icuSources/common/utf_impl.c

   1 /*
   2 ******************************************************************************
   3 *
   4 *   Copyright (C) 1999-2011, 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 /*
  90  * Handle the non-inline part of the U8_NEXT() macro and its obsolete sibling
  91  * UTF8_NEXT_CHAR_SAFE().
  92  *
  93  * The "strict" parameter controls the error behavior:
  94  * <0  "Safe" behavior of U8_NEXT(): All illegal byte sequences yield a negative
  95  *     code point result.
  96  *  0  Obsolete "safe" behavior of UTF8_NEXT_CHAR_SAFE(..., FALSE):
  97  *     All illegal byte sequences yield a positive code point such that this
  98  *     result code point would be encoded with the same number of bytes as
  99  *     the illegal sequence.
 100  * >0  Obsolete "strict" behavior of UTF8_NEXT_CHAR_SAFE(..., TRUE):
 101  *     Same as the obsolete "safe" behavior, but non-characters are also treated
 102  *     like illegal sequences.
 103  *
 104  * The special negative (<0) value -2 is used for lenient treatment of surrogate
 105  * code points as legal. Some implementations use this for roundtripping of
 106  * Unicode 16-bit strings that are not well-formed UTF-16, that is, they
 107  * contain unpaired surrogates.
 108  *
 109  * Note that a UBool is the same as an int8_t.
 110  */
 111 U_CAPI UChar32 U_EXPORT2
 112 utf8_nextCharSafeBody(const uint8_t *s, int32_t *pi, int32_t length, UChar32 c, UBool strict) {
 113     int32_t i=*pi;
 114     uint8_t count=U8_COUNT_TRAIL_BYTES(c);
 115     U_ASSERT(count >= 0 && count <= 5); /* U8_COUNT_TRAIL_BYTES returns value 0...5 */
 116     if((i)+count<=(length)) {
 117         uint8_t trail, illegal=0;
 118
 119         U8_MASK_LEAD_BYTE((c), count);
 120         /* count==0 for illegally leading trail bytes and the illegal bytes 0xfe and 0xff */
 121         switch(count) {
 122         /* each branch falls through to the next one */
 123         case 5:
 124         case 4:
 125             /* count>=4 is always illegal: no more than 3 trail bytes in Unicode's UTF-8 */
 126             illegal=1;
 127             break;
 128         case 3:
 129             trail=s[(i)++];
 130             (c)=((c)<<6)|(trail&0x3f);
 131             if(c<0x110) {
 132                 illegal|=(trail&0xc0)^0x80;
 133             } else {
 134                 /* code point>0x10ffff, outside Unicode */
 135                 illegal=1;
 136                 break;
 137             }
 138         case 2:
 139             trail=s[(i)++];
 140             (c)=((c)<<6)|(trail&0x3f);
 141             illegal|=(trail&0xc0)^0x80;
 142         case 1:
 143             trail=s[(i)++];
 144             (c)=((c)<<6)|(trail&0x3f);
 145             illegal|=(trail&0xc0)^0x80;
 146             break;
 147         case 0:
 148             if(strict>=0) {
 149                 return UTF8_ERROR_VALUE_1;
 150             } else {
 151                 return U_SENTINEL;
 152             }
 153         /* no default branch to optimize switch()  - all values are covered */
 154         }
 155
 156         /*
 157          * All the error handling should return a value
 158          * that needs count bytes so that UTF8_GET_CHAR_SAFE() works right.
 159          *
 160          * Starting with Unicode 3.0.1, non-shortest forms are illegal.
 161          * Starting with Unicode 3.2, surrogate code points must not be
 162          * encoded in UTF-8, and there are no irregular sequences any more.
 163          *
 164          * U8_ macros (new in ICU 2.4) return negative values for error conditions.
 165          */
 166
 167         /* correct sequence - all trail bytes have (b7..b6)==(10)? */
 168         /* illegal is also set if count>=4 */
 169         if(illegal || (c)<utf8_minLegal[count] || (U_IS_SURROGATE(c) && strict!=-2)) {
 170             /* error handling */
 171             uint8_t errorCount=count;
 172             /* don't go beyond this sequence */
 173             i=*pi;
 174             while(count>0 && U8_IS_TRAIL(s[i])) {
 175                 ++(i);
 176                 --count;
 177             }
 178             if(strict>=0) {
 179                 c=utf8_errorValue[errorCount-count];
 180             } else {
 181                 c=U_SENTINEL;
 182             }
 183         } else if((strict)>0 && U_IS_UNICODE_NONCHAR(c)) {
 184             /* strict: forbid non-characters like U+fffe */
 185             c=utf8_errorValue[count];
 186         }
 187     } else /* too few bytes left */ {
 188         /* error handling */
 189         int32_t i0=i;
 190         /* don't just set (i)=(length) in case there is an illegal sequence */
 191         while((i)<(length) && U8_IS_TRAIL(s[i])) {
 192             ++(i);
 193         }
 194         if(strict>=0) {
 195             c=utf8_errorValue[i-i0];
 196         } else {
 197             c=U_SENTINEL;
 198         }
 199     }
 200     *pi=i;
 201     return c;
 202 }
 203
 204 U_CAPI int32_t U_EXPORT2
 205 utf8_appendCharSafeBody(uint8_t *s, int32_t i, int32_t length, UChar32 c, UBool *pIsError) {
 206     if((uint32_t)(c)<=0x7ff) {
 207         if((i)+1<(length)) {
 208             (s)[(i)++]=(uint8_t)(((c)>>6)|0xc0);
 209             (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80);
 210             return i;
 211         }
 212     } else if((uint32_t)(c)<=0xffff) {
 213         /* Starting with Unicode 3.2, surrogate code points must not be encoded in UTF-8. */
 214         if((i)+2<(length) && !U_IS_SURROGATE(c)) {
 215             (s)[(i)++]=(uint8_t)(((c)>>12)|0xe0);
