u, 0xF702u, 0xF886u}, /* SARA II */
 133     {0x0E36u, 0xF703u, 0xF887u}, /* SARA UE */
 134     {0x0E37u, 0xF704u, 0xF888u}, /* SARA UEE */
 135     {0x0E47u, 0xF712u, 0xF889u}, /* MAITAIKHU */
 136     {0x0E4Du, 0xF711u, 0xF899u}, /* NIKHAHIT */
 137     {0x0000u, 0x0000u, 0x0000u}
 138   };
 139   static const thai_pua_mapping_t RD_mappings[] = {
 140     {0x0E0Du, 0xF70Fu, 0xF89Au}, /* YO YING */
 141     {0x0E10u, 0xF700u, 0xF89Eu}, /* THO THAN */
 142     {0x0000u, 0x0000u, 0x0000u}
 143   };
 144
 145   switch (action) {
 146     case NOP: return u;
 147     case SD:  pua_mappings = SD_mappings; break;
 148     case SDL: pua_mappings = SDL_mappings; break;
 149     case SL:  pua_mappings = SL_mappings; break;
 150     case RD:  pua_mappings = RD_mappings; break;
 151   }
 152   for (; pua_mappings->u; pua_mappings++)
 153     if (pua_mappings->u == u)
 154     {
 155       hb_codepoint_t glyph;
 156       if (hb_font_get_glyph (font, pua_mappings->win_pua, 0, &glyph))
 157         return pua_mappings->win_pua;
 158       if (hb_font_get_glyph (font, pua_mappings->mac_pua, 0, &glyph))
 159         return pua_mappings->mac_pua;
 160       break;
 161     }
 162   return u;
 163 }
 164
 165
 166 static enum thai_above_state_t
 167 {     /* Cluster above looks like: */
 168   T0, /*  ⣤                      */
 169   T1, /*     ⣼                   */
 170   T2, /*        ⣾                */
 171   T3, /*           ⣿             */
 172   NUM_ABOVE_STATES
 173 } thai_above_start_state[NUM_CONSONANT_TYPES + 1/* For NOT_CONSONANT */] =
 174 {
 175   T0, /* NC */
 176   T1, /* AC */
 177   T0, /* RC */
 178   T0, /* DC */
 179   T3, /* NOT_CONSONANT */
 180 };
 181
 182 static const struct thai_above_state_machine_edge_t {
 183   thai_action_t action;
 184   thai_above_state_t next_state;
 185 } thai_above_state_machine[NUM_ABOVE_STATES][NUM_MARK_TYPES] =
 186 {        /*AV*/    /*BV*/    /*T*/
 187 /*T0*/ {{NOP,T3}, {NOP,T0}, {SD, T3}},
 188 /*T1*/ {{SL, T2}, {NOP,T1}, {SDL,T2}},
 189 /*T2*/ {{NOP,T3}, {NOP,T2}, {SL, T3}},
 190 /*T3*/ {{NOP,T3}, {NOP,T3}, {NOP,T3}},
 191 };
 192
 193
 194 static enum thai_below_state_t
 195 {
 196   B0, /* No descender */
 197   B1, /* Removable descender */
 198   B2, /* Strict descender */
 199   NUM_BELOW_STATES
 200 } thai_below_start_state[NUM_CONSONANT_TYPES + 1/* For NOT_CONSONANT */] =
 201 {
 202   B0, /* NC */
 203   B0, /* AC */
 204   B1, /* RC */
 205   B2, /* DC */
 206   B2, /* NOT_CONSONANT */
 207 };
 208
 209 static const struct thai_below_state_machine_edge_t {
 210   thai_action_t action;
 211   thai_below_state_t next_state;
 212 } thai_below_state_machine[NUM_BELOW_STATES][NUM_MARK_TYPES] =
 213 {        /*AV*/    /*BV*/    /*T*/
 214 /*B0*/ {{NOP,B0}, {NOP,B2}, {NOP, B0}},
 215 /*B1*/ {{NOP,B1}, {RD, B2}, {NOP, B1}},
 216 /*B2*/ {{NOP,B2}, {SD, B2}, {NOP, B2}},
 217 };
 218
 219
 220 static void
 221 do_thai_pua_shaping (const hb_ot_shape_plan_t *plan HB_UNUSED,
 222                      hb_buffer_t              *buffer,
 223                      hb_font_t                *font)
 224 {
 225 #ifdef HB_NO_OT_SHAPER_THAI_FALLBACK
 226   return;
 227 #endif
 228
 229   thai_above_state_t above_state = thai_above_start_state[NOT_CONSONANT];
 230   thai_below_state_t below_state = thai_below_start_state[NOT_CONSONANT];
 231   unsigned int base = 0;
 232
