summaryrefslogtreecommitdiff
path: root/src/hb-ot-shape-complex-indic.cc
diff options
context:
space:
mode:
Diffstat (limited to 'src/hb-ot-shape-complex-indic.cc')
-rw-r--r--src/hb-ot-shape-complex-indic.cc213
1 files changed, 88 insertions, 125 deletions
diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc
index 7723600..1e151a7 100644
--- a/src/hb-ot-shape-complex-indic.cc
+++ b/src/hb-ot-shape-complex-indic.cc
@@ -142,7 +142,7 @@ 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 (info.indic_category()) & flags);
+ return !!(FLAG_SAFE (info.indic_category()) & flags);
}
static inline bool
@@ -176,24 +176,8 @@ set_indic_properties (hb_glyph_info_t &info)
* Re-assign category
*/
-
- /* The spec says U+0952 is OT_A. However, testing shows that Uniscribe
- * treats a whole bunch of characters similarly.
- * TESTS: For example, for U+0951:
- * U+092E,U+0947,U+0952
- * U+092E,U+0952,U+0947
- * U+092E,U+0947,U+0951
- * U+092E,U+0951,U+0947
- * U+092E,U+0951,U+0952
- * U+092E,U+0952,U+0951
- */
- if (unlikely (hb_in_ranges (u, 0x0951u, 0x0952u,
- 0x1CD0u, 0x1CD2u,
- 0x1CD4u, 0x1CE1u) ||
- u == 0x1CF4u))
- cat = OT_A;
/* The following act more like the Bindus. */
- else if (unlikely (hb_in_range (u, 0x0953u, 0x0954u)))
+ if (unlikely (hb_in_range (u, 0x0953u, 0x0954u)))
cat = OT_SM;
/* The following act like consonants. */
else if (unlikely (hb_in_ranges (u, 0x0A72u, 0x0A73u,
@@ -216,15 +200,12 @@ set_indic_properties (hb_glyph_info_t &info)
cat = OT_Symbol;
ASSERT_STATIC ((int) INDIC_SYLLABIC_CATEGORY_AVAGRAHA == OT_Symbol);
}
- else if (unlikely (hb_in_range (u, 0x17CDu, 0x17D1u) ||
- u == 0x17CBu || u == 0x17D3u || u == 0x17DDu)) /* Khmer Various signs */
+ else if (unlikely (u == 0x17DDu)) /* https://github.com/roozbehp/unicode-data/issues/2 */
{
- /* These are like Top Matras. */
cat = OT_M;
pos = POS_ABOVE_C;
}
else if (unlikely (u == 0x17C6u)) cat = OT_N; /* Khmer Bindu doesn't like to be repositioned. */
- else if (unlikely (u == 0x17D2u)) cat = OT_Coeng; /* Khmer coeng */
else if (unlikely (hb_in_range (u, 0x2010u, 0x2011u)))
cat = OT_PLACEHOLDER;
else if (unlikely (u == 0x25CCu)) cat = OT_DOTTEDCIRCLE;
@@ -237,7 +218,7 @@ set_indic_properties (hb_glyph_info_t &info)
* Re-assign position.
*/
- if ((FLAG (cat) & CONSONANT_FLAGS))
+ if ((FLAG_SAFE (cat) & CONSONANT_FLAGS))
{
pos = POS_BASE_C;
if (is_ra (u))
@@ -247,7 +228,7 @@ set_indic_properties (hb_glyph_info_t &info)
{
pos = matra_position (u, pos);
}
- else if ((FLAG (cat) & (FLAG (OT_SM) | FLAG (OT_VD) | FLAG (OT_A) | FLAG (OT_Symbol))))
+ else if ((FLAG_SAFE (cat) & (FLAG (OT_SM) | FLAG (OT_VD) | FLAG (OT_A) | FLAG (OT_Symbol))))
{
pos = POS_SMVD;
}
@@ -512,12 +493,12 @@ struct indic_shape_plan_t
hb_codepoint_t glyph = virama_glyph;
if (unlikely (virama_glyph == (hb_codepoint_t) -1))
{
- if (!config->virama || !font->get_glyph (config->virama, 0, &glyph))
+ if (!config->virama || !font->get_nominal_glyph (config->virama, &glyph))
glyph = 0;
/* Technically speaking, the spec says we should apply 'locl' to virama too.
