| File: | root/firefox-clang/intl/icu/source/common/normalizer2impl.h |
| Warning: | line 641, column 25 Addition of a null pointer (via field 'extraData') and a probably nonzero integer value may result in undefined behavior |
Press '?' to see keyboard shortcuts
Keyboard shortcuts:
| 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) 2009-2014, International Business Machines | ||||||||
| 7 | * Corporation and others. All Rights Reserved. | ||||||||
| 8 | * | ||||||||
| 9 | ******************************************************************************* | ||||||||
| 10 | * file name: normalizer2impl.cpp | ||||||||
| 11 | * encoding: UTF-8 | ||||||||
| 12 | * tab size: 8 (not used) | ||||||||
| 13 | * indentation:4 | ||||||||
| 14 | * | ||||||||
| 15 | * created on: 2009nov22 | ||||||||
| 16 | * created by: Markus W. Scherer | ||||||||
| 17 | */ | ||||||||
| 18 | |||||||||
| 19 | // #define UCPTRIE_DEBUG | ||||||||
| 20 | |||||||||
| 21 | #include "unicode/utypes.h" | ||||||||
| 22 | |||||||||
| 23 | #if !UCONFIG_NO_NORMALIZATION0 | ||||||||
| 24 | |||||||||
| 25 | #include "unicode/bytestream.h" | ||||||||
| 26 | #include "unicode/edits.h" | ||||||||
| 27 | #include "unicode/normalizer2.h" | ||||||||
| 28 | #include "unicode/stringoptions.h" | ||||||||
| 29 | #include "unicode/ucptrie.h" | ||||||||
| 30 | #include "unicode/udata.h" | ||||||||
| 31 | #include "unicode/umutablecptrie.h" | ||||||||
| 32 | #include "unicode/ustring.h" | ||||||||
| 33 | #include "unicode/utf16.h" | ||||||||
| 34 | #include "unicode/utf8.h" | ||||||||
| 35 | #include "bytesinkutil.h" | ||||||||
| 36 | #include "cmemory.h" | ||||||||
| 37 | #include "mutex.h" | ||||||||
| 38 | #include "normalizer2impl.h" | ||||||||
| 39 | #include "putilimp.h" | ||||||||
| 40 | #include "uassert.h" | ||||||||
| 41 | #include "ucptrie_impl.h" | ||||||||
| 42 | #include "uset_imp.h" | ||||||||
| 43 | #include "uvector.h" | ||||||||
| 44 | |||||||||
| 45 | U_NAMESPACE_BEGINnamespace icu_77 { | ||||||||
| 46 | |||||||||
| 47 | namespace { | ||||||||
| 48 | |||||||||
| 49 | /** | ||||||||
| 50 | * UTF-8 lead byte for minNoMaybeCP. | ||||||||
| 51 | * Can be lower than the actual lead byte for c. | ||||||||
| 52 | * Typically U+0300 for NFC/NFD, U+00A0 for NFKC/NFKD, U+0041 for NFKC_Casefold. | ||||||||
| 53 | */ | ||||||||
| 54 | inline uint8_t leadByteForCP(UChar32 c) { | ||||||||
| 55 | if (c <= 0x7f) { | ||||||||
| 56 | return static_cast<uint8_t>(c); | ||||||||
| 57 | } else if (c <= 0x7ff) { | ||||||||
| 58 | return static_cast<uint8_t>(0xc0 + (c >> 6)); | ||||||||
| 59 | } else { | ||||||||
| 60 | // Should not occur because ccc(U+0300)!=0. | ||||||||
| 61 | return 0xe0; | ||||||||
| 62 | } | ||||||||
| 63 | } | ||||||||
| 64 | |||||||||
| 65 | /** | ||||||||
| 66 | * Returns the code point from one single well-formed UTF-8 byte sequence | ||||||||
| 67 | * between cpStart and cpLimit. | ||||||||
| 68 | * | ||||||||
| 69 | * Trie UTF-8 macros do not assemble whole code points (for efficiency). | ||||||||
| 70 | * When we do need the code point, we call this function. | ||||||||
| 71 | * We should not need it for normalization-inert data (norm16==0). | ||||||||
| 72 | * Illegal sequences yield the error value norm16==0 just like real normalization-inert code points. | ||||||||
| 73 | */ | ||||||||
| 74 | UChar32 codePointFromValidUTF8(const uint8_t *cpStart, const uint8_t *cpLimit) { | ||||||||
| 75 | // Similar to U8_NEXT_UNSAFE(s, i, c). | ||||||||
| 76 | U_ASSERT(cpStart < cpLimit)(static_cast <bool> (cpStart < cpLimit) ? void (0) : __assert_fail ("cpStart < cpLimit", __builtin_FILE (), __builtin_LINE (), __extension__ __PRETTY_FUNCTION__)); | ||||||||
| 77 | uint8_t c = *cpStart; | ||||||||
| 78 | switch(cpLimit-cpStart) { | ||||||||
| 79 | case 1: | ||||||||
| 80 | return c; | ||||||||
| 81 | case 2: | ||||||||
| 82 | return ((c&0x1f)<<6) | (cpStart[1]&0x3f); | ||||||||
| 83 | case 3: | ||||||||
| 84 | // no need for (c&0xf) because the upper bits are truncated after <<12 in the cast to (char16_t) | ||||||||
| 85 | return static_cast<char16_t>((c << 12) | ((cpStart[1] & 0x3f) << 6) | (cpStart[2] & 0x3f)); | ||||||||
| 86 | case 4: | ||||||||
| 87 | return ((c&7)<<18) | ((cpStart[1]&0x3f)<<12) | ((cpStart[2]&0x3f)<<6) | (cpStart[3]&0x3f); | ||||||||
| 88 | default: | ||||||||
| 89 | UPRV_UNREACHABLE_EXITabort(); // Should not occur. | ||||||||
| 90 | } | ||||||||
| 91 | } | ||||||||
| 92 | |||||||||
| 93 | /** | ||||||||
| 94 | * Returns the last code point in [start, p[ if it is valid and in U+1000..U+D7FF. | ||||||||
| 95 | * Otherwise returns a negative value. | ||||||||
| 96 | */ | ||||||||
| 97 | UChar32 previousHangulOrJamo(const uint8_t *start, const uint8_t *p) { | ||||||||
| 98 | if ((p - start) >= 3) { | ||||||||
| 99 | p -= 3; | ||||||||
| 100 | uint8_t l = *p; | ||||||||
| 101 | uint8_t t1, t2; | ||||||||
| 102 | if (0xe1 <= l && l <= 0xed && | ||||||||
| 103 | (t1 = static_cast<uint8_t>(p[1] - 0x80)) <= 0x3f && | ||||||||
| 104 | (t2 = static_cast<uint8_t>(p[2] - 0x80)) <= 0x3f && | ||||||||
| 105 | (l < 0xed || t1 <= 0x1f)) { | ||||||||
| 106 | return ((l & 0xf) << 12) | (t1 << 6) | t2; | ||||||||
| 107 | } | ||||||||
| 108 | } | ||||||||
| 109 | return U_SENTINEL(-1); | ||||||||
| 110 | } | ||||||||
| 111 | |||||||||
| 112 | /** | ||||||||
| 113 | * Returns the offset from the Jamo T base if [src, limit[ starts with a single Jamo T code point. | ||||||||
| 114 | * Otherwise returns a negative value. | ||||||||
| 115 | */ | ||||||||
| 116 | int32_t getJamoTMinusBase(const uint8_t *src, const uint8_t *limit) { | ||||||||
| 117 | // Jamo T: E1 86 A8..E1 87 82 | ||||||||
| 118 | if ((limit - src) >= 3 && *src == 0xe1) { | ||||||||
| 119 | if (src[1] == 0x86) { | ||||||||
| 120 | uint8_t t = src[2]; | ||||||||
| 121 | // The first Jamo T is U+11A8 but JAMO_T_BASE is 11A7. | ||||||||
| 122 | // Offset 0 does not correspond to any conjoining Jamo. | ||||||||
| 123 | if (0xa8 <= t && t <= 0xbf) { | ||||||||
| 124 | return t - 0xa7; | ||||||||
| 125 | } | ||||||||
| 126 | } else if (src[1] == 0x87) { | ||||||||
| 127 | uint8_t t = src[2]; | ||||||||
| 128 | if (static_cast<int8_t>(t) <= static_cast<int8_t>(0x82u)) { | ||||||||
| 129 | return t - (0xa7 - 0x40); | ||||||||
| 130 | } | ||||||||
| 131 | } | ||||||||
| 132 | } | ||||||||
| 133 | return -1; | ||||||||
| 134 | } | ||||||||
| 135 | |||||||||
| 136 | void | ||||||||
| 137 | appendCodePointDelta(const uint8_t *cpStart, const uint8_t *cpLimit, int32_t delta, | ||||||||
| 138 | ByteSink &sink, Edits *edits) { | ||||||||
| 139 | char buffer[U8_MAX_LENGTH4]; | ||||||||
| 140 | int32_t length; | ||||||||
| 141 | int32_t cpLength = static_cast<int32_t>(cpLimit - cpStart); | ||||||||
| 142 | if (cpLength == 1) { | ||||||||
| 143 | // The builder makes ASCII map to ASCII. | ||||||||
| 144 | buffer[0] = static_cast<uint8_t>(*cpStart + delta); | ||||||||
| 145 | length = 1; | ||||||||
| 146 | } else { | ||||||||
| 147 | int32_t trail = *(cpLimit-1) + delta; | ||||||||
| 148 | if (0x80 <= trail && trail <= 0xbf) { | ||||||||
| 149 | // The delta only changes the last trail byte. | ||||||||
| 150 | --cpLimit; | ||||||||
| 151 | length = 0; | ||||||||
| 152 | do { buffer[length++] = *cpStart++; } while (cpStart < cpLimit); | ||||||||
| 153 | buffer[length++] = static_cast<uint8_t>(trail); | ||||||||
| 154 | } else { | ||||||||
| 155 | // Decode the code point, add the delta, re-encode. | ||||||||
| 156 | UChar32 c = codePointFromValidUTF8(cpStart, cpLimit) + delta; | ||||||||
| 157 | length = 0; | ||||||||
| 158 | U8_APPEND_UNSAFE(buffer, length, c)do { uint32_t __uc=(c); if(__uc<=0x7f) { (buffer)[(length) ++]=(uint8_t)__uc; } else { if(__uc<=0x7ff) { (buffer)[(length )++]=(uint8_t)((__uc>>6)|0xc0); } else { if(__uc<=0xffff ) { (buffer)[(length)++]=(uint8_t)((__uc>>12)|0xe0); } else { (buffer)[(length)++]=(uint8_t)((__uc>>18)|0xf0); (buffer )[(length)++]=(uint8_t)(((__uc>>12)&0x3f)|0x80); } ( buffer)[(length)++]=(uint8_t)(((__uc>>6)&0x3f)|0x80 ); } (buffer)[(length)++]=(uint8_t)((__uc&0x3f)|0x80); } } while (false); | ||||||||
| 159 | } | ||||||||
| 160 | } | ||||||||
| 161 | if (edits != nullptr) { | ||||||||
| 162 | edits->addReplace(cpLength, length); | ||||||||
| 163 | } | ||||||||
| 164 | sink.Append(buffer, length); | ||||||||
| 165 | } | ||||||||
| 166 | |||||||||
| 167 | } // namespace | ||||||||
| 168 | |||||||||
| 169 | // ReorderingBuffer -------------------------------------------------------- *** | ||||||||
| 170 | |||||||||
| 171 | ReorderingBuffer::ReorderingBuffer(const Normalizer2Impl &ni, UnicodeString &dest, | ||||||||
| 172 | UErrorCode &errorCode) : | ||||||||
| 173 | impl(ni), str(dest), | ||||||||
| 174 | start(str.getBuffer(8)), reorderStart(start), limit(start), | ||||||||
| 175 | remainingCapacity(str.getCapacity()), lastCC(0) { | ||||||||
| 176 | if (start == nullptr && U_SUCCESS(errorCode)) { | ||||||||
| 177 | // getBuffer() already did str.setToBogus() | ||||||||
| 178 | errorCode = U_MEMORY_ALLOCATION_ERROR; | ||||||||
| 179 | } | ||||||||
| 180 | } | ||||||||
| 181 | |||||||||
| 182 | UBool ReorderingBuffer::init(int32_t destCapacity, UErrorCode &errorCode) { | ||||||||
| 183 | int32_t length=str.length(); | ||||||||
| 184 | start=str.getBuffer(destCapacity); | ||||||||
| 185 | if(start==nullptr) { | ||||||||
| 186 | // getBuffer() already did str.setToBogus() | ||||||||
| 187 | errorCode=U_MEMORY_ALLOCATION_ERROR; | ||||||||
| 188 | return false; | ||||||||
| 189 | } | ||||||||
