icuSources/common/utf_impl.c

   1 /*
   2 ******************************************************************************
   3 *
   4 *   Copyright (C) 1999-2012, 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 /* set import/export definitions */
  21 #ifndef U_UTF8_IMPL
  22 #   define U_UTF8_IMPL
  23 #endif
  24
  25 #include "unicode/utypes.h"
  26 #include "unicode/utf.h"
  27 #include "unicode/utf8.h"
  28 #include "unicode/utf_old.h"
  29 #include "uassert.h"
  30
  31 /*
  32  * This table could be replaced on many machines by
  33  * a few lines of assembler code using an
  34  * "index of first 0-bit from msb" instruction and
  35  * one or two more integer instructions.
  36  *
  37  * For example, on an i386, do something like
  38  * - MOV AL, leadByte
  39  * - NOT AL         (8-bit, leave b15..b8==0..0, reverse only b7..b0)
  40  * - MOV AH, 0
  41  * - BSR BX, AX     (16-bit)
  42  * - MOV AX, 6      (result)
  43  * - JZ finish      (ZF==1 if leadByte==0xff)
  44  * - SUB AX, BX (result)
  45  * -finish:
  46  * (BSR: Bit Scan Reverse, scans for a 1-bit, starting from the MSB)
  47  *
  48  * In Unicode, all UTF-8 byte sequences with more than 4 bytes are illegal;
  49  * lead bytes above 0xf4 are illegal.
  50  * We keep them in this table for skipping long ISO 10646-UTF-8 sequences.
  51  */
  52 U_EXPORT const uint8_t
  53 utf8_countTrailBytes[256]={
  54     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  55     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  56     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  57     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  58
  59     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  60     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  61     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  62     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  63
  64     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  65     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  66     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  67     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  68
  69     1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  70     1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  71
  72     2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  73     3, 3, 3, 3, 3,
  74     3, 3, 3,    /* illegal in Unicode */
  75     4, 4, 4, 4, /* illegal in Unicode */
  76     5, 5,       /* illegal in Unicode */
  77     0, 0        /* illegal bytes 0xfe and 0xff */
  78 };
  79
  80 static const UChar32
  81 utf8_minLegal[4]={ 0, 0x80, 0x800, 0x10000 };
  82
  83 static const UChar32
  84 utf8_errorValue[6]={
  85     UTF8_ERROR_VALUE_1, UTF8_ERROR_VALUE_2, UTF_ERROR_VALUE, 0x10ffff,
  86     0x3ffffff, 0x7fffffff
  87 };
  88
  89 static UChar32
  90 errorValue(int32_t count, int8_t strict) {
  91     if(strict>=0) {
  92         return utf8_errorValue[count];
  93     } else if(strict==-3) {
  94         return 0xfffd;
  95     } else {
  96         return U_SENTINEL;
  97     }
  98 }
  99
 100 /*
 101  * Handle the non-inline part of the U8_NEXT() and U8_NEXT_FFFD() macros
 102  * and their obsolete sibling UTF8_NEXT_CHAR_SAFE().
 103  *
 104  * U8_NEXT() supports NUL-terminated strings indicated via length<0.
 105  *
 106  * The "strict" parameter controls the error behavior:
 107  * <0  "Safe" behavior of U8_NEXT():
 108  *     -1: All illegal byte sequences yield U_SENTINEL=-1.
 109  *     -2: Same as -1, except for lenient treatment of surrogate code points as legal.
 110  *         Some implementations use this for roundtripping of
 111  *         Unicode 16-bit strings that are not well-formed UTF-16, that is, they
 112  *         contain unpaired surrogates.
 113  *     -3: All illegal byte sequences yield U+FFFD.
 114  *  0  Obsolete "safe" behavior of UTF8_NEXT_CHAR_SAFE(..., FALSE):
 115  *     All illegal byte sequences yield a positive code point such that this
 116  *     result code point would be encoded with the same number of bytes as
 117  *     the illegal sequence.
 118  * >0  Obsolete "strict" behavior of UTF8_NEXT_CHAR_SAFE(..., TRUE):
 119  *     Same as the obsolete "safe" behavior, but non-characters are also treated
 120  *     like illegal sequences.
