1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/gfx/harfbuzz/src/hb-ot-shape-complex-thai.cc Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,378 @@
1.4 +/*
1.5 + * Copyright © 2010,2012 Google, Inc.
1.6 + *
1.7 + * This is part of HarfBuzz, a text shaping library.
1.8 + *
1.9 + * Permission is hereby granted, without written agreement and without
1.10 + * license or royalty fees, to use, copy, modify, and distribute this
1.11 + * software and its documentation for any purpose, provided that the
1.12 + * above copyright notice and the following two paragraphs appear in
1.13 + * all copies of this software.
1.14 + *
1.15 + * IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE TO ANY PARTY FOR
1.16 + * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES
1.17 + * ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN
1.18 + * IF THE COPYRIGHT HOLDER HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
1.19 + * DAMAGE.
1.20 + *
1.21 + * THE COPYRIGHT HOLDER SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING,
1.22 + * BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
1.23 + * FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS
1.24 + * ON AN "AS IS" BASIS, AND THE COPYRIGHT HOLDER HAS NO OBLIGATION TO
1.25 + * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
1.26 + *
1.27 + * Google Author(s): Behdad Esfahbod
1.28 + */
1.29 +
1.30 +#include "hb-ot-shape-complex-private.hh"
1.31 +
1.32 +
1.33 +/* Thai / Lao shaper */
1.34 +
1.35 +
1.36 +/* PUA shaping */
1.37 +
1.38 +
1.39 +enum thai_consonant_type_t
1.40 +{
1.41 + NC,
1.42 + AC,
1.43 + RC,
1.44 + DC,
1.45 + NOT_CONSONANT,
1.46 + NUM_CONSONANT_TYPES = NOT_CONSONANT
1.47 +};
1.48 +
1.49 +static thai_consonant_type_t
1.50 +get_consonant_type (hb_codepoint_t u)
1.51 +{
1.52 + if (u == 0x0E1B || u == 0x0E1D || u == 0x0E1F/* || u == 0x0E2C*/)
1.53 + return AC;
1.54 + if (u == 0x0E0D || u == 0x0E10)
1.55 + return RC;
1.56 + if (u == 0x0E0E || u == 0x0E0F)
1.57 + return DC;
1.58 + if (hb_in_range<hb_codepoint_t> (u, 0x0E01, 0x0E2E))
1.59 + return NC;
1.60 + return NOT_CONSONANT;
1.61 +}
1.62 +
1.63 +
1.64 +enum thai_mark_type_t
1.65 +{
1.66 + AV,
1.67 + BV,
1.68 + T,
1.69 + NOT_MARK,
1.70 + NUM_MARK_TYPES = NOT_MARK
1.71 +};
1.72 +
1.73 +static thai_mark_type_t
1.74 +get_mark_type (hb_codepoint_t u)
1.75 +{
1.76 + if (u == 0x0E31 || hb_in_range<hb_codepoint_t> (u, 0x0E34, 0x0E37) ||
1.77 + u == 0x0E47 || hb_in_range<hb_codepoint_t> (u, 0x0E4D, 0x0E4E))
1.78 + return AV;
1.79 + if (hb_in_range<hb_codepoint_t> (u, 0x0E38, 0x0E3A))
1.80 + return BV;
1.81 + if (hb_in_range<hb_codepoint_t> (u, 0x0E48, 0x0E4C))
1.82 + return T;
1.83 + return NOT_MARK;
1.84 +}
1.85 +
1.86 +
1.87 +enum thai_action_t
1.88 +{
1.89 + NOP,
1.90 + SD, /* Shift combining-mark down */
1.91 + SL, /* Shift combining-mark left */
1.92 + SDL, /* Shift combining-mark down-left */
1.93 + RD /* Remove descender from base */
1.94 +};
1.95 +
1.96 +static hb_codepoint_t
1.97 +thai_pua_shape (hb_codepoint_t u, thai_action_t action, hb_font_t *font)
1.98 +{
1.99 + struct thai_pua_mapping_t {
1.100 + hb_codepoint_t u;
1.101 + hb_codepoint_t win_pua;
1.102 + hb_codepoint_t mac_pua;
1.103 + } const *pua_mappings = NULL;
1.104 + static const thai_pua_mapping_t SD_mappings[] = {
1.105 + {0x0E48, 0xF70A, 0xF88B}, /* MAI EK */
1.106 + {0x0E49, 0xF70B, 0xF88E}, /* MAI THO */