| 190 | limit=start+length; | ||||||||
| 191 | remainingCapacity=str.getCapacity()-length; | ||||||||
| 192 | reorderStart=start; | ||||||||
| 193 | if(start==limit) { | ||||||||
| 194 | lastCC=0; | ||||||||
| 195 | } else { | ||||||||
| 196 | setIterator(); | ||||||||
| 197 | lastCC=previousCC(); | ||||||||
| 198 | // Set reorderStart after the last code point with cc<=1 if there is one. | ||||||||
| 199 | if(lastCC>1) { | ||||||||
| 200 | while(previousCC()>1) {} | ||||||||
| 201 | } | ||||||||
| 202 | reorderStart=codePointLimit; | ||||||||
| 203 | } | ||||||||
| 204 | return true; | ||||||||
| 205 | } | ||||||||
| 206 | |||||||||
| 207 | UBool ReorderingBuffer::equals(const char16_t *otherStart, const char16_t *otherLimit) const { | ||||||||
| 208 | int32_t length = static_cast<int32_t>(limit - start); | ||||||||
| 209 | return | ||||||||
| 210 | length == static_cast<int32_t>(otherLimit - otherStart) && | ||||||||
| 211 | 0==u_memcmpu_memcmp_77(start, otherStart, length); | ||||||||
| 212 | } | ||||||||
| 213 | |||||||||
| 214 | UBool ReorderingBuffer::equals(const uint8_t *otherStart, const uint8_t *otherLimit) const { | ||||||||
| 215 | U_ASSERT((otherLimit - otherStart) <= INT32_MAX)(static_cast <bool> ((otherLimit - otherStart) <= (2147483647 )) ? void (0) : __assert_fail ("(otherLimit - otherStart) <= (2147483647)" , __builtin_FILE (), __builtin_LINE (), __extension__ __PRETTY_FUNCTION__ )); // ensured by caller | ||||||||
| 216 | int32_t length = static_cast<int32_t>(limit - start); | ||||||||
| 217 | int32_t otherLength = static_cast<int32_t>(otherLimit - otherStart); | ||||||||
| 218 | // For equal strings, UTF-8 is at least as long as UTF-16, and at most three times as long. | ||||||||
| 219 | if (otherLength < length || (otherLength / 3) > length) { | ||||||||
| 220 | return false; | ||||||||
| 221 | } | ||||||||
| 222 | // Compare valid strings from between normalization boundaries. | ||||||||
| 223 | // (Invalid sequences are normalization-inert.) | ||||||||
| 224 | for (int32_t i = 0, j = 0;;) { | ||||||||
| 225 | if (i >= length) { | ||||||||
| 226 | return j >= otherLength; | ||||||||
| 227 | } else if (j >= otherLength) { | ||||||||
| 228 | return false; | ||||||||
| 229 | } | ||||||||
| 230 | // Not at the end of either string yet. | ||||||||
| 231 | UChar32 c, other; | ||||||||
| 232 | U16_NEXT_UNSAFE(start, i, c)do { (c)=(start)[(i)++]; if((((c)&0xfffffc00)==0xd800)) { (c)=(((UChar32)((c))<<10UL)+(UChar32)((start)[(i)++])- ((0xd800<<10UL)+0xdc00-0x10000)); } } while (false); | ||||||||
| 233 | U8_NEXT_UNSAFE(otherStart, j, other)do { (other)=(uint8_t)(otherStart)[(j)++]; if(!(((other)& 0x80)==0)) { if((other)<0xe0) { (other)=(((other)&0x1f )<<6)|((otherStart)[(j)++]&0x3f); } else if((other) <0xf0) { (other)=(UChar)(((other)<<12)|(((otherStart )[j]&0x3f)<<6)|((otherStart)[(j)+1]&0x3f)); (j) +=2; } else { (other)=(((other)&7)<<18)|(((otherStart )[j]&0x3f)<<12)|(((otherStart)[(j)+1]&0x3f)<< 6)|((otherStart)[(j)+2]&0x3f); (j)+=3; } } } while (false ); | ||||||||
| 234 | if (c != other) { | ||||||||
| 235 | return false; | ||||||||
| 236 | } | ||||||||
| 237 | } | ||||||||
| 238 | } | ||||||||
| 239 | |||||||||
| 240 | UBool ReorderingBuffer::appendSupplementary(UChar32 c, uint8_t cc, UErrorCode &errorCode) { | ||||||||
| 241 | if(remainingCapacity<2 && !resize(2, errorCode)) { | ||||||||
| 242 | return false; | ||||||||
| 243 | } | ||||||||
| 244 | if(lastCC<=cc || cc==0) { | ||||||||
| 245 | limit[0]=U16_LEAD(c)(UChar)(((c)>>10)+0xd7c0); | ||||||||
| 246 | limit[1]=U16_TRAIL(c)(UChar)(((c)&0x3ff)|0xdc00); | ||||||||
| 247 | limit+=2; | ||||||||
| 248 | lastCC=cc; | ||||||||
| 249 | if(cc<=1) { | ||||||||
| 250 | reorderStart=limit; | ||||||||
| 251 | } | ||||||||
| 252 | } else { | ||||||||
| 253 | insert(c, cc); | ||||||||
| 254 | } | ||||||||
| 255 | remainingCapacity-=2; | ||||||||
| 256 | return true; | ||||||||
| 257 | } | ||||||||
| 258 | |||||||||
| 259 | UBool ReorderingBuffer::append(const char16_t *s, int32_t length, UBool isNFD, | ||||||||
| 260 | uint8_t leadCC, uint8_t trailCC, | ||||||||
| 261 | UErrorCode &errorCode) { | ||||||||
| 262 | if(length==0) { | ||||||||
| 263 | return true; | ||||||||
| 264 | } | ||||||||
| 265 | if(remainingCapacity<length && !resize(length, errorCode)) { | ||||||||
| 266 | return false; | ||||||||
| 267 | } | ||||||||
| 268 | remainingCapacity-=length; | ||||||||
| 269 | if(lastCC<=leadCC || leadCC==0) { | ||||||||
| 270 | if(trailCC<=1) { | ||||||||
| 271 | reorderStart=limit+length; | ||||||||
| 272 | } else if(leadCC<=1) { | ||||||||
| 273 | reorderStart=limit+1; // Ok if not a code point boundary. | ||||||||
| 274 | } | ||||||||
| 275 | const char16_t *sLimit=s+length; | ||||||||
| 276 | do { *limit++=*s++; } while(s!=sLimit); | ||||||||
| 277 | lastCC=trailCC; | ||||||||
| 278 | } else { | ||||||||
| 279 | int32_t i=0; | ||||||||
| 280 | UChar32 c; | ||||||||
| 281 | U16_NEXT(s, i, length, c)do { (c)=(s)[(i)++]; if((((c)&0xfffffc00)==0xd800)) { uint16_t __c2; if((i)!=(length) && (((__c2=(s)[(i)])&0xfffffc00 )==0xdc00)) { ++(i); (c)=(((UChar32)((c))<<10UL)+(UChar32 )(__c2)-((0xd800<<10UL)+0xdc00-0x10000)); } } } while ( false); | ||||||||
| 282 | insert(c, leadCC); // insert first code point | ||||||||
| 283 | while(i<length) { | ||||||||
| 284 | U16_NEXT(s, i, length, c)do { (c)=(s)[(i)++]; if((((c)&0xfffffc00)==0xd800)) { uint16_t __c2; if((i)!=(length) && (((__c2=(s)[(i)])&0xfffffc00 )==0xdc00)) { ++(i); (c)=(((UChar32)((c))<<10UL)+(UChar32 )(__c2)-((0xd800<<10UL)+0xdc00-0x10000)); } } } while ( false); | ||||||||
| 285 | if(i<length) { | ||||||||
| 286 | if (isNFD) { | ||||||||
| 287 | leadCC = Normalizer2Impl::getCCFromYesOrMaybeYes(impl.getRawNorm16(c)); | ||||||||
| 288 | } else { | ||||||||
| 289 | leadCC = impl.getCC(impl.getNorm16(c)); | ||||||||
| 290 | } | ||||||||
| 291 | } else { | ||||||||
| 292 | leadCC=trailCC; | ||||||||
| 293 | } | ||||||||
| 294 | append(c, leadCC, errorCode); | ||||||||
| 295 | } | ||||||||
| 296 | } | ||||||||
| 297 | return true; | ||||||||
| 298 | } | ||||||||
| 299 | |||||||||
| 300 | UBool ReorderingBuffer::appendZeroCC(UChar32 c, UErrorCode &errorCode) { | ||||||||
| 301 | int32_t cpLength=U16_LENGTH(c)((uint32_t)(c)<=0xffff ? 1 : 2); | ||||||||
| 302 | if(remainingCapacity<cpLength && !resize(cpLength, errorCode)) { | ||||||||
| 303 | return false; | ||||||||
| 304 | } | ||||||||
| 305 | remainingCapacity-=cpLength; | ||||||||
| 306 | if(cpLength==1) { | ||||||||
| 307 | *limit++ = static_cast<char16_t>(c); | ||||||||
| 308 | } else { | ||||||||
| 309 | limit[0]=U16_LEAD(c)(UChar)(((c)>>10)+0xd7c0); | ||||||||
| 310 | limit[1]=U16_TRAIL(c)(UChar)(((c)&0x3ff)|0xdc00); | ||||||||
| 311 | limit+=2; | ||||||||
| 312 | } | ||||||||
| 313 | lastCC=0; | ||||||||
| 314 | reorderStart=limit; | ||||||||
| 315 | return true; | ||||||||
| 316 | } | ||||||||
| 317 | |||||||||
| 318 | UBool ReorderingBuffer::appendZeroCC(const char16_t *s, const char16_t *sLimit, UErrorCode &errorCode) { | ||||||||
| 319 | if(s==sLimit) { | ||||||||
| 320 | return true; | ||||||||
| 321 | } | ||||||||
| 322 | int32_t length = static_cast<int32_t>(sLimit - s); | ||||||||
| 323 | if(remainingCapacity<length && !resize(length, errorCode)) { | ||||||||
| 324 | return false; | ||||||||
| 325 | } | ||||||||
| 326 | u_memcpyu_memcpy_77(limit, s, length); | ||||||||
| 327 | limit+=length; | ||||||||
| 328 | remainingCapacity-=length; | ||||||||
| 329 | lastCC=0; | ||||||||
| 330 | reorderStart=limit; | ||||||||
| 331 | return true; | ||||||||
| 332 | } | ||||||||
| 333 | |||||||||
| 334 | void ReorderingBuffer::remove() { | ||||||||
| 335 | reorderStart=limit=start; | ||||||||
| 336 | remainingCapacity=str.getCapacity(); | ||||||||
| 337 | lastCC=0; | ||||||||
| 338 | } | ||||||||
| 339 | |||||||||
| 340 | void ReorderingBuffer::removeSuffix(int32_t suffixLength) { | ||||||||
| 341 | if(suffixLength<(limit-start)) { | ||||||||
| 342 | limit-=suffixLength; | ||||||||
| 343 | remainingCapacity+=suffixLength; | ||||||||
| 344 | } else { | ||||||||
| 345 | limit=start; | ||||||||
| 346 | remainingCapacity=str.getCapacity(); | ||||||||
| 347 | } | ||||||||
| 348 | lastCC=0; | ||||||||
| 349 | reorderStart=limit; | ||||||||
| 350 | } | ||||||||
| 351 | |||||||||
| 352 | UBool ReorderingBuffer::resize(int32_t appendLength, UErrorCode &errorCode) { | ||||||||
| 353 | int32_t reorderStartIndex = static_cast<int32_t>(reorderStart - start); | ||||||||
| 354 | int32_t length = static_cast<int32_t>(limit - start); | ||||||||
| 355 | str.releaseBuffer(length); | ||||||||
| 356 | int32_t newCapacity=length+appendLength; | ||||||||
| 357 | int32_t doubleCapacity=2*str.getCapacity(); | ||||||||
| 358 | if(newCapacity<doubleCapacity) { | ||||||||
| 359 | newCapacity=doubleCapacity; | ||||||||
| 360 | } | ||||||||
| 361 | if(newCapacity<256) { | ||||||||
| 362 | newCapacity=256; | ||||||||
| 363 | } | ||||||||
| 364 | start=str.getBuffer(newCapacity); | ||||||||
| 365 | if(start==nullptr) { | ||||||||
| 366 | // getBuffer() already did str.setToBogus() | ||||||||
| 367 | errorCode=U_MEMORY_ALLOCATION_ERROR; | ||||||||
| 368 | return false; | ||||||||
| 369 | } | ||||||||
| 370 | reorderStart=start+reorderStartIndex; | ||||||||
| 371 | limit=start+length; | ||||||||
| 372 | remainingCapacity=str.getCapacity()-length; | ||||||||
| 373 | return true; | ||||||||
| 374 | } | ||||||||
| 375 | |||||||||
| 376 | void ReorderingBuffer::skipPrevious() { | ||||||||
| 377 | codePointLimit=codePointStart; | ||||||||
