icuSources/common/utf_impl.cpp

   1 // © 2016 and later: Unicode, Inc. and others.
   2 // License & terms of use: http://www.unicode.org/copyright.html
   3 /*
   4 ******************************************************************************
   5 *
   6 *   Copyright (C) 1999-2012, International Business Machines
   7 *   Corporation and others.  All Rights Reserved.
   8 *
   9 ******************************************************************************
  10 *   file name:  utf_impl.c
  11 *   encoding:   UTF-8
  12 *   tab size:   8 (not used)
  13 *   indentation:4
  14 *
  15 *   created on: 1999sep13
  16 *   created by: Markus W. Scherer
  17 *
  18 *   This file provides implementation functions for macros in the utfXX.h
  19 *   that would otherwise be too long as macros.
  20 */
  21
  22 /* set import/export definitions */
  23 #ifndef U_UTF8_IMPL
  24 #   define U_UTF8_IMPL
  25 #endif
  26
  27 #include "unicode/utypes.h"
  28 #include "unicode/utf.h"
  29 #include "unicode/utf8.h"
  30 #include "unicode/utf_old.h"
  31 #include "uassert.h"
  32
  33 /*
  34  * Table of the number of utf8 trail bytes, indexed by the lead byte.
  35  * Used by the deprecated macro UTF8_COUNT_TRAIL_BYTES, defined in utf_old.h
  36  *
  37  * The current macro, U8_COUNT_TRAIL_BYTES, does _not_ use this table.
  38  *
  39  * Note that this table cannot be removed, even if UTF8_COUNT_TRAIL_BYTES were
  40  * changed to no longer use it. References to the table from expansions of UTF8_COUNT_TRAIL_BYTES
  41  * may exist in old client code that must continue to run with newer icu library versions.
  42  *
  43  * This table could be replaced on many machines by
  44  * a few lines of assembler code using an
  45  * "index of first 0-bit from msb" instruction and
  46  * one or two more integer instructions.
  47  *
  48  * For example, on an i386, do something like
  49  * - MOV AL, leadByte
  50  * - NOT AL         (8-bit, leave b15..b8==0..0, reverse only b7..b0)
  51  * - MOV AH, 0
  52  * - BSR BX, AX     (16-bit)
  53  * - MOV AX, 6      (result)
  54  * - JZ finish      (ZF==1 if leadByte==0xff)
  55  * - SUB AX, BX (result)
  56  * -finish:
  57  * (BSR: Bit Scan Reverse, scans for a 1-bit, starting from the MSB)
  58  *
  59  * In Unicode, all UTF-8 byte sequences with more than 4 bytes are illegal;
  60  * lead bytes above 0xf4 are illegal.
  61  * We keep them in this table for skipping long ISO 10646-UTF-8 sequences.
  62  */
  63 extern "C" U_EXPORT const uint8_t
  64 utf8_countTrailBytes[256]={
  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     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  69
  70     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  71     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  72     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  73     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  74
  75     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  76     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  77     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  78     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  79
  80     1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  81     1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  82
  83     2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  84     3, 3, 3, 3, 3,
  85     3, 3, 3,    /* illegal in Unicode */
  86     4, 4, 4, 4, /* illegal in Unicode */
  87     5, 5,       /* illegal in Unicode */
  88     0, 0        /* illegal bytes 0xfe and 0xff */
  89 };
  90
  91 static const UChar32
  92 utf8_minLegal[4]={ 0, 0x80, 0x800, 0x10000 };
  93
  94 static const UChar32
  95 utf8_errorValue[6]={
  96     UTF8_ERROR_VALUE_1, UTF8_ERROR_VALUE_2, UTF_ERROR_VALUE, 0x10ffff,
  97     0x3ffffff, 0x7fffffff
  98 };
  99
 100 static UChar32
 101 errorValue(int32_t count, int8_t strict) {
 102     if(strict>=0) {
 103         return utf8_errorValue[count];
 104     } else if(strict==-3) {
 105         return 0xfffd;
 106     } else {
 107         return U_SENTINEL;
 108     }
 109 }
 110
 111 /*
 112  * Handle the non-inline part of the U8_NEXT() and U8_NEXT_FFFD() macros
 113  * and their obsolete sibling UTF8_NEXT_CHAR_SAFE().
 114  *
 115  * U8_NEXT() supports NUL-terminated strings indicated via length<0.
 116  *
 117  * The "strict" parameter controls the error behavior:
 118  * <0  "Safe" behavior of U8_NEXT():
 119  *     -1: All illegal byte sequences yield U_SENTINEL=-1.
