[HarfBuzz] harfbuzz-ng: Branch 'master' - 5 commits

Behdad Esfahbod behdad at kemper.freedesktop.org
Tue Apr 10 07:52:40 PDT 2012


 TODO                                                                  |   10 
 src/hb-buffer-private.hh                                              |    5 
 src/hb-buffer.cc                                                      |   24 ++
 src/hb-ot-shape-complex-indic-machine.rl                              |   14 -
 src/hb-ot-shape-complex-misc.cc                                       |  104 +++++++++-
 src/hb-ot-shape-complex-private.hh                                    |    8 
 test/shaping/texts/in-tree/MANIFEST                                   |    1 
 test/shaping/texts/in-tree/shaper-default/MANIFEST                    |    1 
 test/shaping/texts/in-tree/shaper-default/script-hangul/MANIFEST      |    1 
 test/shaping/texts/in-tree/shaper-default/script-hangul/misc/MANIFEST |    1 
 test/shaping/texts/in-tree/shaper-default/script-hangul/misc/misc.txt |    2 
 test/shaping/texts/in-tree/shaper-hangul/MANIFEST                     |    1 
 test/shaping/texts/in-tree/shaper-hangul/script-hangul/MANIFEST       |    1 
 test/shaping/texts/in-tree/shaper-hangul/script-hangul/misc/MANIFEST  |    1 
 test/shaping/texts/in-tree/shaper-hangul/script-hangul/misc/misc.txt  |    2 
 test/shaping/texts/in-tree/shaper-thai/misc/misc.txt                  |    5 
 16 files changed, 156 insertions(+), 25 deletions(-)

New commits:
commit 3b26f96ebe859570d14c6902afc23462bca40712
Author: Behdad Esfahbod <behdad at behdad.org>
Date:   Tue Apr 10 10:52:07 2012 -0400

    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
--- 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
--- 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
--- 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 */
 
commit 0b6d2ac6a1d04877ae4542fc2a3b920185547053
Author: Behdad Esfahbod <behdad at behdad.org>
Date:   Tue Apr 10 10:52:03 2012 -0400

    Minor

diff --git a/TODO b/TODO
index 7a99cd2..1436337 100644
--- a/TODO
+++ b/TODO
@@ -58,13 +58,11 @@ API to add (maybe after 1.0):
 - Add hb-fribidi?
 
 
-hb-view enhancements:
-====================
+hb-view / hb-shape enhancements:
+===============================
 
-- Add --format
 - Add --width, --height, --auto-size, --align, etc?
-- Port to GOption, --help
-- Add XML and JSON formats
+- Add XML and JSON formats to hb-shape
 
 
 Tests to write:
@@ -82,4 +80,4 @@ Optimizations:
 
 - Avoid allocating blob objects internally for for_data() faces?
 
-- Add caching layer to hb-ft
+- Add caching layer to hb-ft?
commit e099dd6592b4ea887696330f4718efb572494d93
Author: Behdad Esfahbod <behdad at behdad.org>
Date:   Tue Apr 10 10:47:33 2012 -0400

    Add Thai test case for SARA AM decomposition

diff --git a/test/shaping/texts/in-tree/shaper-hangul/MANIFEST b/test/shaping/texts/in-tree/shaper-hangul/MANIFEST
new file mode 100644
index 0000000..ea81716
--- /dev/null
+++ b/test/shaping/texts/in-tree/shaper-hangul/MANIFEST
@@ -0,0 +1 @@
+script-hangul
diff --git a/test/shaping/texts/in-tree/shaper-thai/misc/misc.txt b/test/shaping/texts/in-tree/shaper-thai/misc/misc.txt
new file mode 100644
index 0000000..fc2dba9
--- /dev/null
+++ b/test/shaping/texts/in-tree/shaper-thai/misc/misc.txt
@@ -0,0 +1,5 @@
+ด๋ำ
+ดํ๋า
+ด๋ํา
+ดำ
+ำ
commit 4450dc9354b18cd68980b0891b24ea8efa4f38b6
Author: Behdad Esfahbod <behdad at behdad.org>
Date:   Sat Apr 7 22:07:23 2012 -0400

