1 /*
2 ******************************************************************************
3 *
4 * Copyright (C) 1999-2006, International Business Machines
5 * Corporation and others. All Rights Reserved.
6 *
7 ******************************************************************************
8 * file name: utf_impl.c
9 * encoding: US-ASCII
10 * tab size: 8 (not used)
11 * indentation:4
12 *
13 * created on: 1999sep13
14 * created by: Markus W. Scherer
15 *
16 * This file provides implementation functions for macros in the utfXX.h
17 * that would otherwise be too long as macros.
18 */
19
20 #include "base/third_party/icu/icu_utf.h"
21
22 namespace base_icu {
23
24 /**
25 * UTF8_ERROR_VALUE_1 and UTF8_ERROR_VALUE_2 are special error values for UTF-8,
26 * which need 1 or 2 bytes in UTF-8:
27 * \code
28 * U+0015 = NAK = Negative Acknowledge, C0 control character
29 * U+009f = highest C1 control character
30 * \endcode
31 *
32 * These are used by UTF8_..._SAFE macros so that they can return an error value
33 * that needs the same number of code units (bytes) as were seen by
34 * a macro. They should be tested with UTF_IS_ERROR() or UTF_IS_VALID().
35 *
36 * @deprecated ICU 2.4. Obsolete, see utf_old.h.
37 */
38 #define CBUTF8_ERROR_VALUE_1 0x15
39
40 /**
41 * See documentation on UTF8_ERROR_VALUE_1 for details.
42 *
43 * @deprecated ICU 2.4. Obsolete, see utf_old.h.
44 */
45 #define CBUTF8_ERROR_VALUE_2 0x9f
46
47
48 /**
49 * Error value for all UTFs. This code point value will be set by macros with e>
50 * checking if an error is detected.
51 *
52 * @deprecated ICU 2.4. Obsolete, see utf_old.h.
53 */
54 #define CBUTF_ERROR_VALUE 0xffff
55
56 /*
57 * This table could be replaced on many machines by
58 * a few lines of assembler code using an
59 * "index of first 0-bit from msb" instruction and
60 * one or two more integer instructions.
61 *
62 * For example, on an i386, do something like
63 * - MOV AL, leadByte
64 * - NOT AL (8-bit, leave b15..b8==0..0, reverse only b7..b0)
65 * - MOV AH, 0
66 * - BSR BX, AX (16-bit)
67 * - MOV AX, 6 (result)
68 * - JZ finish (ZF==1 if leadByte==0xff)
69 * - SUB AX, BX (result)
70 * -finish:
71 * (BSR: Bit Scan Reverse, scans for a 1-bit, starting from the MSB)
72 *
73 * In Unicode, all UTF-8 byte sequences with more than 4 bytes are illegal;
74 * lead bytes above 0xf4 are illegal.
75 * We keep them in this table for skipping long ISO 10646-UTF-8 sequences.
76 */
77 const uint8_t utf8_countTrailBytes[256] =
78 {
79 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
80 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
81 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
82
83 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
84 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
85 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
86
87 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
88 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
89 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
90
91 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
92 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
93
94 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3,
95 3, 3, /* illegal in Unicode */
96 4, 4, 4, 4, /* illegal in Unicode */
97 5, 5, /* illegal in Unicode */
98 0, 0 /* illegal bytes 0xfe and 0xff */
99 };
100
101 static const UChar32
102 utf8_minLegal[4]={ 0, 0x80, 0x800, 0x10000 };
103
104 static const UChar32
105 utf8_errorValue[6]={
106 CBUTF8_ERROR_VALUE_1, CBUTF8_ERROR_VALUE_2, CBUTF_ERROR_VALUE, 0x10ffff,
107 0x3ffffff, 0x7fffffff
108 };
109
110 /*
111 * Handle the non-inline part of the U8_NEXT() macro and its obsolete sibling
112 * UTF8_NEXT_CHAR_SAFE().
113 *
114 * The "strict" parameter controls the error behavior:
115 * <0 "Safe" behavior of U8_NEXT(): All illegal byte sequences yield a negative
116 * code point result.
117 * 0 Obsolete "safe" behavior of UTF8_NEXT_CHAR_SAFE(..., FALSE):
118 * All illegal byte sequences yield a positive code point such that this
119 * result code point would be encoded with the same number of bytes as
120 * the illegal sequence.
