1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/gfx/harfbuzz/src/hb-ot-shape-complex-hangul.cc Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,417 @@
1.4 +/*
1.5 + * Copyright © 2013 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 +/* Hangul shaper */
1.34 +
1.35 +
1.36 +/* Same order as the feature array below */
1.37 +enum {
1.38 + NONE,
1.39 +
1.40 + LJMO,
1.41 + VJMO,
1.42 + TJMO,
1.43 +
1.44 + FIRST_HANGUL_FEATURE = LJMO,
1.45 + HANGUL_FEATURE_COUNT = TJMO + 1
1.46 +};
1.47 +
1.48 +static const hb_tag_t hangul_features[HANGUL_FEATURE_COUNT] =
1.49 +{
1.50 + HB_TAG_NONE,
1.51 + HB_TAG('l','j','m','o'),
1.52 + HB_TAG('v','j','m','o'),
1.53 + HB_TAG('t','j','m','o')
1.54 +};
1.55 +
1.56 +static void
1.57 +collect_features_hangul (hb_ot_shape_planner_t *plan)
1.58 +{
1.59 + hb_ot_map_builder_t *map = &plan->map;
1.60 +
1.61 + for (unsigned int i = FIRST_HANGUL_FEATURE; i < HANGUL_FEATURE_COUNT; i++)
1.62 + map->add_feature (hangul_features[i], 1, F_NONE);
1.63 +}
1.64 +
1.65 +struct hangul_shape_plan_t
1.66 +{
1.67 + ASSERT_POD ();
1.68 +
1.69 + hb_mask_t mask_array[HANGUL_FEATURE_COUNT];
1.70 +};
1.71 +
1.72 +static void *
1.73 +data_create_hangul (const hb_ot_shape_plan_t *plan)
1.74 +{
1.75 + hangul_shape_plan_t *hangul_plan = (hangul_shape_plan_t *) calloc (1, sizeof (hangul_shape_plan_t));
1.76 + if (unlikely (!hangul_plan))
1.77 + return NULL;
1.78 +
1.79 + for (unsigned int i = 0; i < HANGUL_FEATURE_COUNT; i++)
1.80 + hangul_plan->mask_array[i] = plan->map.get_1_mask (hangul_features[i]);
1.81 +
1.82 + return hangul_plan;
1.83 +}
1.84 +
1.85 +static void
1.86 +data_destroy_hangul (void *data)
1.87 +{
1.88 + free (data);
1.89 +}
1.90 +
1.91 +/* Constants for algorithmic hangul syllable [de]composition. */
1.92 +#define LBase 0x1100
1.93 +#define VBase 0x1161
1.94 +#define TBase 0x11A7
1.95 +#define LCount 19
1.96 +#define VCount 21
1.97 +#define TCount 28
1.98 +#define SBase 0xAC00
1.99 +#define NCount (VCount * TCount)
1.100 +#define SCount (LCount * NCount)
1.101 +
1.102 +#define isCombiningL(u) (hb_in_range<hb_codepoint_t> ((u), LBase, LBase+LCount-1))
1.103 +#define isCombiningV(u) (hb_in_range<hb_codepoint_t> ((u), VBase, VBase+VCount-1))
1.104 +#define isCombiningT(u) (hb_in_range<hb_codepoint_t> ((u), TBase+1, TBase+TCount-1))
1.105 +#define isCombinedS(u) (hb_in_range<hb_codepoint_t> ((u), SBase, SBase+SCount-1))
1.106 +
1.107 +#define isL(u) (hb_in_ranges<hb_codepoint_t> ((u), 0x1100, 0x115F, 0xA960, 0xA97C))
1.108 +#define isV(u) (hb_in_ranges<hb_codepoint_t> ((u), 0x1160, 0x11A7, 0xD7B0, 0xD7C6))
1.109 +#define isT(u) (hb_in_ranges<hb_codepoint_t> ((u), 0x11A8, 0x11FF, 0xD7CB, 0xD7FB))
1.110 +
1.111 +#define isHangulTone(u) (hb_in_range<hb_codepoint_t> ((u), 0x302e, 0x302f))
1.112 +
1.113 +/* buffer var allocations */
1.114 +#define hangul_shaping_feature() complex_var_u8_0() /* hangul jamo shaping feature */
1.115 +
1.116 +static bool
1.117 +is_zero_width_char (hb_font_t *font,
1.118 + hb_codepoint_t unicode)
1.119 +{
1.120 + hb_codepoint_t glyph;
1.121 + return hb_font_get_glyph (font, unicode, 0, &glyph) && hb_font_get_glyph_h_advance (font, glyph) == 0;
1.122 +}
1.123 +
1.124 +static void
1.125 +preprocess_text_hangul (const hb_ot_shape_plan_t *plan,
1.126 + hb_buffer_t *buffer,
1.127 + hb_font_t *font)
1.128 +{
1.129 + HB_BUFFER_ALLOCATE_VAR (buffer, hangul_shaping_feature);
1.130 +
1.131 + /* Hangul syllables come in two shapes: LV, and LVT. Of those:
1.132 + *
1.133 + * - LV can be precomposed, or decomposed. Lets call those
1.134 + * <LV> and <L,V>,
1.135 + * - LVT can be fully precomposed, partically precomposed, or
1.136 + * fully decomposed. Ie. <LVT>, <LV,T>, or <L,V,T>.
