[Indic] Refactor

Move all the logic that needs to eventually move into the indic table
into hb-ot-shape-complex-indic-private.hh.

src/hb-ot-shape-complex-indic-private.hh

@@ -31,6 +31,7 @@
 
 
 #include "hb-ot-shape-complex-private.hh"
+#include "hb-ot-shape-private.hh" /* XXX Remove */
 
 
 /* buffer var allocations */
@@ -152,6 +153,79 @@ enum indic_matra_category_t {
 #include "hb-ot-shape-complex-indic-table.hh"
 
 
+#define IN_HALF_BLOCK(u, Base) (((u) & ~0x7F) == (Base))
+
+#define IS_DEVA(u) (IN_HALF_BLOCK (u, 0x0900))
+#define IS_BENG(u) (IN_HALF_BLOCK (u, 0x0980))
+#define IS_GURM(u) (IN_HALF_BLOCK (u, 0x0A00))
+#define IS_GUJA(u) (IN_HALF_BLOCK (u, 0x0A80))
+#define IS_ORYA(u) (IN_HALF_BLOCK (u, 0x0B00))
+#define IS_TAML(u) (IN_HALF_BLOCK (u, 0x0B80))
+#define IS_TELU(u) (IN_HALF_BLOCK (u, 0x0C00))
+#define IS_KNDA(u) (IN_HALF_BLOCK (u, 0x0C80))
+#define IS_MLYM(u) (IN_HALF_BLOCK (u, 0x0D00))
+#define IS_SINH(u) (IN_HALF_BLOCK (u, 0x0D80))
+#define IS_KHMR(u) (IN_HALF_BLOCK (u, 0x1780))
+
+
+#define MATRA_POS_LEFT(u)	POS_PRE_M
+#define MATRA_POS_RIGHT(u)	( \
+				  IS_DEVA(u) ? POS_AFTER_SUB  : \
+				  IS_BENG(u) ? POS_AFTER_POST : \
+				  IS_GURM(u) ? POS_AFTER_POST : \
+				  IS_GUJA(u) ? POS_AFTER_POST : \
+				  IS_ORYA(u) ? POS_AFTER_POST : \
+				  IS_TAML(u) ? POS_AFTER_POST : \
+				  IS_TELU(u) ? (u <= 0x0C42 ? POS_BEFORE_SUB : POS_AFTER_SUB) : \
+				  IS_KNDA(u) ? (u < 0x0CC3 || u > 0xCD6 ? POS_BEFORE_SUB : POS_AFTER_SUB) : \
+				  IS_MLYM(u) ? POS_AFTER_POST : \
+				  IS_SINH(u) ? POS_AFTER_SUB  : \
+				  IS_KHMR(u) ? POS_AFTER_POST : \
+				  /*default*/  POS_AFTER_SUB    \
+				)
+#define MATRA_POS_TOP(u)	( /* BENG and MLYM don't have top matras. */ \
+				  IS_DEVA(u) ? POS_AFTER_SUB  : \
+				  IS_GURM(u) ? POS_AFTER_POST : /* Deviate from spec */ \
+				  IS_GUJA(u) ? POS_AFTER_SUB  : \
+				  IS_ORYA(u) ? POS_AFTER_MAIN : \
+				  IS_TAML(u) ? POS_AFTER_SUB  : \
+				  IS_TELU(u) ? POS_BEFORE_SUB : \
+				  IS_KNDA(u) ? POS_BEFORE_SUB : \
+				  IS_SINH(u) ? POS_AFTER_SUB  : \
+				  IS_KHMR(u) ? POS_AFTER_POST : \
+				  /*default*/  POS_AFTER_SUB    \
+				)
+#define MATRA_POS_BOTTOM(u)	( \
+				  IS_DEVA(u) ? POS_AFTER_SUB  : \
+				  IS_BENG(u) ? POS_AFTER_SUB  : \
+				  IS_GURM(u) ? POS_AFTER_POST : \
+				  IS_GUJA(u) ? POS_AFTER_POST : \
+				  IS_ORYA(u) ? POS_AFTER_SUB  : \
+				  IS_TAML(u) ? POS_AFTER_POST : \
+				  IS_TELU(u) ? POS_BEFORE_SUB : \
+				  IS_KNDA(u) ? POS_BEFORE_SUB : \
+				  IS_MLYM(u) ? POS_AFTER_POST : \
+				  IS_SINH(u) ? POS_AFTER_SUB  : \
+				  IS_KHMR(u) ? POS_AFTER_POST : \
+				  /*default*/  POS_AFTER_SUB    \
+				)
+
+
+static inline indic_position_t
+matra_position (hb_codepoint_t u, indic_position_t side)
+{
+  switch ((int) side)
+  {
+    case POS_PRE_C:	return MATRA_POS_LEFT (u);
+    case POS_POST_C:	return MATRA_POS_RIGHT (u);
+    case POS_ABOVE_C:	return MATRA_POS_TOP (u);
+    case POS_BELOW_C:	return MATRA_POS_BOTTOM (u);
+  };
+  abort ();
+}
+
+
+
 /* XXX
  * This is a hack for now.  We should move this data into the main Indic table.
  * Or completely remove it and just check in the tables.
@@ -173,5 +247,141 @@ static const hb_codepoint_t ra_chars[] = {
   0x179A, /* Khmer */		/* No Reph, Visual Repha */
 };
 
