[ruby-dev:36292] Re: 合成文字の2コードポイント目
From:
Tanaka Akira <akr@...>
Date:
2008-09-14 02:08:15 UTC
List:
ruby-dev #36292
In article <E1KeRWe-00037N-Vb@x61.netlab.jp>,
Yukihiro Matsumoto <matz@ruby-lang.org> writes:
> そんな気はします。Unicodeでは文字を定義することをあきらめて
> るような気がしますが。
完全に定義することをあきらめるからといって、コードポイントの
ままで済ますわけではないようなので。
> その辺には踏み込みませんが、将来のことを考えると、そういう仕
> 組みを用意しておくのが親切なのかもしれません。
>
> しかし、エンコーディングを定義するのにrb_enc_precise_mbclenと
> rb_enc_precise_mbcharlenの両方を必要とするのはあんまりうれし
> くありませんねえ。
まぁ、each_char と each_codepoint を両方実装するとすれば結局
どちらも必要になりますが、似ていてわかりにくいというのはあり
ますね。あと、oniguruma をそのまま動かすにはコードポイント単
位に見せておくのが簡単なようです。
コードポイントの情報は mbc_to_code に返させるというのはでき
るかな。
あと、left_adjust_char_head も両方が必要ですね。
% svn diff --diff-cmd diff -x '-u -p'
Index: encoding.c
===================================================================
--- encoding.c (revision 19335)
+++ encoding.c (working copy)
@@ -1248,3 +1248,28 @@ rb_toupper(int c)
return rb_isascii(c) ? ONIGENC_ASCII_CODE_TO_UPPER_CASE(c) : c;
}
+char*
+rb_enc_prev_char(const char* start, const char* s, const char* end, OnigEncoding enc)
+{
+ if (s <= start)
+ return (char *)NULL;
+
+ return (char *)ONIGENC_LEFT_ADJUST_CHAR2_HEAD(enc, (UChar *)start, (UChar *)s - 1, (UChar *)end);
+}
+
+char*
+rb_enc_left_char_head(const char* start, const char* s, const char* end, OnigEncoding enc)
+{
+ return (char *)ONIGENC_LEFT_ADJUST_CHAR2_HEAD(enc, (UChar *)start, (UChar *)s, (UChar *)end);
+}
+
+char*
+rb_enc_right_char_head(const char* start, const char* s, const char* end, OnigEncoding enc)
+{
+ char *p = (char *)ONIGENC_LEFT_ADJUST_CHAR2_HEAD(enc, (UChar *)start, (UChar *)s, (UChar *)end);
+ if (p < s) {
+ p += rb_enc_mbclen(p, end, enc);
+ }
+ return p;
+}
+
Index: include/ruby/oniguruma.h
===================================================================
--- include/ruby/oniguruma.h (revision 19335)
+++ include/ruby/oniguruma.h (working copy)
@@ -155,7 +155,7 @@ typedef struct OnigEncodingTypeST {
int max_enc_len;
int min_enc_len;
int (*is_mbc_newline)(const OnigUChar* p, const OnigUChar* end, struct OnigEncodingTypeST* enc);
- OnigCodePoint (*mbc_to_code)(const OnigUChar* p, const OnigUChar* end, struct OnigEncodingTypeST* enc);
+ OnigCodePoint (*mbc_to_code)(const OnigUChar* p, const OnigUChar* end, int *precious_ret, struct OnigEncodingTypeST* enc);
int (*code_to_mbclen)(OnigCodePoint code, struct OnigEncodingTypeST* enc);
int (*code_to_mbc)(OnigCodePoint code, OnigUChar *buf, struct OnigEncodingTypeST* enc);
int (*mbc_case_fold)(OnigCaseFoldType flag, const OnigUChar** pp, const OnigUChar* end, OnigUChar* to, struct OnigEncodingTypeST* enc);
@@ -165,6 +165,7 @@ typedef struct OnigEncodingTypeST {
int (*is_code_ctype)(OnigCodePoint code, OnigCtype ctype, struct OnigEncodingTypeST* enc);
int (*get_ctype_code_range)(OnigCtype ctype, OnigCodePoint* sb_out, const OnigCodePoint* ranges[], struct OnigEncodingTypeST* enc);
OnigUChar* (*left_adjust_char_head)(const OnigUChar* start, const OnigUChar* p, const OnigUChar* end, struct OnigEncodingTypeST* enc);
+ OnigUChar* (*left_adjust_char2_head)(const OnigUChar* start, const OnigUChar* p, const OnigUChar* end, struct OnigEncodingTypeST* enc);
int (*is_allowed_reverse_match)(const OnigUChar* p, const OnigUChar* end, struct OnigEncodingTypeST* enc);
int ruby_encoding_index;
} OnigEncodingType;
@@ -221,6 +222,8 @@ ONIG_EXTERN OnigEncodingType OnigEncodin
(enc)->is_allowed_reverse_match(s,end,enc)
#define ONIGENC_LEFT_ADJUST_CHAR_HEAD(enc,start,s,end) \
(enc)->left_adjust_char_head(start, s, end, enc)
+#define ONIGENC_LEFT_ADJUST_CHAR2_HEAD(enc,start,s,end) \
+ (enc)->left_adjust_char2_head(start, s, end, enc)
#define ONIGENC_APPLY_ALL_CASE_FOLD(enc,case_fold_flag,f,arg) \
(enc)->apply_all_case_fold(case_fold_flag,f,arg,enc)
#define ONIGENC_GET_CASE_FOLD_CODES_BY_STR(enc,case_fold_flag,p,end,acs) \
@@ -242,14 +245,15 @@ ONIG_EXTERN OnigEncodingType OnigEncodin
#define ONIGENC_PRECISE_MBC_ENC_LEN(enc,p,e) (enc)->precise_mbc_enc_len(p,e,enc)
ONIG_EXTERN
-int onigenc_mbclen_approximate P_((const OnigUChar* p,const OnigUChar* e, struct OnigEncodingTypeST* enc));
+int onigenc_codepoint_len P_((const OnigUChar* p, const OnigUChar* e, struct OnigEncodingTypeST* enc));
-#define ONIGENC_MBC_ENC_LEN(enc,p,e) onigenc_mbclen_approximate(p,e,enc)
+#define ONIGENC_MBC_ENC_LEN(enc,p,e) onigenc_codepoint_len(p,e,enc)
#define ONIGENC_MBC_MAXLEN(enc) ((enc)->max_enc_len)
#define ONIGENC_MBC_MAXLEN_DIST(enc) ONIGENC_MBC_MAXLEN(enc)
#define ONIGENC_MBC_MINLEN(enc) ((enc)->min_enc_len)
#define ONIGENC_IS_MBC_NEWLINE(enc,p,end) (enc)->is_mbc_newline((p),(end),enc)
-#define ONIGENC_MBC_TO_CODE(enc,p,end) (enc)->mbc_to_code((p),(end),enc)
+#define ONIGENC_MBC_TO_CODE(enc,p,end) (enc)->mbc_to_code((p),(end),NULL,enc)
+#define ONIGENC_MBC_PRECIOUS_CODEPOINT(enc,p,end,prec_ret) (enc)->mbc_to_code((p),(end),(prec_ret),enc)
#define ONIGENC_CODE_TO_MBCLEN(enc,code) (enc)->code_to_mbclen(code,enc)
#define ONIGENC_CODE_TO_MBC(enc,code,buf) (enc)->code_to_mbc(code,buf,enc)
#define ONIGENC_PROPERTY_NAME_TO_CTYPE(enc,p,end) \
Index: include/ruby/encoding.h
===================================================================
--- include/ruby/encoding.h (revision 19335)
+++ include/ruby/encoding.h (working copy)
@@ -122,6 +122,7 @@ int rb_enc_ascget(const char *p, const c
/* -> code or raise exception */
unsigned int rb_enc_codepoint(const char *p, const char *e, rb_encoding *enc);
#define rb_enc_mbc_to_codepoint(p, e, enc) ONIGENC_MBC_TO_CODE(enc,(UChar*)(p),(UChar*)(e))
+#define rb_enc_mbc_precious_codepoint(p, e, prec_ret, enc) ONIGENC_MBC_PRECIOUS_CODEPOINT(enc,(UChar*)(p),(UChar*)(e),(prec_ret))
/* -> codelen>0 or raise exception */
int rb_enc_codelen(int code, rb_encoding *enc);
@@ -130,10 +131,16 @@ int rb_enc_codelen(int code, rb_encoding
#define rb_enc_mbcput(c,buf,enc) ONIGENC_CODE_TO_MBC(enc,c,(UChar*)(buf))
/* start, ptr, end, encoding -> prev_char */
-#define rb_enc_prev_char(s,p,e,enc) (char *)onigenc_get_prev_char_head(enc,(UChar*)(s),(UChar*)(p),(UChar*)(e))
+char* rb_enc_prev_char(const char* start, const char* s, const char* end, OnigEncoding enc);
/* start, ptr, end, encoding -> next_char */
-#define rb_enc_left_char_head(s,p,e,enc) (char *)onigenc_get_left_adjust_char_head(enc,(UChar*)(s),(UChar*)(p),(UChar*)(e))
-#define rb_enc_right_char_head(s,p,e,enc) (char *)onigenc_get_right_adjust_char_head(enc,(UChar*)(s),(UChar*)(p),(UChar*)(e))
+char* rb_enc_left_char_head(const char* start, const char* s, const char* end, OnigEncoding enc);
+char* rb_enc_right_char_head(const char* start, const char* s, const char* end, OnigEncoding enc);
+
+/* ptr, ptr, encoding -> prev_codepoint */
+#define rb_enc_prev_codepoint(s,p,e,enc) (char *)onigenc_get_prev_char_head(enc,(UChar*)(s),(UChar*)(p),(UChar*)(e))
+/* ptr, ptr, encoding -> next_codepoint */
+#define rb_enc_left_codepoint_head(s,p,e,enc) (char *)onigenc_get_left_adjust_char_head(enc,(UChar*)(s),(UChar*)(p),(UChar*)(e))
+#define rb_enc_right_codepoint_head(s,p,e,enc) (char *)onigenc_get_right_adjust_char_head(enc,(UChar*)(s),(UChar*)(p),(UChar*)(e))
/* ptr, ptr, encoding -> newline_or_not */
#define rb_enc_is_newline(p,end,enc) ONIGENC_IS_MBC_NEWLINE(enc,(UChar*)(p),(UChar*)(end))
Index: regenc.c
===================================================================
--- regenc.c (revision 19335)
+++ regenc.c (working copy)
@@ -51,9 +51,10 @@ onigenc_set_default_encoding(OnigEncodin
}
extern int
-onigenc_mbclen_approximate(const OnigUChar* p,const OnigUChar* e, struct OnigEncodingTypeST* enc)
+onigenc_codepoint_len(const OnigUChar* p,const OnigUChar* e, struct OnigEncodingTypeST* enc)
{
- int ret = ONIGENC_PRECISE_MBC_ENC_LEN(enc,p,e);
+ int ret;