    Move around

diff --git a/test/shaping/texts/in-tree/MANIFEST b/test/shaping/texts/in-tree/MANIFEST
index 2448c4b..0119909 100644
--- a/test/shaping/texts/in-tree/MANIFEST
+++ b/test/shaping/texts/in-tree/MANIFEST
@@ -1,3 +1,4 @@
 shaper-arabic
 shaper-default
+shaper-hangul
 shaper-indic
diff --git a/test/shaping/texts/in-tree/shaper-default/MANIFEST b/test/shaping/texts/in-tree/shaper-default/MANIFEST
index 623eae4..5139532 100644
--- a/test/shaping/texts/in-tree/shaper-default/MANIFEST
+++ b/test/shaping/texts/in-tree/shaper-default/MANIFEST
@@ -1,3 +1,2 @@
-script-hangul
 script-hebrew
 script-linear-b
diff --git a/test/shaping/texts/in-tree/shaper-default/script-hangul/MANIFEST b/test/shaping/texts/in-tree/shaper-default/script-hangul/MANIFEST
deleted file mode 100644
index b8752e7..0000000
--- a/test/shaping/texts/in-tree/shaper-default/script-hangul/MANIFEST
+++ /dev/null
@@ -1 +0,0 @@
-misc
diff --git a/test/shaping/texts/in-tree/shaper-default/script-hangul/misc/MANIFEST b/test/shaping/texts/in-tree/shaper-default/script-hangul/misc/MANIFEST
deleted file mode 100644
index 29cfb2f..0000000
--- a/test/shaping/texts/in-tree/shaper-default/script-hangul/misc/MANIFEST
+++ /dev/null
@@ -1 +0,0 @@
-misc.txt
diff --git a/test/shaping/texts/in-tree/shaper-default/script-hangul/misc/misc.txt b/test/shaping/texts/in-tree/shaper-default/script-hangul/misc/misc.txt
deleted file mode 100644
index 9c374b9..0000000
--- a/test/shaping/texts/in-tree/shaper-default/script-hangul/misc/misc.txt
+++ /dev/null
@@ -1,2 +0,0 @@
-휴가 가-- (오--)
-휴가 가-- (오--)
diff --git a/test/shaping/texts/in-tree/shaper-hangul/script-hangul/MANIFEST b/test/shaping/texts/in-tree/shaper-hangul/script-hangul/MANIFEST
new file mode 100644
index 0000000..b8752e7
--- /dev/null
+++ b/test/shaping/texts/in-tree/shaper-hangul/script-hangul/MANIFEST
@@ -0,0 +1 @@
+misc
diff --git a/test/shaping/texts/in-tree/shaper-hangul/script-hangul/misc/MANIFEST b/test/shaping/texts/in-tree/shaper-hangul/script-hangul/misc/MANIFEST
new file mode 100644
index 0000000..29cfb2f
--- /dev/null
+++ b/test/shaping/texts/in-tree/shaper-hangul/script-hangul/misc/MANIFEST
@@ -0,0 +1 @@
+misc.txt
diff --git a/test/shaping/texts/in-tree/shaper-hangul/script-hangul/misc/misc.txt b/test/shaping/texts/in-tree/shaper-hangul/script-hangul/misc/misc.txt
new file mode 100644
index 0000000..9c374b9
--- /dev/null
+++ b/test/shaping/texts/in-tree/shaper-hangul/script-hangul/misc/misc.txt
@@ -0,0 +1,2 @@
+휴가 가-- (오--)
+휴가 가-- (오--)
commit d4cc44716c1e098f8abbc0e495404598026ef242
Author: Behdad Esfahbod <behdad at behdad.org>
Date:   Sat Apr 7 21:52:28 2012 -0400