| 378 | char16_t c=*--codePointStart; | ||||||||
| 379 | if(U16_IS_TRAIL(c)(((c)&0xfffffc00)==0xdc00) && start<codePointStart && U16_IS_LEAD(*(codePointStart-1))(((*(codePointStart-1))&0xfffffc00)==0xd800)) { | ||||||||
| 380 | --codePointStart; | ||||||||
| 381 | } | ||||||||
| 382 | } | ||||||||
| 383 | |||||||||
| 384 | uint8_t ReorderingBuffer::previousCC() { | ||||||||
| 385 | codePointLimit=codePointStart; | ||||||||
| 386 | if(reorderStart>=codePointStart) { | ||||||||
| 387 | return 0; | ||||||||
| 388 | } | ||||||||
| 389 | UChar32 c=*--codePointStart; | ||||||||
| 390 | char16_t c2; | ||||||||
| 391 | if(U16_IS_TRAIL(c)(((c)&0xfffffc00)==0xdc00) && start<codePointStart && U16_IS_LEAD(c2=*(codePointStart-1))(((c2=*(codePointStart-1))&0xfffffc00)==0xd800)) { | ||||||||
| 392 | --codePointStart; | ||||||||
| 393 | c=U16_GET_SUPPLEMENTARY(c2, c)(((UChar32)(c2)<<10UL)+(UChar32)(c)-((0xd800<<10UL )+0xdc00-0x10000)); | ||||||||
| 394 | } | ||||||||
| 395 | return impl.getCCFromYesOrMaybeYesCP(c); | ||||||||
| 396 | } | ||||||||
| 397 | |||||||||
| 398 | // Inserts c somewhere before the last character. | ||||||||
| 399 | // Requires 0<cc<lastCC which implies reorderStart<limit. | ||||||||
| 400 | void ReorderingBuffer::insert(UChar32 c, uint8_t cc) { | ||||||||
| 401 | for(setIterator(), skipPrevious(); previousCC()>cc;) {} | ||||||||
| 402 | // insert c at codePointLimit, after the character with prevCC<=cc | ||||||||
| 403 | char16_t *q=limit; | ||||||||
| 404 | char16_t *r=limit+=U16_LENGTH(c)((uint32_t)(c)<=0xffff ? 1 : 2); | ||||||||
| 405 | do { | ||||||||
| 406 | *--r=*--q; | ||||||||
| 407 | } while(codePointLimit!=q); | ||||||||
| 408 | writeCodePoint(q, c); | ||||||||
| 409 | if(cc<=1) { | ||||||||
| 410 | reorderStart=r; | ||||||||
| 411 | } | ||||||||
| 412 | } | ||||||||
| 413 | |||||||||
| 414 | // Normalizer2Impl --------------------------------------------------------- *** | ||||||||
| 415 | |||||||||
| 416 | struct CanonIterData : public UMemory { | ||||||||
| 417 | CanonIterData(UErrorCode &errorCode); | ||||||||
| 418 | ~CanonIterData(); | ||||||||
| 419 | void addToStartSet(UChar32 origin, UChar32 decompLead, UErrorCode &errorCode); | ||||||||
| 420 | UMutableCPTrie *mutableTrie; | ||||||||
| 421 | UCPTrie *trie; | ||||||||
| 422 | UVector canonStartSets; // contains UnicodeSet * | ||||||||
| 423 | }; | ||||||||
| 424 | |||||||||
| 425 | Normalizer2Impl::~Normalizer2Impl() { | ||||||||
| 426 | delete fCanonIterData; | ||||||||
| 427 | } | ||||||||
| 428 | |||||||||
| 429 | void | ||||||||
| 430 | Normalizer2Impl::init(const int32_t *inIndexes, const UCPTrie *inTrie, | ||||||||
| 431 | const uint16_t *inExtraData, const uint8_t *inSmallFCD) { | ||||||||
| 432 | minDecompNoCP = static_cast<char16_t>(inIndexes[IX_MIN_DECOMP_NO_CP]); | ||||||||
| 433 | minCompNoMaybeCP = static_cast<char16_t>(inIndexes[IX_MIN_COMP_NO_MAYBE_CP]); | ||||||||
| 434 | minLcccCP = static_cast<char16_t>(inIndexes[IX_MIN_LCCC_CP]); | ||||||||
| 435 | |||||||||
| 436 | minYesNo = static_cast<uint16_t>(inIndexes[IX_MIN_YES_NO]); | ||||||||
| 437 | minYesNoMappingsOnly = static_cast<uint16_t>(inIndexes[IX_MIN_YES_NO_MAPPINGS_ONLY]); | ||||||||
| 438 | minNoNo = static_cast<uint16_t>(inIndexes[IX_MIN_NO_NO]); | ||||||||
| 439 | minNoNoCompBoundaryBefore = static_cast<uint16_t>(inIndexes[IX_MIN_NO_NO_COMP_BOUNDARY_BEFORE]); | ||||||||
| 440 | minNoNoCompNoMaybeCC = static_cast<uint16_t>(inIndexes[IX_MIN_NO_NO_COMP_NO_MAYBE_CC]); | ||||||||
| 441 | minNoNoEmpty = static_cast<uint16_t>(inIndexes[IX_MIN_NO_NO_EMPTY]); | ||||||||
| 442 | limitNoNo = static_cast<uint16_t>(inIndexes[IX_LIMIT_NO_NO]); | ||||||||
| 443 | minMaybeNo = static_cast<uint16_t>(inIndexes[IX_MIN_MAYBE_NO]); | ||||||||
| 444 | minMaybeNoCombinesFwd = static_cast<uint16_t>(inIndexes[IX_MIN_MAYBE_NO_COMBINES_FWD]); | ||||||||
| 445 | minMaybeYes = static_cast<uint16_t>(inIndexes[IX_MIN_MAYBE_YES]); | ||||||||
| 446 | U_ASSERT((minMaybeNo & 7) == 0)(static_cast <bool> ((minMaybeNo & 7) == 0) ? void ( 0) : __assert_fail ("(minMaybeNo & 7) == 0", __builtin_FILE (), __builtin_LINE (), __extension__ __PRETTY_FUNCTION__)); // 8-aligned for noNoDelta bit fields | ||||||||
| 447 | centerNoNoDelta = (minMaybeNo >> DELTA_SHIFT) - MAX_DELTA - 1; | ||||||||
| 448 | |||||||||
| 449 | normTrie=inTrie; | ||||||||
| 450 | extraData=inExtraData; | ||||||||
| 451 | smallFCD=inSmallFCD; | ||||||||
| 452 | } | ||||||||
| 453 | |||||||||
| 454 | U_CDECL_BEGINextern "C" { | ||||||||
| 455 | |||||||||
| 456 | static uint32_t U_CALLCONV | ||||||||
| 457 | segmentStarterMapper(const void * /*context*/, uint32_t value) { | ||||||||
| 458 | return value&CANON_NOT_SEGMENT_STARTER0x80000000; | ||||||||
| 459 | } | ||||||||
| 460 | |||||||||
| 461 | U_CDECL_END} | ||||||||
| 462 | |||||||||
| 463 | void | ||||||||
| 464 | Normalizer2Impl::addLcccChars(UnicodeSet &set) const { | ||||||||
| 465 | UChar32 start = 0, end; | ||||||||
| 466 | uint32_t norm16; | ||||||||
| 467 | while ((end = ucptrie_getRangeucptrie_getRange_77(normTrie, start, UCPMAP_RANGE_FIXED_LEAD_SURROGATES, INERT, | ||||||||
| 468 | nullptr, nullptr, &norm16)) >= 0) { | ||||||||
| 469 | if (norm16 > Normalizer2Impl::MIN_NORMAL_MAYBE_YES && | ||||||||
| 470 | norm16 != Normalizer2Impl::JAMO_VT) { | ||||||||
| 471 | set.add(start, end); | ||||||||
| 472 | } else if (minNoNoCompNoMaybeCC <= norm16 && norm16 < limitNoNo) { | ||||||||
| 473 | uint16_t fcd16 = getFCD16(start); | ||||||||
| 474 | if (fcd16 > 0xff) { set.add(start, end); } | ||||||||
| 475 | } | ||||||||
| 476 | start = end + 1; | ||||||||
| 477 | } | ||||||||
| 478 | } | ||||||||
| 479 | |||||||||
| 480 | void | ||||||||
| 481 | Normalizer2Impl::addPropertyStarts(const USetAdder *sa, UErrorCode & /*errorCode*/) const { | ||||||||
| 482 | // Add the start code point of each same-value range of the trie. | ||||||||
| 483 | UChar32 start = 0, end; | ||||||||
| 484 | uint32_t value; | ||||||||
| 485 | while ((end = ucptrie_getRangeucptrie_getRange_77(normTrie, start, UCPMAP_RANGE_FIXED_LEAD_SURROGATES, INERT, | ||||||||
| 486 | nullptr, nullptr, &value)) >= 0) { | ||||||||
| 487 | sa->add(sa->set, start); | ||||||||
| 488 | if (start != end && isAlgorithmicNoNo(static_cast<uint16_t>(value)) && | ||||||||
| 489 | (value & Normalizer2Impl::DELTA_TCCC_MASK) > Normalizer2Impl::DELTA_TCCC_1) { | ||||||||
| 490 | // Range of code points with same-norm16-value algorithmic decompositions. | ||||||||
| 491 | // They might have different non-zero FCD16 values. | ||||||||
| 492 | uint16_t prevFCD16 = getFCD16(start); | ||||||||
| 493 | while (++start <= end) { | ||||||||
| 494 | uint16_t fcd16 = getFCD16(start); | ||||||||
| 495 | if (fcd16 != prevFCD16) { | ||||||||
| 496 | sa->add(sa->set, start); | ||||||||
| 497 | prevFCD16 = fcd16; | ||||||||
| 498 | } | ||||||||
| 499 | } | ||||||||
| 500 | } | ||||||||
| 501 | start = end + 1; | ||||||||
| 502 | } | ||||||||
| 503 | |||||||||
| 504 | /* add Hangul LV syllables and LV+1 because of skippables */ | ||||||||
| 505 | for(char16_t c=Hangul::HANGUL_BASE; c<Hangul::HANGUL_LIMIT; c+=Hangul::JAMO_T_COUNT) { | ||||||||
| 506 | sa->add(sa->set, c); | ||||||||
| 507 | sa->add(sa->set, c+1); | ||||||||
| 508 | } | ||||||||
| 509 | sa->add(sa->set, Hangul::HANGUL_LIMIT); /* add Hangul+1 to continue with other properties */ | ||||||||
| 510 | } | ||||||||
| 511 | |||||||||
| 512 | void | ||||||||
| 513 | Normalizer2Impl::addCanonIterPropertyStarts(const USetAdder *sa, UErrorCode &errorCode) const { | ||||||||
| 514 | // Add the start code point of each same-value range of the canonical iterator data trie. | ||||||||
| 515 | if (!ensureCanonIterData(errorCode)) { return; } | ||||||||
| 516 | // Currently only used for the SEGMENT_STARTER property. | ||||||||
| 517 | UChar32 start = 0, end; | ||||||||
| 518 | uint32_t value; | ||||||||
| 519 | while ((end = ucptrie_getRangeucptrie_getRange_77(fCanonIterData->trie, start, UCPMAP_RANGE_NORMAL, 0, | ||||||||
| 520 | segmentStarterMapper, nullptr, &value)) >= 0) { | ||||||||
| 521 | sa->add(sa->set, start); | ||||||||
| 522 | start = end + 1; | ||||||||
| 523 | } | ||||||||
| 524 | } | ||||||||
| 525 | |||||||||
| 526 | const char16_t * | ||||||||
| 527 | Normalizer2Impl::copyLowPrefixFromNulTerminated(const char16_t *src, | ||||||||
| 528 | UChar32 minNeedDataCP, | ||||||||
| 529 | ReorderingBuffer *buffer, | ||||||||
| 530 | UErrorCode &errorCode) const { | ||||||||
| 531 | // Make some effort to support NUL-terminated strings reasonably. | ||||||||
| 532 | // Take the part of the fast quick check loop that does not look up | ||||||||
| 533 | // data and check the first part of the string. | ||||||||
| 534 | // After this prefix, determine the string length to simplify the rest | ||||||||
| 535 | // of the code. | ||||||||
| 536 | const char16_t *prevSrc=src; | ||||||||
| 537 | char16_t c; | ||||||||
| 538 | while((c=*src++)<minNeedDataCP && c!=0) {} | ||||||||
| 539 | // Back out the last character for full processing. | ||||||||
| 540 | // Copy this prefix. | ||||||||
| 541 | if(--src!=prevSrc) { | ||||||||
| 542 | if(buffer!=nullptr) { | ||||||||
| 543 | buffer->appendZeroCC(prevSrc, src, errorCode); | ||||||||
| 544 | } | ||||||||
| 545 | } | ||||||||
| 546 | return src; | ||||||||
| 547 | } | ||||||||
| 548 | |||||||||
| 549 | UnicodeString & | ||||||||
| 550 | Normalizer2Impl::decompose(const UnicodeString &src, UnicodeString &dest, | ||||||||
| 551 | UErrorCode &errorCode) const { | ||||||||
| 552 | if(U_FAILURE(errorCode)) { | ||||||||