+static inline indic_position_t
+consonant_position (hb_codepoint_t  u)
+{
+  if ((u & ~0x007F) == 0x1780)
+    return POS_BELOW_C; /* In Khmer coeng model, post and below forms should not be reordered. */
+  return POS_BASE_C; /* Will recategorize later based on font lookups. */
+}
+
+static inline bool
+is_ra (hb_codepoint_t u)
+{
+  for (unsigned int i = 0; i < ARRAY_LENGTH (ra_chars); i++)
+    if (u == ra_chars[i])
+      return true;
+  return false;
+}
+
+
+static inline bool
+is_one_of (const hb_glyph_info_t &info, unsigned int flags)
+{
+  /* If it ligated, all bets are off. */
+  if (is_a_ligature (info)) return false;
+  return !!(FLAG (info.indic_category()) & flags);
+}
+
+#define JOINER_FLAGS (FLAG (OT_ZWJ) | FLAG (OT_ZWNJ))
+static inline bool
+is_joiner (const hb_glyph_info_t &info)
+{
+  return is_one_of (info, JOINER_FLAGS);
+}
+
+/* Note:
+ *
+ * We treat Vowels and placeholders as if they were consonants.  This is safe because Vowels
+ * cannot happen in a consonant syllable.  The plus side however is, we can call the
+ * consonant syllable logic from the vowel syllable function and get it all right! */
+#define CONSONANT_FLAGS (FLAG (OT_C) | FLAG (OT_Ra) | FLAG (OT_V) | FLAG (OT_NBSP) | FLAG (OT_DOTTEDCIRCLE))
+static inline bool
+is_consonant (const hb_glyph_info_t &info)
+{
+  return is_one_of (info, CONSONANT_FLAGS);
+}
+
+#define HALANT_OR_COENG_FLAGS (FLAG (OT_H) | FLAG (OT_Coeng))
+static inline bool
+is_halant_or_coeng (const hb_glyph_info_t &info)
+{
+  return is_one_of (info, HALANT_OR_COENG_FLAGS);
+}
+
+static inline void
+set_indic_properties (hb_glyph_info_t   &info)
+{
+  hb_codepoint_t u = info.codepoint;
+  unsigned int type = get_indic_categories (u);
+  indic_category_t cat = (indic_category_t) (type & 0x0F);
+  indic_position_t pos = (indic_position_t) (type >> 4);
+
+
+  /*
+   * Re-assign category
+   */
+
+
+  /* The spec says U+0952 is OT_A.  However, testing shows that Uniscribe
+   * treats U+0951..U+0952 all as OT_VD.
+   * TESTS:
+   * U+092E,U+0947,U+0952
+   * U+092E,U+0952,U+0947
+   * U+092E,U+0947,U+0951
+   * U+092E,U+0951,U+0947
+   * */
+  if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x0951, 0x0954)))
+    cat = OT_VD;
+
+  if (unlikely (u == 0x17D1))
+    cat = OT_X;
+  if (cat == OT_X &&
+      unlikely (hb_in_range<hb_codepoint_t> (u, 0x17CB, 0x17D3))) /* Khmer Various signs */
+  {
+    /* These are like Top Matras. */
+    cat = OT_M;
+    pos = POS_ABOVE_C;
+  }
+  if (u == 0x17C6) /* Khmer Bindu doesn't like to be repositioned. */
+    cat = OT_N;
+
+  if (unlikely (u == 0x17D2)) cat = OT_Coeng; /* Khmer coeng */
+  else if (unlikely (u == 0x200C)) cat = OT_ZWNJ;
+  else if (unlikely (u == 0x200D)) cat = OT_ZWJ;
+  else if (unlikely (u == 0x25CC)) cat = OT_DOTTEDCIRCLE;
+  else if (unlikely (u == 0x0A71)) cat = OT_SM; /* GURMUKHI ADDAK.  More like consonant medial. like 0A75. */
+
+  if (cat == OT_Repha) {
+    /* There are two kinds of characters marked as Repha:
+     * - The ones that are GenCat=Mn are already positioned visually, ie. after base. (eg. Khmer)
+     * - The ones that are GenCat=Lo is encoded logically, ie. beginning of syllable. (eg. Malayalam)
+     *
+     * We recategorize the first kind to look like a Nukta and attached to the base directly.
+     */
+    if (_hb_glyph_info_get_general_category (&info) == HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK)
+      cat = OT_N;
+  }
+
+
+
+  /*
+   * Re-assign position.
+   */
+
+  if ((FLAG (cat) & CONSONANT_FLAGS))
+  {
+    pos = consonant_position (u);
+    if (is_ra (u))
+      cat = OT_Ra;
+  }
+  else if (cat == OT_M)
+  {
+    pos = matra_position (u, pos);
+  }
+  else if (cat == OT_SM || cat == OT_VD)
+  {
+    pos = POS_SMVD;
+  }
+
+  if (unlikely (u == 0x0B01)) pos = POS_BEFORE_SUB; /* Oriya Bindu is BeforeSub in the spec. */
+
+
+
+  info.indic_category() = cat;
+  info.indic_position() = pos;
+}
+
+
 