 120  *     -2: Same as -1, except for lenient treatment of surrogate code points as legal.
 121  *         Some implementations use this for roundtripping of
 122  *         Unicode 16-bit strings that are not well-formed UTF-16, that is, they
 123  *         contain unpaired surrogates.
 124  *     -3: All illegal byte sequences yield U+FFFD.
 125  *  0  Obsolete "safe" behavior of UTF8_NEXT_CHAR_SAFE(..., FALSE):
 126  *     All illegal byte sequences yield a positive code point such that this
 127  *     result code point would be encoded with the same number of bytes as
 128  *     the illegal sequence.
 129  * >0  Obsolete "strict" behavior of UTF8_NEXT_CHAR_SAFE(..., TRUE):
 130  *     Same as the obsolete "safe" behavior, but non-characters are also treated
 131  *     like illegal sequences.
 132  *
 133  * Note that a UBool is the same as an int8_t.
 134  */
 135 U_CAPI UChar32 U_EXPORT2
 136 utf8_nextCharSafeBody(const uint8_t *s, int32_t *pi, int32_t length, UChar32 c, UBool strict) {
 137     int32_t i=*pi;
 138     uint8_t count=U8_COUNT_TRAIL_BYTES(c);
 139     U_ASSERT(count <= 5); /* U8_COUNT_TRAIL_BYTES returns value 0...5 */
 140     if(i+count<=length || length<0) {
 141         uint8_t trail;
 142
 143         U8_MASK_LEAD_BYTE(c, count);
 144         /* support NUL-terminated strings: do not read beyond the first non-trail byte */
 145         switch(count) {
 146         /* each branch falls through to the next one */
 147         case 0:
 148             /* count==0 for illegally leading trail bytes and the illegal bytes 0xfe and 0xff */
 149         case 5:
 150         case 4:
 151             /* count>=4 is always illegal: no more than 3 trail bytes in Unicode's UTF-8 */
 152             break;
 153         case 3:
 154             trail=s[i++]-0x80;
 155             c=(c<<6)|trail;
 156             /* c>=0x110 would result in code point>0x10ffff, outside Unicode */
 157             if(c>=0x110 || trail>0x3f) { break; }
 158             U_FALLTHROUGH;
 159         case 2:
 160             trail=s[i++]-0x80;
 161             c=(c<<6)|trail;
 162             /*
 163              * test for a surrogate d800..dfff unless we are lenient:
 164              * before the last (c<<6), a surrogate is c=360..37f
 165              */
 166             if(((c&0xffe0)==0x360 && strict!=-2) || trail>0x3f) { break; }
 167             U_FALLTHROUGH;
 168         case 1:
 169             trail=s[i++]-0x80;
 170             c=(c<<6)|trail;
 171             if(trail>0x3f) { break; }
 172             /* correct sequence - all trail bytes have (b7..b6)==(10) */
 173             if(c>=utf8_minLegal[count] &&
 174                     /* strict: forbid non-characters like U+fffe */
 175                     (strict<=0 || !U_IS_UNICODE_NONCHAR(c))) {
 176                 *pi=i;
 177                 return c;
 178             }
 179         /* no default branch to optimize switch()  - all values are covered */
 180         }
 181     } else {
 182         /* too few bytes left */
 183         count=length-i;
 184     }
 185
 186     /* error handling */
 187     i=*pi;
 188     while(count>0 && U8_IS_TRAIL(s[i])) {
 189         ++i;
 190         --count;
 191     }
 192     c=errorValue(i-*pi, strict);
 193     *pi=i;
 194     return c;
 195 }
 196
 197 U_CAPI int32_t U_EXPORT2
 198 utf8_appendCharSafeBody(uint8_t *s, int32_t i, int32_t length, UChar32 c, UBool *pIsError) {
 199     if((uint32_t)(c)<=0x7ff) {
 200         if((i)+1<(length)) {
 201             (s)[(i)++]=(uint8_t)(((c)>>6)|0xc0);
 202             (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80);
 203             return i;
 204         }
 205     } else if((uint32_t)(c)<=0xffff) {
 206         /* Starting with Unicode 3.2, surrogate code points must not be encoded in UTF-8. */
 207         if((i)+2<(length) && !U_IS_SURROGATE(c)) {
 208             (s)[(i)++]=(uint8_t)(((c)>>12)|0xe0);
 209             (s)[(i)++]=(uint8_t)((((c)>>6)&0x3f)|0x80);
 210             (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80);
 211             return i;
 212         }
 213     } else if((uint32_t)(c)<=0x10ffff) {
 214         if((i)+3<(length)) {
 215             (s)[(i)++]=(uint8_t)(((c)>>18)|0xf0);
 216             (s)[(i)++]=(uint8_t)((((c)>>12)&0x3f)|0x80);