* Maybe one day... */
- /* Our get_glyph() function needs a font, so we can't get the virama glyph
+ /* 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;
}
@@ -557,8 +538,15 @@ data_create_indic (const hb_ot_shape_plan_t *plan)
indic_plan->virama_glyph = (hb_codepoint_t) -1;
/* Use zero-context would_substitute() matching for new-spec of the main
- * Indic scripts, and scripts with one spec only, but not for old-specs. */
- bool zero_context = !indic_plan->is_old_spec;
+ * Indic scripts, and scripts with one spec only, but not for old-specs.
+ * The new-spec for all dual-spec scripts says zero-context matching happens.
+ *
+ * However, testing with Malayalam shows that old and new spec both allow
+ * context. Testing with Bengali new-spec however shows that it doesn't.
+ * So, the heuristic here is the way it is. It should *only* be changed,
+ * as we discover more cases of what Windows does. DON'T TOUCH OTHERWISE.
+ */
+ bool zero_context = !indic_plan->is_old_spec && plan->props.script != HB_SCRIPT_MALAYALAM;
indic_plan->rphf.init (&plan->map, HB_TAG('r','p','h','f'), zero_context);
indic_plan->pref.init (&plan->map, HB_TAG('p','r','e','f'), zero_context);
indic_plan->blwf.init (&plan->map, HB_TAG('b','l','w','f'), zero_context);
@@ -754,10 +742,6 @@ initial_reordering_consonant_syllable (const hb_ot_shape_plan_t *plan,
switch (indic_plan->config->base_pos)
{
- default:
- assert (false);
- /* fallthrough */
-
case BASE_POS_LAST:
{
/* -> starting from the end of the syllable, move backwards */
@@ -963,7 +947,7 @@ initial_reordering_consonant_syllable (const hb_ot_shape_plan_t *plan,
indic_position_t last_pos = POS_START;
for (unsigned int i = start; i < end; i++)
{
- if ((FLAG (info[i].indic_category()) & (JOINER_FLAGS | FLAG (OT_N) | FLAG (OT_RS) | MEDIAL_FLAGS | HALANT_OR_COENG_FLAGS)))
+ if ((FLAG_SAFE (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 &&
@@ -1012,7 +996,7 @@ initial_reordering_consonant_syllable (const hb_ot_shape_plan_t *plan,
info[i].syllable() = i - start;
/* Sit tight, rock 'n roll! */
- hb_bubble_sort (info + start, end - start, compare_indic_order);
+ hb_stable_sort (info + start, end - start, compare_indic_order);
/* Find base again */
base = end;
for (unsigned int i = start; i < end; i++)
@@ -1025,7 +1009,11 @@ initial_reordering_consonant_syllable (const hb_ot_shape_plan_t *plan,
* around like crazy. In old-spec mode, we move halants around, so in
* that case merge all clusters after base. Otherwise, check the sort
* order and merge as needed.
- * For pre-base stuff, we handle cluster issues in final reordering. */
+ * For pre-base stuff, we handle cluster issues in final reordering.
+ *
+ * We could use buffer->sort() for this, if there was no special
+ * reordering of pre-base stuff happening later...