 #endif /* HB_OT_SHAPE_COMPLEX_INDIC_PRIVATE_HH */

src/hb-ot-shape-complex-indic.cc

@@ -29,21 +29,6 @@
 #include "hb-ot-layout-private.hh"
 
 
-#define IN_HALF_BLOCK(u, Base) (((u) & ~0x7F) == (Base))
-
-#define IS_DEVA(u) (IN_HALF_BLOCK (u, 0x0900))
-#define IS_BENG(u) (IN_HALF_BLOCK (u, 0x0980))
-#define IS_GURM(u) (IN_HALF_BLOCK (u, 0x0A00))
-#define IS_GUJA(u) (IN_HALF_BLOCK (u, 0x0A80))
-#define IS_ORYA(u) (IN_HALF_BLOCK (u, 0x0B00))
-#define IS_TAML(u) (IN_HALF_BLOCK (u, 0x0B80))
-#define IS_TELU(u) (IN_HALF_BLOCK (u, 0x0C00))
-#define IS_KNDA(u) (IN_HALF_BLOCK (u, 0x0C80))
-#define IS_MLYM(u) (IN_HALF_BLOCK (u, 0x0D00))
-#define IS_SINH(u) (IN_HALF_BLOCK (u, 0x0D80))
-#define IS_KHMR(u) (IN_HALF_BLOCK (u, 0x1780))
-
-
 #define OLD_INDIC_TAG(script) (((hb_tag_t) script) | 0x20000000)
 #define IS_OLD_INDIC_TAG(tag) ( \
 				(tag) == OLD_INDIC_TAG (HB_SCRIPT_BENGALI)	|| \
@@ -97,15 +82,6 @@ indic_options (void)
 }
 
 
-static int
-compare_codepoint (const void *pa, const void *pb)
-{
-  hb_codepoint_t a = * (hb_codepoint_t *) pa;
-  hb_codepoint_t b = * (hb_codepoint_t *) pb;
-
-  return a < b ? -1 : a == b ? 0 : +1;
-}
-
 struct indic_shape_plan_t
 {
   struct would_apply_feature_t
@@ -147,13 +123,10 @@ struct indic_shape_plan_t
 };
 
 static indic_position_t
-consonant_position (hb_codepoint_t  u,
-		    const indic_shape_plan_t *indic_plan,
-		    hb_font_t      *font)
+consonant_position_from_font (hb_codepoint_t  u,
+			      const indic_shape_plan_t *indic_plan,
+			      hb_font_t      *font)
 {
-  if ((u & ~0x007F) == 0x1780)
-    return POS_BELOW_C; /* In Khmer coeng model, post and below forms should not be reordered. */
-
   hb_codepoint_t virama = (u & ~0x007F) | 0x004D;
   if ((u & ~0x007F) == 0x0D80) virama = 0x0DCA; /* Sinahla */
   if ((u & ~0x007F) == 0x1780) virama = 0x17D2; /* Khmaer */
@@ -170,194 +143,6 @@ consonant_position (hb_codepoint_t  u,
   return POS_BASE_C;
 }
 