1.107 + {0x0E4A, 0xF70C, 0xF891}, /* MAI TRI */
1.108 + {0x0E4B, 0xF70D, 0xF894}, /* MAI CHATTAWA */
1.109 + {0x0E4C, 0xF70E, 0xF897}, /* THANTHAKHAT */
1.110 + {0x0E38, 0xF718, 0xF89B}, /* SARA U */
1.111 + {0x0E39, 0xF719, 0xF89C}, /* SARA UU */
1.112 + {0x0E3A, 0xF71A, 0xF89D}, /* PHINTHU */
1.113 + {0x0000, 0x0000, 0x0000}
1.114 + };
1.115 + static const thai_pua_mapping_t SDL_mappings[] = {
1.116 + {0x0E48, 0xF705, 0xF88C}, /* MAI EK */
1.117 + {0x0E49, 0xF706, 0xF88F}, /* MAI THO */
1.118 + {0x0E4A, 0xF707, 0xF892}, /* MAI TRI */
1.119 + {0x0E4B, 0xF708, 0xF895}, /* MAI CHATTAWA */
1.120 + {0x0E4C, 0xF709, 0xF898}, /* THANTHAKHAT */
1.121 + {0x0000, 0x0000, 0x0000}
1.122 + };
1.123 + static const thai_pua_mapping_t SL_mappings[] = {
1.124 + {0x0E48, 0xF713, 0xF88A}, /* MAI EK */
1.125 + {0x0E49, 0xF714, 0xF88D}, /* MAI THO */
1.126 + {0x0E4A, 0xF715, 0xF890}, /* MAI TRI */
1.127 + {0x0E4B, 0xF716, 0xF893}, /* MAI CHATTAWA */
1.128 + {0x0E4C, 0xF717, 0xF896}, /* THANTHAKHAT */
1.129 + {0x0E31, 0xF710, 0xF884}, /* MAI HAN-AKAT */
1.130 + {0x0E34, 0xF701, 0xF885}, /* SARA I */
1.131 + {0x0E35, 0xF702, 0xF886}, /* SARA II */
1.132 + {0x0E36, 0xF703, 0xF887}, /* SARA UE */
1.133 + {0x0E37, 0xF704, 0xF888}, /* SARA UEE */
1.134 + {0x0E47, 0xF712, 0xF889}, /* MAITAIKHU */
1.135 + {0x0E4D, 0xF711, 0xF899}, /* NIKHAHIT */
1.136 + {0x0000, 0x0000, 0x0000}
1.137 + };
1.138 + static const thai_pua_mapping_t RD_mappings[] = {
1.139 + {0x0E0D, 0xF70F, 0xF89A}, /* YO YING */
1.140 + {0x0E10, 0xF700, 0xF89E}, /* THO THAN */
1.141 + {0x0000, 0x0000, 0x0000}
1.142 + };
1.143 +
1.144 + switch (action) {
1.145 + default: assert (false); /* Fallthrough */
1.146 + case NOP: return u;
1.147 + case SD: pua_mappings = SD_mappings; break;
1.148 + case SDL: pua_mappings = SDL_mappings; break;
1.149 + case SL: pua_mappings = SL_mappings; break;
1.150 + case RD: pua_mappings = RD_mappings; break;
1.151 + }
1.152 + for (; pua_mappings->u; pua_mappings++)
1.153 + if (pua_mappings->u == u)
1.154 + {
1.155 + hb_codepoint_t glyph;
1.156 + if (hb_font_get_glyph (font, pua_mappings->win_pua, 0, &glyph))
1.157 + return pua_mappings->win_pua;
1.158 + if (hb_font_get_glyph (font, pua_mappings->mac_pua, 0, &glyph))
1.159 + return pua_mappings->mac_pua;
1.160 + break;
1.161 + }
1.162 + return u;
1.163 +}
1.164 +
1.165 +
1.166 +static enum thai_above_state_t
1.167 +{ /* Cluster above looks like: */
1.168 + T0, /* ⣤ */
1.169 + T1, /* ⣼ */
1.170 + T2, /* ⣾ */
1.171 + T3, /* ⣿ */
1.172 + NUM_ABOVE_STATES
1.173 +} thai_above_start_state[NUM_CONSONANT_TYPES + 1/* For NOT_CONSONANT */] =
1.174 +{
1.175 + T0, /* NC */
1.176 + T1, /* AC */
1.177 + T0, /* RC */
1.178 + T0, /* DC */
1.179 + T3, /* NOT_CONSONANT */
1.180 +};
1.181 +
1.182 +static const struct thai_above_state_machine_edge_t {
1.183 + thai_action_t action;
1.184 + thai_above_state_t next_state;
1.185 +} thai_above_state_machine[NUM_ABOVE_STATES][NUM_MARK_TYPES] =
1.186 +{ /*AV*/ /*BV*/ /*T*/
1.187 +/*T0*/ {{NOP,T3}, {NOP,T0}, {SD, T3}},
1.188 +/*T1*/ {{SL, T2}, {NOP,T1}, {SDL,T2}},
1.189 +/*T2*/ {{NOP,T3}, {NOP,T2}, {SL, T3}},
1.190 +/*T3*/ {{NOP,T3}, {NOP,T3}, {NOP,T3}},
1.191 +};
1.192 +
1.193 +
1.194 +static enum thai_below_state_t
1.195 +{
1.196 + B0, /* No descender */
1.197 + B1, /* Removable descender */
1.198 + B2, /* Strict descender */
1.199 + NUM_BELOW_STATES
