Use a lookup table for modified_combining_class

src/hb-unicode-private.hh

 /*
  * Copyright © 2009  Red Hat, Inc.
  * Copyright © 2011  Codethink Limited
- * Copyright © 2010,2011  Google, Inc.
+ * Copyright © 2010,2011,2012  Google, Inc.
  *
  *  This is part of HarfBuzz, a text shaping library.
  *
@@ -37,6 +37,7 @@
 #include "hb-object-private.hh"
 
 
+extern HB_INTERNAL const uint8_t _hb_modified_combining_class[256];
 
 /*
  * hb_unicode_funcs_t
@@ -143,8 +144,11 @@ HB_UNICODE_FUNCS_IMPLEMENT_CALLBACKS_SIMPLE
   }
 
 
-  HB_INTERNAL unsigned int
-  modified_combining_class (hb_codepoint_t unicode);
+  unsigned int
+  modified_combining_class (hb_codepoint_t unicode)
+  {
+    return _hb_modified_combining_class[combining_class (unicode)];
+  }
 
   inline hb_bool_t
   is_variation_selector (hb_codepoint_t unicode)

src/hb-unicode.cc

 /*
  * Copyright © 2009  Red Hat, Inc.
- * Copyright © 2011 Codethink Limited
- * Copyright © 2010,2011  Google, Inc.
+ * Copyright © 2011  Codethink Limited
+ * Copyright © 2010,2011,2012  Google, Inc.
  *
  *  This is part of HarfBuzz, a text shaping library.
  *
@@ -287,69 +287,148 @@ hb_unicode_decompose_compatibility (hb_unicode_funcs_t *ufuncs,
 }
 