 121  *
 122  * Note that a UBool is the same as an int8_t.
 123  */
 124 U_CAPI UChar32 U_EXPORT2
 125 utf8_nextCharSafeBody(const uint8_t *s, int32_t *pi, int32_t length, UChar32 c, UBool strict) {
 126     int32_t i=*pi;
 127     uint8_t count=U8_COUNT_TRAIL_BYTES(c);
 128     U_ASSERT(count <= 5); /* U8_COUNT_TRAIL_BYTES returns value 0...5 */
 129     if(i+count<=length || length<0) {
 130         uint8_t trail;
 131
 132         U8_MASK_LEAD_BYTE(c, count);
 133         /* support NUL-terminated strings: do not read beyond the first non-trail byte */
 134         switch(count) {
 135         /* each branch falls through to the next one */
 136         case 0:
 137             /* count==0 for illegally leading trail bytes and the illegal bytes 0xfe and 0xff */
 138         case 5:
 139         case 4:
 140             /* count>=4 is always illegal: no more than 3 trail bytes in Unicode's UTF-8 */
 141             break;
 142         case 3:
 143             trail=s[i++]-0x80;
 144             c=(c<<6)|trail;
 145             /* c>=0x110 would result in code point>0x10ffff, outside Unicode */
 146             if(c>=0x110 || trail>0x3f) { break; }
 147         case 2:
 148             trail=s[i++]-0x80;
 149             c=(c<<6)|trail;
 150             /*
 151              * test for a surrogate d800..dfff unless we are lenient:
 152              * before the last (c<<6), a surrogate is c=360..37f
 153              */
 154             if(((c&0xffe0)==0x360 && strict!=-2) || trail>0x3f) { break; }
 155         case 1:
 156             trail=s[i++]-0x80;
 157             c=(c<<6)|trail;
 158             if(trail>0x3f) { break; }
 159             /* correct sequence - all trail bytes have (b7..b6)==(10) */
 160             if(c>=utf8_minLegal[count] &&
 161                     /* strict: forbid non-characters like U+fffe */
 162                     (strict<=0 || !U_IS_UNICODE_NONCHAR(c))) {
 163                 *pi=i;
 164                 return c;
 165             }
 166         /* no default branch to optimize switch()  - all values are covered */
 167         }
 168     } else {
 169         /* too few bytes left */
 170         count=length-i;
 171     }
 172
 173     /* error handling */
 174     i=*pi;
 175     while(count>0 && U8_IS_TRAIL(s[i])) {
 176         ++i;
 177         --count;
 178     }
 179     c=errorValue(i-*pi, strict);
 180     *pi=i;
 181     return c;
 182 }
 183
 184 U_CAPI int32_t U_EXPORT2
 185 utf8_appendCharSafeBody(uint8_t *s, int32_t i, int32_t length, UChar32 c, UBool *pIsError) {
 186     if((uint32_t)(c)<=0x7ff) {
 187         if((i)+1<(length)) {
 188             (s)[(i)++]=(uint8_t)(((c)>>6)|0xc0);
 189             (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80);
 190             return i;
 191         }
 192     } else if((uint32_t)(c)<=0xffff) {
 193         /* Starting with Unicode 3.2, surrogate code points must not be encoded in UTF-8. */
 194         if((i)+2<(length) && !U_IS_SURROGATE(c)) {
 195             (s)[(i)++]=(uint8_t)(((c)>>12)|0xe0);
 196             (s)[(i)++]=(uint8_t)((((c)>>6)&0x3f)|0x80);
 197             (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80);
 198             return i;
 199         }
 200     } else if((uint32_t)(c)<=0x10ffff) {
 201         if((i)+3<(length)) {
 202             (s)[(i)++]=(uint8_t)(((c)>>18)|0xf0);
 203             (s)[(i)++]=(uint8_t)((((c)>>12)&0x3f)|0x80);
 204             (s)[(i)++]=(uint8_t)((((c)>>6)&0x3f)|0x80);
 205             (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80);
 206             return i;
 207         }
 208     }
 209     /* c>0x10ffff or not enough space, write an error value */
 210     if(pIsError!=NULL) {
 211         *pIsError=TRUE;
 212     } else {
 213         length-=i;
 214         if(length>0) {
 215             int32_t offset;
 216             if(length>3) {
 217                 length=3;
