< prev index next >
src/java.desktop/share/native/libfontmanager/harfbuzz/hb-ot-shape-complex-indic.cc
Print this page
@@ -140,11 +140,11 @@
static inline bool
is_one_of (const hb_glyph_info_t &info, unsigned int flags)
{
/* If it ligated, all bets are off. */
if (_hb_glyph_info_ligated (&info)) return false;
- return !!(FLAG_SAFE (info.indic_category()) & flags);
+ return !!(FLAG_UNSAFE (info.indic_category()) & flags);
}
static inline bool
is_joiner (const hb_glyph_info_t &info)
{
@@ -175,54 +175,77 @@
/*
* Re-assign category
*/
/* The following act more like the Bindus. */
- if (unlikely (hb_in_range (u, 0x0953u, 0x0954u)))
+ if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x0953u, 0x0954u)))
cat = OT_SM;
/* The following act like consonants. */
- else if (unlikely (hb_in_ranges (u, 0x0A72u, 0x0A73u,
+ else if (unlikely (hb_in_ranges<hb_codepoint_t> (u, 0x0A72u, 0x0A73u,
0x1CF5u, 0x1CF6u)))
cat = OT_C;
/* TODO: The following should only be allowed after a Visarga.
* For now, just treat them like regular tone marks. */
- else if (unlikely (hb_in_range (u, 0x1CE2u, 0x1CE8u)))
+ else if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x1CE2u, 0x1CE8u)))
cat = OT_A;
/* TODO: The following should only be allowed after some of
* the nasalization marks, maybe only for U+1CE9..U+1CF1.
* For now, just treat them like tone marks. */
else if (unlikely (u == 0x1CEDu))
cat = OT_A;
/* The following take marks in standalone clusters, similar to Avagraha. */
- else if (unlikely (hb_in_ranges (u, 0xA8F2u, 0xA8F7u,
+ else if (unlikely (hb_in_ranges<hb_codepoint_t> (u, 0xA8F2u, 0xA8F7u,
0x1CE9u, 0x1CECu,
0x1CEEu, 0x1CF1u)))
{
cat = OT_Symbol;
- ASSERT_STATIC ((int) INDIC_SYLLABIC_CATEGORY_AVAGRAHA == OT_Symbol);
+ static_assert (((int) INDIC_SYLLABIC_CATEGORY_AVAGRAHA == OT_Symbol), "");
}
+ else if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x17CDu, 0x17D1u) ||
+ u == 0x17CBu || u == 0x17D3u || u == 0x17DDu)) /* Khmer Various signs */
+ {
+ /* These can occur mid-syllable (eg. before matras), even though Unicode marks them as Syllable_Modifier.
+ * https://github.com/roozbehp/unicode-data/issues/5 */
+ cat = OT_M;
+ pos = POS_ABOVE_C;
+ }
+ else if (unlikely (u == 0x0A51u))
+ {
+ /* https://github.com/behdad/harfbuzz/issues/524 */
+ cat = OT_M;
+ pos = POS_BELOW_C;
+ }
+
+ /* According to ScriptExtensions.txt, these Grantha marks may also be used in Tamil,
+ * so the Indic shaper needs to know their categories. */
+ else if (unlikely (u == 0x11301u || u == 0x11303u)) cat = OT_SM;
+ else if (unlikely (u == 0x1133cu)) cat = OT_N;
+
+ else if (unlikely (u == 0x0AFBu)) cat = OT_N; /* https://github.com/behdad/harfbuzz/issues/552 */
+
+ else if (unlikely (u == 0x0980u)) cat = OT_PLACEHOLDER; /* https://github.com/behdad/harfbuzz/issues/538 */
else if (unlikely (u == 0x17C6u)) cat = OT_N; /* Khmer Bindu doesn't like to be repositioned. */
- else if (unlikely (hb_in_range (u, 0x2010u, 0x2011u)))
+ else if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x2010u, 0x2011u)))
cat = OT_PLACEHOLDER;
else if (unlikely (u == 0x25CCu)) cat = OT_DOTTEDCIRCLE;
/*
* Re-assign position.