+ ONIGENC_MBC_PRECIOUS_CODEPOINT(enc, p, e, &ret);
if (ONIGENC_MBCLEN_CHARFOUND_P(ret))
return ONIGENC_MBCLEN_CHARFOUND_LEN(ret);
else if (ONIGENC_MBCLEN_NEEDMORE_P(ret))
@@ -615,8 +616,11 @@ onigenc_single_byte_mbc_enc_len(const UC
extern OnigCodePoint
onigenc_single_byte_mbc_to_code(const UChar* p, const UChar* end ARG_UNUSED,
+ int *precious_ret,
OnigEncoding enc ARG_UNUSED)
{
+ if (precious_ret)
+ *precious_ret = ONIGENC_CONSTRUCT_MBCLEN_CHARFOUND(1);
return (OnigCodePoint )(*p);
}
@@ -668,12 +672,20 @@ onigenc_ascii_is_code_ctype(OnigCodePoin
}
extern OnigCodePoint
-onigenc_mbn_mbc_to_code(OnigEncoding enc, const UChar* p, const UChar* end)
+onigenc_mbn_mbc_to_code(OnigEncoding enc, const UChar* p, const UChar* end, int *precious_ret)
{
- int c, i, len;
+ int c, i, len, ret;
OnigCodePoint n;
- len = enclen(enc, p, end);
+ ret = ONIGENC_PRECISE_MBC_ENC_LEN(enc, p, end);
+ if (precious_ret)
+ *precious_ret = ret;
+ if (ONIGENC_MBCLEN_CHARFOUND_P(ret))
+ len = ONIGENC_MBCLEN_CHARFOUND_LEN(ret);
+ else if (ONIGENC_MBCLEN_NEEDMORE_P(ret))
+ len = end-p+ONIGENC_MBCLEN_NEEDMORE_LEN(ret);
+ else
+ len = 1;
n = (OnigCodePoint )(*p++);
if (len == 1) return n;
Index: regenc.h
===================================================================
--- regenc.h (revision 19335)
+++ regenc.h (working copy)
@@ -70,7 +70,7 @@ typedef struct {
#define ONIG_CHECK_NULL_RETURN(p) if (ONIG_IS_NULL(p)) return NULL
#define ONIG_CHECK_NULL_RETURN_VAL(p,val) if (ONIG_IS_NULL(p)) return (val)
-#define enclen(enc,p,e) ((enc->max_enc_len == enc->min_enc_len) ? enc->min_enc_len : ONIGENC_MBC_ENC_LEN(enc,p,e))
+#define enclen(enc,p,e) ((enc->max_enc_len == enc->min_enc_len) ? enc->min_enc_len : onigenc_codepoint_len(p,e,enc))
/* character types bit flag */
#define BIT_CTYPE_NEWLINE (1<< ONIGENC_CTYPE_NEWLINE)
@@ -122,7 +122,7 @@ ONIG_EXTERN int onigenc_is_mbc_newline_0
/* methods for single byte encoding */
ONIG_EXTERN int onigenc_ascii_mbc_case_fold P_((OnigCaseFoldType flag, const UChar** p, const UChar* end, UChar* lower, OnigEncoding enc));
ONIG_EXTERN int onigenc_single_byte_mbc_enc_len P_((const UChar* p, const UChar* e, OnigEncoding enc));
-ONIG_EXTERN OnigCodePoint onigenc_single_byte_mbc_to_code P_((const UChar* p, const UChar* end, OnigEncoding enc));
+ONIG_EXTERN OnigCodePoint onigenc_single_byte_mbc_to_code P_((const UChar* p, const UChar* end, int *precious_ret, OnigEncoding enc));
ONIG_EXTERN int onigenc_single_byte_code_to_mbclen P_((OnigCodePoint code, OnigEncoding enc));
ONIG_EXTERN int onigenc_single_byte_code_to_mbc P_((OnigCodePoint code, UChar *buf, OnigEncoding enc));
ONIG_EXTERN UChar* onigenc_single_byte_left_adjust_char_head P_((const UChar* start, const UChar* s, const OnigUChar* end, OnigEncoding enc));
@@ -131,7 +131,7 @@ ONIG_EXTERN int onigenc_always_false_is_
ONIG_EXTERN int onigenc_ascii_is_code_ctype P_((OnigCodePoint code, unsigned int ctype, OnigEncoding enc));
/* methods for multi byte encoding */
-ONIG_EXTERN OnigCodePoint onigenc_mbn_mbc_to_code P_((OnigEncoding enc, const UChar* p, const UChar* end));
+ONIG_EXTERN OnigCodePoint onigenc_mbn_mbc_to_code P_((OnigEncoding enc, const UChar* p, const UChar* end, int *precious_ret));
ONIG_EXTERN int onigenc_mbn_mbc_case_fold P_((OnigEncoding enc, OnigCaseFoldType flag, const UChar** p, const UChar* end, UChar* lower));
ONIG_EXTERN int onigenc_mb2_code_to_mbclen P_((OnigCodePoint code, OnigEncoding enc));
ONIG_EXTERN int onigenc_mb2_code_to_mbc P_((OnigEncoding enc, OnigCodePoint code, UChar *buf));
Index: enc/koi8_u.c
===================================================================
--- enc/koi8_u.c (revision 19335)
+++ enc/koi8_u.c (working copy)
@@ -217,5 +217,6 @@ OnigEncodingDefine(koi8_u, KOI8_U) = {
koi8_u_is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
Index: enc/gbk.c
===================================================================
--- enc/gbk.c (revision 19335)
+++ enc/gbk.c (working copy)
@@ -130,9 +130,9 @@ gbk_mbc_enc_len(const UChar* p, const UC
}
static OnigCodePoint
-gbk_mbc_to_code(const UChar* p, const UChar* end, OnigEncoding enc)
+gbk_mbc_to_code(const UChar* p, const UChar* end, int *precious_ret, OnigEncoding enc)
{
- return onigenc_mbn_mbc_to_code(enc, p, end);
+ return onigenc_mbn_mbc_to_code(enc, p, end, precious_ret);
}
static int
@@ -210,6 +210,7 @@ OnigEncodingDefine(gbk, GBK) = {
gbk_is_code_ctype,
onigenc_not_support_get_ctype_code_range,
gbk_left_adjust_char_head,
+ gbk_left_adjust_char_head,
gbk_is_allowed_reverse_match
};
/*
Index: enc/euc_jp.c
===================================================================
--- enc/euc_jp.c (revision 19335)
+++ enc/euc_jp.c (working copy)
@@ -133,21 +133,9 @@ mbc_enc_len(const UChar* p, const UChar*
}
static OnigCodePoint
-mbc_to_code(const UChar* p, const UChar* end, OnigEncoding enc)
+mbc_to_code(const UChar* p, const UChar* end, int *precious_ret, OnigEncoding enc)
{
- int c, i, len;
- OnigCodePoint n;
-
- len = enclen(enc, p, end);
- n = (OnigCodePoint )*p++;
- if (len == 1) return n;
-
- for (i = 1; i < len; i++) {
- if (p >= end) break;
- c = *p++;
- n <<= 8; n += c;
- }
- return n;
+ return onigenc_mbn_mbc_to_code(enc, p, end, precious_ret);
}
static int
@@ -359,6 +347,7 @@ OnigEncodingDefine(euc_jp, EUC_JP) = {
is_code_ctype,
get_ctype_code_range,
left_adjust_char_head,
+ left_adjust_char_head,
is_allowed_reverse_match,
0
};
Index: enc/cp949.c
===================================================================
--- enc/cp949.c (revision 19335)
+++ enc/cp949.c (working copy)
@@ -130,9 +130,9 @@ cp949_mbc_enc_len(const UChar* p, const
}
static OnigCodePoint
-cp949_mbc_to_code(const UChar* p, const UChar* end, OnigEncoding enc)
+cp949_mbc_to_code(const UChar* p, const UChar* end, int *precious_ret, OnigEncoding enc)
{
- return onigenc_mbn_mbc_to_code(enc, p, end);
+ return onigenc_mbn_mbc_to_code(enc, p, end, precious_ret);
}
static int
@@ -210,6 +210,7 @@ OnigEncodingDefine(cp949, CP949) = {
cp949_is_code_ctype,
onigenc_not_support_get_ctype_code_range,
cp949_left_adjust_char_head,
+ cp949_left_adjust_char_head,
cp949_is_allowed_reverse_match
};
/*
Index: enc/shift_jis.c
===================================================================
--- enc/shift_jis.c (revision 19335)
+++ enc/shift_jis.c (working copy)
@@ -145,22 +145,9 @@ code_to_mbclen(OnigCodePoint code, OnigE
}
static OnigCodePoint
-mbc_to_code(const UChar* p, const UChar* end, OnigEncoding enc)
+mbc_to_code(const UChar* p, const UChar* end, int *precious_ret, OnigEncoding enc)
{
- int c, i, len;
- OnigCodePoint n;
-
- len = enclen(enc, p, end);
- c = *p++;
- n = c;
- if (len == 1) return n;
-
- for (i = 1; i < len; i++) {
- if (p >= end) break;
- c = *p++;
- n <<= 8; n += c;
- }
- return n;
+ return onigenc_mbn_mbc_to_code(enc, p, end, precious_ret);
}
static int
@@ -367,6 +354,7 @@ OnigEncodingDefine(shift_jis, Shift_JIS)
is_code_ctype,
get_ctype_code_range,
left_adjust_char_head,
+ left_adjust_char_head,
is_allowed_reverse_match,
0
};
Index: enc/utf_8.c
===================================================================
--- enc/utf_8.c (revision 19335)
+++ enc/utf_8.c (working copy)
@@ -241,6 +241,657 @@ mbc_enc_len(const UChar* p, const UChar*
ONIGENC_CONSTRUCT_MBCLEN_INVALID();
}
+static OnigCodePoint mbc_to_code(const UChar* p, const UChar* end, int *precious_ret, OnigEncoding enc);
+
+/* generated from GraphemeBreakProperty-5.1.0.txt
+ * Since CR LF is handled in another layer such as IO with text mode,
+ * CR and LF are merged into CONTROL. */
+#define GRAPHEME_BIT_CONTROL 0x001
+#define GRAPHEME_BIT_EXTEND 0x002
+#define GRAPHEME_BIT_PREPEND 0x004
+#define GRAPHEME_BIT_SPACINGMARK 0x008
+#define GRAPHEME_BIT_L 0x010
+#define GRAPHEME_BIT_V 0x020