 
-unsigned int
-hb_unicode_funcs_t::modified_combining_class (hb_codepoint_t unicode)
+const uint8_t
+_hb_modified_combining_class[256] =
 {
-  int c = combining_class (unicode);
-
-  if (unlikely (hb_in_range<int> (c, 27, 33)))
-  {
-    /* Modify the combining-class to suit Arabic better.  See:
-     * http://unicode.org/faq/normalization.html#8
-     * http://unicode.org/faq/normalization.html#9
-     */
-    c = c == 33 ? 27 : c + 1;
-  }
-  else if (unlikely (hb_in_range<int> (c, 10, 26)))
-  {
-    /* The equivalent fix for Hebrew is more complex.
-     *
-     * We permute the "fixed-position" classes 10-26 into the order
-     * described in the SBL Hebrew manual:
-     *
-     * http://www.sbl-site.org/Fonts/SBLHebrewUserManual1.5x.pdf
-     *
-     * (as recommended by:
-     *  http://forum.fontlab.com/archive-old-microsoft-volt-group/vista-and-diacritic-ordering-t6751.0.html)
-     *
-     * More details here:
-     * https://bugzilla.mozilla.org/show_bug.cgi?id=662055
-     */
-    static const int permuted_hebrew_classes[26 - 10 + 1] = {
-      /* 10 sheva */        22,
-      /* 11 hataf segol */  15,
-      /* 12 hataf patah */  16,
-      /* 13 hataf qamats */ 17,
-      /* 14 hiriq */        23,
-      /* 15 tsere */        18,
-      /* 16 segol */        19,
-      /* 17 patah */        20,
-      /* 18 qamats */       21,
-      /* 19 holam */        14,
-      /* 20 qubuts */       24,
-      /* 21 dagesh */       12,
-      /* 22 meteg */        25,
-      /* 23 rafe */         13,
-      /* 24 shin dot */     10,
-      /* 25 sin dot */      11,
-      /* 26 point varika */ 26,
-    };
-    c = permuted_hebrew_classes[c - 10];
-  }
-  else if (unlikely (unicode == 0x0E3A)) /* THAI VOWEL SIGN PHINTHU */
-  {
-    /* Assign 104, so it reorders after the THAI ccc=103 marks.
-     * Uniscribe does this. */
-    c = 104;
-  }
-  else if (unlikely (hb_in_range<hb_codepoint_t> (unicode, 0x0C55, 0x0C56)))
-  {
-    /* Telugu length marks.
-     * These are the only matras in the main Indic script range that have
-     * a non-zero ccc.  That makes them reorder with the Halant that is
-     * ccc=9.  Just zero them, we don't need them in our Indic shaper. */
-    c = 0;
-  }
-
-  return c;
-}
+  0, /* HB_UNICODE_COMBINING_CLASS_NOT_REORDERED */
+  1, /* HB_UNICODE_COMBINING_CLASS_OVERLAY */
+  2, 3, 4, 5, 6,
+  7, /* HB_UNICODE_COMBINING_CLASS_NUKTA */
+  8, /* HB_UNICODE_COMBINING_CLASS_KANA_VOICING */
+  9, /* HB_UNICODE_COMBINING_CLASS_VIRAMA */
+
+  /* Hebrew */
+
+  /*
+   * We permute the "fixed-position" classes 10-26 into the order
+   * described in the SBL Hebrew manual:
+   *
+   * http://www.sbl-site.org/Fonts/SBLHebrewUserManual1.5x.pdf
+   *
+   * (as recommended by:
+   *  http://forum.fontlab.com/archive-old-microsoft-volt-group/vista-and-diacritic-ordering-t6751.0.html)
+   *
+   * More details here:
+   * https://bugzilla.mozilla.org/show_bug.cgi?id=662055
+   */
+  22, /* HB_UNICODE_COMBINING_CLASS_CCC10 sheva */
+  15, /* HB_UNICODE_COMBINING_CLASS_CCC11 hataf segol */
+  16, /* HB_UNICODE_COMBINING_CLASS_CCC12 hataf patah*/
+  17, /* HB_UNICODE_COMBINING_CLASS_CCC13 hataf qamats */
+  23, /* HB_UNICODE_COMBINING_CLASS_CCC14 hiriq */
+  18, /* HB_UNICODE_COMBINING_CLASS_CCC15 tsere */
+  19, /* HB_UNICODE_COMBINING_CLASS_CCC16 segol */
+  20, /* HB_UNICODE_COMBINING_CLASS_CCC17 patah */
+  21, /* HB_UNICODE_COMBINING_CLASS_CCC18 qamats */
+  14, /* HB_UNICODE_COMBINING_CLASS_CCC19 holam */
+  24, /* HB_UNICODE_COMBINING_CLASS_CCC20 qubuts */
+  12, /* HB_UNICODE_COMBINING_CLASS_CCC21 dagesh */
+  25, /* HB_UNICODE_COMBINING_CLASS_CCC22 meteg */
+  13, /* HB_UNICODE_COMBINING_CLASS_CCC23 rafe */
+  10, /* HB_UNICODE_COMBINING_CLASS_CCC24 shin dot */
+  11, /* HB_UNICODE_COMBINING_CLASS_CCC25 sin dot */
+