*/
- if ((FLAG_SAFE (cat) & CONSONANT_FLAGS))
+ if ((FLAG_UNSAFE (cat) & CONSONANT_FLAGS))
{
pos = POS_BASE_C;
if (is_ra (u))
cat = OT_Ra;
}
else if (cat == OT_M)
{
pos = matra_position (u, pos);
}
- else if ((FLAG_SAFE (cat) & (FLAG (OT_SM) | FLAG (OT_VD) | FLAG (OT_A) | FLAG (OT_Symbol))))
+ else if ((FLAG_UNSAFE (cat) & (FLAG (OT_SM) | FLAG (OT_VD) | FLAG (OT_A) | FLAG (OT_Symbol))))
{
pos = POS_SMVD;
}
if (unlikely (u == 0x0B01u)) pos = POS_BEFORE_SUB; /* Oriya Bindu is BeforeSub in the spec. */
@@ -409,16 +432,16 @@
unsigned int i = 0;
map->add_gsub_pause (initial_reordering);
for (; i < INDIC_BASIC_FEATURES; i++) {
- map->add_feature (indic_features[i].tag, 1, indic_features[i].flags | F_MANUAL_ZWJ);
- map->add_gsub_pause (NULL);
+ map->add_feature (indic_features[i].tag, 1, indic_features[i].flags | F_MANUAL_ZWJ | F_MANUAL_ZWNJ);
+ map->add_gsub_pause (nullptr);
}
map->add_gsub_pause (final_reordering);
for (; i < INDIC_NUM_FEATURES; i++) {
- map->add_feature (indic_features[i].tag, 1, indic_features[i].flags | F_MANUAL_ZWJ);
+ map->add_feature (indic_features[i].tag, 1, indic_features[i].flags | F_MANUAL_ZWJ | F_MANUAL_ZWNJ);
}
map->add_global_bool_feature (HB_TAG('c','a','l','t'));
map->add_global_bool_feature (HB_TAG('c','l','i','g'));
@@ -483,21 +506,21 @@
/* Technically speaking, the spec says we should apply 'locl' to virama too.
* Maybe one day... */
/* Our get_nominal_glyph() function needs a font, so we can't get the virama glyph
* during shape planning... Instead, overwrite it here. It's safe. Don't worry! */
- (const_cast<indic_shape_plan_t *> (this))->virama_glyph = glyph;
+ virama_glyph = glyph;
}
*pglyph = glyph;
return glyph != 0;
}
const indic_config_t *config;
bool is_old_spec;
- hb_codepoint_t virama_glyph;
+ mutable hb_codepoint_t virama_glyph;
would_substitute_feature_t rphf;
would_substitute_feature_t pref;
would_substitute_feature_t blwf;
would_substitute_feature_t pstf;
@@ -508,11 +531,11 @@
static void *
data_create_indic (const hb_ot_shape_plan_t *plan)
{
indic_shape_plan_t *indic_plan = (indic_shape_plan_t *) calloc (1, sizeof (indic_shape_plan_t));
if (unlikely (!indic_plan))
- return NULL;
+ return nullptr;
indic_plan->config = &indic_configs[0];
for (unsigned int i = 1; i < ARRAY_LENGTH (indic_configs); i++)
if (plan->props.script == indic_configs[i].script) {
indic_plan->config = &indic_configs[i];
@@ -613,10 +636,12 @@
setup_syllables (const hb_ot_shape_plan_t *plan HB_UNUSED,
hb_font_t *font HB_UNUSED,
hb_buffer_t *buffer)
{
find_syllables (buffer);
+ foreach_syllable (buffer, start, end)
+ buffer->unsafe_to_break (start, end);
}
static int
compare_indic_order (const hb_glyph_info_t *pa, const hb_glyph_info_t *pb)
{
@@ -664,10 +689,25 @@
unsigned int start, unsigned int end)
{
const indic_shape_plan_t *indic_plan = (const indic_shape_plan_t *) plan->data;
hb_glyph_info_t *info = buffer->info;
+ /* https://github.com/behdad/harfbuzz/issues/435#issuecomment-335560167
+ * // For compatibility with legacy usage in Kannada,
+ * // Ra+h+ZWJ must behave like Ra+ZWJ+h...
+ */
+ if (buffer->props.script == HB_SCRIPT_KANNADA &&
+ start + 3 <= end &&
+ is_one_of (info[start ], FLAG (OT_Ra)) &&
+ is_one_of (info[start+1], FLAG (OT_H)) &&
+ is_one_of (info[start+2], FLAG (OT_ZWJ)))
+ {
+ buffer->merge_clusters (start+1, start+3);
+ hb_glyph_info_t tmp = info[start+1];
+ info[start+1] = info[start+2];
+ info[start+2] = tmp;
+ }
/* 1. Find base consonant:
*
* The shaping engine finds the base consonant of the syllable, using the
* following algorithm: starting from the end of the syllable, move backwards
@@ -829,12 +869,12 @@
}
/* 2. Decompose and reorder Matras:
*
- * Each matra and any syllable modifier sign in the cluster are moved to the
- * appropriate position relative to the consonant(s) in the cluster. The
+ * Each matra and any syllable modifier sign in the syllable are moved to the
+ * appropriate position relative to the consonant(s) in the syllable. The
* shaping engine decomposes two- or three-part matras into their constituent
* parts before any repositioning. Matra characters are classified by which
* consonant in a conjunct they have affinity for and are reordered to the
* following positions:
*
@@ -926,11 +966,11 @@
/* Attach misc marks to previous char to move with them. */
{
indic_position_t last_pos = POS_START;
for (unsigned int i = start; i < end; i++)
{
- if ((FLAG_SAFE (info[i].indic_category()) & (JOINER_FLAGS | FLAG (OT_N) | FLAG (OT_RS) | MEDIAL_FLAGS | HALANT_OR_COENG_FLAGS)))
+ if ((FLAG_UNSAFE (info[i].indic_category()) & (JOINER_FLAGS | FLAG (OT_N) | FLAG (OT_RS) | MEDIAL_FLAGS | HALANT_OR_COENG_FLAGS)))
{
info[i].indic_position() = last_pos;
if (unlikely (info[i].indic_category() == OT_H &&
info[i].indic_position() == POS_PRE_M))
{
@@ -1256,11 +1296,11 @@
const indic_shape_plan_t *indic_plan = (const indic_shape_plan_t *) plan->data;
hb_glyph_info_t *info = buffer->info;
/* This function relies heavily on halant glyphs. Lots of ligation
- * and possibly multiplication substitutions happened prior to this
+ * and possibly multiple substitutions happened prior to this
* phase, and that might have messed up our properties. Recover
* from a particular case of that where we're fairly sure that a
* class of OT_H is desired but has been lost. */
if (indic_plan->virama_glyph)
{
@@ -1280,11 +1320,11 @@
/* 4. Final reordering:
*
* After the localized forms and basic shaping forms GSUB features have been
* applied (see below), the shaping engine performs some final glyph
* reordering before applying all the remaining font features to the entire
- * cluster.