1.137 + *
1.138 + * The composition / decomposition is mechanical. However, not
1.139 + * all <L,V> sequences compose, and not all <LV,T> sequences
1.140 + * compose.
1.141 + *
1.142 + * Here are the specifics:
1.143 + *
1.144 + * - <L>: U+1100..115F, U+A960..A97F
1.145 + * - <V>: U+1160..11A7, U+D7B0..D7C7
1.146 + * - <T>: U+11A8..11FF, U+D7CB..D7FB
1.147 + *
1.148 + * - Only the <L,V> sequences for the 11xx ranges combine.
1.149 + * - Only <LV,T> sequences for T in U+11A8..11C3 combine.
1.150 + *
1.151 + * Here is what we want to accomplish in this shaper:
1.152 + *
1.153 + * - If the whole syllable can be precomposed, do that,
1.154 + * - Otherwise, fully decompose and apply ljmo/vjmo/tjmo features.
1.155 + * - If a valid syllable is followed by a Hangul tone mark, reorder the tone
1.156 + * mark to precede the whole syllable - unless it is a zero-width glyph, in
1.157 + * which case we leave it untouched, assuming it's designed to overstrike.
1.158 + *
1.159 + * That is, of the different possible syllables:
1.160 + *
1.161 + * <L>
1.162 + * <L,V>
1.163 + * <L,V,T>
1.164 + * <LV>
1.165 + * <LVT>
1.166 + * <LV, T>
1.167 + *
1.168 + * - <L> needs no work.
1.169 + *
1.170 + * - <LV> and <LVT> can stay the way they are if the font supports them, otherwise we
1.171 + * should fully decompose them if font supports.
1.172 + *
1.173 + * - <L,V> and <L,V,T> we should compose if the whole thing can be composed.
1.174 + *
1.175 + * - <LV,T> we should compose if the whole thing can be composed, otherwise we should
1.176 + * decompose.
1.177 + */
1.178 +
1.179 + buffer->clear_output ();
1.180 + unsigned int start = 0, end = 0; /* Extent of most recently seen syllable;
1.181 + * valid only if start < end
1.182 + */
1.183 + unsigned int count = buffer->len;
1.184 +
1.185 + for (buffer->idx = 0; buffer->idx < count;)
1.186 + {
1.187 + hb_codepoint_t u = buffer->cur().codepoint;
1.188 +
1.189 + if (isHangulTone (u))
1.190 + {
1.191 + /*
1.192 + * We could cache the width of the tone marks and the existence of dotted-circle,
1.193 + * but the use of the Hangul tone mark characters seems to be rare enough that
1.194 + * I didn't bother for now.
1.195 + */
1.196 + if (start < end && end == buffer->out_len)
1.197 + {
1.198 + /* Tone mark follows a valid syllable; move it in front, unless it's zero width. */
1.199 + buffer->next_glyph ();
1.200 + if (!is_zero_width_char (font, u))
1.201 + {
1.202 + hb_glyph_info_t *info = buffer->out_info;
1.203 + hb_glyph_info_t tone = info[end];
1.204 + memmove (&info[start + 1], &info[start], (end - start) * sizeof (hb_glyph_info_t));
1.205 + info[start] = tone;
1.206 + }
1.207 + /* Merge clusters across the (possibly reordered) syllable+tone.
1.208 + * We want to merge even in the zero-width tone mark case here,
1.209 + * so that clustering behavior isn't dependent on how the tone mark
1.210 + * is handled by the font.