 217             (s)[(i)++]=(uint8_t)((((c)>>6)&0x3f)|0x80);
 218             (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80);
 219             return i;
 220         }
 221     }
 222     /* c>0x10ffff or not enough space, write an error value */
 223     if(pIsError!=NULL) {
 224         *pIsError=TRUE;
 225     } else {
 226         length-=i;
 227         if(length>0) {
 228             int32_t offset;
 229             if(length>3) {
 230                 length=3;
 231             }
 232             s+=i;
 233             offset=0;
 234             c=utf8_errorValue[length-1];
 235             UTF8_APPEND_CHAR_UNSAFE(s, offset, c);
 236             i=i+offset;
 237         }
 238     }
 239     return i;
 240 }
 241
 242 U_CAPI UChar32 U_EXPORT2
 243 utf8_prevCharSafeBody(const uint8_t *s, int32_t start, int32_t *pi, UChar32 c, UBool strict) {
 244     int32_t i=*pi;
 245     uint8_t b, count=1, shift=6;
 246
 247     if(!U8_IS_TRAIL(c)) { return errorValue(0, strict); }
 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             return errorValue(0, strict);
 256         }
 257
 258         /* read another previous byte */
 259         b=s[--i];
 260         if((uint8_t)(b-0x80)<0x7e) { /* 0x80<=b<0xfe */
 261             if(b&0x40) {
 262                 /* lead byte, this will always end the loop */
 263                 uint8_t shouldCount=U8_COUNT_TRAIL_BYTES(b);
 264
 265                 if(count==shouldCount) {
 266                     /* set the new position */
 267                     *pi=i;
 268                     U8_MASK_LEAD_BYTE(b, count);
 269                     c|=(UChar32)b<<shift;
 270                     if(count>=4 || c>0x10ffff || c<utf8_minLegal[count] || (U_IS_SURROGATE(c) && strict!=-2) || (strict>0 && U_IS_UNICODE_NONCHAR(c))) {
 271                         /* illegal sequence or (strict and non-character) */
 272                         if(count>=4) {
 273                             count=3;
 274                         }
 275                         c=errorValue(count, strict);
 276                     } else {
 277                         /* exit with correct c */
 278                     }
 279                 } else {
 280                     /* the lead byte does not match the number of trail bytes */
 281                     /* only set the position to the lead byte if it would
 282                        include the trail byte that we started with */
 283                     if(count<shouldCount) {
 284                         *pi=i;
 285                         c=errorValue(count, strict);
 286                     } else {
 287                         c=errorValue(0, strict);
 288                     }
 289                 }
 290                 break;
 291             } else if(count<5) {
 292                 /* trail byte */
 293                 c|=(UChar32)(b&0x3f)<<shift;
 294                 ++count;
 295                 shift+=6;
 296             } else {
 297                 /* more than 5 trail bytes is illegal */
 298                 c=errorValue(0, strict);
 299                 break;
 300             }
 301         } else {
 302             /* single-byte character precedes trailing bytes */
 303             c=errorValue(0, strict);
 304             break;
 305         }
 306     }
 307     return c;
 308 }
 309
 310 U_CAPI int32_t U_EXPORT2
 311 utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i) {
 312     /* i had been decremented once before the function call */
 313     int32_t I=i, Z;
 314     uint8_t b;
 315
 316     /* read at most the 6 bytes s[Z] to s[i], inclusively */
 317     if(I-5>start) {
 318         Z=I-5;
 319     } else {
 320         Z=start;
 321     }
 322
 323     /* return I if the sequence starting there is long enough to include i */
 324     do {
 325         b=s[I];
 326         if((uint8_t)(b-0x80)>=0x7e) { /* not 0x80<=b<0xfe */
 327             break;
 328         } else if(b>=0xc0) {
 329             if(U8_COUNT_TRAIL_BYTES(b)>=(i-I)) {
 330                 return I;
 331             } else {
 332                 break;
 333             }
 334         }
 335     } while(Z<=--I);
 336
 337     /* return i itself to be consistent with the FWD_1 macro */
 338     return i;
 339 }