| 553 | dest.setToBogus(); | ||||||||
| 554 | return dest; | ||||||||
| 555 | } | ||||||||
| 556 | const char16_t *sArray=src.getBuffer(); | ||||||||
| 557 | if(&dest==&src || sArray==nullptr) { | ||||||||
| 558 | errorCode=U_ILLEGAL_ARGUMENT_ERROR; | ||||||||
| 559 | dest.setToBogus(); | ||||||||
| 560 | return dest; | ||||||||
| 561 | } | ||||||||
| 562 | decompose(sArray, sArray+src.length(), dest, src.length(), errorCode); | ||||||||
| 563 | return dest; | ||||||||
| 564 | } | ||||||||
| 565 | |||||||||
| 566 | void | ||||||||
| 567 | Normalizer2Impl::decompose(const char16_t *src, const char16_t *limit, | ||||||||
| 568 | UnicodeString &dest, | ||||||||
| 569 | int32_t destLengthEstimate, | ||||||||
| 570 | UErrorCode &errorCode) const { | ||||||||
| 571 | if(destLengthEstimate<0 && limit!=nullptr) { | ||||||||
| 572 | destLengthEstimate = static_cast<int32_t>(limit - src); | ||||||||
| 573 | } | ||||||||
| 574 | dest.remove(); | ||||||||
| 575 | ReorderingBuffer buffer(*this, dest); | ||||||||
| 576 | if(buffer.init(destLengthEstimate, errorCode)) { | ||||||||
| 577 | decompose(src, limit, &buffer, errorCode); | ||||||||
| 578 | } | ||||||||
| 579 | } | ||||||||
| 580 | |||||||||
| 581 | // Dual functionality: | ||||||||
| 582 | // buffer!=nullptr: normalize | ||||||||
| 583 | // buffer==nullptr: isNormalized/spanQuickCheckYes | ||||||||
| 584 | const char16_t * | ||||||||
| 585 | Normalizer2Impl::decompose(const char16_t *src, const char16_t *limit, | ||||||||
| 586 | ReorderingBuffer *buffer, | ||||||||
| 587 | UErrorCode &errorCode) const { | ||||||||
| 588 | UChar32 minNoCP=minDecompNoCP; | ||||||||
| 589 | if(limit==nullptr) { | ||||||||
| 590 | src=copyLowPrefixFromNulTerminated(src, minNoCP, buffer, errorCode); | ||||||||
| 591 | if(U_FAILURE(errorCode)) { | ||||||||
| 592 | return src; | ||||||||
| 593 | } | ||||||||
| 594 | limit=u_strchru_strchr_77(src, 0); | ||||||||
| 595 | } | ||||||||
| 596 | |||||||||
| 597 | const char16_t *prevSrc; | ||||||||
| 598 | UChar32 c=0; | ||||||||
| 599 | uint16_t norm16=0; | ||||||||
| 600 | |||||||||
| 601 | // only for quick check | ||||||||
| 602 | const char16_t *prevBoundary=src; | ||||||||
| 603 | uint8_t prevCC=0; | ||||||||
| 604 | |||||||||
| 605 | for(;;) { | ||||||||
| 606 | // count code units below the minimum or with irrelevant data for the quick check | ||||||||
| 607 | for(prevSrc=src; src!=limit;) { | ||||||||
| 608 | if( (c=*src)<minNoCP || | ||||||||
| 609 | isMostDecompYesAndZeroCC(norm16=UCPTRIE_FAST_BMP_GET(normTrie, UCPTRIE_16, c)((normTrie)->data.ptr16[((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT] + ((c) & UCPTRIE_FAST_DATA_MASK ))])) | ||||||||
| 610 | ) { | ||||||||
| 611 | ++src; | ||||||||
| 612 | } else if(!U16_IS_LEAD(c)(((c)&0xfffffc00)==0xd800)) { | ||||||||
| 613 | break; | ||||||||
| 614 | } else { | ||||||||
| 615 | char16_t c2; | ||||||||
| 616 | if((src+1)!=limit && U16_IS_TRAIL(c2=src[1])(((c2=src[1])&0xfffffc00)==0xdc00)) { | ||||||||
| 617 | c=U16_GET_SUPPLEMENTARY(c, c2)(((UChar32)(c)<<10UL)+(UChar32)(c2)-((0xd800<<10UL )+0xdc00-0x10000)); | ||||||||
| 618 | norm16=UCPTRIE_FAST_SUPP_GET(normTrie, UCPTRIE_16, c)((normTrie)->data.ptr16[((c) >= (normTrie)->highStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77(normTrie, c))]); | ||||||||
| 619 | if(isMostDecompYesAndZeroCC(norm16)) { | ||||||||
| 620 | src+=2; | ||||||||
| 621 | } else { | ||||||||
| 622 | break; | ||||||||
| 623 | } | ||||||||
| 624 | } else { | ||||||||
| 625 | ++src; // unpaired lead surrogate: inert | ||||||||
| 626 | } | ||||||||
| 627 | } | ||||||||
| 628 | } | ||||||||
| 629 | // copy these code units all at once | ||||||||
| 630 | if(src!=prevSrc) { | ||||||||
| 631 | if(buffer!=nullptr) { | ||||||||
| 632 | if(!buffer->appendZeroCC(prevSrc, src, errorCode)) { | ||||||||
| 633 | break; | ||||||||
| 634 | } | ||||||||
| 635 | } else { | ||||||||
| 636 | prevCC=0; | ||||||||
| 637 | prevBoundary=src; | ||||||||
| 638 | } | ||||||||
| 639 | } | ||||||||
| 640 | if(src==limit) { | ||||||||
| 641 | break; | ||||||||
| 642 | } | ||||||||
| 643 | |||||||||
| 644 | // Check one above-minimum, relevant code point. | ||||||||
| 645 | src+=U16_LENGTH(c)((uint32_t)(c)<=0xffff ? 1 : 2); | ||||||||
| 646 | if(buffer!=nullptr) { | ||||||||
| 647 | if(!decompose(c, norm16, *buffer, errorCode)) { | ||||||||
| 648 | break; | ||||||||
| 649 | } | ||||||||
| 650 | } else { | ||||||||
| 651 | if(isDecompYes(norm16)) { | ||||||||
| 652 | uint8_t cc=getCCFromYesOrMaybeYes(norm16); | ||||||||
| 653 | if(prevCC<=cc || cc==0) { | ||||||||
| 654 | prevCC=cc; | ||||||||
| 655 | if(cc<=1) { | ||||||||
| 656 | prevBoundary=src; | ||||||||
| 657 | } | ||||||||
| 658 | continue; | ||||||||
| 659 | } | ||||||||
| 660 | } | ||||||||
| 661 | return prevBoundary; // "no" or cc out of order | ||||||||
| 662 | } | ||||||||
| 663 | } | ||||||||
| 664 | return src; | ||||||||
| 665 | } | ||||||||
| 666 | |||||||||
| 667 | // Decompose a short piece of text which is likely to contain characters that | ||||||||
| 668 | // fail the quick check loop and/or where the quick check loop's overhead | ||||||||
| 669 | // is unlikely to be amortized. | ||||||||
| 670 | // Called by the compose() and makeFCD() implementations. | ||||||||
| 671 | const char16_t * | ||||||||
| 672 | Normalizer2Impl::decomposeShort(const char16_t *src, const char16_t *limit, | ||||||||
| 673 | UBool stopAtCompBoundary, UBool onlyContiguous, | ||||||||
| 674 | ReorderingBuffer &buffer, UErrorCode &errorCode) const { | ||||||||
| 675 | if (U_FAILURE(errorCode)) { | ||||||||
| 676 | return nullptr; | ||||||||
| 677 | } | ||||||||
| 678 | while(src<limit) { | ||||||||
| 679 | if (stopAtCompBoundary && *src < minCompNoMaybeCP) { | ||||||||
| 680 | return src; | ||||||||
| 681 | } | ||||||||
| 682 | const char16_t *prevSrc = src; | ||||||||
| 683 | UChar32 c; | ||||||||
| 684 | uint16_t norm16; | ||||||||
| 685 | UCPTRIE_FAST_U16_NEXT(normTrie, UCPTRIE_16, src, limit, c, norm16)do { (c) = *(src)++; int32_t __index; if (!(((c)&0xfffff800 )==0xd800)) { __index = ((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT] + ((c) & UCPTRIE_FAST_DATA_MASK)); } else { uint16_t __c2; if ((((c)&0x400)==0) && (src ) != (limit) && (((__c2 = *(src))&0xfffffc00)==0xdc00 )) { ++(src); (c) = (((UChar32)((c))<<10UL)+(UChar32)(__c2 )-((0xd800<<10UL)+0xdc00-0x10000)); __index = ((c) >= (normTrie)->highStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77(normTrie, c)); } else { __index = (normTrie)->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET ; } } (norm16) = ((normTrie)->data.ptr16[__index]); } while (false); | ||||||||
| 686 | if (stopAtCompBoundary && norm16HasCompBoundaryBefore(norm16)) { | ||||||||
| 687 | return prevSrc; | ||||||||
| 688 | } | ||||||||
| 689 | if(!decompose(c, norm16, buffer, errorCode)) { | ||||||||
| 690 | return nullptr; | ||||||||
| 691 | } | ||||||||
| 692 | if (stopAtCompBoundary && norm16HasCompBoundaryAfter(norm16, onlyContiguous)) { | ||||||||
| 693 | return src; | ||||||||
| 694 | } | ||||||||
| 695 | } | ||||||||
| 696 | return src; | ||||||||
| 697 | } | ||||||||
| 698 | |||||||||
| 699 | UBool Normalizer2Impl::decompose(UChar32 c, uint16_t norm16, | ||||||||
| 700 | ReorderingBuffer &buffer, | ||||||||
| 701 | UErrorCode &errorCode) const { | ||||||||
| 702 | // get the decomposition and the lead and trail cc's | ||||||||
| 703 | if (norm16 >= limitNoNo) { | ||||||||
| 704 | if (isMaybeYesOrNonZeroCC(norm16)) { | ||||||||
| 705 | return buffer.append(c, getCCFromYesOrMaybeYes(norm16), errorCode); | ||||||||
| 706 | } else if (norm16 < minMaybeNo) { | ||||||||
| 707 | // Maps to an isCompYesAndZeroCC. | ||||||||
| 708 | c=mapAlgorithmic(c, norm16); | ||||||||
| 709 | norm16=getRawNorm16(c); | ||||||||
| 710 | } | ||||||||
| 711 | } | ||||||||
| 712 | if (norm16 < minYesNo) { | ||||||||
| 713 | // c does not decompose | ||||||||
| 714 | return buffer.append(c, 0, errorCode); | ||||||||
| 715 | } else if(isHangulLV(norm16) || isHangulLVT(norm16)) { | ||||||||
| 716 | // Hangul syllable: decompose algorithmically | ||||||||
| 717 | char16_t jamos[3]; | ||||||||
| 718 | return buffer.appendZeroCC(jamos, jamos+Hangul::decompose(c, jamos), errorCode); | ||||||||
| 719 | } | ||||||||
| 720 | // c decomposes, get everything from the variable-length extra data | ||||||||
| 721 | const uint16_t *mapping=getData(norm16); | ||||||||
| 722 | uint16_t firstUnit=*mapping; | ||||||||
| 723 | int32_t length=firstUnit&MAPPING_LENGTH_MASK; | ||||||||
| 724 | uint8_t leadCC, trailCC; | ||||||||
| 725 | trailCC = static_cast<uint8_t>(firstUnit >> 8); | ||||||||
| 726 | if(firstUnit&MAPPING_HAS_CCC_LCCC_WORD) { | ||||||||
| 727 | leadCC = static_cast<uint8_t>(*(mapping - 1) >> 8); | ||||||||
| 728 | } else { | ||||||||
| 729 | leadCC=0; | ||||||||
| 730 | } | ||||||||
| 731 | return buffer.append(reinterpret_cast<const char16_t*>(mapping) + 1, length, true, leadCC, trailCC, errorCode); | ||||||||
| 732 | } | ||||||||
| 733 | |||||||||
| 734 | // Dual functionality: | ||||||||
| 735 | // sink != nullptr: normalize | ||||||||
| 736 | // sink == nullptr: isNormalized/spanQuickCheckYes | ||||||||
| 737 | const uint8_t * | ||||||||
| 738 | Normalizer2Impl::decomposeUTF8(uint32_t options, | ||||||||
| 739 | const uint8_t *src, const uint8_t *limit, | ||||||||
| 740 | ByteSink *sink, Edits *edits, UErrorCode &errorCode) const { | ||||||||