+#define GRAPHEME_BIT_T 0x040
+#define GRAPHEME_BIT_LV 0x080
+#define GRAPHEME_BIT_LVT 0x100
+const struct graphme_table_t {
+ OnigCodePoint codepoint_min;
+ unsigned short num_codepoints;
+ unsigned short properties;
+} graphme_table[] = {
+ {0x00000,32,0x001}, {0x0007F,33,0x001},
+ {0x000AD,1,0x001}, {0x00300,112,0x002},
+ {0x00483,7,0x002}, {0x00591,45,0x002},
+ {0x005BF,1,0x002}, {0x005C1,2,0x002},
+ {0x005C4,2,0x002}, {0x005C7,1,0x002},
+ {0x00600,4,0x001}, {0x00610,11,0x002},
+ {0x0064B,20,0x002}, {0x00670,1,0x002},
+ {0x006D6,7,0x002}, {0x006DD,1,0x001},
+ {0x006DE,7,0x002}, {0x006E7,2,0x002},
+ {0x006EA,4,0x002}, {0x0070F,1,0x001},
+ {0x00711,1,0x002}, {0x00730,27,0x002},
+ {0x007A6,11,0x002}, {0x007EB,9,0x002},
+ {0x00901,2,0x002}, {0x00903,1,0x008},
+ {0x0093C,1,0x002}, {0x0093E,3,0x008},
+ {0x00941,8,0x002}, {0x00949,4,0x008},
+ {0x0094D,1,0x002}, {0x00951,4,0x002},
+ {0x00962,2,0x002}, {0x00981,1,0x002},
+ {0x00982,2,0x008}, {0x009BC,1,0x002},
+ {0x009BE,1,0x002}, {0x009BF,2,0x008},
+ {0x009C1,4,0x002}, {0x009C7,2,0x008},
+ {0x009CB,2,0x008}, {0x009CD,1,0x002},
+ {0x009D7,1,0x002}, {0x009E2,2,0x002},
+ {0x00A01,2,0x002}, {0x00A03,1,0x008},
+ {0x00A3C,1,0x002}, {0x00A3E,3,0x008},
+ {0x00A41,2,0x002}, {0x00A47,2,0x002},
+ {0x00A4B,3,0x002}, {0x00A51,1,0x002},
+ {0x00A70,2,0x002}, {0x00A75,1,0x002},
+ {0x00A81,2,0x002}, {0x00A83,1,0x008},
+ {0x00ABC,1,0x002}, {0x00ABE,3,0x008},
+ {0x00AC1,5,0x002}, {0x00AC7,2,0x002},
+ {0x00AC9,1,0x008}, {0x00ACB,2,0x008},
+ {0x00ACD,1,0x002}, {0x00AE2,2,0x002},
+ {0x00B01,1,0x002}, {0x00B02,2,0x008},
+ {0x00B3C,1,0x002}, {0x00B3E,2,0x002},
+ {0x00B40,1,0x008}, {0x00B41,4,0x002},
+ {0x00B47,2,0x008}, {0x00B4B,2,0x008},
+ {0x00B4D,1,0x002}, {0x00B56,2,0x002},
+ {0x00B62,2,0x002}, {0x00B82,1,0x002},
+ {0x00BBE,1,0x002}, {0x00BBF,1,0x008},
+ {0x00BC0,1,0x002}, {0x00BC1,2,0x008},
+ {0x00BC6,3,0x008}, {0x00BCA,3,0x008},
+ {0x00BCD,1,0x002}, {0x00BD7,1,0x002},
+ {0x00C01,3,0x008}, {0x00C3E,3,0x002},
+ {0x00C41,4,0x008}, {0x00C46,3,0x002},
+ {0x00C4A,4,0x002}, {0x00C55,2,0x002},
+ {0x00C62,2,0x002}, {0x00C82,2,0x008},
+ {0x00CBC,1,0x002}, {0x00CBE,1,0x008},
+ {0x00CBF,1,0x002}, {0x00CC0,2,0x008},
+ {0x00CC2,1,0x002}, {0x00CC3,2,0x008},
+ {0x00CC6,1,0x002}, {0x00CC7,2,0x008},
+ {0x00CCA,2,0x008}, {0x00CCC,2,0x002},
+ {0x00CD5,2,0x002}, {0x00CE2,2,0x002},
+ {0x00D02,2,0x008}, {0x00D3E,1,0x002},
+ {0x00D3F,2,0x008}, {0x00D41,4,0x002},
+ {0x00D46,3,0x008}, {0x00D4A,3,0x008},
+ {0x00D4D,1,0x002}, {0x00D57,1,0x002},
+ {0x00D62,2,0x002}, {0x00D82,2,0x008},
+ {0x00DCA,1,0x002}, {0x00DCF,1,0x002},
+ {0x00DD0,2,0x008}, {0x00DD2,3,0x002},
+ {0x00DD6,1,0x002}, {0x00DD8,7,0x008},
+ {0x00DDF,1,0x002}, {0x00DF2,2,0x008},
+ {0x00E30,11,0x002}, {0x00E40,5,0x004},
+ {0x00E45,1,0x002}, {0x00E47,8,0x002},
+ {0x00EB0,10,0x002}, {0x00EBB,2,0x002},
+ {0x00EC0,5,0x004}, {0x00EC8,6,0x002},
+ {0x00F18,2,0x002}, {0x00F35,1,0x002},
+ {0x00F37,1,0x002}, {0x00F39,1,0x002},
+ {0x00F3E,2,0x008}, {0x00F71,14,0x002},
+ {0x00F7F,1,0x008}, {0x00F80,5,0x002},
+ {0x00F86,2,0x002}, {0x00F90,8,0x002},
+ {0x00F99,36,0x002}, {0x00FC6,1,0x002},
+ {0x0102B,2,0x008}, {0x0102D,4,0x002},
+ {0x01031,1,0x008}, {0x01032,6,0x002},
+ {0x01038,1,0x008}, {0x01039,2,0x002},
+ {0x0103B,2,0x008}, {0x0103D,2,0x002},
+ {0x01056,2,0x008}, {0x01058,2,0x002},
+ {0x0105E,3,0x002}, {0x01062,3,0x008},
+ {0x01067,7,0x008}, {0x01071,4,0x002},
+ {0x01082,1,0x002}, {0x01083,2,0x008},
+ {0x01085,2,0x002}, {0x01087,6,0x008},
+ {0x0108D,1,0x002}, {0x0108F,1,0x008},
+ {0x01100,90,0x010}, {0x0115F,1,0x010},
+ {0x01160,67,0x020}, {0x011A8,82,0x040},
+ {0x0135F,1,0x002}, {0x01712,3,0x002},
+ {0x01732,3,0x002}, {0x01752,2,0x002},
+ {0x01772,2,0x002}, {0x017B4,2,0x001},
+ {0x017B6,1,0x008}, {0x017B7,7,0x002},
+ {0x017BE,8,0x008}, {0x017C6,1,0x002},
+ {0x017C7,2,0x008}, {0x017C9,11,0x002},
+ {0x017DD,1,0x002}, {0x0180B,3,0x002},
+ {0x018A9,1,0x002}, {0x01920,3,0x002},
+ {0x01923,4,0x008}, {0x01927,2,0x002},
+ {0x01929,3,0x008}, {0x01930,2,0x008},
+ {0x01932,1,0x002}, {0x01933,6,0x008},
+ {0x01939,3,0x002}, {0x019B0,17,0x008},
+ {0x019C8,2,0x008}, {0x01A17,2,0x002},
+ {0x01A19,3,0x008}, {0x01B00,4,0x002},
+ {0x01B04,1,0x008}, {0x01B34,1,0x002},
+ {0x01B35,1,0x008}, {0x01B36,5,0x002},
+ {0x01B3B,1,0x008}, {0x01B3C,1,0x002},
+ {0x01B3D,5,0x008}, {0x01B42,1,0x002},
+ {0x01B43,2,0x008}, {0x01B6B,9,0x002},
+ {0x01B80,2,0x002}, {0x01B82,1,0x008},
+ {0x01BA1,1,0x008}, {0x01BA2,4,0x002},
+ {0x01BA6,2,0x008}, {0x01BA8,2,0x002},
+ {0x01BAA,1,0x008}, {0x01C24,8,0x008},
+ {0x01C2C,8,0x002}, {0x01C34,2,0x008},
+ {0x01C36,2,0x002}, {0x01DC0,39,0x002},
+ {0x01DFE,2,0x002}, {0x0200B,1,0x001},
+ {0x0200C,2,0x002}, {0x0200E,2,0x001},
+ {0x02028,7,0x001}, {0x02060,5,0x001},
+ {0x0206A,6,0x001}, {0x020D0,33,0x002},
+ {0x02DE0,32,0x002}, {0x0302A,6,0x002},
+ {0x03099,2,0x002}, {0x0A66F,4,0x002},
+ {0x0A67C,2,0x002}, {0x0A802,1,0x002},
+ {0x0A806,1,0x002}, {0x0A80B,1,0x002},
+ {0x0A823,2,0x008}, {0x0A825,2,0x002},
+ {0x0A827,1,0x008}, {0x0A880,2,0x008},
+ {0x0A8B4,16,0x008}, {0x0A8C4,1,0x002},
+ {0x0A926,8,0x002}, {0x0A947,11,0x002},
+ {0x0A952,2,0x008}, {0x0AA29,6,0x002},
+ {0x0AA2F,2,0x008}, {0x0AA31,2,0x002},
+ {0x0AA33,2,0x008}, {0x0AA35,2,0x002},
+ {0x0AA43,1,0x002}, {0x0AA4C,1,0x002},
+ {0x0AA4D,1,0x008}, {0x0AC00,1,0x080},
+ {0x0AC01,27,0x100}, {0x0AC1C,1,0x080},
+ {0x0AC1D,27,0x100}, {0x0AC38,1,0x080},
+ {0x0AC39,27,0x100}, {0x0AC54,1,0x080},
+ {0x0AC55,27,0x100}, {0x0AC70,1,0x080},
+ {0x0AC71,27,0x100}, {0x0AC8C,1,0x080},
+ {0x0AC8D,27,0x100}, {0x0ACA8,1,0x080},
+ {0x0ACA9,27,0x100}, {0x0ACC4,1,0x080},
+ {0x0ACC5,27,0x100}, {0x0ACE0,1,0x080},
+ {0x0ACE1,27,0x100}, {0x0ACFC,1,0x080},
+ {0x0ACFD,27,0x100}, {0x0AD18,1,0x080},
+ {0x0AD19,27,0x100}, {0x0AD34,1,0x080},
+ {0x0AD35,27,0x100}, {0x0AD50,1,0x080},
+ {0x0AD51,27,0x100}, {0x0AD6C,1,0x080},
+ {0x0AD6D,27,0x100}, {0x0AD88,1,0x080},
+ {0x0AD89,27,0x100}, {0x0ADA4,1,0x080},
+ {0x0ADA5,27,0x100}, {0x0ADC0,1,0x080},
+ {0x0ADC1,27,0x100}, {0x0ADDC,1,0x080},
+ {0x0ADDD,27,0x100}, {0x0ADF8,1,0x080},
+ {0x0ADF9,27,0x100}, {0x0AE14,1,0x080},
+ {0x0AE15,27,0x100}, {0x0AE30,1,0x080},
+ {0x0AE31,27,0x100}, {0x0AE4C,1,0x080},
+ {0x0AE4D,27,0x100}, {0x0AE68,1,0x080},
+ {0x0AE69,27,0x100}, {0x0AE84,1,0x080},
+ {0x0AE85,27,0x100}, {0x0AEA0,1,0x080},
+ {0x0AEA1,27,0x100}, {0x0AEBC,1,0x080},
+ {0x0AEBD,27,0x100}, {0x0AED8,1,0x080},
+ {0x0AED9,27,0x100}, {0x0AEF4,1,0x080},
+ {0x0AEF5,27,0x100}, {0x0AF10,1,0x080},
+ {0x0AF11,27,0x100}, {0x0AF2C,1,0x080},
+ {0x0AF2D,27,0x100}, {0x0AF48,1,0x080},
+ {0x0AF49,27,0x100}, {0x0AF64,1,0x080},
+ {0x0AF65,27,0x100}, {0x0AF80,1,0x080},
+ {0x0AF81,27,0x100}, {0x0AF9C,1,0x080},
+ {0x0AF9D,27,0x100}, {0x0AFB8,1,0x080},
+ {0x0AFB9,27,0x100}, {0x0AFD4,1,0x080},
+ {0x0AFD5,27,0x100}, {0x0AFF0,1,0x080},
+ {0x0AFF1,27,0x100}, {0x0B00C,1,0x080},
+ {0x0B00D,27,0x100}, {0x0B028,1,0x080},
+ {0x0B029,27,0x100}, {0x0B044,1,0x080},
+ {0x0B045,27,0x100}, {0x0B060,1,0x080},
+ {0x0B061,27,0x100}, {0x0B07C,1,0x080},
+ {0x0B07D,27,0x100}, {0x0B098,1,0x080},
+ {0x0B099,27,0x100}, {0x0B0B4,1,0x080},
+ {0x0B0B5,27,0x100}, {0x0B0D0,1,0x080},
+ {0x0B0D1,27,0x100}, {0x0B0EC,1,0x080},
+ {0x0B0ED,27,0x100}, {0x0B108,1,0x080},
+ {0x0B109,27,0x100}, {0x0B124,1,0x080},
+ {0x0B125,27,0x100}, {0x0B140,1,0x080},
+ {0x0B141,27,0x100}, {0x0B15C,1,0x080},
+ {0x0B15D,27,0x100}, {0x0B178,1,0x080},
+ {0x0B179,27,0x100}, {0x0B194,1,0x080},
+ {0x0B195,27,0x100}, {0x0B1B0,1,0x080},
+ {0x0B1B1,27,0x100}, {0x0B1CC,1,0x080},
+ {0x0B1CD,27,0x100}, {0x0B1E8,1,0x080},
+ {0x0B1E9,27,0x100}, {0x0B204,1,0x080},
+ {0x0B205,27,0x100}, {0x0B220,1,0x080},
+ {0x0B221,27,0x100}, {0x0B23C,1,0x080},
+ {0x0B23D,27,0x100}, {0x0B258,1,0x080},
+ {0x0B259,27,0x100}, {0x0B274,1,0x080},