    Move code around, in prep for Thai/Lao shaper

diff --git a/src/hb-buffer-private.hh b/src/hb-buffer-private.hh
index 4292222..06d4912 100644
--- a/src/hb-buffer-private.hh
+++ b/src/hb-buffer-private.hh
@@ -131,6 +131,11 @@ struct _hb_buffer_t {
 			      unsigned int cluster_start,
 			      unsigned int cluster_end);
 
+  HB_INTERNAL void merge_clusters (unsigned int start,
+				   unsigned int end);
+  HB_INTERNAL void merge_out_clusters (unsigned int start,
+				       unsigned int end);
+
   /* Internal methods */
   HB_INTERNAL bool enlarge (unsigned int size);
 
diff --git a/src/hb-buffer.cc b/src/hb-buffer.cc
index e8bdfb1..6b562aa 100644
--- a/src/hb-buffer.cc
+++ b/src/hb-buffer.cc
@@ -431,6 +431,29 @@ hb_buffer_t::reverse_clusters (void)
 }
 
 void
+hb_buffer_t::merge_clusters (unsigned int start,
+			     unsigned int end)
+{
+  unsigned int cluster = this->info[start].cluster;
+
+  for (unsigned int i = start + 1; i < end; i++)
+    cluster = MIN (cluster, this->info[i].cluster);
+  for (unsigned int i = start; i < end; i++)
+    this->info[i].cluster = cluster;
+}
+void
+hb_buffer_t::merge_out_clusters (unsigned int start,
+				 unsigned int end)
+{
+  unsigned int cluster = this->out_info[start].cluster;
+
+  for (unsigned int i = start + 1; i < end; i++)
+    cluster = MIN (cluster, this->out_info[i].cluster);
+  for (unsigned int i = start; i < end; i++)
+    this->out_info[i].cluster = cluster;
+}
+
+void
 hb_buffer_t::guess_properties (void)
 {
   /* If script is set to INVALID, guess from buffer contents */
diff --git a/src/hb-ot-shape-complex-indic-machine.rl b/src/hb-ot-shape-complex-indic-machine.rl
index bcca6da..7af23c1 100644
--- a/src/hb-ot-shape-complex-indic-machine.rl
+++ b/src/hb-ot-shape-complex-indic-machine.rl
@@ -64,7 +64,7 @@ action found_vowel_syllable { found_vowel_syllable (map, buffer, mask_array, las
 action found_standalone_cluster { found_standalone_cluster (map, buffer, mask_array, last, p); }
 action found_non_indic { found_non_indic (map, buffer, mask_array, last, p); }
 
-action next_syllable { set_cluster (buffer, last, p); last = p; }
+action next_syllable { buffer->merge_clusters (last, p); last = p; }
 
 consonant_syllable =	(c.N? (z.H|H.z?))* c.N? A? (H.z? | matra_group*)? syllable_tail %(found_consonant_syllable);
 vowel_syllable =	(Ra H)? V N? (z.H.c | ZWJ.c)? matra_group* syllable_tail %(found_vowel_syllable);
@@ -84,18 +84,6 @@ main := (syllable %(next_syllable))**;
 
 
 static void
-set_cluster (hb_buffer_t *buffer,
-	     unsigned int start, unsigned int end)
-{
-  unsigned int cluster = buffer->info[start].cluster;
-
-  for (unsigned int i = start + 1; i < end; i++)
-    cluster = MIN (cluster, buffer->info[i].cluster);
-  for (unsigned int i = start; i < end; i++)
-    buffer->info[i].cluster = cluster;
-}
-
-static void
 find_syllables (const hb_ot_map_t *map, hb_buffer_t *buffer, hb_mask_t *mask_array)
 {
   unsigned int p, pe, eof;



More information about the HarfBuzz mailing list