| 741 | U_ASSERT(limit != nullptr)(static_cast <bool> (limit != nullptr) ? void (0) : __assert_fail ("limit != nullptr", __builtin_FILE (), __builtin_LINE (), __extension__ __PRETTY_FUNCTION__)); | ||||||||
| 742 | UnicodeString s16; | ||||||||
| 743 | uint8_t minNoLead = leadByteForCP(minDecompNoCP); | ||||||||
| 744 | |||||||||
| 745 | const uint8_t *prevBoundary = src; | ||||||||
| 746 | // only for quick check | ||||||||
| 747 | uint8_t prevCC = 0; | ||||||||
| 748 | |||||||||
| 749 | for (;;) { | ||||||||
| 750 | // Fast path: Scan over a sequence of characters below the minimum "no" code point, | ||||||||
| 751 | // or with (decompYes && ccc==0) properties. | ||||||||
| 752 | const uint8_t *fastStart = src; | ||||||||
| 753 | const uint8_t *prevSrc; | ||||||||
| 754 | uint16_t norm16 = 0; | ||||||||
| 755 | |||||||||
| 756 | for (;;) { | ||||||||
| 757 | if (src == limit) { | ||||||||
| 758 | if (prevBoundary != limit && sink != nullptr) { | ||||||||
| 759 | ByteSinkUtil::appendUnchanged(prevBoundary, limit, | ||||||||
| 760 | *sink, options, edits, errorCode); | ||||||||
| 761 | } | ||||||||
| 762 | return src; | ||||||||
| 763 | } | ||||||||
| 764 | if (*src < minNoLead) { | ||||||||
| 765 | ++src; | ||||||||
| 766 | } else { | ||||||||
| 767 | prevSrc = src; | ||||||||
| 768 | UCPTRIE_FAST_U8_NEXT(normTrie, UCPTRIE_16, src, limit, norm16)do { int32_t __lead = (uint8_t)*(src)++; if (!(((__lead)& 0x80)==0)) { uint8_t __t1, __t2, __t3; if ((src) != (limit) && (__lead >= 0xe0 ? __lead < 0xf0 ? "\x20\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x10\x30\x30" [__lead &= 0xf] & (1 << ((__t1 = *(src)) >> 5)) && ++(src) != (limit) && (__t2 = *(src) - 0x80) <= 0x3f && (__lead = ((int32_t)(normTrie)-> index[(__lead << 6) + (__t1 & 0x3f)]) + __t2, 1) : ( __lead -= 0xf0) <= 4 && "\x00\x00\x00\x00\x00\x00\x00\x00\x1E\x0F\x0F\x0F\x00\x00\x00\x00" [(__t1 = *(src)) >> 4] & (1 << __lead) && (__lead = (__lead << 6) | (__t1 & 0x3f), ++(src) != (limit)) && (__t2 = *(src) - 0x80) <= 0x3f && ++(src) != (limit) && (__t3 = *(src) - 0x80) <= 0x3f && (__lead = __lead >= (normTrie)->shifted12HighStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallU8Index_77((normTrie), __lead, __t2, __t3 ), 1) : __lead >= 0xc2 && (__t1 = *(src) - 0x80) <= 0x3f && (__lead = (int32_t)(normTrie)->index[__lead & 0x1f] + __t1, 1))) { ++(src); } else { __lead = (normTrie )->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET; } } ( norm16) = ((normTrie)->data.ptr16[__lead]); } while (false ); | ||||||||
| 769 | if (!isMostDecompYesAndZeroCC(norm16)) { | ||||||||
| 770 | break; | ||||||||
| 771 | } | ||||||||
| 772 | } | ||||||||
| 773 | } | ||||||||
| 774 | // isMostDecompYesAndZeroCC(norm16) is false, that is, norm16>=minYesNo, | ||||||||
| 775 | // and the current character at [prevSrc..src[ is not a common case with cc=0 | ||||||||
| 776 | // (MIN_NORMAL_MAYBE_YES or JAMO_VT). | ||||||||
| 777 | // It could still be a maybeYes with cc=0. | ||||||||
| 778 | if (prevSrc != fastStart) { | ||||||||
| 779 | // The fast path looped over yes/0 characters before the current one. | ||||||||
| 780 | if (sink != nullptr && | ||||||||
| 781 | !ByteSinkUtil::appendUnchanged(prevBoundary, prevSrc, | ||||||||
| 782 | *sink, options, edits, errorCode)) { | ||||||||
| 783 | break; | ||||||||
| 784 | } | ||||||||
| 785 | prevBoundary = prevSrc; | ||||||||
| 786 | prevCC = 0; | ||||||||
| 787 | } | ||||||||
| 788 | |||||||||
| 789 | // Medium-fast path: Quick check. | ||||||||
| 790 | if (isMaybeYesOrNonZeroCC(norm16)) { | ||||||||
| 791 | // Does not decompose. | ||||||||
| 792 | uint8_t cc = getCCFromYesOrMaybeYes(norm16); | ||||||||
| 793 | if (prevCC <= cc || cc == 0) { | ||||||||
| 794 | prevCC = cc; | ||||||||
| 795 | if (cc <= 1) { | ||||||||
| 796 | if (sink != nullptr && | ||||||||
| 797 | !ByteSinkUtil::appendUnchanged(prevBoundary, src, | ||||||||
| 798 | *sink, options, edits, errorCode)) { | ||||||||
| 799 | break; | ||||||||
| 800 | } | ||||||||
| 801 | prevBoundary = src; | ||||||||
| 802 | } | ||||||||
| 803 | continue; | ||||||||
| 804 | } | ||||||||
| 805 | } | ||||||||
| 806 | if (sink == nullptr) { | ||||||||
| 807 | return prevBoundary; // quick check: "no" or cc out of order | ||||||||
| 808 | } | ||||||||
| 809 | |||||||||
| 810 | // Slow path | ||||||||
| 811 | // Decompose up to and including the current character. | ||||||||
| 812 | if (prevBoundary != prevSrc && norm16HasDecompBoundaryBefore(norm16)) { | ||||||||
| 813 | if (!ByteSinkUtil::appendUnchanged(prevBoundary, prevSrc, | ||||||||
| 814 | *sink, options, edits, errorCode)) { | ||||||||
| 815 | break; | ||||||||
| 816 | } | ||||||||
| 817 | prevBoundary = prevSrc; | ||||||||
| 818 | } | ||||||||
| 819 | ReorderingBuffer buffer(*this, s16, errorCode); | ||||||||
| 820 | if (U_FAILURE(errorCode)) { | ||||||||
| 821 | break; | ||||||||
| 822 | } | ||||||||
| 823 | decomposeShort(prevBoundary, src, STOP_AT_LIMIT, false /* onlyContiguous */, | ||||||||
| 824 | buffer, errorCode); | ||||||||
| 825 | // Decompose until the next boundary. | ||||||||
| 826 | if (buffer.getLastCC() > 1) { | ||||||||
| 827 | src = decomposeShort(src, limit, STOP_AT_DECOMP_BOUNDARY, false /* onlyContiguous */, | ||||||||
| 828 | buffer, errorCode); | ||||||||
| 829 | } | ||||||||
| 830 | if (U_FAILURE(errorCode)) { | ||||||||
| 831 | break; | ||||||||
| 832 | } | ||||||||
| 833 | if ((src - prevSrc) > INT32_MAX(2147483647)) { // guard before buffer.equals() | ||||||||
| 834 | errorCode = U_INDEX_OUTOFBOUNDS_ERROR; | ||||||||
| 835 | break; | ||||||||
| 836 | } | ||||||||
| 837 | // We already know there was a change if the original character decomposed; | ||||||||
| 838 | // otherwise compare. | ||||||||
| 839 | if (isMaybeYesOrNonZeroCC(norm16) && buffer.equals(prevBoundary, src)) { | ||||||||
| 840 | if (!ByteSinkUtil::appendUnchanged(prevBoundary, src, | ||||||||
| 841 | *sink, options, edits, errorCode)) { | ||||||||
| 842 | break; | ||||||||
| 843 | } | ||||||||
| 844 | } else { | ||||||||
| 845 | if (!ByteSinkUtil::appendChange(prevBoundary, src, buffer.getStart(), buffer.length(), | ||||||||
| 846 | *sink, edits, errorCode)) { | ||||||||
| 847 | break; | ||||||||
| 848 | } | ||||||||
| 849 | } | ||||||||
| 850 | prevBoundary = src; | ||||||||
| 851 | prevCC = 0; | ||||||||
| 852 | } | ||||||||
| 853 | return src; | ||||||||
| 854 | } | ||||||||
| 855 | |||||||||
| 856 | const uint8_t * | ||||||||
| 857 | Normalizer2Impl::decomposeShort(const uint8_t *src, const uint8_t *limit, | ||||||||
| 858 | StopAt stopAt, UBool onlyContiguous, | ||||||||
| 859 | ReorderingBuffer &buffer, UErrorCode &errorCode) const { | ||||||||
| 860 | if (U_FAILURE(errorCode)) { | ||||||||
| 861 | return nullptr; | ||||||||
| 862 | } | ||||||||
| 863 | while (src < limit) { | ||||||||
| 864 | const uint8_t *prevSrc = src; | ||||||||
| 865 | uint16_t norm16; | ||||||||
| 866 | UCPTRIE_FAST_U8_NEXT(normTrie, UCPTRIE_16, src, limit, norm16)do { int32_t __lead = (uint8_t)*(src)++; if (!(((__lead)& 0x80)==0)) { uint8_t __t1, __t2, __t3; if ((src) != (limit) && (__lead >= 0xe0 ? __lead < 0xf0 ? "\x20\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x10\x30\x30" [__lead &= 0xf] & (1 << ((__t1 = *(src)) >> 5)) && ++(src) != (limit) && (__t2 = *(src) - 0x80) <= 0x3f && (__lead = ((int32_t)(normTrie)-> index[(__lead << 6) + (__t1 & 0x3f)]) + __t2, 1) : ( __lead -= 0xf0) <= 4 && "\x00\x00\x00\x00\x00\x00\x00\x00\x1E\x0F\x0F\x0F\x00\x00\x00\x00" [(__t1 = *(src)) >> 4] & (1 << __lead) && (__lead = (__lead << 6) | (__t1 & 0x3f), ++(src) != (limit)) && (__t2 = *(src) - 0x80) <= 0x3f && ++(src) != (limit) && (__t3 = *(src) - 0x80) <= 0x3f && (__lead = __lead >= (normTrie)->shifted12HighStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallU8Index_77((normTrie), __lead, __t2, __t3 ), 1) : __lead >= 0xc2 && (__t1 = *(src) - 0x80) <= 0x3f && (__lead = (int32_t)(normTrie)->index[__lead & 0x1f] + __t1, 1))) { ++(src); } else { __lead = (normTrie )->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET; } } ( norm16) = ((normTrie)->data.ptr16[__lead]); } while (false ); | ||||||||
| 867 | // Get the decomposition and the lead and trail cc's. | ||||||||
| 868 | UChar32 c = U_SENTINEL(-1); | ||||||||
| 869 | if (norm16 >= limitNoNo) { | ||||||||
| 870 | if (isMaybeYesOrNonZeroCC(norm16)) { | ||||||||
| 871 | // No comp boundaries around this character. | ||||||||
| 872 | uint8_t cc = getCCFromYesOrMaybeYes(norm16); | ||||||||
| 873 | if (cc == 0 && stopAt == STOP_AT_DECOMP_BOUNDARY) { | ||||||||
| 874 | return prevSrc; | ||||||||
| 875 | } | ||||||||
| 876 | c = codePointFromValidUTF8(prevSrc, src); | ||||||||
| 877 | if (!buffer.append(c, cc, errorCode)) { | ||||||||
| 878 | return nullptr; | ||||||||
| 879 | } | ||||||||
| 880 | if (stopAt == STOP_AT_DECOMP_BOUNDARY && buffer.getLastCC() <= 1) { | ||||||||
| 881 | return src; | ||||||||
| 882 | } | ||||||||
| 883 | continue; | ||||||||
| 884 | } else if (norm16 < minMaybeNo) { | ||||||||
| 885 | // Maps to an isCompYesAndZeroCC. | ||||||||
| 886 | if (stopAt != STOP_AT_LIMIT) { | ||||||||
| 887 | return prevSrc; | ||||||||
| 888 | } | ||||||||
| 889 | c = codePointFromValidUTF8(prevSrc, src); | ||||||||
| 890 | c = mapAlgorithmic(c, norm16); | ||||||||
| 891 | norm16 = getRawNorm16(c); | ||||||||
| 892 | } | ||||||||
| 893 | } else if (stopAt != STOP_AT_LIMIT && norm16 < minNoNoCompNoMaybeCC) { | ||||||||
| 894 | return prevSrc; | ||||||||
| 895 | } | ||||||||
| 896 | // norm16!=INERT guarantees that [prevSrc, src[ is valid UTF-8. | ||||||||
| 897 | // We do not see invalid UTF-8 here because | ||||||||