 233   hb_glyph_info_t *info = buffer->info;
 234   unsigned int count = buffer->len;
 235   for (unsigned int i = 0; i < count; i++)
 236   {
 237     thai_mark_type_t mt = get_mark_type (info[i].codepoint);
 238
 239     if (mt == NOT_MARK) {
 240       thai_consonant_type_t ct = get_consonant_type (info[i].codepoint);
 241       above_state = thai_above_start_state[ct];
 242       below_state = thai_below_start_state[ct];
 243       base = i;
 244       continue;
 245     }
 246
 247     const thai_above_state_machine_edge_t &above_edge = thai_above_state_machine[above_state][mt];
 248     const thai_below_state_machine_edge_t &below_edge = thai_below_state_machine[below_state][mt];
 249     above_state = above_edge.next_state;
 250     below_state = below_edge.next_state;
 251
 252     /* At least one of the above/below actions is NOP. */
 253     thai_action_t action = above_edge.action != NOP ? above_edge.action : below_edge.action;
 254
 255     buffer->unsafe_to_break (base, i);
 256     if (action == RD)
 257       info[base].codepoint = thai_pua_shape (info[base].codepoint, action, font);
 258     else
 259       info[i].codepoint = thai_pua_shape (info[i].codepoint, action, font);
 260   }
 261 }
 262
 263
 264 static void
 265 preprocess_text_thai (const hb_ot_shape_plan_t *plan,
 266                       hb_buffer_t              *buffer,
 267                       hb_font_t                *font)
 268 {
 269   /* This function implements the shaping logic documented here:
 270    *
 271    *   https://linux.thai.net/~thep/th-otf/shaping.html
 272    *
 273    * The first shaping rule listed there is needed even if the font has Thai
 274    * OpenType tables.  The rest do fallback positioning based on PUA codepoints.
 275    * We implement that only if there exist no Thai GSUB in the font.
 276    */
 277
 278   /* The following is NOT specified in the MS OT Thai spec, however, it seems
 279    * to be what Uniscribe and other engines implement.  According to Eric Muller:
 280    *
 281    * When you have a SARA AM, decompose it in NIKHAHIT + SARA AA, *and* move the
 282    * NIKHAHIT backwards over any above-base marks.
 283    *
 284    * <0E14, 0E4B, 0E33> -> <0E14, 0E4D, 0E4B, 0E32>
 285    *
 286    * This reordering is legit only when the NIKHAHIT comes from a SARA AM, not
 287    * when it's there to start with. The string <0E14, 0E4B, 0E4D> is probably
 288    * not what a user wanted, but the rendering is nevertheless nikhahit above
 289    * chattawa.
 290    *
 291    * Same for Lao.
 292    *
 293    * Note:
 294    *
 295    * Uniscribe also does some below-marks reordering.  Namely, it positions U+0E3A
 296    * after U+0E38 and U+0E39.  We do that by modifying the ccc for U+0E3A.
 297    * See unicode->modified_combining_class ().  Lao does NOT have a U+0E3A
 298    * equivalent.
 299    */
 300
 301
 302   /*
 303    * Here are the characters of significance:
 304    *
 305    *                    Thai    Lao
 306    * SARA AM:           U+0E33  U+0EB3
 307    * SARA AA:           U+0E32  U+0EB2
 308    * Nikhahit:          U+0E4D  U+0ECD
 309    *
 310    * Testing shows that Uniscribe reorder the following marks:
 311    * Thai:      <0E31,0E34..0E37,     0E47..0E4E>
 312    * Lao:       <0EB1,0EB4..0EB7,0EBB,0EC8..0ECD>
 313    *
 314    * Note how the Lao versions are the same as Thai + 0x80.