-#define MATRA_POS_LEFT(u)	POS_PRE_M
-#define MATRA_POS_RIGHT(u)	( \
-				  IS_DEVA(u) ? POS_AFTER_SUB  : \
-				  IS_BENG(u) ? POS_AFTER_POST : \
-				  IS_GURM(u) ? POS_AFTER_POST : \
-				  IS_GUJA(u) ? POS_AFTER_POST : \
-				  IS_ORYA(u) ? POS_AFTER_POST : \
-				  IS_TAML(u) ? POS_AFTER_POST : \
-				  IS_TELU(u) ? (u <= 0x0C42 ? POS_BEFORE_SUB : POS_AFTER_SUB) : \
-				  IS_KNDA(u) ? (u < 0x0CC3 || u > 0xCD6 ? POS_BEFORE_SUB : POS_AFTER_SUB) : \
-				  IS_MLYM(u) ? POS_AFTER_POST : \
-				  IS_SINH(u) ? POS_AFTER_SUB  : \
-				  IS_KHMR(u) ? POS_AFTER_POST : \
-				  /*default*/  POS_AFTER_SUB    \
-				)
-#define MATRA_POS_TOP(u)	( /* BENG and MLYM don't have top matras. */ \
-				  IS_DEVA(u) ? POS_AFTER_SUB  : \
-				  IS_GURM(u) ? POS_AFTER_POST : /* Deviate from spec */ \
-				  IS_GUJA(u) ? POS_AFTER_SUB  : \
-				  IS_ORYA(u) ? POS_AFTER_MAIN : \
-				  IS_TAML(u) ? POS_AFTER_SUB  : \
-				  IS_TELU(u) ? POS_BEFORE_SUB : \
-				  IS_KNDA(u) ? POS_BEFORE_SUB : \
-				  IS_SINH(u) ? POS_AFTER_SUB  : \
-				  IS_KHMR(u) ? POS_AFTER_POST : \
-				  /*default*/  POS_AFTER_SUB    \
-				)
-#define MATRA_POS_BOTTOM(u)	( \
-				  IS_DEVA(u) ? POS_AFTER_SUB  : \
-				  IS_BENG(u) ? POS_AFTER_SUB  : \
-				  IS_GURM(u) ? POS_AFTER_POST : \
-				  IS_GUJA(u) ? POS_AFTER_POST : \
-				  IS_ORYA(u) ? POS_AFTER_SUB  : \
-				  IS_TAML(u) ? POS_AFTER_POST : \
-				  IS_TELU(u) ? POS_BEFORE_SUB : \
-				  IS_KNDA(u) ? POS_BEFORE_SUB : \
-				  IS_MLYM(u) ? POS_AFTER_POST : \
-				  IS_SINH(u) ? POS_AFTER_SUB  : \
-				  IS_KHMR(u) ? POS_AFTER_POST : \
-				  /*default*/  POS_AFTER_SUB    \
-				)
-
-
-static indic_position_t
-matra_position (hb_codepoint_t u, indic_position_t side)
-{
-  switch ((int) side)
-  {
-    case POS_PRE_C:	return MATRA_POS_LEFT (u);
-    case POS_POST_C:	return MATRA_POS_RIGHT (u);
-    case POS_ABOVE_C:	return MATRA_POS_TOP (u);
-    case POS_BELOW_C:	return MATRA_POS_BOTTOM (u);
-  };
-  abort ();
-}
-
-static inline bool
-is_ra (hb_codepoint_t u)
-{
-  return !!bsearch (&u, ra_chars,
-		    ARRAY_LENGTH (ra_chars),
-		    sizeof (ra_chars[0]),
-		    compare_codepoint);
-}
-
-static inline bool
-is_one_of (const hb_glyph_info_t &info, unsigned int flags)
-{
-  /* If it ligated, all bets are off. */
-  if (is_a_ligature (info)) return false;
-  return !!(FLAG (info.indic_category()) & flags);
-}
-
-#define JOINER_FLAGS (FLAG (OT_ZWJ) | FLAG (OT_ZWNJ))
-static inline bool
-is_joiner (const hb_glyph_info_t &info)
-{
-  return is_one_of (info, JOINER_FLAGS);
-}
-
-/* Note:
- *
- * We treat Vowels and placeholders as if they were consonants.  This is safe because Vowels
- * cannot happen in a consonant syllable.  The plus side however is, we can call the
- * consonant syllable logic from the vowel syllable function and get it all right! */
-#define CONSONANT_FLAGS (FLAG (OT_C) | FLAG (OT_Ra) | FLAG (OT_V) | FLAG (OT_NBSP) | FLAG (OT_DOTTEDCIRCLE))
-static inline bool
-is_consonant (const hb_glyph_info_t &info)
-{
-  return is_one_of (info, CONSONANT_FLAGS);
-}
-
-#define HALANT_OR_COENG_FLAGS (FLAG (OT_H) | FLAG (OT_Coeng))
-static inline bool