| 898 | // its norm16==INERT is normalization-inert, | ||||||||
| 899 | // so it gets copied unchanged in the fast path, | ||||||||
| 900 | // and we stop the slow path where invalid UTF-8 begins. | ||||||||
| 901 | // c >= 0 is the result of an algorithmic mapping. | ||||||||
| 902 | U_ASSERT(c >= 0 || norm16 != INERT)(static_cast <bool> (c >= 0 || norm16 != INERT) ? void (0) : __assert_fail ("c >= 0 || norm16 != INERT", __builtin_FILE (), __builtin_LINE (), __extension__ __PRETTY_FUNCTION__)); | ||||||||
| 903 | if (norm16 < minYesNo) { | ||||||||
| 904 | if (c < 0) { | ||||||||
| 905 | c = codePointFromValidUTF8(prevSrc, src); | ||||||||
| 906 | } | ||||||||
| 907 | // does not decompose | ||||||||
| 908 | if (!buffer.append(c, 0, errorCode)) { | ||||||||
| 909 | return nullptr; | ||||||||
| 910 | } | ||||||||
| 911 | } else if (isHangulLV(norm16) || isHangulLVT(norm16)) { | ||||||||
| 912 | // Hangul syllable: decompose algorithmically | ||||||||
| 913 | if (c < 0) { | ||||||||
| 914 | c = codePointFromValidUTF8(prevSrc, src); | ||||||||
| 915 | } | ||||||||
| 916 | char16_t jamos[3]; | ||||||||
| 917 | if (!buffer.appendZeroCC(jamos, jamos+Hangul::decompose(c, jamos), errorCode)) { | ||||||||
| 918 | return nullptr; | ||||||||
| 919 | } | ||||||||
| 920 | } else { | ||||||||
| 921 | // The character decomposes, get everything from the variable-length extra data. | ||||||||
| 922 | const uint16_t *mapping = getData(norm16); | ||||||||
| 923 | uint16_t firstUnit = *mapping; | ||||||||
| 924 | int32_t length = firstUnit & MAPPING_LENGTH_MASK; | ||||||||
| 925 | uint8_t trailCC = static_cast<uint8_t>(firstUnit >> 8); | ||||||||
| 926 | uint8_t leadCC; | ||||||||
| 927 | if (firstUnit & MAPPING_HAS_CCC_LCCC_WORD) { | ||||||||
| 928 | leadCC = static_cast<uint8_t>(*(mapping - 1) >> 8); | ||||||||
| 929 | } else { | ||||||||
| 930 | leadCC = 0; | ||||||||
| 931 | } | ||||||||
| 932 | if (leadCC == 0 && stopAt == STOP_AT_DECOMP_BOUNDARY) { | ||||||||
| 933 | return prevSrc; | ||||||||
| 934 | } | ||||||||
| 935 | if (!buffer.append(reinterpret_cast<const char16_t*>(mapping) + 1, length, true, leadCC, trailCC, errorCode)) { | ||||||||
| 936 | return nullptr; | ||||||||
| 937 | } | ||||||||
| 938 | } | ||||||||
| 939 | if ((stopAt == STOP_AT_COMP_BOUNDARY && norm16HasCompBoundaryAfter(norm16, onlyContiguous)) || | ||||||||
| 940 | (stopAt == STOP_AT_DECOMP_BOUNDARY && buffer.getLastCC() <= 1)) { | ||||||||
| 941 | return src; | ||||||||
| 942 | } | ||||||||
| 943 | } | ||||||||
| 944 | return src; | ||||||||
| 945 | } | ||||||||
| 946 | |||||||||
| 947 | const char16_t * | ||||||||
| 948 | Normalizer2Impl::getDecomposition(UChar32 c, char16_t buffer[4], int32_t &length) const { | ||||||||
| 949 | uint16_t norm16; | ||||||||
| 950 | if(c<minDecompNoCP || isMaybeYesOrNonZeroCC(norm16=getNorm16(c))) { | ||||||||
| 951 | // c does not decompose | ||||||||
| 952 | return nullptr; | ||||||||
| 953 | } | ||||||||
| 954 | const char16_t *decomp = nullptr; | ||||||||
| 955 | if(isDecompNoAlgorithmic(norm16)) { | ||||||||
| 956 | // Maps to an isCompYesAndZeroCC. | ||||||||
| 957 | c=mapAlgorithmic(c, norm16); | ||||||||
| 958 | decomp=buffer; | ||||||||
| 959 | length=0; | ||||||||
| 960 | U16_APPEND_UNSAFE(buffer, length, c)do { if((uint32_t)(c)<=0xffff) { (buffer)[(length)++]=(uint16_t )(c); } else { (buffer)[(length)++]=(uint16_t)(((c)>>10 )+0xd7c0); (buffer)[(length)++]=(uint16_t)(((c)&0x3ff)|0xdc00 ); } } while (false); | ||||||||
| 961 | // The mapping might decompose further. | ||||||||
| 962 | norm16 = getRawNorm16(c); | ||||||||
| 963 | } | ||||||||
| 964 | if (norm16 < minYesNo) { | ||||||||
| 965 | return decomp; | ||||||||
| 966 | } else if(isHangulLV(norm16) || isHangulLVT(norm16)) { | ||||||||
| 967 | // Hangul syllable: decompose algorithmically | ||||||||
| 968 | length=Hangul::decompose(c, buffer); | ||||||||
| 969 | return buffer; | ||||||||
| 970 | } | ||||||||
| 971 | // c decomposes, get everything from the variable-length extra data | ||||||||
| 972 | const uint16_t *mapping=getData(norm16); | ||||||||
| 973 | length=*mapping&MAPPING_LENGTH_MASK; | ||||||||
| 974 | return reinterpret_cast<const char16_t*>(mapping) + 1; | ||||||||
| 975 | } | ||||||||
| 976 | |||||||||
| 977 | // The capacity of the buffer must be 30=MAPPING_LENGTH_MASK-1 | ||||||||
| 978 | // so that a raw mapping fits that consists of one unit ("rm0") | ||||||||
| 979 | // plus all but the first two code units of the normal mapping. | ||||||||
| 980 | // The maximum length of a normal mapping is 31=MAPPING_LENGTH_MASK. | ||||||||
| 981 | const char16_t * | ||||||||
| 982 | Normalizer2Impl::getRawDecomposition(UChar32 c, char16_t buffer[30], int32_t &length) const { | ||||||||
| 983 | uint16_t norm16; | ||||||||
| 984 | if(c<minDecompNoCP || isDecompYes(norm16=getNorm16(c))) { | ||||||||
| 985 | // c does not decompose | ||||||||
| 986 | return nullptr; | ||||||||
| 987 | } else if(isHangulLV(norm16) || isHangulLVT(norm16)) { | ||||||||
| 988 | // Hangul syllable: decompose algorithmically | ||||||||
| 989 | Hangul::getRawDecomposition(c, buffer); | ||||||||
| 990 | length=2; | ||||||||
| 991 | return buffer; | ||||||||
| 992 | } else if(isDecompNoAlgorithmic(norm16)) { | ||||||||
| 993 | c=mapAlgorithmic(c, norm16); | ||||||||
| 994 | length=0; | ||||||||
| 995 | U16_APPEND_UNSAFE(buffer, length, c)do { if((uint32_t)(c)<=0xffff) { (buffer)[(length)++]=(uint16_t )(c); } else { (buffer)[(length)++]=(uint16_t)(((c)>>10 )+0xd7c0); (buffer)[(length)++]=(uint16_t)(((c)&0x3ff)|0xdc00 ); } } while (false); | ||||||||
| 996 | return buffer; | ||||||||
| 997 | } | ||||||||
| 998 | // c decomposes, get everything from the variable-length extra data | ||||||||
| 999 | const uint16_t *mapping=getData(norm16); | ||||||||
| 1000 | uint16_t firstUnit=*mapping; | ||||||||
| 1001 | int32_t mLength=firstUnit&MAPPING_LENGTH_MASK; // length of normal mapping | ||||||||
| 1002 | if(firstUnit&MAPPING_HAS_RAW_MAPPING) { | ||||||||
| 1003 | // Read the raw mapping from before the firstUnit and before the optional ccc/lccc word. | ||||||||
| 1004 | // Bit 7=MAPPING_HAS_CCC_LCCC_WORD | ||||||||
| 1005 | const uint16_t *rawMapping=mapping-((firstUnit>>7)&1)-1; | ||||||||
| 1006 | uint16_t rm0=*rawMapping; | ||||||||
| 1007 | if(rm0<=MAPPING_LENGTH_MASK) { | ||||||||
| 1008 | length=rm0; | ||||||||
| 1009 | return reinterpret_cast<const char16_t*>(rawMapping) - rm0; | ||||||||
| 1010 | } else { | ||||||||
| 1011 | // Copy the normal mapping and replace its first two code units with rm0. | ||||||||
| 1012 | buffer[0] = static_cast<char16_t>(rm0); | ||||||||
| 1013 | u_memcpyu_memcpy_77(buffer + 1, reinterpret_cast<const char16_t*>(mapping) + 1 + 2, mLength - 2); | ||||||||
| 1014 | length=mLength-1; | ||||||||
| 1015 | return buffer; | ||||||||
| 1016 | } | ||||||||
| 1017 | } else { | ||||||||
| 1018 | length=mLength; | ||||||||
| 1019 | return reinterpret_cast<const char16_t*>(mapping) + 1; | ||||||||
| 1020 | } | ||||||||
| 1021 | } | ||||||||
| 1022 | |||||||||
| 1023 | void Normalizer2Impl::decomposeAndAppend(const char16_t *src, const char16_t *limit, | ||||||||
| 1024 | UBool doDecompose, | ||||||||
| 1025 | UnicodeString &safeMiddle, | ||||||||
| 1026 | ReorderingBuffer &buffer, | ||||||||
| 1027 | UErrorCode &errorCode) const { | ||||||||
| 1028 | buffer.copyReorderableSuffixTo(safeMiddle); | ||||||||
| 1029 | if(doDecompose) { | ||||||||
| 1030 | decompose(src, limit, &buffer, errorCode); | ||||||||
| 1031 | return; | ||||||||
| 1032 | } | ||||||||
| 1033 | // Just merge the strings at the boundary. | ||||||||
| 1034 | bool isFirst = true; | ||||||||
| 1035 | uint8_t firstCC = 0, prevCC = 0, cc; | ||||||||
| 1036 | const char16_t *p = src; | ||||||||
| 1037 | while (p != limit) { | ||||||||
| 1038 | const char16_t *codePointStart = p; | ||||||||
| 1039 | UChar32 c; | ||||||||
| 1040 | uint16_t norm16; | ||||||||
| 1041 | UCPTRIE_FAST_U16_NEXT(normTrie, UCPTRIE_16, p, limit, c, norm16)do { (c) = *(p)++; int32_t __index; if (!(((c)&0xfffff800 )==0xd800)) { __index = ((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT] + ((c) & UCPTRIE_FAST_DATA_MASK)); } else { uint16_t __c2; if ((((c)&0x400)==0) && (p ) != (limit) && (((__c2 = *(p))&0xfffffc00)==0xdc00 )) { ++(p); (c) = (((UChar32)((c))<<10UL)+(UChar32)(__c2 )-((0xd800<<10UL)+0xdc00-0x10000)); __index = ((c) >= (normTrie)->highStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77(normTrie, c)); } else { __index = (normTrie)->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET ; } } (norm16) = ((normTrie)->data.ptr16[__index]); } while (false); | ||||||||
| 1042 | if ((cc = getCC(norm16)) == 0) { | ||||||||
| 1043 | p = codePointStart; | ||||||||
| 1044 | break; | ||||||||
| 1045 | } | ||||||||
| 1046 | if (isFirst) { | ||||||||
| 1047 | firstCC = cc; | ||||||||
| 1048 | isFirst = false; | ||||||||
| 1049 | } | ||||||||
| 1050 | prevCC = cc; | ||||||||
| 1051 | } | ||||||||
| 1052 | if(limit==nullptr) { // appendZeroCC() needs limit!=nullptr | ||||||||
| 1053 | limit=u_strchru_strchr_77(p, 0); | ||||||||
| 1054 | } | ||||||||
| 1055 | |||||||||
| 1056 | if (buffer.append(src, static_cast<int32_t>(p - src), false, firstCC, prevCC, errorCode)) { | ||||||||
| 1057 | buffer.appendZeroCC(p, limit, errorCode); | ||||||||
| 1058 | } | ||||||||
| 1059 | } | ||||||||
| 1060 | |||||||||
| 1061 | UBool Normalizer2Impl::hasDecompBoundaryBefore(UChar32 c) const { | ||||||||
| 1062 | return c < minLcccCP || (c <= 0xffff && !singleLeadMightHaveNonZeroFCD16(c)) || | ||||||||