 315    */
 316
 317   /* We only get one script at a time, so a script-agnostic implementation
 318    * is adequate here. */
 319 #define IS_SARA_AM(x) (((x) & ~0x0080u) == 0x0E33u)
 320 #define NIKHAHIT_FROM_SARA_AM(x) ((x) - 0x0E33u + 0x0E4Du)
 321 #define SARA_AA_FROM_SARA_AM(x) ((x) - 1)
 322 #define IS_ABOVE_BASE_MARK(x) (hb_in_ranges<hb_codepoint_t> ((x) & ~0x0080u, 0x0E34u, 0x0E37u, 0x0E47u, 0x0E4Eu, 0x0E31u, 0x0E31u, 0x0E3Bu, 0x0E3Bu))
 323
 324   buffer->clear_output ();
 325   unsigned int count = buffer->len;
 326   for (buffer->idx = 0; buffer->idx < count /* No need for: && buffer->successful */;)
 327   {
 328     hb_codepoint_t u = buffer->cur().codepoint;
 329     if (likely (!IS_SARA_AM (u)))
 330     {
 331       if (unlikely (!buffer->next_glyph ())) break;
 332       continue;
 333     }
 334
 335     /* Is SARA AM. Decompose and reorder. */
 336     (void) buffer->output_glyph (NIKHAHIT_FROM_SARA_AM (u));
 337     _hb_glyph_info_set_continuation (&buffer->prev());
 338     if (unlikely (!buffer->replace_glyph (SARA_AA_FROM_SARA_AM (u)))) break;
 339
 340     /* Make Nikhahit be recognized as a ccc=0 mark when zeroing widths. */
 341     unsigned int end = buffer->out_len;
 342     _hb_glyph_info_set_general_category (&buffer->out_info[end - 2], HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK);
 343
 344     /* Ok, let's see... */
 345     unsigned int start = end - 2;
 346     while (start > 0 && IS_ABOVE_BASE_MARK (buffer->out_info[start - 1].codepoint))
 347       start--;
 348
 349     if (start + 2 < end)
 350     {
 351       /* Move Nikhahit (end-2) to the beginning */
 352       buffer->merge_out_clusters (start, end);
 353       hb_glyph_info_t t = buffer->out_info[end - 2];
 354       memmove (buffer->out_info + start + 1,
 355                buffer->out_info + start,
 356                sizeof (buffer->out_info[0]) * (end - start - 2));
 357       buffer->out_info[start] = t;
 358     }
 359     else
 360     {
 361       /* Since we decomposed, and NIKHAHIT is combining, merge clusters with the
 362        * previous cluster. */
 363       if (start && buffer->cluster_level == HB_BUFFER_CLUSTER_LEVEL_MONOTONE_GRAPHEMES)
 364         buffer->merge_out_clusters (start - 1, end);
 365     }
 366   }
 367   buffer->sync ();
 368
 369   /* If font has Thai GSUB, we are done. */
 370   if (plan->props.script == HB_SCRIPT_THAI && !plan->map.found_script[0])
 371     do_thai_pua_shaping (plan, buffer, font);
 372 }
 373
 374 const hb_ot_shaper_t _hb_ot_shaper_thai =
 375 {
 376   nullptr, /* collect_features */
 377   nullptr, /* override_features */
 378   nullptr, /* data_create */
 379   nullptr, /* data_destroy */
 380   preprocess_text_thai,
 381   nullptr, /* postprocess_glyphs */
 382   nullptr, /* decompose */
 383   nullptr, /* compose */
 384   nullptr, /* setup_masks */
 385   nullptr, /* reorder_marks */
 386   HB_TAG_NONE, /* gpos_tag */
 387   HB_OT_SHAPE_NORMALIZATION_MODE_DEFAULT,
 388   HB_OT_SHAPE_ZERO_WIDTH_MARKS_BY_GDEF_LATE,
 389   false,/* fallback_position */
 390 };
 391
 392
 393 #endif