i18npool/source/breakiterator/data/char.txt

   1 #
   2 #   Copyright (C) 2002-2009, International Business Machines Corporation and others.
   3 #       All Rights Reserved.
   4 #
   5 #   file:  char.txt
   6 #
   7 #   ICU Character Break Rules, also known as Grapheme Cluster Boundaries
   8 #      See Unicode Standard Annex #29.
   9 #      These rules are based on TR29 Revision 13, for Unicode Version 5.1
  10 #   Modifications to SpacingMark and Prepend by M. Hosken.
  11 #
  12
  13 #
  14 #  Character Class Definitions.
  15 #
  16 $CR          = [\p{Grapheme_Cluster_Break = CR}];
  17 $LF          = [\p{Grapheme_Cluster_Break = LF}];
  18 $Control     = [\p{Grapheme_Cluster_Break = Control}];
  19 $Prepend     = [\p{Grapheme_Cluster_Break = Prepend}];
  20 $Extend      = [\p{Grapheme_Cluster_Break = Extend}];
  21 $SpacingMark = [\p{Grapheme_Cluster_Break = SpacingMark}];
  22 # True Indic wants to move by syllables. Break up SpacingMark. This based on Unicode 6.0 data
  23 # In effect it is [\p{Grapheme_Cluster_Break = SpacingMark} - \u0E30 \u0E32 \u0E45 \u0EB0 \u0EB2 \u102B \u102C \u1038
  24 #       \u1062-\u1064 \u1067-\u106D \u1083 \u1087-\u108C \u108F \u109A-\u109C \u19B0-\u19B4 \u19B8-\u19C0 \u19C8 \u19C9
  25 #       \u1A61 \u1A63 \u1A64 \u1BE7 \u1BEA-\u1BEC \u1BEE \u1BF2 \u1BF3 \uAA7B
  26 $IndicSpacing = [\u0903 \u093B \u093E-\u0940 \u0949-\u094C \u094E \u094F \u0982 \u0983 \u09BF \u09C0 \u09C7 \u09C8 \u09CB \u09CC \u0A03 \u0A3E-\u0A40 \u0A83 \u0ABE-\u0AC0 \u0AC9 \u0ACB \u0ACC \u0B02 \u0B03 \u0B40 \u0B47 \u0B48 \u0B4B-\u0B4C \u0BBF \u0BC1 \u0BC2 \u0BC6-\u0BC8 \u0BCA-\u0BCC \u0C01-\u0C03 \u0C41-\u0C44 \u0C82 \u0C83 \u0CBE \u0CC0 \u0CC1 \u0CC3 \u0CC4 \u0CC7 \u0CC8 \u0CCA \u0CCB \u0D02 \u0D03 \u0D3F \u0D40 \u0D46-\u0D48 \u0D4A-\u0D4C \u0D82 \u0D83 \u0DD0 \u0DD1 \u0DD8-\u0DDE \u0F3E \u0F3F \u0F7F \u1923-\u1926 \u1929-\u192B \u1930 \u1931 \u1933-\u1938 \u1A19-\u1A1B \u1B04 \u1B35 \u1B3B \u1B3D-\u1B41 \u1B43 \u1B44 \u1B82 \u1BA1 \u1BA6 \u1BA7 \u1BAA \u1C24-\u1C2B \u1C34 \u1C35 \u1CE1 \u1CF2 \uA880 \uA881 \uA8B4-\uA8C3 \uA952 \uA953 \uA983 \uA9B4 \uA9B5 \uA9BA \uA9BB \uA9BD-\uA9C0 \uAA2F \uAA30 \uAA33 \uAA34 \uABE3 \uABE4 \uABE6 \uABE7 \uABE9 \uABEA \uABEC \U00011000 \U00011002 \U00011082 \U000110B0-\U000110B2 \U000110B7 \U000100B8 \U0001D166 \U0001D16D];
  27 # SEAsian (Thai, Lao, Burmese, Tai Lue, Tai Tham, Batak) are cluster based not syllable based
  28 $SEASpacing = [\u0E33 \u0EB3 \u1031 \u103B \u103C \u1056 \u1057 \u1084 \u17B6 \u17BE-\u17C5 \u17C7 \u17C8 \u19B5-\u19B7 \u19BA \u1A55 \u1A57 \u1A6D-\u1A72 \uA823 \uA824 \uA827 \uAA4D];
  29 $BengaliLetter = [\u0985-\u09B9 \u09CE \u09DC-\u09E1 \u09F0-\u09F1];
  30 $BengaliSignVirama = \u09CD;
  31 $GujaratiLetter = [\u0A85-\u0A8C \u0A8F-\u0A90 \u0A93-\u0AB9 \u0AE0-\u0AE1];
  32 $GujaratiSignVirama = \u0ACD;
  33 $DevanagariLetter = [\u0904-\u0939 \u0958-\u0961 \u0972-\u097F];