 218             }
 219             s+=i;
 220             offset=0;
 221             c=utf8_errorValue[length-1];
 222             UTF8_APPEND_CHAR_UNSAFE(s, offset, c);
 223             i=i+offset;
 224         }
 225     }
 226     return i;
 227 }
 228
 229 U_CAPI UChar32 U_EXPORT2
 230 utf8_prevCharSafeBody(const uint8_t *s, int32_t start, int32_t *pi, UChar32 c, UBool strict) {
 231     int32_t i=*pi;
 232     uint8_t b, count=1, shift=6;
 233
 234     if(!U8_IS_TRAIL(c)) { return errorValue(0, strict); }
 235
 236     /* extract value bits from the last trail byte */
 237     c&=0x3f;
 238
 239     for(;;) {
 240         if(i<=start) {
 241             /* no lead byte at all */
 242             return errorValue(0, strict);
 243         }
 244
 245         /* read another previous byte */
 246         b=s[--i];
 247         if((uint8_t)(b-0x80)<0x7e) { /* 0x80<=b<0xfe */
 248             if(b&0x40) {
 249                 /* lead byte, this will always end the loop */
 250                 uint8_t shouldCount=U8_COUNT_TRAIL_BYTES(b);
 251
 252                 if(count==shouldCount) {
 253                     /* set the new position */
 254                     *pi=i;
 255                     U8_MASK_LEAD_BYTE(b, count);
 256                     c|=(UChar32)b<<shift;
 257                     if(count>=4 || c>0x10ffff || c<utf8_minLegal[count] || (U_IS_SURROGATE(c) && strict!=-2) || (strict>0 && U_IS_UNICODE_NONCHAR(c))) {
 258                         /* illegal sequence or (strict and non-character) */
 259                         if(count>=4) {
 260                             count=3;
 261                         }
 262                         c=errorValue(count, strict);
 263                     } else {
 264                         /* exit with correct c */
 265                     }
 266                 } else {
 267                     /* the lead byte does not match the number of trail bytes */
 268                     /* only set the position to the lead byte if it would
 269                        include the trail byte that we started with */
 270                     if(count<shouldCount) {
 271                         *pi=i;
 272                         c=errorValue(count, strict);
 273                     } else {
 274                         c=errorValue(0, strict);
 275                     }
 276                 }
 277                 break;
 278             } else if(count<5) {
 279                 /* trail byte */
 280                 c|=(UChar32)(b&0x3f)<<shift;
 281                 ++count;
 282                 shift+=6;
 283             } else {
 284                 /* more than 5 trail bytes is illegal */
 285                 c=errorValue(0, strict);
 286                 break;
 287             }
 288         } else {
 289             /* single-byte character precedes trailing bytes */
 290             c=errorValue(0, strict);
 291             break;
 292         }
 293     }
 294     return c;
 295 }
 296
 297 U_CAPI int32_t U_EXPORT2
 298 utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i) {
 299     /* i had been decremented once before the function call */
 300     int32_t I=i, Z;
 301     uint8_t b;
 302
 303     /* read at most the 6 bytes s[Z] to s[i], inclusively */
 304     if(I-5>start) {
 305         Z=I-5;
 306     } else {
 307         Z=start;
 308     }
 309
 310     /* return I if the sequence starting there is long enough to include i */
 311     do {
 312         b=s[I];
 313         if((uint8_t)(b-0x80)>=0x7e) { /* not 0x80<=b<0xfe */
 314             break;
 315         } else if(b>=0xc0) {
 316             if(U8_COUNT_TRAIL_BYTES(b)>=(i-I)) {
 317                 return I;
 318             } else {
 319                 break;
 320             }
 321         }
 322     } while(Z<=--I);
 323
 324     /* return i itself to be consistent with the FWD_1 macro */
 325     return i;
 326 }