* matra for the Indic shaper.
*/
+static void
+output_glyph (hb_ot_shape_context_t *c,
+ hb_codepoint_t glyph)
+{
+ hb_buffer_t *buffer = c->buffer;
+
+ buffer->output_glyph (glyph);
+ hb_glyph_info_set_unicode_props (&buffer->out_info[buffer->out_len - 1], buffer->unicode);
+}
static bool
decompose (hb_ot_shape_context_t *c,
bool has_a = hb_font_get_glyph (c->font, a, 0, &glyph);
if (shortest && has_a) {
/* Output a and b */
- c->buffer->output_glyph (a);
+ output_glyph (c, a);
if (b)
- c->buffer->output_glyph (b);
+ output_glyph (c, b);
return TRUE;
}
if (decompose (c, shortest, a)) {
if (b)
- c->buffer->output_glyph (b);
+ output_glyph (c, b);
return TRUE;
}
if (has_a) {
- c->buffer->output_glyph (a);
+ output_glyph (c, a);
if (b)
- c->buffer->output_glyph (b);
+ output_glyph (c, b);
return TRUE;
}
return FALSE;
}
-static bool
+static void
decompose_current_glyph (hb_ot_shape_context_t *c,
bool shortest)
{
- if (decompose (c, shortest, c->buffer->info[c->buffer->idx].codepoint)) {
+ if (decompose (c, shortest, c->buffer->info[c->buffer->idx].codepoint))
c->buffer->skip_glyph ();
- return TRUE;
- } else {
+ else
c->buffer->next_glyph ();
- return FALSE;
- }
}
-static bool
+static void
decompose_single_char_cluster (hb_ot_shape_context_t *c,
bool will_recompose)
{
/* If recomposing and font supports this, we're good to go */
if (will_recompose && hb_font_get_glyph (c->font, c->buffer->info[c->buffer->idx].codepoint, 0, &glyph)) {
c->buffer->next_glyph ();
- return FALSE;
+ return;
}
- return decompose_current_glyph (c, will_recompose);
+ decompose_current_glyph (c, will_recompose);
}
-static bool
+static void
decompose_multi_char_cluster (hb_ot_shape_context_t *c,
unsigned int end)
{
- bool changed = FALSE;
-
/* TODO Currently if there's a variation-selector we give-up, it's just too hard. */
for (unsigned int i = c->buffer->idx; i < end; i++)
if (unlikely (is_variation_selector (c->buffer->info[i].codepoint)))
- return changed;
+ return;
while (c->buffer->idx < end)
- changed |= decompose_current_glyph (c, FALSE);
+ decompose_current_glyph (c, FALSE);
+}
+
+static int
+compare_combining_class (const hb_glyph_info_t *pa, const hb_glyph_info_t *pb)
+{
+ unsigned int a = pa->combining_class();
+ unsigned int b = pb->combining_class();
- return changed;
+ return a < b ? -1 : a == b ? 0 : +1;
}
void
{
hb_buffer_t *buffer = c->buffer;
bool recompose = !hb_ot_shape_complex_prefer_decomposed (c->plan->shaper);
- bool changed = FALSE;
bool has_multichar_clusters = FALSE;
unsigned int count;
- buffer->clear_output ();
+ /* We do a fairly straightforward yet custom normalization process in three
+ * separate rounds: decompose, reorder, recompose (if desired). Currently
+ * this makes two buffer swaps. We can make it faster by moving the last
+ * two rounds into the inner loop for the first round, but it's more readable
+ * this way. */
/* First round, decompose */
+ buffer->clear_output ();
count = buffer->len;
for (buffer->idx = 0; buffer->idx < count;)
{
break;
if (buffer->idx + 1 == end)
- changed |= decompose_single_char_cluster (c, recompose);
+ decompose_single_char_cluster (c, recompose);
else {
- changed |= decompose_multi_char_cluster (c, end);
+ decompose_multi_char_cluster (c, end);
has_multichar_clusters = TRUE;
}
}
/* Technically speaking, two characters with ccc=0 may combine. But all
* those cases are in languages that the indic module handles (which expects
* decomposed), or in Hangul jamo, which again, we want decomposed anyway.
- * So we don't bother combining across cluster boundaries. */
+ * So we don't bother combining across cluster boundaries.
+ *
+ * TODO: Am I right about Hangul? If I am, we should add a Hangul module
+ * that requests decomposed. */
if (!has_multichar_clusters)
return; /* Done! */
- if (changed)
- _hb_set_unicode_props (c->buffer); /* BUFFER: Set general_category and combining_class in var1 */
-
/* Second round, reorder (inplace) */
continue;
}
- unsigned int k = end - i - 1;
- do {
- hb_glyph_info_t *pinfo = buffer->info + i;
- unsigned int new_k = 0;
-
- for (unsigned int j = 0; j < k; j++)
- if (pinfo[j].combining_class() > pinfo[j+1].combining_class()) {
- hb_glyph_info_t t;
- t = pinfo[j];
- pinfo[j] = pinfo[j + 1];
- pinfo[j + 1] = t;
-
- new_k = j;
- }
- k = new_k;
- } while (k);
+ hb_bubble_sort (buffer->info + i, end - i, compare_combining_class);
i = end;
}
+ if (!recompose)
+ return;
+
/* Third round, recompose */
- if (recompose) {
+ /* As noted in the comment earlier, we don't try to combine
+ * ccc=0 chars with their previous Starter. */
+ buffer->clear_output ();
+ count = buffer->len;
+ unsigned int starter = 0;
+ buffer->next_glyph ();
+ while (buffer->idx < count)
+ {
+ if (buffer->info[buffer->idx].combining_class() == 0) {
+ starter = buffer->out_len;
+ buffer->next_glyph ();
+ continue;
+ }
+ hb_codepoint_t composed, glyph;
+ if ((buffer->out_info[buffer->out_len - 1].combining_class() >=
+ buffer->info[buffer->idx].combining_class()) ||
+ !hb_unicode_compose (c->buffer->unicode,
+ buffer->out_info[starter].codepoint,
+ buffer->info[buffer->idx].codepoint,
+ &composed) ||
+ !hb_font_get_glyph (c->font, composed, 0, &glyph))
+ {
+ /* Blocked, or doesn't compose. */
+ buffer->next_glyph ();
+ continue;
+ }
+
+ /* Composes. Modify starter and carry on. */
+ buffer->out_info[starter].codepoint = composed;
+ hb_glyph_info_set_unicode_props (&buffer->out_info[starter], buffer->unicode);
+
+ buffer->skip_glyph ();
}
+ buffer->swap_buffers ();
}