+  26, /* HB_UNICODE_COMBINING_CLASS_CCC26 */
+
+  /* Arabic */
+
+  /*
+   * Modify to move Shadda (ccc=33) before other marks.  See:
+   * http://unicode.org/faq/normalization.html#8
+   * http://unicode.org/faq/normalization.html#9
+   */
+  28, /* HB_UNICODE_COMBINING_CLASS_CCC27 */
+  29, /* HB_UNICODE_COMBINING_CLASS_CCC28 */
+  30, /* HB_UNICODE_COMBINING_CLASS_CCC29 */
+  31, /* HB_UNICODE_COMBINING_CLASS_CCC30 */
+  32, /* HB_UNICODE_COMBINING_CLASS_CCC31 */
+  33, /* HB_UNICODE_COMBINING_CLASS_CCC32 */
+  27, /* HB_UNICODE_COMBINING_CLASS_CCC33 shadda */
+
+  34, /* HB_UNICODE_COMBINING_CLASS_CCC34 */
+  35, /* HB_UNICODE_COMBINING_CLASS_CCC35 */
+
+  /* Syriac */
+  36, /* HB_UNICODE_COMBINING_CLASS_CCC36 */
+
+  37, 38, 39,
+  40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59,
+  60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79,
+  80, 81, 82, 83,
+
+  /* Telugu */
+
+  /*
+   * Modify Telugu length marks (ccc=84, ccc=91).
+   * These are the only matras in the main Indic scripts range that have
+   * a non-zero ccc.  That makes them reorder with the Halant that is
+   * ccc=9.  Just zero them, we don't need them in our Indic shaper.
+   */
+  0, /* HB_UNICODE_COMBINING_CLASS_CCC84 */
+  85, 86, 87, 88, 89, 90,
+  0, /* HB_UNICODE_COMBINING_CLASS_CCC91 */
+  92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102,
+
+  /* Thai */
+
+  /*
+   * Modify U+0E38 and U+0E39 (ccc=104) to be reordered before U+0E3A (ccc=9).
+   * Uniscribe does this too.
+   */
+  3, /* HB_UNICODE_COMBINING_CLASS_CCC103 */
+
+  104, 105, 106,
+  107, /* HB_UNICODE_COMBINING_CLASS_CCC107 */
+  108, 109, 110, 111, 112, 113, 114, 115, 116, 117,
+
+  /* Lao */
+  118, /* HB_UNICODE_COMBINING_CLASS_CCC118 */
+  119, 120, 121,
+  122, /* HB_UNICODE_COMBINING_CLASS_CCC122 */
+  123, 124, 125, 126, 127, 128,
+
+  /* Tibetan */
+  129, /* HB_UNICODE_COMBINING_CLASS_CCC129 */
+  130, /* HB_UNICODE_COMBINING_CLASS_CCC130 */
+  131,
+  132, /* HB_UNICODE_COMBINING_CLASS_CCC133 */
+  133, 134, 135, 136, 137, 138, 139,
+
+
+  140, 141, 142, 143, 144, 145, 146, 147, 148, 149,
+  150, 151, 152, 153, 154, 155, 156, 157, 158, 159,
+  160, 161, 162, 163, 164, 165, 166, 167, 168, 169,
+  170, 171, 172, 173, 174, 175, 176, 177, 178, 179,
+  180, 181, 182, 183, 184, 185, 186, 187, 188, 189,
+  190, 191, 192, 193, 194, 195, 196, 197, 198, 199,
+
+  200, /* HB_UNICODE_COMBINING_CLASS_ATTACHED_BELOW_LEFT */
+  201,
+  202, /* HB_UNICODE_COMBINING_CLASS_ATTACHED_BELOW */
+  203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213,
+  214, /* HB_UNICODE_COMBINING_CLASS_ATTACHED_ABOVE */
+  215,
+  216, /* HB_UNICODE_COMBINING_CLASS_ATTACHED_ABOVE_RIGHT */
+  217,
+  218, /* HB_UNICODE_COMBINING_CLASS_BELOW_LEFT */
+  219,
+  220, /* HB_UNICODE_COMBINING_CLASS_BELOW */
+  221,
+  222, /* HB_UNICODE_COMBINING_CLASS_BELOW_RIGHT */
+  223,
+  224, /* HB_UNICODE_COMBINING_CLASS_LEFT */
+  225,
+  226, /* HB_UNICODE_COMBINING_CLASS_RIGHT */
+  227,
+  228, /* HB_UNICODE_COMBINING_CLASS_ABOVE_LEFT */
+  229,
+  230, /* HB_UNICODE_COMBINING_CLASS_ABOVE */
+  231,
+  232, /* HB_UNICODE_COMBINING_CLASS_ABOVE_RIGHT */
+  233, /* HB_UNICODE_COMBINING_CLASS_DOUBLE_BELOW */
+  234, /* HB_UNICODE_COMBINING_CLASS_DOUBLE_ABOVE */
+  235, 236, 237, 238, 239,
+  240, /* HB_UNICODE_COMBINING_CLASS_IOTA_SUBSCRIPT */
+  241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254,
+  255, /* HB_UNICODE_COMBINING_CLASS_INVALID */
+};

src/hb-unicode.h

@@ -79,6 +79,10 @@ typedef enum
 
 /* hb_unicode_combining_class_t */
 
+/* Note: newer versions of Unicode may add new values.  Clients should be ready to handle
+ * any value in the 0..254 range being returned from hb_unicode_combining_class().
+ */
+
 /* Unicode Character Database property: Canonical_Combining_Class (ccc) */
 typedef enum
 {