+ */
if (indic_plan->is_old_spec || end - base > 127)
buffer->merge_clusters (base, end);
else
@@ -1161,17 +1149,6 @@ initial_reordering_consonant_syllable (const hb_ot_shape_plan_t *plan,
}
}
-
-static void
-initial_reordering_vowel_syllable (const hb_ot_shape_plan_t *plan,
- hb_face_t *face,
- hb_buffer_t *buffer,
- unsigned int start, unsigned int end)
-{
- /* We made the vowels look like consonants. So let's call the consonant logic! */
- initial_reordering_consonant_syllable (plan, face, buffer, start, end);
-}
-
static void
initial_reordering_standalone_cluster (const hb_ot_shape_plan_t *plan,
hb_face_t *face,
@@ -1194,50 +1171,27 @@ initial_reordering_standalone_cluster (const hb_ot_shape_plan_t *plan,
}
static void
-initial_reordering_broken_cluster (const hb_ot_shape_plan_t *plan,
- hb_face_t *face,
- hb_buffer_t *buffer,
- unsigned int start, unsigned int end)
-{
- /* We already inserted dotted-circles, so just call the standalone_cluster. */
- initial_reordering_standalone_cluster (plan, face, buffer, start, end);
-}
-
-static void
-initial_reordering_symbol_cluster (const hb_ot_shape_plan_t *plan HB_UNUSED,
- hb_face_t *face HB_UNUSED,
- hb_buffer_t *buffer HB_UNUSED,
- unsigned int start HB_UNUSED, unsigned int end HB_UNUSED)
-{
- /* Nothing to do right now. If we ever switch to using the output
- * buffer in the reordering process, we'd need to next_glyph() here. */
-}
-
-static void
-initial_reordering_non_indic_cluster (const hb_ot_shape_plan_t *plan HB_UNUSED,
- hb_face_t *face HB_UNUSED,
- hb_buffer_t *buffer HB_UNUSED,
- unsigned int start HB_UNUSED, unsigned int end HB_UNUSED)
-{
- /* Nothing to do right now. If we ever switch to using the output
- * buffer in the reordering process, we'd need to next_glyph() here. */
-}
-
-
-static void
initial_reordering_syllable (const hb_ot_shape_plan_t *plan,
hb_face_t *face,
hb_buffer_t *buffer,
unsigned int start, unsigned int end)
{
syllable_type_t syllable_type = (syllable_type_t) (buffer->info[start].syllable() & 0x0F);
- switch (syllable_type) {
- case consonant_syllable: initial_reordering_consonant_syllable (plan, face, buffer, start, end); return;
- case vowel_syllable: initial_reordering_vowel_syllable (plan, face, buffer, start, end); return;
- case standalone_cluster: initial_reordering_standalone_cluster (plan, face, buffer, start, end); return;
- case symbol_cluster: initial_reordering_symbol_cluster (plan, face, buffer, start, end); return;
- case broken_cluster: initial_reordering_broken_cluster (plan, face, buffer, start, end); return;
- case non_indic_cluster: initial_reordering_non_indic_cluster (plan, face, buffer, start, end); return;
+ switch (syllable_type)
+ {
+ case vowel_syllable: /* We made the vowels look like consonants. So let's call the consonant logic! */
+ case consonant_syllable:
+ initial_reordering_consonant_syllable (plan, face, buffer, start, end);
+ break;
+
+ case broken_cluster: /* We already inserted dotted-circles, so just call the standalone_cluster. */
+ case standalone_cluster:
+ initial_reordering_standalone_cluster (plan, face, buffer, start, end);
+ break;
+
+ case symbol_cluster:
+ case non_indic_cluster:
+ break;
}
}
@@ -1261,7 +1215,7 @@ insert_dotted_circles (const hb_ot_shape_plan_t *plan HB_UNUSED,
hb_codepoint_t dottedcircle_glyph;
