2 * Copyright © 2010,2012 Google, Inc.
4 * This is part of HarfBuzz, a text shaping library.
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.
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
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.
24 * Google Author(s): Behdad Esfahbod
27 #include "hb-ot-shape-complex-private.hh"
30 /* TODO Add kana, and other small shapers here */
33 /* The default shaper *only* adds additional per-script features.*/
35 static const hb_tag_t hangul_features[] =
37 HB_TAG('l','j','m','o'),
38 HB_TAG('v','j','m','o'),
39 HB_TAG('t','j','m','o'),
43 static const hb_tag_t tibetan_features[] =
45 HB_TAG('a','b','v','s'),
46 HB_TAG('b','l','w','s'),
47 HB_TAG('a','b','v','m'),
48 HB_TAG('b','l','w','m'),
53 collect_features_default (hb_ot_shape_planner_t *plan)
55 const hb_tag_t *script_features = NULL;
57 switch ((hb_tag_t) plan->props.script)
59 /* Unicode-1.1 additions */
60 case HB_SCRIPT_HANGUL:
61 script_features = hangul_features;
64 /* Unicode-2.0 additions */
65 case HB_SCRIPT_TIBETAN:
66 script_features = tibetan_features;
70 for (; script_features && *script_features; script_features++)
71 plan->map.add_bool_feature (*script_features);
74 static hb_ot_shape_normalization_mode_t
75 normalization_preference_default (const hb_ot_shape_plan_t *plan)
77 switch ((hb_tag_t) plan->props.script)
79 /* Unicode-1.1 additions */
80 case HB_SCRIPT_HANGUL:
81 return HB_OT_SHAPE_NORMALIZATION_MODE_COMPOSED_FULL;
83 return HB_OT_SHAPE_NORMALIZATION_MODE_COMPOSED_DIACRITICS;
86 const hb_ot_complex_shaper_t _hb_ot_complex_shaper_default =
89 collect_features_default,
90 NULL, /* override_features */
91 NULL, /* data_create */
92 NULL, /* data_destroy */
93 NULL, /* preprocess_text */
94 normalization_preference_default,
95 NULL, /* setup_masks */
96 true, /* zero_width_attached_marks */
100 /* Thai / Lao shaper */
103 preprocess_text_thai (const hb_ot_shape_plan_t *plan HB_UNUSED,
105 hb_font_t *font HB_UNUSED)
107 /* The following is NOT specified in the MS OT Thai spec, however, it seems
108 * to be what Uniscribe and other engines implement. According to Eric Muller:
110 * When you have a SARA AM, decompose it in NIKHAHIT + SARA AA, *and* move the
111 * NIKHAHIT backwards over any tone mark (0E48-0E4B).
113 * <0E14, 0E4B, 0E33> -> <0E14, 0E4D, 0E4B, 0E32>
115 * This reordering is legit only when the NIKHAHIT comes from a SARA AM, not
116 * when it's there to start with. The string <0E14, 0E4B, 0E4D> is probably
117 * not what a user wanted, but the rendering is nevertheless nikhahit above
124 * Uniscribe also does so below-marks reordering. Namely, it positions U+0E3A
125 * after U+0E38 and U+0E39. We do that by modifying the ccc for U+0E3A.
126 * See unicode->modified_combining_class (). Lao does NOT have a U+0E3A
132 * Here are the characters of significance:
135 * SARA AM: U+0E33 U+0EB3
136 * SARA AA: U+0E32 U+0EB2
137 * Nikhahit: U+0E4D U+0ECD
139 * Testing shows that Uniscribe reorder the following marks:
140 * Thai: <0E31,0E34..0E37,0E47..0E4E>
141 * Lao: <0EB1,0EB4..0EB7,0EC7..0ECE>
143 * Note how the Lao versions are the same as Thai + 0x80.
146 /* We only get one script at a time, so a script-agnostic implementation
147 * is adequate here. */
148 #define IS_SARA_AM(x) (((x) & ~0x0080) == 0x0E33)
149 #define NIKHAHIT_FROM_SARA_AM(x) ((x) - 0xE33 + 0xE4D)
150 #define SARA_AA_FROM_SARA_AM(x) ((x) - 1)
151 #define IS_TONE_MARK(x) (hb_in_ranges<hb_codepoint_t> ((x) & ~0x0080, 0x0E34, 0x0E37, 0x0E47, 0x0E4E, 0x0E31, 0x0E31))
153 buffer->clear_output ();
154 unsigned int count = buffer->len;
155 for (buffer->idx = 0; buffer->idx < count;)
157 hb_codepoint_t u = buffer->cur().codepoint;
158 if (likely (!IS_SARA_AM (u))) {
159 buffer->next_glyph ();
163 /* Is SARA AM. Decompose and reorder. */
164 hb_codepoint_t decomposed[2] = {hb_codepoint_t (NIKHAHIT_FROM_SARA_AM (u)),
165 hb_codepoint_t (SARA_AA_FROM_SARA_AM (u))};
166 buffer->replace_glyphs (1, 2, decomposed);
167 if (unlikely (buffer->in_error))
170 /* Ok, let's see... */
171 unsigned int end = buffer->out_len;
172 unsigned int start = end - 2;
173 while (start > 0 && IS_TONE_MARK (buffer->out_info[start - 1].codepoint))
178 /* Move Nikhahit (end-2) to the beginning */
179 buffer->merge_out_clusters (start, end);
180 hb_glyph_info_t t = buffer->out_info[end - 2];
181 memmove (buffer->out_info + start + 1,
182 buffer->out_info + start,
183 sizeof (buffer->out_info[0]) * (end - start - 2));
184 buffer->out_info[start] = t;
188 /* Since we decomposed, and NIKHAHIT is combining, merge clusters with the
189 * previous cluster. */
191 buffer->merge_out_clusters (start - 1, end);
194 buffer->swap_buffers ();
197 const hb_ot_complex_shaper_t _hb_ot_complex_shaper_thai =
200 NULL, /* collect_features */
201 NULL, /* override_features */
202 NULL, /* data_create */
203 NULL, /* data_destroy */
204 preprocess_text_thai,
205 NULL, /* normalization_preference */
206 NULL, /* setup_masks */
207 true, /* zero_width_attached_marks */