icuSources/common/utf_impl.c

   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 /* set import/export definitions */
  21 #ifndef U_UTF8_IMPL
  22 #   define U_UTF8_IMPL
  23 #endif
  24
  25 #include "unicode/utypes.h"
  26
  27 /*
  28  * This table could be replaced on many machines by
  29  * a few lines of assembler code using an
  30  * "index of first 0-bit from msb" instruction and
  31  * one or two more integer instructions.
  32  *
  33  * For example, on an i386, do something like
  34  * - MOV AL, leadByte
  35  * - NOT AL         (8-bit, leave b15..b8==0..0, reverse only b7..b0)
  36  * - MOV AH, 0
  37  * - BSR BX, AX     (16-bit)
  38  * - MOV AX, 6      (result)
  39  * - JZ finish      (ZF==1 if leadByte==0xff)
  40  * - SUB AX, BX (result)
  41  * -finish:
  42  * (BSR: Bit Scan Reverse, scans for a 1-bit, starting from the MSB)
  43  *
  44  * In Unicode, all UTF-8 byte sequences with more than 4 bytes are illegal;
  45  * lead bytes above 0xf4 are illegal.
  46  * We keep them in this table for skipping long ISO 10646-UTF-8 sequences.
  47  */
  48 U_EXPORT const uint8_t
  49 utf8_countTrailBytes[256]={
  50     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  51     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  52     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  53     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  54
  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     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  59
  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     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  64
  65     1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  66     1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  67
  68     2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  69     3, 3, 3, 3, 3,
  70     3, 3, 3,    /* illegal in Unicode */
  71     4, 4, 4, 4, /* illegal in Unicode */
  72     5, 5,       /* illegal in Unicode */
  73     0, 0        /* illegal bytes 0xfe and 0xff */
  74 };
  75
  76 static const UChar32
  77 utf8_minLegal[4]={ 0, 0x80, 0x800, 0x10000 };
  78
  79 static const UChar32
  80 utf8_errorValue[6]={
  81     UTF8_ERROR_VALUE_1, UTF8_ERROR_VALUE_2, UTF_ERROR_VALUE, 0x10ffff,
  82     0x3ffffff, 0x7fffffff
  83 };
  84
  85 /*
  86  * Handle the non-inline part of the U8_NEXT() macro and its obsolete sibling
  87  * UTF8_NEXT_CHAR_SAFE().
  88  *
  89  * The "strict" parameter controls the error behavior:
  90  * <0  "Safe" behavior of U8_NEXT(): All illegal byte sequences yield a negative
  91  *     code point result.
  92  *  0  Obsolete "safe" behavior of UTF8_NEXT_CHAR_SAFE(..., FALSE):
  93  *     All illegal byte sequences yield a positive code point such that this
  94  *     result code point would be encoded with the same number of bytes as
  95  *     the illegal sequence.
  96  * >0  Obsolete "strict" behavior of UTF8_NEXT_CHAR_SAFE(..., TRUE):
  97  *     Same as the obsolete "safe" behavior, but non-characters are also treated
  98  *     like illegal sequences.
  99  *
 100  * The special negative (<0) value -2 is used for lenient treatment of surrogate
 101  * code points as legal. Some implementations use this for roundtripping of
 102  * Unicode 16-bit strings that are not well-formed UTF-16, that is, they
 103  * contain unpaired surrogates.
 104  *
 105  * Note that a UBool is the same as an int8_t.
 106  */
 107 U_CAPI UChar32 U_EXPORT2
 108 utf8_nextCharSafeBody(const uint8_t *s, int32_t *pi, int32_t length, UChar32 c, UBool strict) {
 109     int32_t i=*pi;
 110     uint8_t count=UTF8_COUNT_TRAIL_BYTES(c);
 111     if((i)+count<=(length)) {
 112         uint8_t trail, illegal=0;
 113
 114         UTF8_MASK_LEAD_BYTE((c), count);
 115         /* count==0 for illegally leading trail bytes and the illegal bytes 0xfe and 0xff */
 116         switch(count) {
 117         /* each branch falls through to the next one */
 118         case 5:
 119         case 4:
 120             /* count>=4 is always illegal: no more than 3 trail bytes in Unicode's UTF-8 */
 121             illegal=1;
 122             break;
 123         case 3:
 124             trail=s[(i)++];
 125             (c)=((c)<<6)|(trail&0x3f);
 126             if(c<0x110) {
 127                 illegal|=(trail&0xc0)^0x80;
 128             } else {
 129                 /* code point>0x10ffff, outside Unicode */
 130                 illegal=1;
 131                 break;
 132             }
 133         case 2:
 134             trail=s[(i)++];
 135             (c)=((c)<<6)|(trail&0x3f);
 136             illegal|=(trail&0xc0)^0x80;
 137         case 1:
 138             trail=s[(i)++];
 139             (c)=((c)<<6)|(trail&0x3f);
 140             illegal|=(trail&0xc0)^0x80;
 141             break;
 142         case 0:
 143             if(strict>=0) {
 144                 return UTF8_ERROR_VALUE_1;
 145             } else {
 146                 return U_SENTINEL;
 147             }
 148         /* no default branch to optimize switch()  - all values are covered */
 149         }
 150
 151         /*
 152          * All the error handling should return a value
 153          * that needs count bytes so that UTF8_GET_CHAR_SAFE() works right.