+ {0x0B275,27,0x100}, {0x0B290,1,0x080},
+ {0x0B291,27,0x100}, {0x0B2AC,1,0x080},
+ {0x0B2AD,27,0x100}, {0x0B2C8,1,0x080},
+ {0x0B2C9,27,0x100}, {0x0B2E4,1,0x080},
+ {0x0B2E5,27,0x100}, {0x0B300,1,0x080},
+ {0x0B301,27,0x100}, {0x0B31C,1,0x080},
+ {0x0B31D,27,0x100}, {0x0B338,1,0x080},
+ {0x0B339,27,0x100}, {0x0B354,1,0x080},
+ {0x0B355,27,0x100}, {0x0B370,1,0x080},
+ {0x0B371,27,0x100}, {0x0B38C,1,0x080},
+ {0x0B38D,27,0x100}, {0x0B3A8,1,0x080},
+ {0x0B3A9,27,0x100}, {0x0B3C4,1,0x080},
+ {0x0B3C5,27,0x100}, {0x0B3E0,1,0x080},
+ {0x0B3E1,27,0x100}, {0x0B3FC,1,0x080},
+ {0x0B3FD,27,0x100}, {0x0B418,1,0x080},
+ {0x0B419,27,0x100}, {0x0B434,1,0x080},
+ {0x0B435,27,0x100}, {0x0B450,1,0x080},
+ {0x0B451,27,0x100}, {0x0B46C,1,0x080},
+ {0x0B46D,27,0x100}, {0x0B488,1,0x080},
+ {0x0B489,27,0x100}, {0x0B4A4,1,0x080},
+ {0x0B4A5,27,0x100}, {0x0B4C0,1,0x080},
+ {0x0B4C1,27,0x100}, {0x0B4DC,1,0x080},
+ {0x0B4DD,27,0x100}, {0x0B4F8,1,0x080},
+ {0x0B4F9,27,0x100}, {0x0B514,1,0x080},
+ {0x0B515,27,0x100}, {0x0B530,1,0x080},
+ {0x0B531,27,0x100}, {0x0B54C,1,0x080},
+ {0x0B54D,27,0x100}, {0x0B568,1,0x080},
+ {0x0B569,27,0x100}, {0x0B584,1,0x080},
+ {0x0B585,27,0x100}, {0x0B5A0,1,0x080},
+ {0x0B5A1,27,0x100}, {0x0B5BC,1,0x080},
+ {0x0B5BD,27,0x100}, {0x0B5D8,1,0x080},
+ {0x0B5D9,27,0x100}, {0x0B5F4,1,0x080},
+ {0x0B5F5,27,0x100}, {0x0B610,1,0x080},
+ {0x0B611,27,0x100}, {0x0B62C,1,0x080},
+ {0x0B62D,27,0x100}, {0x0B648,1,0x080},
+ {0x0B649,27,0x100}, {0x0B664,1,0x080},
+ {0x0B665,27,0x100}, {0x0B680,1,0x080},
+ {0x0B681,27,0x100}, {0x0B69C,1,0x080},
+ {0x0B69D,27,0x100}, {0x0B6B8,1,0x080},
+ {0x0B6B9,27,0x100}, {0x0B6D4,1,0x080},
+ {0x0B6D5,27,0x100}, {0x0B6F0,1,0x080},
+ {0x0B6F1,27,0x100}, {0x0B70C,1,0x080},
+ {0x0B70D,27,0x100}, {0x0B728,1,0x080},
+ {0x0B729,27,0x100}, {0x0B744,1,0x080},
+ {0x0B745,27,0x100}, {0x0B760,1,0x080},
+ {0x0B761,27,0x100}, {0x0B77C,1,0x080},
+ {0x0B77D,27,0x100}, {0x0B798,1,0x080},
+ {0x0B799,27,0x100}, {0x0B7B4,1,0x080},
+ {0x0B7B5,27,0x100}, {0x0B7D0,1,0x080},
+ {0x0B7D1,27,0x100}, {0x0B7EC,1,0x080},
+ {0x0B7ED,27,0x100}, {0x0B808,1,0x080},
+ {0x0B809,27,0x100}, {0x0B824,1,0x080},
+ {0x0B825,27,0x100}, {0x0B840,1,0x080},
+ {0x0B841,27,0x100}, {0x0B85C,1,0x080},
+ {0x0B85D,27,0x100}, {0x0B878,1,0x080},
+ {0x0B879,27,0x100}, {0x0B894,1,0x080},
+ {0x0B895,27,0x100}, {0x0B8B0,1,0x080},
+ {0x0B8B1,27,0x100}, {0x0B8CC,1,0x080},
+ {0x0B8CD,27,0x100}, {0x0B8E8,1,0x080},
+ {0x0B8E9,27,0x100}, {0x0B904,1,0x080},
+ {0x0B905,27,0x100}, {0x0B920,1,0x080},
+ {0x0B921,27,0x100}, {0x0B93C,1,0x080},
+ {0x0B93D,27,0x100}, {0x0B958,1,0x080},
+ {0x0B959,27,0x100}, {0x0B974,1,0x080},
+ {0x0B975,27,0x100}, {0x0B990,1,0x080},
+ {0x0B991,27,0x100}, {0x0B9AC,1,0x080},
+ {0x0B9AD,27,0x100}, {0x0B9C8,1,0x080},
+ {0x0B9C9,27,0x100}, {0x0B9E4,1,0x080},
+ {0x0B9E5,27,0x100}, {0x0BA00,1,0x080},
+ {0x0BA01,27,0x100}, {0x0BA1C,1,0x080},
+ {0x0BA1D,27,0x100}, {0x0BA38,1,0x080},
+ {0x0BA39,27,0x100}, {0x0BA54,1,0x080},
+ {0x0BA55,27,0x100}, {0x0BA70,1,0x080},
+ {0x0BA71,27,0x100}, {0x0BA8C,1,0x080},
+ {0x0BA8D,27,0x100}, {0x0BAA8,1,0x080},
+ {0x0BAA9,27,0x100}, {0x0BAC4,1,0x080},
+ {0x0BAC5,27,0x100}, {0x0BAE0,1,0x080},
+ {0x0BAE1,27,0x100}, {0x0BAFC,1,0x080},
+ {0x0BAFD,27,0x100}, {0x0BB18,1,0x080},
+ {0x0BB19,27,0x100}, {0x0BB34,1,0x080},
+ {0x0BB35,27,0x100}, {0x0BB50,1,0x080},
+ {0x0BB51,27,0x100}, {0x0BB6C,1,0x080},
+ {0x0BB6D,27,0x100}, {0x0BB88,1,0x080},
+ {0x0BB89,27,0x100}, {0x0BBA4,1,0x080},
+ {0x0BBA5,27,0x100}, {0x0BBC0,1,0x080},
+ {0x0BBC1,27,0x100}, {0x0BBDC,1,0x080},
+ {0x0BBDD,27,0x100}, {0x0BBF8,1,0x080},
+ {0x0BBF9,27,0x100}, {0x0BC14,1,0x080},
+ {0x0BC15,27,0x100}, {0x0BC30,1,0x080},
+ {0x0BC31,27,0x100}, {0x0BC4C,1,0x080},
+ {0x0BC4D,27,0x100}, {0x0BC68,1,0x080},
+ {0x0BC69,27,0x100}, {0x0BC84,1,0x080},
+ {0x0BC85,27,0x100}, {0x0BCA0,1,0x080},
+ {0x0BCA1,27,0x100}, {0x0BCBC,1,0x080},
+ {0x0BCBD,27,0x100}, {0x0BCD8,1,0x080},
+ {0x0BCD9,27,0x100}, {0x0BCF4,1,0x080},
+ {0x0BCF5,27,0x100}, {0x0BD10,1,0x080},
+ {0x0BD11,27,0x100}, {0x0BD2C,1,0x080},
+ {0x0BD2D,27,0x100}, {0x0BD48,1,0x080},
+ {0x0BD49,27,0x100}, {0x0BD64,1,0x080},
+ {0x0BD65,27,0x100}, {0x0BD80,1,0x080},
+ {0x0BD81,27,0x100}, {0x0BD9C,1,0x080},
+ {0x0BD9D,27,0x100}, {0x0BDB8,1,0x080},
+ {0x0BDB9,27,0x100}, {0x0BDD4,1,0x080},
+ {0x0BDD5,27,0x100}, {0x0BDF0,1,0x080},
+ {0x0BDF1,27,0x100}, {0x0BE0C,1,0x080},
+ {0x0BE0D,27,0x100}, {0x0BE28,1,0x080},
+ {0x0BE29,27,0x100}, {0x0BE44,1,0x080},
+ {0x0BE45,27,0x100}, {0x0BE60,1,0x080},
+ {0x0BE61,27,0x100}, {0x0BE7C,1,0x080},
+ {0x0BE7D,27,0x100}, {0x0BE98,1,0x080},
+ {0x0BE99,27,0x100}, {0x0BEB4,1,0x080},
+ {0x0BEB5,27,0x100}, {0x0BED0,1,0x080},
+ {0x0BED1,27,0x100}, {0x0BEEC,1,0x080},
+ {0x0BEED,27,0x100}, {0x0BF08,1,0x080},
+ {0x0BF09,27,0x100}, {0x0BF24,1,0x080},
+ {0x0BF25,27,0x100}, {0x0BF40,1,0x080},
+ {0x0BF41,27,0x100}, {0x0BF5C,1,0x080},
+ {0x0BF5D,27,0x100}, {0x0BF78,1,0x080},
+ {0x0BF79,27,0x100}, {0x0BF94,1,0x080},
+ {0x0BF95,27,0x100}, {0x0BFB0,1,0x080},
+ {0x0BFB1,27,0x100}, {0x0BFCC,1,0x080},
+ {0x0BFCD,27,0x100}, {0x0BFE8,1,0x080},
+ {0x0BFE9,27,0x100}, {0x0C004,1,0x080},
+ {0x0C005,27,0x100}, {0x0C020,1,0x080},
+ {0x0C021,27,0x100}, {0x0C03C,1,0x080},
+ {0x0C03D,27,0x100}, {0x0C058,1,0x080},
+ {0x0C059,27,0x100}, {0x0C074,1,0x080},
+ {0x0C075,27,0x100}, {0x0C090,1,0x080},
+ {0x0C091,27,0x100}, {0x0C0AC,1,0x080},
+ {0x0C0AD,27,0x100}, {0x0C0C8,1,0x080},
+ {0x0C0C9,27,0x100}, {0x0C0E4,1,0x080},
+ {0x0C0E5,27,0x100}, {0x0C100,1,0x080},
+ {0x0C101,27,0x100}, {0x0C11C,1,0x080},
+ {0x0C11D,27,0x100}, {0x0C138,1,0x080},
+ {0x0C139,27,0x100}, {0x0C154,1,0x080},
+ {0x0C155,27,0x100}, {0x0C170,1,0x080},
+ {0x0C171,27,0x100}, {0x0C18C,1,0x080},
+ {0x0C18D,27,0x100}, {0x0C1A8,1,0x080},
+ {0x0C1A9,27,0x100}, {0x0C1C4,1,0x080},
+ {0x0C1C5,27,0x100}, {0x0C1E0,1,0x080},
+ {0x0C1E1,27,0x100}, {0x0C1FC,1,0x080},
+ {0x0C1FD,27,0x100}, {0x0C218,1,0x080},
+ {0x0C219,27,0x100}, {0x0C234,1,0x080},
+ {0x0C235,27,0x100}, {0x0C250,1,0x080},
+ {0x0C251,27,0x100}, {0x0C26C,1,0x080},
+ {0x0C26D,27,0x100}, {0x0C288,1,0x080},
+ {0x0C289,27,0x100}, {0x0C2A4,1,0x080},
+ {0x0C2A5,27,0x100}, {0x0C2C0,1,0x080},
+ {0x0C2C1,27,0x100}, {0x0C2DC,1,0x080},
+ {0x0C2DD,27,0x100}, {0x0C2F8,1,0x080},
+ {0x0C2F9,27,0x100}, {0x0C314,1,0x080},
+ {0x0C315,27,0x100}, {0x0C330,1,0x080},
+ {0x0C331,27,0x100}, {0x0C34C,1,0x080},
+ {0x0C34D,27,0x100}, {0x0C368,1,0x080},
+ {0x0C369,27,0x100}, {0x0C384,1,0x080},
+ {0x0C385,27,0x100}, {0x0C3A0,1,0x080},
+ {0x0C3A1,27,0x100}, {0x0C3BC,1,0x080},
+ {0x0C3BD,27,0x100}, {0x0C3D8,1,0x080},
+ {0x0C3D9,27,0x100}, {0x0C3F4,1,0x080},
+ {0x0C3F5,27,0x100}, {0x0C410,1,0x080},
+ {0x0C411,27,0x100}, {0x0C42C,1,0x080},
+ {0x0C42D,27,0x100}, {0x0C448,1,0x080},
+ {0x0C449,27,0x100}, {0x0C464,1,0x080},
+ {0x0C465,27,0x100}, {0x0C480,1,0x080},
+ {0x0C481,27,0x100}, {0x0C49C,1,0x080},
+ {0x0C49D,27,0x100}, {0x0C4B8,1,0x080},
+ {0x0C4B9,27,0x100}, {0x0C4D4,1,0x080},
+ {0x0C4D5,27,0x100}, {0x0C4F0,1,0x080},
+ {0x0C4F1,27,0x100}, {0x0C50C,1,0x080},
+ {0x0C50D,27,0x100}, {0x0C528,1,0x080},
+ {0x0C529,27,0x100}, {0x0C544,1,0x080},
+ {0x0C545,27,0x100}, {0x0C560,1,0x080},
+ {0x0C561,27,0x100}, {0x0C57C,1,0x080},
+ {0x0C57D,27,0x100}, {0x0C598,1,0x080},
+ {0x0C599,27,0x100}, {0x0C5B4,1,0x080},
+ {0x0C5B5,27,0x100}, {0x0C5D0,1,0x080},
+ {0x0C5D1,27,0x100}, {0x0C5EC,1,0x080},
+ {0x0C5ED,27,0x100}, {0x0C608,1,0x080},
+ {0x0C609,27,0x100}, {0x0C624,1,0x080},
+ {0x0C625,27,0x100}, {0x0C640,1,0x080},
+ {0x0C641,27,0x100}, {0x0C65C,1,0x080},
+ {0x0C65D,27,0x100}, {0x0C678,1,0x080},
+ {0x0C679,27,0x100}, {0x0C694,1,0x080},
+ {0x0C695,27,0x100}, {0x0C6B0,1,0x080},
+ {0x0C6B1,27,0x100}, {0x0C6CC,1,0x080},
+ {0x0C6CD,27,0x100}, {0x0C6E8,1,0x080},
+ {0x0C6E9,27,0x100}, {0x0C704,1,0x080},