-is_halant_or_coeng (const hb_glyph_info_t &info)
-{
-  return is_one_of (info, HALANT_OR_COENG_FLAGS);
-}
-
-static inline void
-set_indic_properties (hb_glyph_info_t   &info,
-		      const indic_shape_plan_t *closure,
-		      hb_font_t         *font)
-{
-  hb_codepoint_t u = info.codepoint;
-  unsigned int type = get_indic_categories (u);
-  indic_category_t cat = (indic_category_t) (type & 0x0F);
-  indic_position_t pos = (indic_position_t) (type >> 4);
-
-
-  /*
-   * Re-assign category
-   */
-
-
-  /* The spec says U+0952 is OT_A.  However, testing shows that Uniscribe
-   * treats U+0951..U+0952 all as OT_VD.
-   * TESTS:
-   * U+092E,U+0947,U+0952
-   * U+092E,U+0952,U+0947
-   * U+092E,U+0947,U+0951
-   * U+092E,U+0951,U+0947
-   * */
-  if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x0951, 0x0954)))
-    cat = OT_VD;
-
-  if (unlikely (u == 0x17D1))
-    cat = OT_X;
-  if (cat == OT_X &&
-      unlikely (hb_in_range<hb_codepoint_t> (u, 0x17CB, 0x17D3))) /* Khmer Various signs */
-  {
-    /* These are like Top Matras. */
-    cat = OT_M;
-    pos = POS_ABOVE_C;
-  }
-  if (u == 0x17C6) /* Khmer Bindu doesn't like to be repositioned. */
-    cat = OT_N;
-
-  if (unlikely (u == 0x17D2)) cat = OT_Coeng; /* Khmer coeng */
-  else if (unlikely (u == 0x200C)) cat = OT_ZWNJ;
-  else if (unlikely (u == 0x200D)) cat = OT_ZWJ;
-  else if (unlikely (u == 0x25CC)) cat = OT_DOTTEDCIRCLE;
-  else if (unlikely (u == 0x0A71)) cat = OT_SM; /* GURMUKHI ADDAK.  More like consonant medial. like 0A75. */
-
-  if (cat == OT_Repha) {
-    /* There are two kinds of characters marked as Repha:
-     * - The ones that are GenCat=Mn are already positioned visually, ie. after base. (eg. Khmer)
-     * - The ones that are GenCat=Lo is encoded logically, ie. beginning of syllable. (eg. Malayalam)
-     *
-     * We recategorize the first kind to look like a Nukta and attached to the base directly.
-     */
-    if (_hb_glyph_info_get_general_category (&info) == HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK)
-      cat = OT_N;
-  }
-
-
-
-  /*
-   * Re-assign position.
-   */
-
-  if ((FLAG (cat) & CONSONANT_FLAGS))
-  {
-    pos = consonant_position (u, closure, font);
-    if (is_ra (u))
-      cat = OT_Ra;
-  }
-  else if (cat == OT_M)
-  {
-    pos = matra_position (u, pos);
-  }
-  else if (cat == OT_SM || cat == OT_VD)
-  {
-    pos = POS_SMVD;
-  }
-
-  if (unlikely (u == 0x0B01)) pos = POS_BEFORE_SUB; /* Oriya Bindu is BeforeSub in the spec. */
-
-
-
-  info.indic_category() = cat;
-  info.indic_position() = pos;
-}
-
-
-
-
-
 
 
 struct feature_list_t {
@@ -477,7 +262,11 @@ setup_masks_indic (const hb_ot_complex_shaper_t *shaper,
 
   unsigned int count = buffer->len;
   for (unsigned int i = 0; i < count; i++)
-    set_indic_properties (buffer->info[i], &indic_plan, font);
+    set_indic_properties (buffer->info[i]);
+
+  for (unsigned int i = 0; i < count; i++)
+    if (buffer->info[i].indic_position() == POS_BASE_C)
+      buffer->info[i].indic_position() = consonant_position_from_font (buffer->info[i].codepoint, &indic_plan, font);
 }
 
 static int