diff options
Diffstat (limited to 'lib/unilbrk/lbrktables.h')
-rw-r--r-- | lib/unilbrk/lbrktables.h | 102 |
1 files changed, 52 insertions, 50 deletions
diff --git a/lib/unilbrk/lbrktables.h b/lib/unilbrk/lbrktables.h index 865bea16..ea2114c5 100644 --- a/lib/unilbrk/lbrktables.h +++ b/lib/unilbrk/lbrktables.h @@ -1,5 +1,5 @@ /* Line breaking auxiliary tables. - Copyright (C) 2001-2003, 2006-2024 Free Software Foundation, Inc. + Copyright (C) 2001-2003, 2006-2025 Free Software Foundation, Inc. Written by Bruno Haible <bruno@clisp.org>, 2001. This file is free software. @@ -34,64 +34,66 @@ extern "C" { enum { - /* Values >= 41 are resolved at run time. */ - LBP_BK = 41, /* mandatory break */ - LBP_CR = 42, /* carriage return */ - LBP_LF = 43, /* line feed */ - LBP_CM = 44, /* attached characters and combining marks */ + /* Values >= 43 are resolved at run time. */ + LBP_BK = 43, /* mandatory break */ + LBP_CR = 44, /* carriage return */ + LBP_LF = 45, /* line feed */ + LBP_CM = 46, /* attached characters and combining marks */ /*LBP_NL, next line - not used here because it's equivalent to LBP_BK */ /*LBP_SG, surrogates - not used here because they are not characters */ LBP_WJ = 0, /* word joiner */ - LBP_ZW = 45, /* zero width space */ + LBP_ZW = 47, /* zero width space */ LBP_GL = 1, /* non-breaking (glue) */ - LBP_SP = 46, /* space */ + LBP_SP = 48, /* space */ LBP_B2 = 2, /* break opportunity before and after */ LBP_BA = 3, /* break opportunity after */ LBP_BB = 4, /* break opportunity before */ - LBP_HY = 5, /* hyphen */ - LBP_CB = 47, /* contingent break opportunity */ - LBP_CL = 6, /* closing punctuation */ - LBP_CP1 = 7, /* closing parenthesis, non-EastAsian character */ - LBP_CP2 = 8, /* closing parenthesis, EastAsian character */ - LBP_EX = 9, /* exclamation/interrogation */ - LBP_IN = 10, /* inseparable */ - LBP_NS = 11, /* non starter */ - LBP_OP1 = 12, /* opening punctuation, non-EastAsian character */ - LBP_OP2 = 13, /* opening punctuation, EastAsian character */ - LBP_QU1 = 14, /* ambiguous quotation, neither initial nor final punctuation */ - LBP_QU2 = 15, /* ambiguous quotation, initial punctuation */ - LBP_QU3 = 16, /* ambiguous quotation, final punctuation */ - LBP_IS = 17, /* infix separator (numeric) */ - LBP_NU = 18, /* numeric */ - LBP_PO = 19, /* postfix (numeric) */ - LBP_PR = 20, /* prefix (numeric) */ - LBP_SY = 21, /* symbols allowing breaks */ - LBP_AI = 48, /* ambiguous (alphabetic or ideograph) */ - LBP_AL1 = 22, /* ordinary alphabetic and symbol characters, != U+25CC */ - LBP_AL2 = 23, /* ordinary alphabetic and symbol characters, == U+25CC */ + LBP_HH = 5, /* unambiguous hyphen */ + LBP_HY = 6, /* hyphen */ + LBP_CB = 49, /* contingent break opportunity */ + LBP_CL = 7, /* closing punctuation */ + LBP_CP1 = 8, /* closing parenthesis, non-EastAsian character */ + LBP_CP2 = 9, /* closing parenthesis, EastAsian character */ + LBP_EX = 10, /* exclamation/interrogation */ + LBP_IN = 11, /* inseparable */ + LBP_NS = 12, /* non starter */ + LBP_OP1 = 13, /* opening punctuation, non-EastAsian