- if (!font->get_glyph (0x25CCu, 0, &dottedcircle_glyph))
+ if (!font->get_nominal_glyph (0x25CCu, &dottedcircle_glyph))
return;
hb_glyph_info_t dottedcircle = {0};
@@ -1273,7 +1227,7 @@ insert_dotted_circles (const hb_ot_shape_plan_t *plan HB_UNUSED,
buffer->idx = 0;
unsigned int last_syllable = 0;
- while (buffer->idx < buffer->len)
+ while (buffer->idx < buffer->len && !buffer->in_error)
{
unsigned int syllable = buffer->cur().syllable();
syllable_type_t syllable_type = (syllable_type_t) (syllable & 0x0F);
@@ -1281,19 +1235,19 @@ insert_dotted_circles (const hb_ot_shape_plan_t *plan HB_UNUSED,
{
last_syllable = syllable;
- hb_glyph_info_t info = dottedcircle;
- info.cluster = buffer->cur().cluster;
- info.mask = buffer->cur().mask;
- info.syllable() = buffer->cur().syllable();
+ hb_glyph_info_t ginfo = dottedcircle;
+ ginfo.cluster = buffer->cur().cluster;
+ ginfo.mask = buffer->cur().mask;
+ ginfo.syllable() = buffer->cur().syllable();
/* TODO Set glyph_props? */
/* Insert dottedcircle after possible Repha. */
- while (buffer->idx < buffer->len &&
+ while (buffer->idx < buffer->len && !buffer->in_error &&
last_syllable == buffer->cur().syllable() &&
buffer->cur().indic_category() == OT_Repha)
buffer->next_glyph ();
- buffer->output_info (info);
+ buffer->output_info (ginfo);
}
else
buffer->next_glyph ();
@@ -1310,18 +1264,8 @@ initial_reordering (const hb_ot_shape_plan_t *plan,
update_consonant_positions (plan, font, buffer);
insert_dotted_circles (plan, font, buffer);
- hb_glyph_info_t *info = buffer->info;
- unsigned int count = buffer->len;
- if (unlikely (!count)) return;
- unsigned int last = 0;
- unsigned int last_syllable = info[0].syllable();
- for (unsigned int i = 1; i < count; i++)
- if (last_syllable != info[i].syllable()) {
- initial_reordering_syllable (plan, font->face, buffer, last, i);
- last = i;
- last_syllable = info[last].syllable();
- }
- initial_reordering_syllable (plan, font->face, buffer, last, count);
+ foreach_syllable (buffer, start, end)
+ initial_reordering_syllable (plan, font->face, buffer, start, end);
}
static void
@@ -1388,6 +1332,25 @@ final_reordering_syllable (const hb_ot_shape_plan_t *plan,
break;
}
}
+ /* For Malayalam, skip over unformed below- (but NOT post-) forms. */
+ if (buffer->props.script == HB_SCRIPT_MALAYALAM)
+ {
+ for (unsigned int i = base + 1; i < end; i++)
+ {
+ while (i < end && is_joiner (info[i]))
+ i++;
+ if (i == end || !is_halant_or_coeng (info[i]))
+ break;
+ i++; /* Skip halant. */
+ while (i < end && is_joiner (info[i]))
+ i++;
+ if (i < end && is_consonant (info[i]) && info[i].indic_position() == POS_BELOW_C)
+ {
+ base = i;
+ info[base].indic_position() = POS_BASE_C;
+ }
+ }
+ }
if (start < base && info[base].indic_position() > POS_BASE_C)
base--;
@@ -1448,12 +1411,17 @@ final_reordering_syllable (const hb_ot_shape_plan_t *plan,
if (info[i - 1].indic_position () == POS_PRE_M)
{
unsigned int old_pos = i - 1;
+ if (old_pos < base && base <= new_pos) /* Shouldn't actually happen. */
+ base--;
+
hb_glyph_info_t tmp = info[old_pos];
memmove (&info[old_pos], &info[old_pos + 1], (new_pos - old_pos) * sizeof (info[0]));
info[new_pos] = tmp;
- if (old_pos < base && base <= new_pos) /* Shouldn't actually happen. */
- base--;
+
+ /* Note: this merge_clusters() is intentionally *after* the reordering.