+ * syllable.
*/
bool try_pref = !!indic_plan->mask_array[PREF];
/* Find base again */
@@ -1495,12 +1535,12 @@
*/
/* This is our take on what step 4 is trying to say (and failing, BADLY). */
if (reph_pos == REPH_POS_AFTER_SUB)
{
new_reph_pos = base;
- while (new_reph_pos < end &&
- !( FLAG_SAFE (info[new_reph_pos + 1].indic_position()) & (FLAG (POS_POST_C) | FLAG (POS_AFTER_POST) | FLAG (POS_SMVD))))
+ while (new_reph_pos + 1 < end &&
+ !( FLAG_UNSAFE (info[new_reph_pos + 1].indic_position()) & (FLAG (POS_POST_C) | FLAG (POS_AFTER_POST) | FLAG (POS_SMVD))))
new_reph_pos++;
if (new_reph_pos < end)
goto reph_move;
}
@@ -1644,15 +1684,19 @@
}
}
/* Apply 'init' to the Left Matra if it's a word start. */
- if (info[start].indic_position () == POS_PRE_M &&
- (!start ||
- !(FLAG_SAFE (_hb_glyph_info_get_general_category (&info[start - 1])) &
- FLAG_RANGE (HB_UNICODE_GENERAL_CATEGORY_FORMAT, HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK))))
+ if (info[start].indic_position () == POS_PRE_M)
+ {
+ if (!start ||
+ !(FLAG_UNSAFE (_hb_glyph_info_get_general_category (&info[start - 1])) &
+ FLAG_RANGE (HB_UNICODE_GENERAL_CATEGORY_FORMAT, HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK)))
info[start].mask |= indic_plan->mask_array[INIT];
+ else
+ buffer->unsafe_to_break (start - 1, start + 1);
+ }
/*
* Finish off the clusters and go home!
*/
@@ -1663,12 +1707,12 @@
case HB_SCRIPT_TAMIL:
case HB_SCRIPT_SINHALA:
break;
default:
- /* Uniscribe merges the entire cluster... Except for Tamil & Sinhala.
- * This means, half forms are submerged into the main consonants cluster.
+ /* Uniscribe merges the entire syllable into a single cluster... Except for Tamil & Sinhala.
+ * This means, half forms are submerged into the main consonant's cluster.
* This is unnecessary, and makes cursor positioning harder, but that's what
* Uniscribe does. */
buffer->merge_clusters (start, end);
break;
}
@@ -1736,11 +1780,11 @@
/* Oriya */
case 0x0B57u : *a = no decomp, -> RIGHT; return true;
#endif
}
- if ((ab == 0x0DDAu || hb_in_range (ab, 0x0DDCu, 0x0DDEu)))
+ if ((ab == 0x0DDAu || hb_in_range<hb_codepoint_t> (ab, 0x0DDCu, 0x0DDEu)))
{
/*
* Sinhala split matras... Let the fun begin.
*
* These four characters have Unicode decompositions. However, Uniscribe
@@ -1801,20 +1845,20 @@
}
const hb_ot_complex_shaper_t _hb_ot_complex_shaper_indic =
{
- "indic",
collect_features_indic,
override_features_indic,
data_create_indic,
data_destroy_indic,
- NULL, /* preprocess_text */
- NULL, /* postprocess_glyphs */
+ nullptr, /* preprocess_text */
+ nullptr, /* postprocess_glyphs */
HB_OT_SHAPE_NORMALIZATION_MODE_COMPOSED_DIACRITICS_NO_SHORT_CIRCUIT,
decompose_indic,
compose_indic,
setup_masks_indic,
- NULL, /* disable_otl */
+ nullptr, /* disable_otl */
+ nullptr, /* reorder_marks */
HB_OT_SHAPE_ZERO_WIDTH_MARKS_NONE,
false, /* fallback_position */
};
< prev index next >