+ {0x0C705,27,0x100}, {0x0C720,1,0x080},
+ {0x0C721,27,0x100}, {0x0C73C,1,0x080},
+ {0x0C73D,27,0x100}, {0x0C758,1,0x080},
+ {0x0C759,27,0x100}, {0x0C774,1,0x080},
+ {0x0C775,27,0x100}, {0x0C790,1,0x080},
+ {0x0C791,27,0x100}, {0x0C7AC,1,0x080},
+ {0x0C7AD,27,0x100}, {0x0C7C8,1,0x080},
+ {0x0C7C9,27,0x100}, {0x0C7E4,1,0x080},
+ {0x0C7E5,27,0x100}, {0x0C800,1,0x080},
+ {0x0C801,27,0x100}, {0x0C81C,1,0x080},
+ {0x0C81D,27,0x100}, {0x0C838,1,0x080},
+ {0x0C839,27,0x100}, {0x0C854,1,0x080},
+ {0x0C855,27,0x100}, {0x0C870,1,0x080},
+ {0x0C871,27,0x100}, {0x0C88C,1,0x080},
+ {0x0C88D,27,0x100}, {0x0C8A8,1,0x080},
+ {0x0C8A9,27,0x100}, {0x0C8C4,1,0x080},
+ {0x0C8C5,27,0x100}, {0x0C8E0,1,0x080},
+ {0x0C8E1,27,0x100}, {0x0C8FC,1,0x080},
+ {0x0C8FD,27,0x100}, {0x0C918,1,0x080},
+ {0x0C919,27,0x100}, {0x0C934,1,0x080},
+ {0x0C935,27,0x100}, {0x0C950,1,0x080},
+ {0x0C951,27,0x100}, {0x0C96C,1,0x080},
+ {0x0C96D,27,0x100}, {0x0C988,1,0x080},
+ {0x0C989,27,0x100}, {0x0C9A4,1,0x080},
+ {0x0C9A5,27,0x100}, {0x0C9C0,1,0x080},
+ {0x0C9C1,27,0x100}, {0x0C9DC,1,0x080},
+ {0x0C9DD,27,0x100}, {0x0C9F8,1,0x080},
+ {0x0C9F9,27,0x100}, {0x0CA14,1,0x080},
+ {0x0CA15,27,0x100}, {0x0CA30,1,0x080},
+ {0x0CA31,27,0x100}, {0x0CA4C,1,0x080},
+ {0x0CA4D,27,0x100}, {0x0CA68,1,0x080},
+ {0x0CA69,27,0x100}, {0x0CA84,1,0x080},
+ {0x0CA85,27,0x100}, {0x0CAA0,1,0x080},
+ {0x0CAA1,27,0x100}, {0x0CABC,1,0x080},
+ {0x0CABD,27,0x100}, {0x0CAD8,1,0x080},
+ {0x0CAD9,27,0x100}, {0x0CAF4,1,0x080},
+ {0x0CAF5,27,0x100}, {0x0CB10,1,0x080},
+ {0x0CB11,27,0x100}, {0x0CB2C,1,0x080},
+ {0x0CB2D,27,0x100}, {0x0CB48,1,0x080},
+ {0x0CB49,27,0x100}, {0x0CB64,1,0x080},
+ {0x0CB65,27,0x100}, {0x0CB80,1,0x080},
+ {0x0CB81,27,0x100}, {0x0CB9C,1,0x080},
+ {0x0CB9D,27,0x100}, {0x0CBB8,1,0x080},
+ {0x0CBB9,27,0x100}, {0x0CBD4,1,0x080},
+ {0x0CBD5,27,0x100}, {0x0CBF0,1,0x080},
+ {0x0CBF1,27,0x100}, {0x0CC0C,1,0x080},
+ {0x0CC0D,27,0x100}, {0x0CC28,1,0x080},
+ {0x0CC29,27,0x100}, {0x0CC44,1,0x080},
+ {0x0CC45,27,0x100}, {0x0CC60,1,0x080},
+ {0x0CC61,27,0x100}, {0x0CC7C,1,0x080},
+ {0x0CC7D,27,0x100}, {0x0CC98,1,0x080},
+ {0x0CC99,27,0x100}, {0x0CCB4,1,0x080},
+ {0x0CCB5,27,0x100}, {0x0CCD0,1,0x080},
+ {0x0CCD1,27,0x100}, {0x0CCEC,1,0x080},
+ {0x0CCED,27,0x100}, {0x0CD08,1,0x080},
+ {0x0CD09,27,0x100}, {0x0CD24,1,0x080},
+ {0x0CD25,27,0x100}, {0x0CD40,1,0x080},
+ {0x0CD41,27,0x100}, {0x0CD5C,1,0x080},
+ {0x0CD5D,27,0x100}, {0x0CD78,1,0x080},
+ {0x0CD79,27,0x100}, {0x0CD94,1,0x080},
+ {0x0CD95,27,0x100}, {0x0CDB0,1,0x080},
+ {0x0CDB1,27,0x100}, {0x0CDCC,1,0x080},
+ {0x0CDCD,27,0x100}, {0x0CDE8,1,0x080},
+ {0x0CDE9,27,0x100}, {0x0CE04,1,0x080},
+ {0x0CE05,27,0x100}, {0x0CE20,1,0x080},
+ {0x0CE21,27,0x100}, {0x0CE3C,1,0x080},
+ {0x0CE3D,27,0x100}, {0x0CE58,1,0x080},
+ {0x0CE59,27,0x100}, {0x0CE74,1,0x080},
+ {0x0CE75,27,0x100}, {0x0CE90,1,0x080},
+ {0x0CE91,27,0x100}, {0x0CEAC,1,0x080},
+ {0x0CEAD,27,0x100}, {0x0CEC8,1,0x080},
+ {0x0CEC9,27,0x100}, {0x0CEE4,1,0x080},
+ {0x0CEE5,27,0x100}, {0x0CF00,1,0x080},
+ {0x0CF01,27,0x100}, {0x0CF1C,1,0x080},
+ {0x0CF1D,27,0x100}, {0x0CF38,1,0x080},
+ {0x0CF39,27,0x100}, {0x0CF54,1,0x080},
+ {0x0CF55,27,0x100}, {0x0CF70,1,0x080},
+ {0x0CF71,27,0x100}, {0x0CF8C,1,0x080},
+ {0x0CF8D,27,0x100}, {0x0CFA8,1,0x080},
+ {0x0CFA9,27,0x100}, {0x0CFC4,1,0x080},
+ {0x0CFC5,27,0x100}, {0x0CFE0,1,0x080},
+ {0x0CFE1,27,0x100}, {0x0CFFC,1,0x080},
+ {0x0CFFD,27,0x100}, {0x0D018,1,0x080},
+ {0x0D019,27,0x100}, {0x0D034,1,0x080},
+ {0x0D035,27,0x100}, {0x0D050,1,0x080},
+ {0x0D051,27,0x100}, {0x0D06C,1,0x080},
+ {0x0D06D,27,0x100}, {0x0D088,1,0x080},
+ {0x0D089,27,0x100}, {0x0D0A4,1,0x080},
+ {0x0D0A5,27,0x100}, {0x0D0C0,1,0x080},
+ {0x0D0C1,27,0x100}, {0x0D0DC,1,0x080},
+ {0x0D0DD,27,0x100}, {0x0D0F8,1,0x080},
+ {0x0D0F9,27,0x100}, {0x0D114,1,0x080},
+ {0x0D115,27,0x100}, {0x0D130,1,0x080},
+ {0x0D131,27,0x100}, {0x0D14C,1,0x080},
+ {0x0D14D,27,0x100}, {0x0D168,1,0x080},
+ {0x0D169,27,0x100}, {0x0D184,1,0x080},
+ {0x0D185,27,0x100}, {0x0D1A0,1,0x080},
+ {0x0D1A1,27,0x100}, {0x0D1BC,1,0x080},
+ {0x0D1BD,27,0x100}, {0x0D1D8,1,0x080},
+ {0x0D1D9,27,0x100}, {0x0D1F4,1,0x080},
+ {0x0D1F5,27,0x100}, {0x0D210,1,0x080},
+ {0x0D211,27,0x100}, {0x0D22C,1,0x080},
+ {0x0D22D,27,0x100}, {0x0D248,1,0x080},
+ {0x0D249,27,0x100}, {0x0D264,1,0x080},
+ {0x0D265,27,0x100}, {0x0D280,1,0x080},
+ {0x0D281,27,0x100}, {0x0D29C,1,0x080},
+ {0x0D29D,27,0x100}, {0x0D2B8,1,0x080},
+ {0x0D2B9,27,0x100}, {0x0D2D4,1,0x080},
+ {0x0D2D5,27,0x100}, {0x0D2F0,1,0x080},
+ {0x0D2F1,27,0x100}, {0x0D30C,1,0x080},
+ {0x0D30D,27,0x100}, {0x0D328,1,0x080},
+ {0x0D329,27,0x100}, {0x0D344,1,0x080},
+ {0x0D345,27,0x100}, {0x0D360,1,0x080},
+ {0x0D361,27,0x100}, {0x0D37C,1,0x080},
+ {0x0D37D,27,0x100}, {0x0D398,1,0x080},
+ {0x0D399,27,0x100}, {0x0D3B4,1,0x080},
+ {0x0D3B5,27,0x100}, {0x0D3D0,1,0x080},
+ {0x0D3D1,27,0x100}, {0x0D3EC,1,0x080},
+ {0x0D3ED,27,0x100}, {0x0D408,1,0x080},
+ {0x0D409,27,0x100}, {0x0D424,1,0x080},
+ {0x0D425,27,0x100}, {0x0D440,1,0x080},
+ {0x0D441,27,0x100}, {0x0D45C,1,0x080},
+ {0x0D45D,27,0x100}, {0x0D478,1,0x080},
+ {0x0D479,27,0x100}, {0x0D494,1,0x080},
+ {0x0D495,27,0x100}, {0x0D4B0,1,0x080},
+ {0x0D4B1,27,0x100}, {0x0D4CC,1,0x080},
+ {0x0D4CD,27,0x100}, {0x0D4E8,1,0x080},
+ {0x0D4E9,27,0x100}, {0x0D504,1,0x080},
+ {0x0D505,27,0x100}, {0x0D520,1,0x080},
+ {0x0D521,27,0x100}, {0x0D53C,1,0x080},
+ {0x0D53D,27,0x100}, {0x0D558,1,0x080},
+ {0x0D559,27,0x100}, {0x0D574,1,0x080},
+ {0x0D575,27,0x100}, {0x0D590,1,0x080},
+ {0x0D591,27,0x100}, {0x0D5AC,1,0x080},
+ {0x0D5AD,27,0x100}, {0x0D5C8,1,0x080},
+ {0x0D5C9,27,0x100}, {0x0D5E4,1,0x080},
+ {0x0D5E5,27,0x100}, {0x0D600,1,0x080},
+ {0x0D601,27,0x100}, {0x0D61C,1,0x080},
+ {0x0D61D,27,0x100}, {0x0D638,1,0x080},
+ {0x0D639,27,0x100}, {0x0D654,1,0x080},
+ {0x0D655,27,0x100}, {0x0D670,1,0x080},
+ {0x0D671,27,0x100}, {0x0D68C,1,0x080},
+ {0x0D68D,27,0x100}, {0x0D6A8,1,0x080},
+ {0x0D6A9,27,0x100}, {0x0D6C4,1,0x080},
+ {0x0D6C5,27,0x100}, {0x0D6E0,1,0x080},
+ {0x0D6E1,27,0x100}, {0x0D6FC,1,0x080},
+ {0x0D6FD,27,0x100}, {0x0D718,1,0x080},
+ {0x0D719,27,0x100}, {0x0D734,1,0x080},
+ {0x0D735,27,0x100}, {0x0D750,1,0x080},
+ {0x0D751,27,0x100}, {0x0D76C,1,0x080},
+ {0x0D76D,27,0x100}, {0x0D788,1,0x080},
+ {0x0D789,27,0x100}, {0x0FB1E,1,0x002},
+ {0x0FE00,16,0x002}, {0x0FE20,7,0x002},
+ {0x0FEFF,1,0x001}, {0x0FF9E,2,0x002},
+ {0x0FFF9,3,0x001}, {0x101FD,1,0x002},
+ {0x10A01,3,0x002}, {0x10A05,2,0x002},
+ {0x10A0C,4,0x002}, {0x10A38,3,0x002},
+ {0x10A3F,1,0x002}, {0x1D165,1,0x002},
+ {0x1D166,1,0x008}, {0x1D167,3,0x002},
+ {0x1D16D,1,0x008}, {0x1D16E,5,0x002},
+ {0x1D173,8,0x001}, {0x1D17B,8,0x002},
+ {0x1D185,7,0x002}, {0x1D1AA,4,0x002},
+ {0x1D242,3,0x002}, {0xE0001,1,0x001},
+ {0xE0020,96,0x001}, {0xE0100,240,0x002},
+};
+
+static int
+grapheme_cmp(const void *p1, const void *p2)
+{
+ const struct graphme_table_t *k = p1;
+ const struct graphme_table_t *v = p2;
+ OnigCodePoint c = k->codepoint_min;
+ if (c < v->codepoint_min)
+ return -1;
+ if (v->codepoint_min + v->num_codepoints <= c)
+ return 1;
+ return 0;
+}
+
+static unsigned int
+get_grapheme_properties(OnigCodePoint c)
+{
+ struct graphme_table_t key, *found;
+ key.codepoint_min = c;
+ found = bsearch(&key, graphme_table, sizeof(graphme_table)/sizeof(*graphme_table),
+ sizeof(*graphme_table), grapheme_cmp);
+ if (found)
+ return found->properties;
+ return 0;
+}
+
+/* Stream-Safe Text Format assumed
+ * http://unicode.org/reports/tr15/ */
+#define MAX_BYTES_LENGTH 128
+
+static OnigCodePoint mbc_to_code0(const UChar* p, const UChar* end, int len);
+
+static int
+grapheme_boundary_p(int props1, int props2)
+{
+ if (props2 & GRAPHEME_BIT_CONTROL)
+ return 1;
+ if (((props1 & GRAPHEME_BIT_L) && (props2 & (GRAPHEME_BIT_L|