 154          *
 155          * Starting with Unicode 3.0.1, non-shortest forms are illegal.
 156          * Starting with Unicode 3.2, surrogate code points must not be
 157          * encoded in UTF-8, and there are no irregular sequences any more.
 158          *
 159          * U8_ macros (new in ICU 2.4) return negative values for error conditions.
 160          */
 161
 162         /* correct sequence - all trail bytes have (b7..b6)==(10)? */
 163         /* illegal is also set if count>=4 */
 164         if(illegal || (c)<utf8_minLegal[count] || (UTF_IS_SURROGATE(c) && strict!=-2)) {
 165             /* error handling */
 166             uint8_t errorCount=count;
 167             /* don't go beyond this sequence */
 168             i=*pi;
 169             while(count>0 && UTF8_IS_TRAIL(s[i])) {
 170                 ++(i);
 171                 --count;
 172             }
 173             if(strict>=0) {
 174                 c=utf8_errorValue[errorCount-count];
 175             } else {
 176                 c=U_SENTINEL;
 177             }
 178         } else if((strict)>0 && UTF_IS_UNICODE_NONCHAR(c)) {
 179             /* strict: forbid non-characters like U+fffe */
 180             c=utf8_errorValue[count];
 181         }
 182     } else /* too few bytes left */ {
 183         /* error handling */
 184         int32_t i0=i;
 185         /* don't just set (i)=(length) in case there is an illegal sequence */
 186         while((i)<(length) && UTF8_IS_TRAIL(s[i])) {
 187             ++(i);
 188         }
 189         if(strict>=0) {
 190             c=utf8_errorValue[i-i0];
 191         } else {
 192             c=U_SENTINEL;
 193         }
 194     }
 195     *pi=i;
 196     return c;
 197 }
 198
 199 U_CAPI int32_t U_EXPORT2
 200 utf8_appendCharSafeBody(uint8_t *s, int32_t i, int32_t length, UChar32 c, UBool *pIsError) {
 201     if((uint32_t)(c)<=0x7ff) {
 202         if((i)+1<(length)) {
 203             (s)[(i)++]=(uint8_t)(((c)>>6)|0xc0);
 204             (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80);
 205             return i;
 206         }
 207     } else if((uint32_t)(c)<=0xffff) {
 208         /* Starting with Unicode 3.2, surrogate code points must not be encoded in UTF-8. */
 209         if((i)+2<(length) && !U_IS_SURROGATE(c)) {
 210             (s)[(i)++]=(uint8_t)(((c)>>12)|0xe0);
 211             (s)[(i)++]=(uint8_t)((((c)>>6)&0x3f)|0x80);
 212             (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80);
 213             return i;
 214         }
 215     } else if((uint32_t)(c)<=0x10ffff) {
 216         if((i)+3<(length)) {
 217             (s)[(i)++]=(uint8_t)(((c)>>18)|0xf0);
 218             (s)[(i)++]=(uint8_t)((((c)>>12)&0x3f)|0x80);
 219             (s)[(i)++]=(uint8_t)((((c)>>6)&0x3f)|0x80);
 220             (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80);
 221             return i;
 222         }
 223     }
 224     /* c>0x10ffff or not enough space, write an error value */
 225     if(pIsError!=NULL) {
 226         *pIsError=TRUE;
 227     } else {
 228         length-=i;
 229         if(length>0) {
 230             int32_t offset;
 231             if(length>3) {
 232                 length=3;
 233             }
 234             s+=i;
 235             offset=0;
 236             c=utf8_errorValue[length-1];
 237             UTF8_APPEND_CHAR_UNSAFE(s, offset, c);
 238             i=i+offset;
 239         }
 240     }
 241     return i;
 242 }
 243
 244 U_CAPI UChar32 U_EXPORT2
 245 utf8_prevCharSafeBody(const uint8_t *s, int32_t start, int32_t *pi, UChar32 c, UBool strict) {
 246     int32_t i=*pi;
 247     uint8_t b, count=1, shift=6;
 248
 249     /* extract value bits from the last trail byte */
 250     c&=0x3f;
 251
 252     for(;;) {
 253         if(i<=start) {
 254             /* no lead byte at all */
 255             if(strict>=0) {
 256                 return UTF8_ERROR_VALUE_1;
 257             } else {
 258                 return U_SENTINEL;