| 1063 | norm16HasDecompBoundaryBefore(getNorm16(c)); | ||||||||
| 1064 | } | ||||||||
| 1065 | |||||||||
| 1066 | UBool Normalizer2Impl::norm16HasDecompBoundaryBefore(uint16_t norm16) const { | ||||||||
| 1067 | if (norm16 < minNoNoCompNoMaybeCC) { | ||||||||
| 1068 | return true; | ||||||||
| 1069 | } | ||||||||
| 1070 | if (norm16 >= limitNoNo) { | ||||||||
| 1071 | return norm16 <= MIN_NORMAL_MAYBE_YES || norm16 == JAMO_VT; | ||||||||
| 1072 | } | ||||||||
| 1073 | // c decomposes, get everything from the variable-length extra data | ||||||||
| 1074 | const uint16_t *mapping=getDataForYesOrNo(norm16); | ||||||||
| 1075 | uint16_t firstUnit=*mapping; | ||||||||
| 1076 | // true if leadCC==0 (hasFCDBoundaryBefore()) | ||||||||
| 1077 | return (firstUnit&MAPPING_HAS_CCC_LCCC_WORD)==0 || (*(mapping-1)&0xff00)==0; | ||||||||
| 1078 | } | ||||||||
| 1079 | |||||||||
| 1080 | UBool Normalizer2Impl::hasDecompBoundaryAfter(UChar32 c) const { | ||||||||
| 1081 | if (c < minDecompNoCP) { | ||||||||
| 1082 | return true; | ||||||||
| 1083 | } | ||||||||
| 1084 | if (c <= 0xffff && !singleLeadMightHaveNonZeroFCD16(c)) { | ||||||||
| 1085 | return true; | ||||||||
| 1086 | } | ||||||||
| 1087 | return norm16HasDecompBoundaryAfter(getNorm16(c)); | ||||||||
| 1088 | } | ||||||||
| 1089 | |||||||||
| 1090 | UBool Normalizer2Impl::norm16HasDecompBoundaryAfter(uint16_t norm16) const { | ||||||||
| 1091 | if(norm16 <= minYesNo || isHangulLVT(norm16)) { | ||||||||
| 1092 | return true; | ||||||||
| 1093 | } | ||||||||
| 1094 | if (norm16 >= limitNoNo) { | ||||||||
| 1095 | if (isMaybeYesOrNonZeroCC(norm16)) { | ||||||||
| 1096 | return norm16 <= MIN_NORMAL_MAYBE_YES || norm16 == JAMO_VT; | ||||||||
| 1097 | } else if (norm16 < minMaybeNo) { | ||||||||
| 1098 | // Maps to an isCompYesAndZeroCC. | ||||||||
| 1099 | return (norm16 & DELTA_TCCC_MASK) <= DELTA_TCCC_1; | ||||||||
| 1100 | } | ||||||||
| 1101 | } | ||||||||
| 1102 | // c decomposes, get everything from the variable-length extra data | ||||||||
| 1103 | const uint16_t *mapping=getData(norm16); | ||||||||
| 1104 | uint16_t firstUnit=*mapping; | ||||||||
| 1105 | // decomp after-boundary: same as hasFCDBoundaryAfter(), | ||||||||
| 1106 | // fcd16<=1 || trailCC==0 | ||||||||
| 1107 | if(firstUnit>0x1ff) { | ||||||||
| 1108 | return false; // trailCC>1 | ||||||||
| 1109 | } | ||||||||
| 1110 | if(firstUnit<=0xff) { | ||||||||
| 1111 | return true; // trailCC==0 | ||||||||
| 1112 | } | ||||||||
| 1113 | // if(trailCC==1) test leadCC==0, same as checking for before-boundary | ||||||||
| 1114 | // true if leadCC==0 (hasFCDBoundaryBefore()) | ||||||||
| 1115 | return (firstUnit&MAPPING_HAS_CCC_LCCC_WORD)==0 || (*(mapping-1)&0xff00)==0; | ||||||||
| 1116 | } | ||||||||
| 1117 | |||||||||
| 1118 | /* | ||||||||
| 1119 | * Finds the recomposition result for | ||||||||
| 1120 | * a forward-combining "lead" character, | ||||||||
| 1121 | * specified with a pointer to its compositions list, | ||||||||
| 1122 | * and a backward-combining "trail" character. | ||||||||
| 1123 | * | ||||||||
| 1124 | * If the lead and trail characters combine, then this function returns | ||||||||
| 1125 | * the following "compositeAndFwd" value: | ||||||||
| 1126 | * Bits 21..1 composite character | ||||||||
| 1127 | * Bit 0 set if the composite is a forward-combining starter | ||||||||
| 1128 | * otherwise it returns -1. | ||||||||
| 1129 | * | ||||||||
| 1130 | * The compositions list has (trail, compositeAndFwd) pair entries, | ||||||||
| 1131 | * encoded as either pairs or triples of 16-bit units. | ||||||||
| 1132 | * The last entry has the high bit of its first unit set. | ||||||||
| 1133 | * | ||||||||
| 1134 | * The list is sorted by ascending trail characters (there are no duplicates). | ||||||||
| 1135 | * A linear search is used. | ||||||||
| 1136 | * | ||||||||
| 1137 | * See normalizer2impl.h for a more detailed description | ||||||||
| 1138 | * of the compositions list format. | ||||||||
| 1139 | */ | ||||||||
| 1140 | int32_t Normalizer2Impl::combine(const uint16_t *list, UChar32 trail) { | ||||||||
| 1141 | uint16_t key1, firstUnit; | ||||||||
| 1142 | if(trail<COMP_1_TRAIL_LIMIT) { | ||||||||
| 1143 | // trail character is 0..33FF | ||||||||
| 1144 | // result entry may have 2 or 3 units | ||||||||
| 1145 | key1 = static_cast<uint16_t>(trail << 1); | ||||||||
| 1146 | while(key1>(firstUnit=*list)) { | ||||||||
| 1147 | list+=2+(firstUnit&COMP_1_TRIPLE); | ||||||||
| 1148 | } | ||||||||
| 1149 | if(key1==(firstUnit&COMP_1_TRAIL_MASK)) { | ||||||||
| 1150 | if(firstUnit&COMP_1_TRIPLE) { | ||||||||
| 1151 | return (static_cast<int32_t>(list[1]) << 16) | list[2]; | ||||||||
| 1152 | } else { | ||||||||
| 1153 | return list[1]; | ||||||||
| 1154 | } | ||||||||
| 1155 | } | ||||||||
| 1156 | } else { | ||||||||
| 1157 | // trail character is 3400..10FFFF | ||||||||
| 1158 | // result entry has 3 units | ||||||||
| 1159 | key1 = static_cast<uint16_t>(COMP_1_TRAIL_LIMIT + | ||||||||
| 1160 | (((trail>>COMP_1_TRAIL_SHIFT))& | ||||||||
| 1161 | ~COMP_1_TRIPLE)); | ||||||||
| 1162 | uint16_t key2 = static_cast<uint16_t>(trail << COMP_2_TRAIL_SHIFT); | ||||||||
| 1163 | uint16_t secondUnit; | ||||||||
| 1164 | for(;;) { | ||||||||
| 1165 | if(key1>(firstUnit=*list)) { | ||||||||
| 1166 | list+=2+(firstUnit&COMP_1_TRIPLE); | ||||||||
| 1167 | } else if(key1==(firstUnit&COMP_1_TRAIL_MASK)) { | ||||||||
| 1168 | if(key2>(secondUnit=list[1])) { | ||||||||
| 1169 | if(firstUnit&COMP_1_LAST_TUPLE) { | ||||||||
| 1170 | break; | ||||||||
| 1171 | } else { | ||||||||
| 1172 | list+=3; | ||||||||
| 1173 | } | ||||||||
| 1174 | } else if(key2==(secondUnit&COMP_2_TRAIL_MASK)) { | ||||||||
| 1175 | return (static_cast<int32_t>(secondUnit & ~COMP_2_TRAIL_MASK) << 16) | list[2]; | ||||||||
| 1176 | } else { | ||||||||
| 1177 | break; | ||||||||
| 1178 | } | ||||||||
| 1179 | } else { | ||||||||
| 1180 | break; | ||||||||
| 1181 | } | ||||||||
| 1182 | } | ||||||||
| 1183 | } | ||||||||
| 1184 | return -1; | ||||||||
| 1185 | } | ||||||||
| 1186 | |||||||||
| 1187 | /** | ||||||||
| 1188 | * @param list some character's compositions list | ||||||||
| 1189 | * @param set recursively receives the composites from these compositions | ||||||||
| 1190 | */ | ||||||||
| 1191 | void Normalizer2Impl::addComposites(const uint16_t *list, UnicodeSet &set) const { | ||||||||
| 1192 | uint16_t firstUnit; | ||||||||
| 1193 | int32_t compositeAndFwd; | ||||||||
| 1194 | do { | ||||||||
| 1195 | firstUnit=*list; | ||||||||
| 1196 | if((firstUnit&COMP_1_TRIPLE)==0) { | ||||||||
| 1197 | compositeAndFwd=list[1]; | ||||||||
| 1198 | list+=2; | ||||||||
| 1199 | } else { | ||||||||
| 1200 | compositeAndFwd = ((static_cast<int32_t>(list[1]) & ~COMP_2_TRAIL_MASK) << 16) | list[2]; | ||||||||
| 1201 | list+=3; | ||||||||
| 1202 | } | ||||||||
| 1203 | UChar32 composite=compositeAndFwd>>1; | ||||||||
| 1204 | if((compositeAndFwd&1)!=0) { | ||||||||
| 1205 | addComposites(getCompositionsListForComposite(getRawNorm16(composite)), set); | ||||||||
| 1206 | } | ||||||||
| 1207 | set.add(composite); | ||||||||
| 1208 | } while((firstUnit&COMP_1_LAST_TUPLE)==0); | ||||||||
| 1209 | } | ||||||||
| 1210 | |||||||||
| 1211 | /* | ||||||||
| 1212 | * Recomposes the buffer text starting at recomposeStartIndex | ||||||||
| 1213 | * (which is in NFD - decomposed and canonically ordered), | ||||||||
| 1214 | * and truncates the buffer contents. | ||||||||
| 1215 | * | ||||||||
| 1216 | * Note that recomposition never lengthens the text: | ||||||||
| 1217 | * Any character consists of either one or two code units; | ||||||||
| 1218 | * a composition may contain at most one more code unit than the original starter, | ||||||||
| 1219 | * while the combining mark that is removed has at least one code unit. | ||||||||
| 1220 | */ | ||||||||
| 1221 | void Normalizer2Impl::recompose(ReorderingBuffer &buffer, int32_t recomposeStartIndex, | ||||||||
| 1222 | UBool onlyContiguous) const { | ||||||||
| 1223 | char16_t *p=buffer.getStart()+recomposeStartIndex; | ||||||||
| 1224 | char16_t *limit=buffer.getLimit(); | ||||||||
| 1225 | if(p==limit) { | ||||||||
| 1226 | return; | ||||||||
| 1227 | } | ||||||||
| 1228 | |||||||||
| 1229 | char16_t *starter, *pRemove, *q, *r; | ||||||||
| 1230 | const uint16_t *compositionsList; | ||||||||
| 1231 | UChar32 c, compositeAndFwd; | ||||||||
| 1232 | uint16_t norm16; | ||||||||
| 1233 | uint8_t cc, prevCC; | ||||||||
| 1234 | UBool starterIsSupplementary; | ||||||||
| 1235 | |||||||||
| 1236 | // Some of the following variables are not used until we have a forward-combining starter | ||||||||
| 1237 | // and are only initialized now to avoid compiler warnings. | ||||||||
| 1238 | compositionsList=nullptr; // used as indicator for whether we have a forward-combining starter | ||||||||
| 1239 | starter=nullptr; | ||||||||
| 1240 | starterIsSupplementary=false; | ||||||||
| 1241 | prevCC=0; | ||||||||
| 1242 | |||||||||
| 1243 | for(;;) { | ||||||||
| 1244 | UCPTRIE_FAST_U16_NEXT(normTrie, UCPTRIE_16, p, limit, c, norm16)do { (c) = *(p)++; int32_t __index; if (!(((c)&0xfffff800 )==0xd800)) { __index = ((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT] + ((c) & UCPTRIE_FAST_DATA_MASK)); } else { uint16_t __c2; if ((((c)&0x400)==0) && (p ) != (limit) && (((__c2 = *(p))&0xfffffc00)==0xdc00 )) { ++(p); (c) = (((UChar32)((c))<<10UL)+(UChar32)(__c2 )-((0xd800<<10UL)+0xdc00-0x10000)); __index = ((c) >= (normTrie)->highStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77(normTrie, c)); } else { __index = (normTrie)->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET ; } } (norm16) = ((normTrie)->data.ptr16[__index]); } while (false); | ||||||||