+ GRAPHEME_BIT_V|
+ GRAPHEME_BIT_LV|
+ GRAPHEME_BIT_LVT))) ||
+ ((props1 & (GRAPHEME_BIT_LV|
+ GRAPHEME_BIT_V)) && (props2 & (GRAPHEME_BIT_V|
+ GRAPHEME_BIT_T))) ||
+ ((props1 & (GRAPHEME_BIT_LVT|
+ GRAPHEME_BIT_T)) && (props2 & GRAPHEME_BIT_T)))
+ return 0;
+ if (props2 & (GRAPHEME_BIT_EXTEND|
+ GRAPHEME_BIT_SPACINGMARK))
+ return 0;
+ if (props1 & GRAPHEME_BIT_PREPEND)
+ return 0;
+ return 1;
+}
+
+static int
+mbchar_enc_len(const UChar* p, const UChar* e, OnigEncoding enc ARG_UNUSED)
+{
+ /*
+ * this implements extended grapheme clusters ("user-perceived characters")
+ * http://www.unicode.org/reports/tr29/
+ */
+ int r1, l1, r2, l2;
+ OnigCodePoint c1, c2;
+ unsigned int p1, p2;
+ r1 = mbc_enc_len(p, e, enc);
+ if (!ONIGENC_MBCLEN_CHARFOUND_P(r1))
+ return r1;
+ l1 = ONIGENC_MBCLEN_CHARFOUND_LEN(r1);
+ c1 = mbc_to_code0(p, e, l1);
+ p1 = get_grapheme_properties(c1);
+
+ if (p + l1 == e)
+ return r1;
+
+ while (p + l1 < e && l1 < MAX_BYTES_LENGTH-4) {
+ if (p1 & GRAPHEME_BIT_CONTROL)
+ return ONIGENC_CONSTRUCT_MBCLEN_CHARFOUND(l1);
+ r2 = mbc_enc_len(p+l1, e, enc);
+ if (ONIGENC_MBCLEN_INVALID_P(r2))
+ return ONIGENC_CONSTRUCT_MBCLEN_CHARFOUND(l1);
+ if (ONIGENC_MBCLEN_NEEDMORE_P(r2))
+ return r2;
+ l2 = ONIGENC_MBCLEN_CHARFOUND_LEN(r2);
+ c2 = mbc_to_code0(p+l1, e, l2);
+ p2 = get_grapheme_properties(c2);
+ if (grapheme_boundary_p(p1, p2))
+ return ONIGENC_CONSTRUCT_MBCLEN_CHARFOUND(l1);
+ l1 += l2;
+ p1 = p2;
+ }
+ /* if p+l1==e, charfound AND needmore */
+ return ONIGENC_CONSTRUCT_MBCLEN_CHARFOUND(l1);
+}
+
static int
is_mbc_newline(const UChar* p, const UChar* end, OnigEncoding enc)
{
@@ -267,12 +918,11 @@ is_mbc_newline(const UChar* p, const UCh
}
static OnigCodePoint
-mbc_to_code(const UChar* p, const UChar* end, OnigEncoding enc)
+mbc_to_code0(const UChar* p, const UChar* end, int len)
{
- int c, len;
+ int c;
OnigCodePoint n;
- len = enclen(enc, p, end);
c = *p++;
if (len > 1) {
len--;
@@ -293,6 +943,24 @@ mbc_to_code(const UChar* p, const UChar*
}
}
+static OnigCodePoint
+mbc_to_code(const UChar* p, const UChar* end, int *precious_ret, OnigEncoding enc)
+{
+ int len;
+ int ret;
+
+ ret = mbc_enc_len(p, end, enc);
+ if (precious_ret)
+ *precious_ret = ret;
+ if (ONIGENC_MBCLEN_CHARFOUND_P(ret))
+ len = ONIGENC_MBCLEN_CHARFOUND_LEN(ret);
+ else if (ONIGENC_MBCLEN_NEEDMORE_P(ret))
+ len = end-p+ONIGENC_MBCLEN_NEEDMORE_LEN(ret);
+ else
+ len = 1;
+ return mbc_to_code0(p, end, len);
+}
+
static int
code_to_mbclen(OnigCodePoint code, OnigEncoding enc ARG_UNUSED)
{
@@ -416,6 +1084,30 @@ left_adjust_char_head(const UChar* start
return (UChar* )p;
}
+static UChar*
+left_adjust_char2_head(const UChar* start, const UChar* s, const UChar* end, OnigEncoding enc ARG_UNUSED)
+{
+ const UChar *p = left_adjust_char_head(start, s, end, enc);
+ const UChar *q;
+ OnigCodePoint c1, c2;
+ unsigned int p1, p2;
+
+ c2 = mbc_to_code(p, end, NULL, enc);
+ p2 = get_grapheme_properties(c2);
+
+ while (start < p) {
+ q = left_adjust_char_head(start, p-1, end, enc);
+ c1 = mbc_to_code(q, end, NULL, enc);
+ p1 = get_grapheme_properties(c1);
+ if (grapheme_boundary_p(p1, p2))
+ break;
+ c2 = c1;
+ p2 = p1;
+ p = q;
+ }
+ return (UChar *)p;
+}
+
static int
get_case_fold_codes_by_str(OnigCaseFoldType flag,
const OnigUChar* p, const OnigUChar* end, OnigCaseFoldCodeItem items[],
@@ -425,9 +1117,9 @@ get_case_fold_codes_by_str(OnigCaseFoldT
}
OnigEncodingDefine(utf_8, UTF_8) = {
- mbc_enc_len,
+ mbchar_enc_len,
"UTF-8", /* name */
- 6, /* max byte length */
+ MAX_BYTES_LENGTH, /* max byte length */
1, /* min byte length */
is_mbc_newline,
mbc_to_code,
@@ -440,6 +1132,7 @@ OnigEncodingDefine(utf_8, UTF_8) = {
onigenc_unicode_is_code_ctype,
get_ctype_code_range,
left_adjust_char_head,
+ left_adjust_char2_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("CP65001", "UTF-8")
Index: enc/big5.c
===================================================================
--- enc/big5.c (revision 19335)
+++ enc/big5.c (working copy)
@@ -108,9 +108,9 @@ big5_mbc_enc_len(const UChar* p, const U
}
static OnigCodePoint
-big5_mbc_to_code(const UChar* p, const UChar* end, OnigEncoding enc)
+big5_mbc_to_code(const UChar* p, const UChar* end, int *precious_ret, OnigEncoding enc)
{
- return onigenc_mbn_mbc_to_code(enc, p, end);
+ return onigenc_mbn_mbc_to_code(enc, p, end, precious_ret);
}
static int
@@ -211,6 +211,7 @@ OnigEncodingDefine(big5, BIG5) = {
big5_is_code_ctype,
onigenc_not_support_get_ctype_code_range,
big5_left_adjust_char_head,
+ big5_left_adjust_char_head,
big5_is_allowed_reverse_match
};
ENC_ALIAS("CP950", "BIG5")
Index: enc/euc_tw.c
===================================================================
--- enc/euc_tw.c (revision 19335)
+++ enc/euc_tw.c (working copy)
@@ -150,9 +150,9 @@ euctw_mbc_enc_len(const UChar* p, const
}
static OnigCodePoint
-euctw_mbc_to_code(const UChar* p, const UChar* end, OnigEncoding enc ARG_UNUSED)
+euctw_mbc_to_code(const UChar* p, const UChar* end, int *precious_ret, OnigEncoding enc ARG_UNUSED)
{
- return onigenc_mbn_mbc_to_code(enc, p, end);
+ return onigenc_mbn_mbc_to_code(enc, p, end, precious_ret);
}
static int
@@ -229,6 +229,7 @@ OnigEncodingDefine(euc_tw, EUC_TW) = {
euctw_is_code_ctype,
onigenc_not_support_get_ctype_code_range,
euctw_left_adjust_char_head,
+ euctw_left_adjust_char_head,
euctw_is_allowed_reverse_match
};
ENC_ALIAS("eucTW", "EUC-TW")
Index: enc/iso_8859_10.c
===================================================================
--- enc/iso_8859_10.c (revision 19335)
+++ enc/iso_8859_10.c (working copy)
@@ -239,6 +239,7 @@ OnigEncodingDefine(iso_8859_10, ISO_8859
is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("ISO8859-10", "ISO-8859-10")
Index: enc/iso_8859_11.c
===================================================================
--- enc/iso_8859_11.c (revision 19335)
+++ enc/iso_8859_11.c (working copy)
@@ -92,6 +92,7 @@ OnigEncodingDefine(iso_8859_11, ISO_8859
is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("ISO8859-11", "ISO-8859-11")
Index: enc/ascii.c
===================================================================
--- enc/ascii.c (revision 19335)
+++ enc/ascii.c (working copy)
@@ -45,6 +45,7 @@ OnigEncodingDefine(ascii, ASCII) = {
onigenc_ascii_is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("BINARY", "ASCII-8BIT")
Index: enc/iso_8859_13.c
===================================================================
--- enc/iso_8859_13.c (revision 19335)
+++ enc/iso_8859_13.c (working copy)
@@ -228,6 +228,7 @@ OnigEncodingDefine(iso_8859_13, ISO_8859
is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("ISO8859-13", "ISO-8859-13")
Index: enc/iso_8859_14.c
===================================================================
--- enc/iso_8859_14.c (revision 19335)
+++ enc/iso_8859_14.c (working copy)
@@ -241,6 +241,7 @@ OnigEncodingDefine(iso_8859_14, ISO_8859
is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("ISO8859-14", "ISO-8859-14")
Index: enc/iso_8859_15.c
===================================================================
--- enc/iso_8859_15.c (revision 19335)
+++ enc/iso_8859_15.c (working copy)
@@ -235,6 +235,7 @@ OnigEncodingDefine(iso_8859_15, ISO_8859
is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("ISO8859-15", "ISO-8859-15")
Index: enc/iso_8859_16.c
===================================================================
--- enc/iso_8859_16.c (revision 19335)
+++ enc/iso_8859_16.c (working copy)
@@ -237,6 +237,7 @@ OnigEncodingDefine(iso_8859_16, ISO_8859
is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("ISO8859-16", "ISO-8859-16")
Index: enc/us_ascii.c
===================================================================
--- enc/us_ascii.c (revision 19335)