1.211 + */
1.212 + buffer->merge_out_clusters (start, end + 1);
1.213 + }
1.214 + else
1.215 + {
1.216 + /* No valid syllable as base for tone mark; try to insert dotted circle. */
1.217 + if (font->has_glyph (0x25cc))
1.218 + {
1.219 + hb_codepoint_t chars[2];
1.220 + if (!is_zero_width_char (font, u)) {
1.221 + chars[0] = u;
1.222 + chars[1] = 0x25cc;
1.223 + } else {
1.224 + chars[0] = 0x25cc;
1.225 + chars[1] = u;
1.226 + }
1.227 + buffer->replace_glyphs (1, 2, chars);
1.228 + }
1.229 + else
1.230 + {
1.231 + /* No dotted circle available in the font; just leave tone mark untouched. */
1.232 + buffer->next_glyph ();
1.233 + }
1.234 + }
1.235 + start = end = buffer->out_len;
1.236 + continue;
1.237 + }
1.238 +
1.239 + start = buffer->out_len; /* Remember current position as a potential syllable start;
1.240 + * will only be used if we set end to a later position.
1.241 + */
1.242 +
1.243 + if (isL (u) && buffer->idx + 1 < count)
1.244 + {
1.245 + hb_codepoint_t l = u;
1.246 + hb_codepoint_t v = buffer->cur(+1).codepoint;
1.247 + if (isV (v))
1.248 + {
1.249 + /* Have <L,V> or <L,V,T>. */
1.250 + hb_codepoint_t t = 0;
1.251 + unsigned int tindex = 0;
1.252 + if (buffer->idx + 2 < count)
1.253 + {
1.254 + t = buffer->cur(+2).codepoint;
1.255 + if (isT (t))
1.256 + tindex = t - TBase; /* Only used if isCombiningT (t); otherwise invalid. */
1.257 + else
1.258 + t = 0; /* The next character was not a trailing jamo. */
1.259 + }
1.260 +
1.261 + /* We've got a syllable <L,V,T?>; see if it can potentially be composed. */
1.262 + if (isCombiningL (l) && isCombiningV (v) && (t == 0 || isCombiningT (t)))
1.263 + {
1.264 + /* Try to compose; if this succeeds, end is set to start+1. */
1.265 + hb_codepoint_t s = SBase + (l - LBase) * NCount + (v - VBase) * TCount + tindex;
1.266 + if (font->has_glyph (s))
1.267 + {
1.268 + buffer->replace_glyphs (t ? 3 : 2, 1, &s);
1.269 + if (unlikely (buffer->in_error))
1.270 + return;
1.271 + end = start + 1;
1.272 + continue;
1.273 + }
1.274 + }
1.275 +
1.276 + /* We didn't compose, either because it's an Old Hangul syllable without a
1.277 + * precomposed character in Unicode, or because the font didn't support the
1.278 + * necessary precomposed glyph.
1.279 + * Set jamo features on the individual glyphs, and advance past them.
1.280 + */
1.281 + buffer->cur().hangul_shaping_feature() = LJMO;
1.282 + buffer->next_glyph ();
1.283 + buffer->cur().hangul_shaping_feature() = VJMO;
1.284 + buffer->next_glyph ();
1.285 + if (t)
1.286 + {
1.287 + buffer->cur().hangul_shaping_feature() = TJMO;
1.288 + buffer->next_glyph ();
1.289 + end = start + 3;
1.290 + }
1.291 + else
1.292 + end = start + 2;
1.293 + buffer->merge_out_clusters (start, end);
1.294 + continue;
1.295 + }
1.296 + }
1.297 +
1.298 + else if (isCombinedS (u))
1.299 + {
1.300 + /* Have <LV>, <LVT>, or <LV,T> */
1.301 + hb_codepoint_t s = u;
1.302 + bool has_glyph = font->has_glyph (s);
1.303 + unsigned int lindex = (s - SBase) / NCount;
1.304 + unsigned int nindex = (s - SBase) % NCount;
1.305 + unsigned int vindex = nindex / TCount;
1.306 + unsigned int tindex = nindex % TCount;
1.307 +
1.308 + if (!tindex &&
1.309 + buffer->idx + 1 < count &&
1.310 + isCombiningT (buffer->cur(+1).codepoint))
1.311 + {
1.312 + /* <LV,T>, try to combine. */
1.313 + unsigned int new_tindex = buffer->cur(+1).codepoint - TBase;