1.200 +} thai_below_start_state[NUM_CONSONANT_TYPES + 1/* For NOT_CONSONANT */] =
1.201 +{
1.202 + B0, /* NC */
1.203 + B0, /* AC */
1.204 + B1, /* RC */
1.205 + B2, /* DC */
1.206 + B2, /* NOT_CONSONANT */
1.207 +};
1.208 +
1.209 +static const struct thai_below_state_machine_edge_t {
1.210 + thai_action_t action;
1.211 + thai_below_state_t next_state;
1.212 +} thai_below_state_machine[NUM_BELOW_STATES][NUM_MARK_TYPES] =
1.213 +{ /*AV*/ /*BV*/ /*T*/
1.214 +/*B0*/ {{NOP,B0}, {NOP,B2}, {NOP, B0}},
1.215 +/*B1*/ {{NOP,B1}, {RD, B2}, {NOP, B1}},
1.216 +/*B2*/ {{NOP,B2}, {SD, B2}, {NOP, B2}},
1.217 +};
1.218 +
1.219 +
1.220 +static void
1.221 +do_thai_pua_shaping (const hb_ot_shape_plan_t *plan HB_UNUSED,
1.222 + hb_buffer_t *buffer,
1.223 + hb_font_t *font)
1.224 +{
1.225 + thai_above_state_t above_state = thai_above_start_state[NOT_CONSONANT];
1.226 + thai_below_state_t below_state = thai_below_start_state[NOT_CONSONANT];
1.227 + unsigned int base = 0;
1.228 +
1.229 + hb_glyph_info_t *info = buffer->info;
1.230 + unsigned int count = buffer->len;
1.231 + for (unsigned int i = 0; i < count; i++)
1.232 + {
1.233 + thai_mark_type_t mt = get_mark_type (info[i].codepoint);
1.234 +
1.235 + if (mt == NOT_MARK) {
1.236 + thai_consonant_type_t ct = get_consonant_type (info[i].codepoint);
1.237 + above_state = thai_above_start_state[ct];
1.238 + below_state = thai_below_start_state[ct];
1.239 + base = i;
1.240 + continue;
1.241 + }
1.242 +
1.243 + const thai_above_state_machine_edge_t &above_edge = thai_above_state_machine[above_state][mt];
1.244 + const thai_below_state_machine_edge_t &below_edge = thai_below_state_machine[below_state][mt];
1.245 + above_state = above_edge.next_state;
1.246 + below_state = below_edge.next_state;
1.247 +
1.248 + /* At least one of the above/below actions is NOP. */
1.249 + thai_action_t action = above_edge.action != NOP ? above_edge.action : below_edge.action;
1.250 +
1.251 + if (action == RD)
1.252 + info[base].codepoint = thai_pua_shape (info[base].codepoint, action, font);
1.253 + else
1.254 + info[i].codepoint = thai_pua_shape (info[i].codepoint, action, font);
1.255 + }
1.256 +}
1.257 +
1.258 +
1.259 +static void
1.260 +preprocess_text_thai (const hb_ot_shape_plan_t *plan,
1.261 + hb_buffer_t *buffer,
1.262 + hb_font_t *font)
1.263 +{
1.264 + /* This function implements the shaping logic documented here:
1.265 + *
1.266 + * http://linux.thai.net/~thep/th-otf/shaping.html
1.267 + *
1.268 + * The first shaping rule listed there is needed even if the font has Thai
1.269 + * OpenType tables. The rest do fallback positioning based on PUA codepoints.
1.270 + * We implement that only if there exist no Thai GSUB in the font.
1.271 + */
1.272 +
1.273 + /* The following is NOT specified in the MS OT Thai spec, however, it seems
1.274 + * to be what Uniscribe and other engines implement. According to Eric Muller:
1.275 + *
1.276 + * When you have a SARA AM, decompose it in NIKHAHIT + SARA AA, *and* move the
1.277 + * NIKHAHIT backwards over any tone mark (0E48-0E4B).
1.278 + *
1.279 + * <0E14, 0E4B, 0E33> -> <0E14, 0E4D, 0E4B, 0E32>
1.280 + *
1.281 + * This reordering is legit only when the NIKHAHIT comes from a SARA AM, not
1.282 + * when it's there to start with. The string <0E14, 0E4B, 0E4D> is probably
1.283 + * not what a user wanted, but the rendering is nevertheless nikhahit above
1.284 + * chattawa.
1.285 + *
1.286 + * Same for Lao.