 259             }
 260             /*break;*/
 261         }
 262
 263         /* read another previous byte */
 264         b=s[--i];
 265         if((uint8_t)(b-0x80)<0x7e) { /* 0x80<=b<0xfe */
 266             if(b&0x40) {
 267                 /* lead byte, this will always end the loop */
 268                 uint8_t shouldCount=UTF8_COUNT_TRAIL_BYTES(b);
 269
 270                 if(count==shouldCount) {
 271                     /* set the new position */
 272                     *pi=i;
 273                     UTF8_MASK_LEAD_BYTE(b, count);
 274                     c|=(UChar32)b<<shift;
 275                     if(count>=4 || c>0x10ffff || c<utf8_minLegal[count] || (UTF_IS_SURROGATE(c) && strict!=-2) || (strict>0 && UTF_IS_UNICODE_NONCHAR(c))) {
 276                         /* illegal sequence or (strict and non-character) */
 277                         if(count>=4) {
 278                             count=3;
 279                         }
 280                         if(strict>=0) {
 281                             c=utf8_errorValue[count];
 282                         } else {
 283                             c=U_SENTINEL;
 284                         }
 285                     } else {
 286                         /* exit with correct c */
 287                     }
 288                 } else {
 289                     /* the lead byte does not match the number of trail bytes */
 290                     /* only set the position to the lead byte if it would
 291                        include the trail byte that we started with */
 292                     if(count<shouldCount) {
 293                         *pi=i;
 294                         if(strict>=0) {
 295                             c=utf8_errorValue[count];
 296                         } else {
 297                             c=U_SENTINEL;
 298                         }
 299                     } else {
 300                         if(strict>=0) {
 301                             c=UTF8_ERROR_VALUE_1;
 302                         } else {
 303                             c=U_SENTINEL;
 304                         }
 305                     }
 306                 }
 307                 break;
 308             } else if(count<5) {
 309                 /* trail byte */
 310                 c|=(UChar32)(b&0x3f)<<shift;
 311                 ++count;
 312                 shift+=6;
 313             } else {
 314                 /* more than 5 trail bytes is illegal */
 315                 if(strict>=0) {
 316                     c=UTF8_ERROR_VALUE_1;
 317                 } else {
 318                     c=U_SENTINEL;
 319                 }
 320                 break;
 321             }
 322         } else {
 323             /* single-byte character precedes trailing bytes */
 324             if(strict>=0) {
 325                 c=UTF8_ERROR_VALUE_1;
 326             } else {
 327                 c=U_SENTINEL;
 328             }
 329             break;
 330         }
 331     }
 332     return c;
 333 }
 334
 335 U_CAPI int32_t U_EXPORT2
 336 utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i) {
 337     /* i had been decremented once before the function call */
 338     int32_t I=i, Z;
 339     uint8_t b;
 340
 341     /* read at most the 6 bytes s[Z] to s[i], inclusively */
 342     if(I-5>start) {
 343         Z=I-5;
 344     } else {
 345         Z=start;
 346     }
 347
 348     /* return I if the sequence starting there is long enough to include i */
 349     do {
 350         b=s[I];
 351         if((uint8_t)(b-0x80)>=0x7e) { /* not 0x80<=b<0xfe */
 352             break;
 353         } else if(b>=0xc0) {
 354             if(UTF8_COUNT_TRAIL_BYTES(b)>=(i-I)) {
 355                 return I;
 356             } else {
 357                 break;
 358             }
 359         }
 360     } while(Z<=--I);
 361
 362     /* return i itself to be consistent with the FWD_1 macro */
 363     return i;
 364 }