character */ + LBP_OP2 = 14, /* opening punctuation, EastAsian character */ + LBP_QU1 = 15, /* ambiguous quotation, neither initial nor final punctuation */ + LBP_QU2 = 16, /* ambiguous quotation, initial punctuation */ + LBP_QU3 = 17, /* ambiguous quotation, final punctuation */ + LBP_IS = 18, /* infix separator (numeric) */ + LBP_NU = 19, /* numeric */ + LBP_PO = 20, /* postfix (numeric) */ + LBP_PR = 21, /* prefix (numeric) */ + LBP_SY = 22, /* symbols allowing breaks */ + LBP_AI = 50, /* ambiguous (alphabetic or ideograph) */ + LBP_AL1 = 23, /* ordinary alphabetic and symbol characters, != U+25CC */ + LBP_AL2 = 24, /* ordinary alphabetic and symbol characters, == U+25CC */ /*LBP_CJ, conditional Japanese starter, resolved to NS */ - LBP_H2 = 24, /* Hangul LV syllable */ - LBP_H3 = 25, /* Hangul LVT syllable */ - LBP_HL = 31, /* Hebrew letter */ - LBP_ID1 = 26, /* ideographic */ - LBP_ID2 = 27, /* ideographic and potential future emoji */ - LBP_JL = 28, /* Hangul L Jamo */ - LBP_JV = 29, /* Hangul V Jamo */ - LBP_JT = 30, /* Hangul T Jamo */ - LBP_AP = 32, /* Brahmic scripts: pre-base repha */ - LBP_AK = 33, /* Brahmic scripts: consonants */ - LBP_AS = 34, /* Brahmic scripts: independent vowels */ - LBP_VI = 35, /* Brahmic scripts: conjoining viramas */ - LBP_VF = 36, /* Brahmic scripts: viramas for final consonants */ - LBP_RI = 37, /* regional indicator */ - LBP_SA = 49, /* complex context (South East Asian) */ - LBP_ZWJ = 38, /* zero width joiner */ - LBP_EB = 39, /* emoji base */ - LBP_EM = 40, /* emoji modifier */ - LBP_XX = 50, /* unknown */ + LBP_H2 = 25, /* Hangul LV syllable */ + LBP_H3 = 26, /* Hangul LVT syllable */ + LBP_HL = 32, /* Hebrew letter */ + LBP_ID = 27, /* ideographic */ + LBP_EBF = 28, /* future emoji base */ + LBP_JL = 29, /* Hangul L Jamo */ + LBP_JV = 30, /* Hangul V Jamo */ + LBP_JT = 31, /* Hangul T Jamo */ + LBP_AP = 33, /* Brahmic scripts: pre-base repha */ + LBP_AK = 34, /* Brahmic scripts: consonants */ + LBP_AS = 35, /* Brahmic scripts: independent vowels */ + LBP_VI = 36, /* Brahmic scripts: conjoining viramas */ + LBP_VF = 37, /* Brahmic scripts: viramas for final consonants */ + LBP_RI = 38, /* regional indicator */ + LBP_SA1 = 51, /* complex context (South East Asian), not combining mark */ + LBP_SA2 = 39, /* complex context (South East Asian), combining mark */ + LBP_ZWJ = 40, /* zero width joiner */ + LBP_EB = 41, /* emoji base */ + LBP_EM = 42, /* emoji modifier */ + LBP_XX = 52, /* unknown */ /* Artificial values that exist only at runtime, not in the tables. */ LBP_AKLS_VI = 100, - LBP_HL_BA = 101 + LBP_HL_HY = 101 }; #include "lbrkprop1.h" @@ -134,7 +136,7 @@ unilbrkprop_lookup (ucs4_t uc) #define I 2 /* indirect break opportunity, '%' in table 7.3 of UTR #14 */ #define P 3 /* prohibited break, '^' in table 7.3 of UTR #14 */ -extern const unsigned char unilbrk_table[41][41]; +extern const unsigned char unilbrk_table[43][43]; /* We don't support line breaking of complex-context dependent characters (Thai, Lao, Myanmar, Khmer) yet, because it requires dictionary lookup. */ |