 216             (s)[(i)++]=(uint8_t)((((c)>>6)&0x3f)|0x80);
 217             (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80);
 218             return i;
 219         }
 220     } else if((uint32_t)(c)<=0x10ffff) {
 221         if((i)+3<(length)) {
 222             (s)[(i)++]=(uint8_t)(((c)>>18)|0xf0);
 223             (s)[(i)++]=(uint8_t)((((c)>>12)&0x3f)|0x80);
 224             (s)[(i)++]=(uint8_t)((((c)>>6)&0x3f)|0x80);
 225             (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80);
 226             return i;
 227         }
 228     }
 229     /* c>0x10ffff or not enough space, write an error value */
 230     if(pIsError!=NULL) {
 231         *pIsError=TRUE;
 232     } else {
 233         length-=i;
 234         if(length>0) {
 235             int32_t offset;
 236             if(length>3) {
 237                 length=3;
 238             }
 239             s+=i;
 240             offset=0;
 241             c=utf8_errorValue[length-1];
 242             UTF8_APPEND_CHAR_UNSAFE(s, offset, c);
 243             i=i+offset;
 244         }
 245     }
 246     return i;
 247 }
 248
 249 U_CAPI UChar32 U_EXPORT2
 250 utf8_prevCharSafeBody(const uint8_t *s, int32_t start, int32_t *pi, UChar32 c, UBool strict) {
 251     int32_t i=*pi;
 252     uint8_t b, count=1, shift=6;
 253
 254     /* extract value bits from the last trail byte */
 255     c&=0x3f;
 256
 257     for(;;) {
 258         if(i<=start) {
 259             /* no lead byte at all */
 260             if(strict>=0) {
 261                 return UTF8_ERROR_VALUE_1;
 262             } else {
 263                 return U_SENTINEL;
 264             }
 265             /*break;*/
 266         }
 267
 268         /* read another previous byte */
 269         b=s[--i];
 270         if((uint8_t)(b-0x80)<0x7e) { /* 0x80<=b<0xfe */
 271             if(b&0x40) {
 272                 /* lead byte, this will always end the loop */
 273                 uint8_t shouldCount=U8_COUNT_TRAIL_BYTES(b);
 274
 275                 if(count==shouldCount) {
 276                     /* set the new position */
 277                     *pi=i;
 278                     U8_MASK_LEAD_BYTE(b, count);
 279                     c|=(UChar32)b<<shift;
 280                     if(count>=4 || c>0x10ffff || c<utf8_minLegal[count] || (U_IS_SURROGATE(c) && strict!=-2) || (strict>0 && U_IS_UNICODE_NONCHAR(c))) {
 281                         /* illegal sequence or (strict and non-character) */
 282                         if(count>=4) {
 283                             count=3;
 284                         }
 285                         if(strict>=0) {
 286                             c=utf8_errorValue[count];
 287                         } else {
 288                             c=U_SENTINEL;
 289                         }
 290                     } else {
 291                         /* exit with correct c */
 292                     }
 293                 } else {
 294                     /* the lead byte does not match the number of trail bytes */
 295                     /* only set the position to the lead byte if it would
 296                        include the trail byte that we started with */
 297                     if(count<shouldCount) {
 298                         *pi=i;
 299                         if(strict>=0) {
 300                             c=utf8_errorValue[count];
 301                         } else {
 302                             c=U_SENTINEL;
 303                         }
 304                     } else {
 305                         if(strict>=0) {
 306                             c=UTF8_ERROR_VALUE_1;
 307                         } else {
 308                             c=U_SENTINEL;
 309                         }
 310                     }
 311                 }
 312                 break;
 313             } else if(count<5) {
 314                 /* trail byte */
 315                 c|=(UChar32)(b&0x3f)<<shift;
 316                 ++count;
 317                 shift+=6;
 318             } else {
 319                 /* more than 5 trail bytes is illegal */
 320                 if(strict>=0) {
 321                     c=UTF8_ERROR_VALUE_1;
 322                 } else {
 323                     c=U_SENTINEL;
 324                 }
 325                 break;
 326             }
 327         } else {
 328             /* single-byte character precedes trailing bytes */
 329             if(strict>=0) {
 330                 c=UTF8_ERROR_VALUE_1;
 331             } else {
 332                 c=U_SENTINEL;
 333             }
 334             break;
 335         }
 336     }
 337     return c;
 338 }
 339
 340 U_CAPI int32_t U_EXPORT2
 341 utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i) {
 342     /* i had been decremented once before the function call */
 343     int32_t I=i, Z;
 344     uint8_t b;
 345
 346     /* read at most the 6 bytes s[Z] to s[i], inclusively */
 347     if(I-5>start) {
 348         Z=I-5;
 349     } else {
 350         Z=start;
 351     }
 352
 353     /* return I if the sequence starting there is long enough to include i */
 354     do {
 355         b=s[I];
 356         if((uint8_t)(b-0x80)>=0x7e) { /* not 0x80<=b<0xfe */
 357             break;
 358         } else if(b>=0xc0) {
 359             if(U8_COUNT_TRAIL_BYTES(b)>=(i-I)) {
 360                 return I;
 361             } else {
 362                 break;
 363             }
 364         }
 365     } while(Z<=--I);
 366
 367     /* return i itself to be consistent with the FWD_1 macro */
 368     return i;
 369 }