1.314 + hb_codepoint_t new_s = s + new_tindex;
1.315 + if (font->has_glyph (new_s))
1.316 + {
1.317 + buffer->replace_glyphs (2, 1, &new_s);
1.318 + if (unlikely (buffer->in_error))
1.319 + return;
1.320 + end = start + 1;
1.321 + continue;
1.322 + }
1.323 + }
1.324 +
1.325 + /* Otherwise, decompose if font doesn't support <LV> or <LVT>,
1.326 + * or if having non-combining <LV,T>. Note that we already handled
1.327 + * combining <LV,T> above. */
1.328 + if (!has_glyph ||
1.329 + (!tindex &&
1.330 + buffer->idx + 1 < count &&
1.331 + isT (buffer->cur(+1).codepoint)))
1.332 + {
1.333 + hb_codepoint_t decomposed[3] = {LBase + lindex,
1.334 + VBase + vindex,
1.335 + TBase + tindex};
1.336 + if (font->has_glyph (decomposed[0]) &&
1.337 + font->has_glyph (decomposed[1]) &&
1.338 + (!tindex || font->has_glyph (decomposed[2])))
1.339 + {
1.340 + unsigned int s_len = tindex ? 3 : 2;
1.341 + buffer->replace_glyphs (1, s_len, decomposed);
1.342 + if (unlikely (buffer->in_error))
1.343 + return;
1.344 +
1.345 + /* We decomposed S: apply jamo features to the individual glyphs
1.346 + * that are now in buffer->out_info.
1.347 + */
1.348 + hb_glyph_info_t *info = buffer->out_info;
1.349 +
1.350 + /* If we decomposed an LV because of a non-combining T following,
1.351 + * we want to include this T in the syllable.
1.352 + */
1.353 + if (has_glyph && !tindex)
1.354 + {
1.355 + buffer->next_glyph ();
1.356 + s_len++;
1.357 + }
1.358 + end = start + s_len;
1.359 +
1.360 + unsigned int i = start;
1.361 + info[i++].hangul_shaping_feature() = LJMO;
1.362 + info[i++].hangul_shaping_feature() = VJMO;
1.363 + if (i < end)
1.364 + info[i++].hangul_shaping_feature() = TJMO;
1.365 + buffer->merge_out_clusters (start, end);
1.366 + continue;
1.367 + }
1.368 + }
1.369 +
1.370 + if (has_glyph)
1.371 + {
1.372 + /* We didn't decompose the S, so just advance past it. */
1.373 + end = start + 1;
1.374 + buffer->next_glyph ();
1.375 + continue;
1.376 + }
1.377 + }
1.378 +
1.379 + /* Didn't find a recognizable syllable, so we leave end <= start;
1.380 + * this will prevent tone-mark reordering happening.
1.381 + */
1.382 + buffer->next_glyph ();
1.383 + }
1.384 + buffer->swap_buffers ();
1.385 +}
1.386 +
1.387 +static void
1.388 +setup_masks_hangul (const hb_ot_shape_plan_t *plan,
1.389 + hb_buffer_t *buffer,
1.390 + hb_font_t *font HB_UNUSED)
1.391 +{
1.392 + const hangul_shape_plan_t *hangul_plan = (const hangul_shape_plan_t *) plan->data;
1.393 +
1.394 + if (likely (hangul_plan))
1.395 + {
1.396 + unsigned int count = buffer->len;
1.397 + hb_glyph_info_t *info = buffer->info;
1.398 + for (unsigned int i = 0; i < count; i++, info++)
1.399 + info->mask |= hangul_plan->mask_array[info->hangul_shaping_feature()];
1.400 + }
1.401 +
1.402 + HB_BUFFER_DEALLOCATE_VAR (buffer, hangul_shaping_feature);
1.403 +}
1.404 +
1.405 +
1.406 +const hb_ot_complex_shaper_t _hb_ot_complex_shaper_hangul =
1.407 +{
1.408 + "hangul",
1.409 + collect_features_hangul,
1.410 + NULL, /* override_features */
1.411 + data_create_hangul, /* data_create */
1.412 + data_destroy_hangul, /* data_destroy */
1.413 + preprocess_text_hangul,
1.414 + HB_OT_SHAPE_NORMALIZATION_MODE_NONE,
1.415 + NULL, /* decompose */
1.416 + NULL, /* compose */
1.417 + setup_masks_hangul, /* setup_masks */
1.418 + HB_OT_SHAPE_ZERO_WIDTH_MARKS_NONE,
1.419 + false, /* fallback_position */
1.420 +};