+ * Indic matra reordering is special and tricky... */
buffer->merge_clusters (new_pos, MIN (end, base + 1));
+
new_pos--;
}
} else {
@@ -1550,7 +1518,7 @@ final_reordering_syllable (const hb_ot_shape_plan_t *plan,
{
new_reph_pos = base;
while (new_reph_pos < end &&
- !( FLAG (info[new_reph_pos + 1].indic_position()) & (FLAG (POS_POST_C) | FLAG (POS_AFTER_POST) | FLAG (POS_SMVD))))
+ !( FLAG_SAFE (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;
@@ -1606,12 +1574,12 @@ final_reordering_syllable (const hb_ot_shape_plan_t *plan,
reph_move:
{
- buffer->merge_clusters (start, new_reph_pos + 1);
-
/* Move */
+ buffer->merge_clusters (start, new_reph_pos + 1);
hb_glyph_info_t reph = info[start];
memmove (&info[start], &info[start + 1], (new_reph_pos - start) * sizeof (info[0]));
info[new_reph_pos] = reph;
+
if (start < base && base <= new_reph_pos)
base--;
}
@@ -1666,8 +1634,8 @@ final_reordering_syllable (const hb_ot_shape_plan_t *plan,
if (new_pos > start && info[new_pos - 1].indic_category() == OT_M)
{
unsigned int old_pos = i;
- for (unsigned int i = base + 1; i < old_pos; i++)
- if (info[i].indic_category() == OT_M)
+ for (unsigned int j = base + 1; j < old_pos; j++)
+ if (info[j].indic_category() == OT_M)
{
new_pos--;
break;
@@ -1684,10 +1652,12 @@ final_reordering_syllable (const hb_ot_shape_plan_t *plan,
{
unsigned int old_pos = i;
+
buffer->merge_clusters (new_pos, old_pos + 1);
hb_glyph_info_t tmp = info[old_pos];
memmove (&info[new_pos + 1], &info[new_pos], (old_pos - new_pos) * sizeof (info[0]));
info[new_pos] = tmp;
+
if (new_pos <= base && base < old_pos)
base++;
}
@@ -1701,7 +1671,7 @@ final_reordering_syllable (const hb_ot_shape_plan_t *plan,
/* Apply 'init' to the Left Matra if it's a word start. */
if (info[start].indic_position () == POS_PRE_M &&
(!start ||
- !(FLAG (_hb_glyph_info_get_general_category (&info[start - 1])) &
+ !(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))))
info[start].mask |= indic_plan->mask_array[INIT];
@@ -1737,16 +1707,8 @@ final_reordering (const hb_ot_shape_plan_t *plan,
unsigned int count = buffer->len;
if (unlikely (!count)) return;
- hb_glyph_info_t *info = buffer->info;
- unsigned int last = 0;
- unsigned int last_syllable = info[0].syllable();
- for (unsigned int i = 1; i < count; i++)
- if (last_syllable != info[i].syllable()) {
- final_reordering_syllable (plan, buffer, last, i);
- last = i;
- last_syllable = info[last].syllable();
- }
- final_reordering_syllable (plan, buffer, last, count);
+ foreach_syllable (buffer, start, end)
+ final_reordering_syllable (plan, buffer, start, end);
HB_BUFFER_DEALLOCATE_VAR (buffer, indic_category);
HB_BUFFER_DEALLOCATE_VAR (buffer, indic_position);
@@ -1837,7 +1799,7 @@ decompose_indic (const hb_ot_shape_normalize_context_t *c,
hb_codepoint_t glyph;
if (hb_options ().uniscribe_bug_compatible ||
- (c->font->get_glyph (ab, 0, &glyph) &&
+ (c->font->get_nominal_glyph (ab, &glyph) &&
indic_plan->pstf.would_substitute (&glyph, 1, c->font->face)))
{
/* Ok, safe to use Uniscribe-style decomposition. */
@@ -1847,7 +1809,7 @@ decompose_indic (const hb_ot_shape_normalize_context_t *c,
}
}
- return c->unicode->decompose (ab, a, b);
+ return (bool) c->unicode->decompose (ab, a, b);
}
static bool
@@ -1863,7 +1825,7 @@ compose_indic (const hb_ot_shape_normalize_context_t *c,
/* Composition-exclusion exceptions that we want to recompose. */
if (a == 0x09AFu && b == 0x09BCu) { *ab = 0x09DFu; return true; }
- return c->unicode->compose (a, b, ab);
+ return (bool) c->unicode->compose (a, b, ab);
}
@@ -1875,6 +1837,7 @@ const hb_ot_complex_shaper_t _hb_ot_complex_shaper_indic =
data_create_indic,
data_destroy_indic,
NULL, /* preprocess_text */
+ NULL, /* postprocess_glyphs */
HB_OT_SHAPE_NORMALIZATION_MODE_COMPOSED_DIACRITICS_NO_SHORT_CIRCUIT,
decompose_indic,
compose_indic,
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-03 09:19:30 +0000