121 * >0 Obsolete "strict" behavior of UTF8_NEXT_CHAR_SAFE(..., TRUE):
122 * Same as the obsolete "safe" behavior, but non-characters are also treated
123 * like illegal sequences.
124 *
125 * The special negative (<0) value -2 is used for lenient treatment of surrogate
126 * code points as legal. Some implementations use this for roundtripping of
127 * Unicode 16-bit strings that are not well-formed UTF-16, that is, they
128 * contain unpaired surrogates.
129 *
130 * Note that a UBool is the same as an int8_t.
131 */
utf8_nextCharSafeBody(const uint8_t * s,int32_t * pi,int32_t length,UChar32 c,UBool strict)132 UChar32 utf8_nextCharSafeBody(const uint8_t* s,
133 int32_t* pi,
134 int32_t length,
135 UChar32 c,
136 UBool strict) {
137 int32_t i = *pi;
138 uint8_t count = CBU8_COUNT_TRAIL_BYTES(c);
139 if((i)+count<=(length)) {
140 uint8_t trail, illegal = 0;
141
142 CBU8_MASK_LEAD_BYTE((c), count);
143 /* count==0 for illegally leading trail bytes and the illegal bytes 0xfe and 0xff */
144 switch(count) {
145 /* each branch falls through to the next one */
146 case 5:
147 case 4:
148 /* count>=4 is always illegal: no more than 3 trail bytes in Unicode's UTF-8 */
149 illegal=1;
150 break;
151 case 3:
152 trail=s[(i)++];
153 (c)=((c)<<6)|(trail&0x3f);
154 if(c<0x110) {
155 illegal|=(trail&0xc0)^0x80;
156 } else {
157 /* code point>0x10ffff, outside Unicode */
158 illegal=1;
159 break;
160 }
161 case 2:
162 trail=s[(i)++];
163 (c)=((c)<<6)|(trail&0x3f);
164 illegal|=(trail&0xc0)^0x80;
165 case 1:
166 trail=s[(i)++];
167 (c)=((c)<<6)|(trail&0x3f);
168 illegal|=(trail&0xc0)^0x80;
169 break;
170 case 0:
171 if(strict>=0) {
172 return CBUTF8_ERROR_VALUE_1;
173 } else {
174 return CBU_SENTINEL;
175 }
176 /* no default branch to optimize switch() - all values are covered */
177 }
178
179 /*
180 * All the error handling should return a value
181 * that needs count bytes so that UTF8_GET_CHAR_SAFE() works right.
182 *
183 * Starting with Unicode 3.0.1, non-shortest forms are illegal.
184 * Starting with Unicode 3.2, surrogate code points must not be
185 * encoded in UTF-8, and there are no irregular sequences any more.
186 *
187 * U8_ macros (new in ICU 2.4) return negative values for error conditions.
188 */
189
190 /* correct sequence - all trail bytes have (b7..b6)==(10)? */
191 /* illegal is also set if count>=4 */
192 if(illegal || (c)<utf8_minLegal[count] || (CBU_IS_SURROGATE(c) && strict!=-2)) {
193 /* error handling */
194 uint8_t errorCount = count;
195 /* don't go beyond this sequence */
196 i=*pi;
197 while(count>0 && CBU8_IS_TRAIL(s[i])) {
198 ++(i);
199 --count;
200 }
201 if(strict>=0) {
202 c=utf8_errorValue[errorCount-count];
203 } else {
204 c=CBU_SENTINEL;
205 }
206 } else if((strict)>0 && CBU_IS_UNICODE_NONCHAR(c)) {
207 /* strict: forbid non-characters like U+fffe */
208 c=utf8_errorValue[count];
209 }
210 } else /* too few bytes left */ {
211 /* error handling */
212 int32_t i0 = i;
213 /* don't just set (i)=(length) in case there is an illegal sequence */
214 while((i)<(length) && CBU8_IS_TRAIL(s[i])) {
215 ++(i);
216 }
217 if(strict>=0) {
218 c=utf8_errorValue[i-i0];
219 } else {
220 c=CBU_SENTINEL;
221 }
222 }
223 *pi=i;
224 return c;
225 }
226
227 } // namespace base_icu
228