+++ enc/us_ascii.c (working copy)
@@ -24,6 +24,7 @@ OnigEncodingDefine(us_ascii, US_ASCII) =
onigenc_ascii_is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("ASCII", "US-ASCII")
Index: enc/windows_1251.c
===================================================================
--- enc/windows_1251.c (revision 19335)
+++ enc/windows_1251.c (working copy)
@@ -196,6 +196,7 @@ OnigEncodingDefine(windows_1251, Windows
cp1251_is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
/*
Index: enc/iso_8859_1.c
===================================================================
--- enc/iso_8859_1.c (revision 19335)
+++ enc/iso_8859_1.c (working copy)
@@ -270,6 +270,7 @@ OnigEncodingDefine(iso_8859_1, ISO_8859_
is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("ISO8859-1", "ISO-8859-1")
Index: enc/iso_8859_2.c
===================================================================
--- enc/iso_8859_2.c (revision 19335)
+++ enc/iso_8859_2.c (working copy)
@@ -235,6 +235,7 @@ OnigEncodingDefine(iso_8859_2, ISO_8859_
is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("ISO8859-2", "ISO-8859-2")
Index: enc/euc_kr.c
===================================================================
--- enc/euc_kr.c (revision 19335)
+++ enc/euc_kr.c (working copy)
@@ -108,9 +108,9 @@ euckr_mbc_enc_len(const UChar* p, const
}
static OnigCodePoint
-euckr_mbc_to_code(const UChar* p, const UChar* end, OnigEncoding enc)
+euckr_mbc_to_code(const UChar* p, const UChar* end, int *precious_ret, OnigEncoding enc)
{
- return onigenc_mbn_mbc_to_code(enc, p, end);
+ return onigenc_mbn_mbc_to_code(enc, p, end, precious_ret);
}
static int
@@ -187,6 +187,7 @@ OnigEncodingDefine(euc_kr, EUC_KR) = {
euckr_is_code_ctype,
onigenc_not_support_get_ctype_code_range,
euckr_left_adjust_char_head,
+ euckr_left_adjust_char_head,
euckr_is_allowed_reverse_match
};
ENC_ALIAS("eucKR", "EUC-KR")
Index: enc/iso_8859_3.c
===================================================================
--- enc/iso_8859_3.c (revision 19335)
+++ enc/iso_8859_3.c (working copy)
@@ -235,6 +235,7 @@ OnigEncodingDefine(iso_8859_3, ISO_8859_
is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("ISO8859-3", "ISO-8859-3")
Index: enc/utf_32be.c
===================================================================
--- enc/utf_32be.c (revision 19335)
+++ enc/utf_32be.c (working copy)
@@ -61,8 +61,11 @@ utf32be_is_mbc_newline(const UChar* p, c
static OnigCodePoint
utf32be_mbc_to_code(const UChar* p, const UChar* end ARG_UNUSED,
+ int *precious_ret,
OnigEncoding enc ARG_UNUSED)
{
+ if (precious_ret)
+ *precious_ret = ONIGENC_CONSTRUCT_MBCLEN_CHARFOUND(4);
return (OnigCodePoint )(((p[0] * 256 + p[1]) * 256 + p[2]) * 256 + p[3]);
}
@@ -189,6 +192,7 @@ OnigEncodingDefine(utf_32be, UTF_32BE) =
onigenc_unicode_is_code_ctype,
onigenc_utf16_32_get_ctype_code_range,
utf32be_left_adjust_char_head,
+ utf32be_left_adjust_char_head,
onigenc_always_false_is_allowed_reverse_match
};
ENC_ALIAS("UCS-4BE", "UTF-32BE")
Index: enc/iso_8859_4.c
===================================================================
--- enc/iso_8859_4.c (revision 19335)
+++ enc/iso_8859_4.c (working copy)
@@ -237,6 +237,7 @@ OnigEncodingDefine(iso_8859_4, ISO_8859_
is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("ISO8859-4", "ISO-8859-4")
Index: enc/emacs_mule.c
===================================================================
--- enc/emacs_mule.c (revision 19335)
+++ enc/emacs_mule.c (working copy)
@@ -223,21 +223,9 @@ mbc_enc_len(const UChar* p, const UChar*
}
static OnigCodePoint
-mbc_to_code(const UChar* p, const UChar* end, OnigEncoding enc)
+mbc_to_code(const UChar* p, const UChar* end, int *precious_ret, OnigEncoding enc)
{
- int c, i, len;
- OnigCodePoint n;
-
- len = enclen(enc, p, end);
- n = (OnigCodePoint )*p++;
- if (len == 1) return n;
-
- for (i = 1; i < len; i++) {
- if (p >= end) break;
- c = *p++;
- n <<= 8; n += c;
- }
- return n;
+ return onigenc_mbn_mbc_to_code(enc, p, end, precious_ret);
}
static int
@@ -333,6 +321,7 @@ OnigEncodingDefine(emacs_mule, Emacs_Mul
is_code_ctype,
onigenc_not_support_get_ctype_code_range,
left_adjust_char_head,
+ left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match,
0
};
Index: enc/iso_8859_5.c
===================================================================
--- enc/iso_8859_5.c (revision 19335)
+++ enc/iso_8859_5.c (working copy)
@@ -225,6 +225,7 @@ OnigEncodingDefine(iso_8859_5, ISO_8859_
is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("ISO8859-5", "ISO-8859-5")
Index: enc/utf_16be.c
===================================================================
--- enc/utf_16be.c (revision 19335)
+++ enc/utf_16be.c (working copy)
@@ -103,9 +103,15 @@ utf16be_is_mbc_newline(const UChar* p, c
static OnigCodePoint
utf16be_mbc_to_code(const UChar* p, const UChar* end ARG_UNUSED,
+ int *precious_ret,
OnigEncoding enc)
{
OnigCodePoint code;
+ int ret;
+
+ ret = utf16be_mbc_enc_len(p, end, enc);
+ if (precious_ret)
+ *precious_ret = ret;
if (UTF16_IS_SURROGATE_FIRST(*p)) {
code = ((((p[0] - 0xd8) << 2) + ((p[1] & 0xc0) >> 6) + 1) << 16)
@@ -253,6 +259,7 @@ OnigEncodingDefine(utf_16be, UTF_16BE) =
onigenc_unicode_is_code_ctype,
onigenc_utf16_32_get_ctype_code_range,
utf16be_left_adjust_char_head,
+ utf16be_left_adjust_char_head,
onigenc_always_false_is_allowed_reverse_match
};
ENC_ALIAS("UCS-2BE", "UTF-16BE")
Index: enc/iso_8859_6.c
===================================================================
--- enc/iso_8859_6.c (revision 19335)
+++ enc/iso_8859_6.c (working copy)
@@ -92,6 +92,7 @@ OnigEncodingDefine(iso_8859_6, ISO_8859_
is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("ISO8859-6", "ISO-8859-6")
Index: enc/iso_8859_7.c
===================================================================
--- enc/iso_8859_7.c (revision 19335)
+++ enc/iso_8859_7.c (working copy)
@@ -222,6 +222,7 @@ OnigEncodingDefine(iso_8859_7, ISO_8859_
is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("ISO8859-7", "ISO-8859-7")
Index: enc/iso_8859_8.c
===================================================================
--- enc/iso_8859_8.c (revision 19335)
+++ enc/iso_8859_8.c (working copy)
@@ -92,6 +92,7 @@ OnigEncodingDefine(iso_8859_8, ISO_8859_
is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("ISO8859-8", "ISO-8859-8")
Index: enc/iso_8859_9.c
===================================================================
--- enc/iso_8859_9.c (revision 19335)
+++ enc/iso_8859_9.c (working copy)
@@ -228,6 +228,7 @@ OnigEncodingDefine(iso_8859_9, ISO_8859_
is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("ISO8859-9", "ISO-8859-9")
Index: enc/utf_32le.c
===================================================================
--- enc/utf_32le.c (revision 19335)
+++ enc/utf_32le.c (working copy)
@@ -61,8 +61,11 @@ utf32le_is_mbc_newline(const UChar* p, c
static OnigCodePoint
utf32le_mbc_to_code(const UChar* p, const UChar* end ARG_UNUSED,
+ int *precious_ret,
OnigEncoding enc ARG_UNUSED)
{
+ if (precious_ret)
+ *precious_ret = ONIGENC_CONSTRUCT_MBCLEN_CHARFOUND(4);
return (OnigCodePoint )(((p[3] * 256 + p[2]) * 256 + p[1]) * 256 + p[0]);
}
@@ -189,6 +192,7 @@ OnigEncodingDefine(utf_32le, UTF_32LE) =
onigenc_unicode_is_code_ctype,
onigenc_utf16_32_get_ctype_code_range,
utf32le_left_adjust_char_head,
+ utf32le_left_adjust_char_head,
onigenc_always_false_is_allowed_reverse_match
};
ENC_ALIAS("UCS-4LE", "UTF-32LE")
Index: enc/gb18030.c
===================================================================
--- enc/gb18030.c (revision 19335)
+++ enc/gb18030.c (working copy)
@@ -164,21 +164,9 @@ gb18030_mbc_enc_len(const UChar* p, cons
}
static OnigCodePoint
-gb18030_mbc_to_code(const UChar* p, const UChar* end, OnigEncoding enc)