  34 $DevanagariSignVirama = \u094D;
  35 $KannadaLetter = [\u0C85-\u0CB9 \u0CDE-\u0CE1];
  36 $KannadaSignVirama = \u0CCD;
  37 $MalayalamLetter = [\u0D05-\u0D39 \u0D60-\u0D61 \u0D7A-\u0D7F];
  38 $MalayalamSignVirama = \u0D4D;
  39 $OdiaLetter = [\u0B05-\u0B39 \u0B5C-\u0B61 \u0B71];
  40 $OdiaSignVirama = \u0B4D;
  41 $GurmukhiLetter = [\u0A05-\u0A39 \u0A59-\u0A5E];
  42 $GurmukhiSignVirama = \u0A4D;
  43 $TamilKa = \u0B95;
  44 $TamilSignVirama = \u0BCD;
  45 $TamilSsa = \u0BB7;
  46 $TeluguLetter = [\u0C05-\u0C39 \u0C58-\u0C61];
  47 $TeluguSignVirama = \u0C4D;
  48
  49 #
  50 # Korean Syllable Definitions
  51 #
  52 $L       = [\p{Grapheme_Cluster_Break = L}];
  53 $V       = [\p{Grapheme_Cluster_Break = V}];
  54 $T       = [\p{Grapheme_Cluster_Break = T}];
  55
  56 $LV      = [\p{Grapheme_Cluster_Break = LV}];
  57 $LVT     = [\p{Grapheme_Cluster_Break = LVT}];
  58
  59
  60 ## -------------------------------------------------
  61 !!chain;
  62
  63 !!forward;
  64
  65 $CR $LF;
  66
  67 $BengaliLetter ($BengaliSignVirama $BengaliLetter?)+;
  68 $GujaratiLetter ($GujaratiSignVirama $GujaratiLetter?)+;
  69 $DevanagariLetter ($DevanagariSignVirama $DevanagariLetter?)+;
  70 $KannadaLetter ($KannadaSignVirama $KannadaLetter?)+;
  71 $MalayalamLetter ($MalayalamSignVirama $MalayalamLetter?)+;
  72 $OdiaLetter ($OdiaSignVirama $OdiaLetter?)+;
  73 $GurmukhiLetter ($GurmukhiSignVirama $GurmukhiLetter?)+;
  74 $TamilKa $TamilSignVirama $TamilSsa;
  75 $TeluguLetter ($TeluguSignVirama $TeluguLetter?)+;
  76
  77 $L ($L | $V | $LV | $LVT);
  78 ($LV | $V) ($V | $T);
  79 ($LVT | $T) $T;
  80
  81 [^$Control $CR $LF] $Extend;
  82
  83 [^$Control $CR $LF] ($IndicSpacing | $SEASpacing);
  84 #[^$Control $CR $LF] $SpacingMark;
  85 # $Prepend [^$Control $CR $LF];
  86
  87
  88 ## -------------------------------------------------
  89
  90 !!reverse;
  91 $LF $CR;
  92 ($BengaliLetter? $BengaliSignVirama)+ $BengaliLetter;
  93 ($GujaratiLetter? $GujaratiSignVirama)+ $GujaratiLetter;
  94 ($DevanagariLetter? $DevanagariSignVirama)+ $DevanagariLetter;
  95 ($KannadaLetter? $KannadaSignVirama)+ $KannadaLetter;
  96 ($MalayalamLetter? $MalayalamSignVirama)+ $MalayalamLetter;
  97 ($OdiaLetter? $OdiaSignVirama)+ $OdiaLetter;
  98 ($GurmukhiLetter? $GurmukhiSignVirama)+ $GurmukhiLetter;
  99 $TamilSsa $TamilSignVirama $TamilKa;
 100 ($TeluguLetter? $TeluguSignVirama)+ $TeluguLetter;
 101 ($L | $V | $LV | $LVT) $L;
 102 ($V | $T) ($LV | $V);
 103 $T ($LVT | $T);
 104
 105 $Extend      [^$Control $CR $LF];
 106 ($IndicSpacing | $SEASpacing) [^$Control $CR $LF];
 107 #$SpacingMark [^$Control $CR $LF];
 108 # [^$Control $CR $LF] $Prepend;
 109
 110
 111 ## -------------------------------------------------
 112
 113 !!safe_reverse;
 114
 115
 116 ## -------------------------------------------------
 117
 118 !!safe_forward;
 119