1.287 + *
1.288 + * Note:
1.289 + *
1.290 + * Uniscribe also does some below-marks reordering. Namely, it positions U+0E3A
1.291 + * after U+0E38 and U+0E39. We do that by modifying the ccc for U+0E3A.
1.292 + * See unicode->modified_combining_class (). Lao does NOT have a U+0E3A
1.293 + * equivalent.
1.294 + */
1.295 +
1.296 +
1.297 + /*
1.298 + * Here are the characters of significance:
1.299 + *
1.300 + * Thai Lao
1.301 + * SARA AM: U+0E33 U+0EB3
1.302 + * SARA AA: U+0E32 U+0EB2
1.303 + * Nikhahit: U+0E4D U+0ECD
1.304 + *
1.305 + * Testing shows that Uniscribe reorder the following marks:
1.306 + * Thai: <0E31,0E34..0E37,0E47..0E4E>
1.307 + * Lao: <0EB1,0EB4..0EB7,0EC7..0ECE>
1.308 + *
1.309 + * Note how the Lao versions are the same as Thai + 0x80.
1.310 + */
1.311 +
1.312 + /* We only get one script at a time, so a script-agnostic implementation
1.313 + * is adequate here. */
1.314 +#define IS_SARA_AM(x) (((x) & ~0x0080) == 0x0E33)
1.315 +#define NIKHAHIT_FROM_SARA_AM(x) ((x) - 0xE33 + 0xE4D)
1.316 +#define SARA_AA_FROM_SARA_AM(x) ((x) - 1)
1.317 +#define IS_TONE_MARK(x) (hb_in_ranges<hb_codepoint_t> ((x) & ~0x0080, 0x0E34, 0x0E37, 0x0E47, 0x0E4E, 0x0E31, 0x0E31))
1.318 +
1.319 + buffer->clear_output ();
1.320 + unsigned int count = buffer->len;
1.321 + for (buffer->idx = 0; buffer->idx < count;)
1.322 + {
1.323 + hb_codepoint_t u = buffer->cur().codepoint;
1.324 + if (likely (!IS_SARA_AM (u))) {
1.325 + buffer->next_glyph ();
1.326 + continue;
1.327 + }
1.328 +
1.329 + /* Is SARA AM. Decompose and reorder. */
1.330 + hb_codepoint_t decomposed[2] = {hb_codepoint_t (NIKHAHIT_FROM_SARA_AM (u)),
1.331 + hb_codepoint_t (SARA_AA_FROM_SARA_AM (u))};
1.332 + buffer->replace_glyphs (1, 2, decomposed);
1.333 + if (unlikely (buffer->in_error))
1.334 + return;
1.335 +
1.336 + /* Ok, let's see... */
1.337 + unsigned int end = buffer->out_len;
1.338 + unsigned int start = end - 2;
1.339 + while (start > 0 && IS_TONE_MARK (buffer->out_info[start - 1].codepoint))
1.340 + start--;
1.341 +
1.342 + if (start + 2 < end)
1.343 + {
1.344 + /* Move Nikhahit (end-2) to the beginning */
1.345 + buffer->merge_out_clusters (start, end);
1.346 + hb_glyph_info_t t = buffer->out_info[end - 2];
1.347 + memmove (buffer->out_info + start + 1,
1.348 + buffer->out_info + start,
1.349 + sizeof (buffer->out_info[0]) * (end - start - 2));
1.350 + buffer->out_info[start] = t;
1.351 + }
1.352 + else
1.353 + {
1.354 + /* Since we decomposed, and NIKHAHIT is combining, merge clusters with the
1.355 + * previous cluster. */
1.356 + if (start)
1.357 + buffer->merge_out_clusters (start - 1, end);
1.358 + }
1.359 + }
1.360 + buffer->swap_buffers ();
1.361 +
1.362 + /* If font has Thai GSUB, we are done. */
1.363 + if (plan->props.script == HB_SCRIPT_THAI && !plan->map.found_script[0])
1.364 + do_thai_pua_shaping (plan, buffer, font);
1.365 +}
1.366 +
1.367 +const hb_ot_complex_shaper_t _hb_ot_complex_shaper_thai =
1.368 +{
1.369 + "thai",
1.370 + NULL, /* collect_features */
1.371 + NULL, /* override_features */
1.372 + NULL, /* data_create */
1.373 + NULL, /* data_destroy */
1.374 + preprocess_text_thai,
1.375 + HB_OT_SHAPE_NORMALIZATION_MODE_DEFAULT,
1.376 + NULL, /* decompose */
1.377 + NULL, /* compose */
1.378 + NULL, /* setup_masks */
1.379 + HB_OT_SHAPE_ZERO_WIDTH_MARKS_DEFAULT,
1.380 + false,/* fallback_position */
1.381 +};