+gb18030_mbc_to_code(const UChar* p, const UChar* end, int *precious_ret, OnigEncoding enc)
{
- int c, i, len;
- OnigCodePoint n;
-
- len = enclen(enc, p, end);
- n = (OnigCodePoint )(*p++);
- if (len == 1) return n;
-
- for (i = 1; i < len; i++) {
- if (p >= end) break;
- c = *p++;
- n <<= 8; n += c;
- }
- return n;
+ return onigenc_mbn_mbc_to_code(enc, p, end, precious_ret);
}
static int
@@ -591,6 +579,7 @@ OnigEncodingDefine(gb18030, GB18030) = {
gb18030_is_code_ctype,
onigenc_not_support_get_ctype_code_range,
gb18030_left_adjust_char_head,
+ gb18030_left_adjust_char_head,
gb18030_is_allowed_reverse_match
};
Index: enc/utf_16le.c
===================================================================
--- enc/utf_16le.c (revision 19335)
+++ enc/utf_16le.c (working copy)
@@ -95,11 +95,17 @@ utf16le_is_mbc_newline(const UChar* p, c
static OnigCodePoint
utf16le_mbc_to_code(const UChar* p, const UChar* end ARG_UNUSED,
+ int *precious_ret,
OnigEncoding enc ARG_UNUSED)
{
OnigCodePoint code;
UChar c0 = *p;
UChar c1 = *(p+1);
+ int ret;
+
+ ret = utf16le_mbc_enc_len(p, end, enc);
+ if (precious_ret)
+ *precious_ret = ret;
if (UTF16_IS_SURROGATE_FIRST(c1)) {
code = ((((c1 - 0xd8) << 2) + ((c0 & 0xc0) >> 6) + 1) << 16)
@@ -245,5 +251,6 @@ OnigEncodingDefine(utf_16le, UTF_16LE) =
onigenc_unicode_is_code_ctype,
onigenc_utf16_32_get_ctype_code_range,
utf16le_left_adjust_char_head,
+ utf16le_left_adjust_char_head,
onigenc_always_false_is_allowed_reverse_match
};
Index: enc/koi8_r.c
===================================================================
--- enc/koi8_r.c (revision 19335)
+++ enc/koi8_r.c (working copy)
@@ -213,6 +213,7 @@ OnigEncodingDefine(koi8_r, KOI8_R) = {
koi8_r_is_code_ctype,
onigenc_not_support_get_ctype_code_range,
onigenc_single_byte_left_adjust_char_head,
+ onigenc_single_byte_left_adjust_char_head,
onigenc_always_true_is_allowed_reverse_match
};
ENC_ALIAS("CP878", "KOI8-R")
Index: string.c
===================================================================
--- string.c (revision 19335)
+++ string.c (working copy)
@@ -805,24 +805,6 @@ rb_enc_strlen_cr(const char *p, const ch
return c;
}
-#ifdef NONASCII_MASK
-#define is_utf8_lead_byte(c) (((c)&0xC0) != 0x80)
-static inline VALUE
-count_utf8_lead_bytes_with_word(const VALUE *s)
-{
- VALUE d = *s;
- d |= ~(d>>1);
- d >>= 6;
- d &= NONASCII_MASK >> 7;
- d += (d>>8);
- d += (d>>16);
-#if SIZEOF_VALUE == 8
- d += (d>>32);
-#endif
- return (d&0xF);
-}
-#endif
-
static long
str_strlen(VALUE str, rb_encoding *enc)
{
@@ -833,32 +815,6 @@ str_strlen(VALUE str, rb_encoding *enc)
if (!enc) enc = STR_ENC_GET(str);
p = RSTRING_PTR(str);
e = RSTRING_END(str);
-#ifdef NONASCII_MASK
- if (ENC_CODERANGE(str) == ENC_CODERANGE_VALID &&
- enc == rb_utf8_encoding()) {
- VALUE len = 0;
- if (sizeof(VALUE) * 2 < e - p) {
- const VALUE *s, *t;
- const VALUE lowbits = sizeof(VALUE) - 1;
- s = (const VALUE*)(~lowbits & ((VALUE)p + lowbits));
- t = (const VALUE*)(~lowbits & (VALUE)e);
- while (p < (const char *)s) {
- if (is_utf8_lead_byte(*p)) len++;
- p++;
- }
- while (s < t) {
- len += count_utf8_lead_bytes_with_word(s);
- s++;
- }
- p = (const char *)s;
- }
- while (p < e) {
- if (is_utf8_lead_byte(*p)) len++;
- p++;
- }
- return (long)len;
- }
-#endif
n = rb_enc_strlen_cr(p, e, enc, &cr);
if (cr) {
ENC_CODERANGE_SET(str, cr);
@@ -1214,44 +1170,6 @@ str_offset(const char *p, const char *e,
return pp - p;
}
-#ifdef NONASCII_MASK
-static char *
-str_utf8_nth(const char *p, const char *e, int nth)
-{
- if (sizeof(VALUE) * 2 < nth) {
- const VALUE *s, *t;
- const VALUE lowbits = sizeof(VALUE) - 1;
- s = (const VALUE*)(~lowbits & ((VALUE)p + lowbits));
- t = (const VALUE*)(~lowbits & (VALUE)e);
- while (p < (const char *)s) {
- if (is_utf8_lead_byte(*p)) nth--;
- p++;
- }
- do {
- nth -= count_utf8_lead_bytes_with_word(s);
- s++;
- } while (s < t && sizeof(VALUE) <= nth);
- p = (char *)s;
- }
- while (p < e) {
- if (is_utf8_lead_byte(*p)) {
- if (nth == 0) break;
- nth--;
- }
- p++;
- }
- return (char *)p;
-}
-
-static int
-str_utf8_offset(const char *p, const char *e, int nth)
-{
- const char *pp = str_utf8_nth(p, e, nth);
- if (!pp) return e - p;
- return pp - p;
-}
-#endif
-
/* byte offset to char offset */
long
rb_str_sublen(VALUE str, long pos)
@@ -1311,13 +1229,6 @@ rb_str_substr(VALUE str, long beg, long
if (len == 0) {
p = 0;
}
-#ifdef NONASCII_MASK
- else if (ENC_CODERANGE(str) == ENC_CODERANGE_VALID &&
- enc == rb_utf8_encoding()) {
- p = str_utf8_nth(s, e, beg);
- len = str_utf8_offset(p, e, len);
- }
-#endif
else if ((p = str_nth(s, e, beg, enc, singlebyte)) == e) {
len = 0;
}
@@ -2406,7 +2317,7 @@ enc_succ_char(char *p, int len, rb_encod
if (i < 0)
return NEIGHBOR_WRAPPED;
++((unsigned char*)p)[i];
- l = rb_enc_precise_mbclen(p, p+len, enc);
+ rb_enc_mbc_precious_codepoint(p, p+len, &l, enc);
if (MBCLEN_CHARFOUND_P(l)) {
l = MBCLEN_CHARFOUND_LEN(l);
if (l == len) {
@@ -2419,7 +2330,7 @@ enc_succ_char(char *p, int len, rb_encod
if (MBCLEN_INVALID_P(l) && i < len-1) {
int len2, l2;
for (len2 = len-1; 0 < len2; len2--) {
- l2 = rb_enc_precise_mbclen(p, p+len2, enc);
+ rb_enc_mbc_precious_codepoint(p, p+len2, &l2, enc);
if (!MBCLEN_INVALID_P(l2))
break;
}
@@ -2438,7 +2349,7 @@ enc_pred_char(char *p, int len, rb_encod
if (i < 0)
return NEIGHBOR_WRAPPED;
--((unsigned char*)p)[i];
- l = rb_enc_precise_mbclen(p, p+len, enc);
+ rb_enc_mbc_precious_codepoint(p, p+len, &l, enc);
if (MBCLEN_CHARFOUND_P(l)) {
l = MBCLEN_CHARFOUND_LEN(l);
if (l == len) {
@@ -2451,7 +2362,7 @@ enc_pred_char(char *p, int len, rb_encod
if (MBCLEN_INVALID_P(l) && i < len-1) {
int len2, l2;
for (len2 = len-1; 0 < len2; len2--) {
- l2 = rb_enc_precise_mbclen(p, p+len2, enc);
+ rb_enc_mbc_precious_codepoint(p, p+len2, &l2, enc);
if (!MBCLEN_INVALID_P(l2))
break;
}
@@ -2558,7 +2469,7 @@ rb_str_succ(VALUE orig)
VALUE str;
char *sbeg, *s, *e, *last_alnum = 0;
int c = -1;
- long l;
+ int l;
char carry[ONIGENC_CODE_TO_MBC_MAXLEN] = "\1";
int carry_pos = 0, carry_len = 1;
enum neighbor_char neighbor = NEIGHBOR_FOUND;
@@ -2580,7 +2491,8 @@ rb_str_succ(VALUE orig)
break;
}
}
- if ((l = rb_enc_precise_mbclen(s, e, enc)) <= 0) continue;
+ rb_enc_mbc_precious_codepoint(s, e, &l, enc);
+ if (l <= 0) continue;
neighbor = enc_succ_alnum_char(s, l, enc, carry);
switch (neighbor) {
case NEIGHBOR_NOT_CHAR:
@@ -2599,11 +2511,14 @@ rb_str_succ(VALUE orig)
s = e;
while ((s = rb_enc_prev_char(sbeg, s, e, enc)) != 0) {
enum neighbor_char neighbor;
- if ((l = rb_enc_precise_mbclen(s, e, enc)) <= 0) continue;
+ int l2;
+ rb_enc_mbc_precious_codepoint(s, e, &l, enc);
+ if (l <= 0) continue;
neighbor = enc_succ_char(s, l, enc);
if (neighbor == NEIGHBOR_FOUND)
return str;
- if (rb_enc_precise_mbclen(s, s+l, enc) != l) {
+ rb_enc_mbc_precious_codepoint(s, s+l, &l2, enc);
+ if (l2 != l) {
/* wrapped to \0...\0. search next valid char. */
enc_succ_char(s, l, enc);
}
@@ -3790,7 +3705,7 @@ rb_str_to_s(VALUE str)
static void
str_cat_char(VALUE str, int c, rb_encoding *enc)
{
- char s[16];
+ char s[ONIGENC_CODE_TO_MBC_MAXLEN];
int n = rb_enc_codelen(c, enc);
rb_enc_mbcput(c, s, enc);
Index: test/ruby/test_m17n.rb
===================================================================
--- test/ruby/test_m17n.rb (revision 19335)
+++ test/ruby/test_m17n.rb (working copy)
@@ -1287,4 +1287,11 @@ class TestM17N < Test::Unit::TestCase
assert(("".center(1, "\x80".force_encoding("utf-8")); true),
"moved from btest/knownbug, [ruby-dev:33807]")
end
+
+ def test_combchar
+ assert_equal(1, "\u{304B 3099}".length)
+ assert_equal("\u{304B 3099}", "\u{304B 3099}"[0])
+ assert_equal(nil, "\u{304B 3099}"[1])
+ assert_equal(nil, "\u{304B 3099}".index("\u3099"))
+ end
end
--
[田中 哲][たなか あきら][Tanaka Akira]