Add Thai shaper that does SARA AM decomposition / reordering

That's not in the OpenType spec, but it's what MS and Adobe do.

diff --git a/src/hb-buffer.cc b/src/hb-buffer.cc
index 6b562aa..4a37e66 100644 (file)
--- a/src/hb-buffer.cc
+++ b/src/hb-buffer.cc
@@ -223,6 +223,7 @@ hb_buffer_t::swap_buffers (void)
   if (unlikely (in_error)) return;
 
   assert (have_output);
+  have_output = FALSE;
 
   if (out_info != info)
   {

diff --git a/src/hb-ot-shape-complex-misc.cc b/src/hb-ot-shape-complex-misc.cc
index 20e365b..d893eb3 100644 (file)
--- a/src/hb-ot-shape-complex-misc.cc
+++ b/src/hb-ot-shape-complex-misc.cc
@@ -27,7 +27,7 @@
 #include "hb-ot-shape-complex-private.hh"
 
 
-/* TODO Add kana, hangul, and other small shapers here */
+/* TODO Add kana, and other small shapers here */
 
 /* When adding trivial shapers, eg. kana, hangul, etc, we can either
  * add a full shaper enum value for them, or switch on the script in
@@ -54,6 +54,8 @@ _hb_ot_shape_complex_setup_masks_default (hb_ot_map_t *map, hb_buffer_t *buffer)
 
 
 
+/* Hangul shaper */
+
 void
 _hb_ot_shape_complex_collect_features_hangul (hb_ot_map_builder_t *map, const hb_segment_properties_t  *props)
 {
@@ -69,3 +71,103 @@ void
 _hb_ot_shape_complex_setup_masks_hangul (hb_ot_map_t *map, hb_buffer_t *buffer)
 {
 }
+
+
+
+/* Thai / Lao shaper */
+
+void
+_hb_ot_shape_complex_collect_features_thai (hb_ot_map_builder_t *map, const hb_segment_properties_t  *props)
+{
+}
+
+hb_ot_shape_normalization_mode_t
+_hb_ot_shape_complex_normalization_preference_thai (void)
+{
+  return HB_OT_SHAPE_NORMALIZATION_MODE_COMPOSED_FULL;
+}
+
+void
+_hb_ot_shape_complex_setup_masks_thai (hb_ot_map_t *map, hb_buffer_t *buffer)
+{
+  /* The following is NOT specified in the MS OT Thai spec, however, it seems
+   * to be what Uniscribe and other engines implement.  According to Eric Muller:
+   *
+   * When you have a sara am, decompose it in nikhahit + sara a, *and* mode the
+   * nihka hit backwards over any *tone* mark (0E48-0E4B).
+   *
+   * <0E14, 0E4B, 0E33> -> <0E14, 0E4D, 0E4B, 0E32>
+   *
+   * This reordering is legit only when the nikhahit comes from a sara am, not
+   * when it's there to start with. The string <0E14, 0E4B, 0E4D> is probably
+   * not what a u↪ser wanted, but the rendering is nevertheless nikhahit above
+   * chattawa.
+   *
+   * Same for Lao.
+   */
+
+  /*
+   * Here are the characters of significance:
+   *
+   *                   Thai    Lao
+   * SARA AM:          U+0E33  U+0EB3
+   * SARA AA:          U+0E32  U+0EB2
+   * Nikhahit:         U+0E4D  U+0ECD
+   *
+   * Tone marks:
+   * Thai:     <0E48..0E4B> CCC=107
+   * Lao:      <0EC8..0ECB> CCC=122
+   *
+   * Note how the Lao versions are the same as Thai + 0x80.
+   */
+
+  /* We only get one script at a time, so a script-agnostic implementation
+   * is adequate here. */
+#define IS_SARA_AM(x) (((x) & ~0x0080) == 0x0E33)
+#define NIKHAHIT_FROM_SARA_AM(x) ((x) - 0xE33 + 0xE4D)
+#define SARA_AA_FROM_SARA_AM(x) ((x) - 1)
+#define IS_TONE_MARK(x) (((x) & ~0x0083) == 0x0E48)
+
+  buffer->clear_output ();
+  unsigned int count = buffer->len;
+  for (buffer->idx = 0; buffer->idx < count;)
+  {
+    if (!IS_SARA_AM (buffer->info[buffer->idx].codepoint)) {
+      buffer->next_glyph ();
+      continue;
+    }
+
+    /* Is SARA AM. Decompose and reorder. */
+    uint16_t decomposed[2] = {NIKHAHIT_FROM_SARA_AM (buffer->info[buffer->idx].codepoint),
+                             SARA_AA_FROM_SARA_AM (buffer->info[buffer->idx].codepoint)};
+    buffer->replace_glyphs (1, 2, decomposed);
+    if (buffer->in_error)
+      return;
+
+    /* Ok, let's see... */
+    unsigned int end = buffer->out_len;
+    unsigned int start = end - 2;
+    while (start > 0 && IS_TONE_MARK (buffer->out_info[start - 1].codepoint))
+      start--;
+
+    /* Move Nikhahit (end-2) to the beginning */
+    hb_glyph_info_t t = buffer->out_info[end - 2];
+    memmove (buffer->out_info + start + 1,
+            buffer->out_info + start,
+            sizeof (buffer->out_info[0]) * (end - start - 2));
+    buffer->out_info[start] = t;
+
+    /* Make cluster */
+    for (; start > 0 && buffer->out_info[start - 1].cluster == buffer->out_info[start].cluster; start--)
+      ;
+    for (; buffer->idx < count;)
+      if (buffer->info[buffer->idx].cluster == buffer->out_info[buffer->out_len - 1].cluster)
+        buffer->next_glyph ();
+      else
+        break;
+    end = buffer->out_len;
+
+    buffer->merge_out_clusters (start, end);
+  }
+  buffer->swap_buffers ();
+}

diff --git a/src/hb-ot-shape-complex-private.hh b/src/hb-ot-shape-complex-private.hh
index 7c2c7a9..75948c0 100644 (file)
--- a/src/hb-ot-shape-complex-private.hh
+++ b/src/hb-ot-shape-complex-private.hh
@@ -52,6 +52,7 @@
   HB_COMPLEX_SHAPER_IMPLEMENT (arabic) \
   HB_COMPLEX_SHAPER_IMPLEMENT (hangul) \
   HB_COMPLEX_SHAPER_IMPLEMENT (indic) \
+  HB_COMPLEX_SHAPER_IMPLEMENT (thai) \
   /* ^--- Add new shapers here */
 
 enum hb_ot_complex_shaper_t {
@@ -91,6 +92,13 @@ hb_ot_shape_complex_categorize (const hb_segment_properties_t *props)
       return hb_ot_complex_shaper_hangul;
 
 
+    /* Unicode-1.1 additions */
+    case HB_SCRIPT_THAI:
+    case HB_SCRIPT_LAO:
+
+      return hb_ot_complex_shaper_thai;
+
+
 
     /* ^--- Add new shapers here */