| 1245 | cc=getCCFromYesOrMaybeYes(norm16); | ||||||||
| 1246 | if( // this character combines backward and | ||||||||
| 1247 | isMaybe(norm16) && | ||||||||
| 1248 | // we have seen a starter that combines forward and | ||||||||
| 1249 | compositionsList!=nullptr && | ||||||||
| 1250 | // the backward-combining character is not blocked | ||||||||
| 1251 | (prevCC<cc || prevCC==0) | ||||||||
| 1252 | ) { | ||||||||
| 1253 | if(isJamoVT(norm16)) { | ||||||||
| 1254 | // c is a Jamo V/T, see if we can compose it with the previous character. | ||||||||
| 1255 | if(c<Hangul::JAMO_T_BASE) { | ||||||||
| 1256 | // c is a Jamo Vowel, compose with previous Jamo L and following Jamo T. | ||||||||
| 1257 | char16_t prev = static_cast<char16_t>(*starter - Hangul::JAMO_L_BASE); | ||||||||
| 1258 | if(prev<Hangul::JAMO_L_COUNT) { | ||||||||
| 1259 | pRemove=p-1; | ||||||||
| 1260 | char16_t syllable = static_cast<char16_t>( | ||||||||
| 1261 | Hangul::HANGUL_BASE + | ||||||||
| 1262 | (prev*Hangul::JAMO_V_COUNT+(c-Hangul::JAMO_V_BASE))* | ||||||||
| 1263 | Hangul::JAMO_T_COUNT); | ||||||||
| 1264 | char16_t t; | ||||||||
| 1265 | if (p != limit && (t = static_cast<char16_t>(*p - Hangul::JAMO_T_BASE)) < Hangul::JAMO_T_COUNT) { | ||||||||
| 1266 | ++p; | ||||||||
| 1267 | syllable+=t; // The next character was a Jamo T. | ||||||||
| 1268 | } | ||||||||
| 1269 | *starter=syllable; | ||||||||
| 1270 | // remove the Jamo V/T | ||||||||
| 1271 | q=pRemove; | ||||||||
| 1272 | r=p; | ||||||||
| 1273 | while(r<limit) { | ||||||||
| 1274 | *q++=*r++; | ||||||||
| 1275 | } | ||||||||
| 1276 | limit=q; | ||||||||
| 1277 | p=pRemove; | ||||||||
| 1278 | } | ||||||||
| 1279 | } | ||||||||
| 1280 | /* | ||||||||
| 1281 | * No "else" for Jamo T: | ||||||||
| 1282 | * Since the input is in NFD, there are no Hangul LV syllables that | ||||||||
| 1283 | * a Jamo T could combine with. | ||||||||
| 1284 | * All Jamo Ts are combined above when handling Jamo Vs. | ||||||||
| 1285 | */ | ||||||||
| 1286 | if(p==limit) { | ||||||||
| 1287 | break; | ||||||||
| 1288 | } | ||||||||
| 1289 | compositionsList=nullptr; | ||||||||
| 1290 | continue; | ||||||||
| 1291 | } else if((compositeAndFwd=combine(compositionsList, c))>=0) { | ||||||||
| 1292 | // The starter and the combining mark (c) do combine. | ||||||||
| 1293 | UChar32 composite=compositeAndFwd>>1; | ||||||||
| 1294 | |||||||||
| 1295 | // Replace the starter with the composite, remove the combining mark. | ||||||||
| 1296 | pRemove=p-U16_LENGTH(c)((uint32_t)(c)<=0xffff ? 1 : 2); // pRemove & p: start & limit of the combining mark | ||||||||
| 1297 | if(starterIsSupplementary) { | ||||||||
| 1298 | if(U_IS_SUPPLEMENTARY(composite)((uint32_t)((composite)-0x10000)<=0xfffff)) { | ||||||||
| 1299 | // both are supplementary | ||||||||
| 1300 | starter[0]=U16_LEAD(composite)(UChar)(((composite)>>10)+0xd7c0); | ||||||||
| 1301 | starter[1]=U16_TRAIL(composite)(UChar)(((composite)&0x3ff)|0xdc00); | ||||||||
| 1302 | } else { | ||||||||
| 1303 | *starter = static_cast<char16_t>(composite); | ||||||||
| 1304 | // The composite is shorter than the starter, | ||||||||
| 1305 | // move the intermediate characters forward one. | ||||||||
| 1306 | starterIsSupplementary=false; | ||||||||
| 1307 | q=starter+1; | ||||||||
| 1308 | r=q+1; | ||||||||
| 1309 | while(r<pRemove) { | ||||||||
| 1310 | *q++=*r++; | ||||||||
| 1311 | } | ||||||||
| 1312 | --pRemove; | ||||||||
| 1313 | } | ||||||||
| 1314 | } else if(U_IS_SUPPLEMENTARY(composite)((uint32_t)((composite)-0x10000)<=0xfffff)) { | ||||||||
| 1315 | // The composite is longer than the starter, | ||||||||
| 1316 | // move the intermediate characters back one. | ||||||||
| 1317 | starterIsSupplementary=true; | ||||||||
| 1318 | ++starter; // temporarily increment for the loop boundary | ||||||||
| 1319 | q=pRemove; | ||||||||
| 1320 | r=++pRemove; | ||||||||
| 1321 | while(starter<q) { | ||||||||
| 1322 | *--r=*--q; | ||||||||
| 1323 | } | ||||||||
| 1324 | *starter=U16_TRAIL(composite)(UChar)(((composite)&0x3ff)|0xdc00); | ||||||||
| 1325 | *--starter=U16_LEAD(composite)(UChar)(((composite)>>10)+0xd7c0); // undo the temporary increment | ||||||||
| 1326 | } else { | ||||||||
| 1327 | // both are on the BMP | ||||||||
| 1328 | *starter = static_cast<char16_t>(composite); | ||||||||
| 1329 | } | ||||||||
| 1330 | |||||||||
| 1331 | /* remove the combining mark by moving the following text over it */ | ||||||||
| 1332 | if(pRemove<p) { | ||||||||
| 1333 | q=pRemove; | ||||||||
| 1334 | r=p; | ||||||||
| 1335 | while(r<limit) { | ||||||||
| 1336 | *q++=*r++; | ||||||||
| 1337 | } | ||||||||
| 1338 | limit=q; | ||||||||
| 1339 | p=pRemove; | ||||||||
| 1340 | } | ||||||||
| 1341 | // Keep prevCC because we removed the combining mark. | ||||||||
| 1342 | |||||||||
| 1343 | if(p==limit) { | ||||||||
| 1344 | break; | ||||||||
| 1345 | } | ||||||||
| 1346 | // Is the composite a starter that combines forward? | ||||||||
| 1347 | if(compositeAndFwd&1) { | ||||||||
| 1348 | compositionsList= | ||||||||
| 1349 | getCompositionsListForComposite(getRawNorm16(composite)); | ||||||||
| 1350 | } else { | ||||||||
| 1351 | compositionsList=nullptr; | ||||||||
| 1352 | } | ||||||||
| 1353 | |||||||||
| 1354 | // We combined; continue with looking for compositions. | ||||||||
| 1355 | continue; | ||||||||
| 1356 | } | ||||||||
| 1357 | } | ||||||||
| 1358 | |||||||||
| 1359 | // no combination this time | ||||||||
| 1360 | prevCC=cc; | ||||||||
| 1361 | if(p==limit) { | ||||||||
| 1362 | break; | ||||||||
| 1363 | } | ||||||||
| 1364 | |||||||||
| 1365 | // If c did not combine, then check if it is a starter. | ||||||||
| 1366 | if(cc
| ||||||||
| 1367 | // Found a new starter. | ||||||||
| 1368 | if((compositionsList=getCompositionsListForDecompYes(norm16))!=nullptr) { | ||||||||
| 1369 | // It may combine with something, prepare for it. | ||||||||
| 1370 | if(U_IS_BMP(c)((uint32_t)(c)<=0xffff)) { | ||||||||
| 1371 | starterIsSupplementary=false; | ||||||||
| 1372 | starter=p-1; | ||||||||
| 1373 | } else { | ||||||||
| 1374 | starterIsSupplementary=true; | ||||||||
| 1375 | starter=p-2; | ||||||||
| 1376 | } | ||||||||
| 1377 | } | ||||||||
| 1378 | } else if(onlyContiguous) { | ||||||||
| 1379 | // FCC: no discontiguous compositions; any intervening character blocks. | ||||||||
| 1380 | compositionsList=nullptr; | ||||||||
| 1381 | } | ||||||||
| 1382 | } | ||||||||
| 1383 | buffer.setReorderingLimit(limit); | ||||||||
| 1384 | } | ||||||||
| 1385 | |||||||||
| 1386 | UChar32 | ||||||||
| 1387 | Normalizer2Impl::composePair(UChar32 a, UChar32 b) const { | ||||||||
| 1388 | uint16_t norm16=getNorm16(a); // maps an out-of-range 'a' to inert norm16 | ||||||||
| 1389 | const uint16_t *list; | ||||||||
| 1390 | if(isInert(norm16)) { | ||||||||
| 1391 | return U_SENTINEL(-1); | ||||||||
| 1392 | } else if(norm16<minYesNoMappingsOnly) { | ||||||||
| 1393 | // a combines forward. | ||||||||
| 1394 | if(isJamoL(norm16)) { | ||||||||
| 1395 | if (b < Hangul::JAMO_V_BASE) { | ||||||||
| 1396 | return U_SENTINEL(-1); | ||||||||
| 1397 | } | ||||||||
| 1398 | b-=Hangul::JAMO_V_BASE; | ||||||||
| 1399 | if(b<Hangul::JAMO_V_COUNT) { | ||||||||
| 1400 | return | ||||||||
| 1401 | (Hangul::HANGUL_BASE+ | ||||||||
| 1402 | ((a-Hangul::JAMO_L_BASE)*Hangul::JAMO_V_COUNT+b)* | ||||||||
| 1403 | Hangul::JAMO_T_COUNT); | ||||||||
| 1404 | } else { | ||||||||
| 1405 | return U_SENTINEL(-1); | ||||||||
| 1406 | } | ||||||||
| 1407 | } else if(isHangulLV(norm16)) { | ||||||||
| 1408 | if (b <= Hangul::JAMO_T_BASE) { | ||||||||
| 1409 | return U_SENTINEL(-1); | ||||||||
| 1410 | } | ||||||||
| 1411 | b-=Hangul::JAMO_T_BASE; | ||||||||
| 1412 | if(b<Hangul::JAMO_T_COUNT) { // not b==0! | ||||||||
| 1413 | return a+b; | ||||||||
| 1414 | } else { | ||||||||
| 1415 | return U_SENTINEL(-1); | ||||||||
| 1416 | } | ||||||||
| 1417 | } else { | ||||||||
| 1418 | // 'a' has a compositions list in extraData | ||||||||
| 1419 | list=getDataForYesOrNo(norm16); | ||||||||
| 1420 | if(norm16>minYesNo) { // composite 'a' has both mapping & compositions list | ||||||||
| 1421 | list+= // mapping pointer | ||||||||
| 1422 | 1+ // +1 to skip the first unit with the mapping length | ||||||||
| 1423 | (*list&MAPPING_LENGTH_MASK); // + mapping length | ||||||||
| 1424 | } | ||||||||
| 1425 | } | ||||||||
| 1426 | } else if(norm16<minMaybeNoCombinesFwd || MIN_NORMAL_MAYBE_YES<=norm16) { | ||||||||
| 1427 | return U_SENTINEL(-1); | ||||||||
| 1428 | } else { | ||||||||
| 1429 | list=getDataForMaybe(norm16); | ||||||||
| 1430 | if(norm16<minMaybeYes) { // composite 'a' has both mapping & compositions list | ||||||||
| 1431 | list+= // mapping pointer | ||||||||
| 1432 | 1+ // +1 to skip the first unit with the mapping length | ||||||||
| 1433 | (*list&MAPPING_LENGTH_MASK); // + mapping length | ||||||||
| 1434 | } | ||||||||
| 1435 | } | ||||||||
| 1436 | if(b<0 || 0x10ffff<b) { // combine(list, b) requires a valid code point b | ||||||||
| 1437 | return U_SENTINEL(-1); | ||||||||
| 1438 | } | ||||||||
| 1439 | #if U_SIGNED_RIGHT_SHIFT_IS_ARITHMETIC1 | ||||||||
| 1440 | return combine(list, b)>>1; | ||||||||
| 1441 | #else | ||||||||
| 1442 | int32_t compositeAndFwd=combine(list, b); | ||||||||
| 1443 | return compositeAndFwd>=0 ? compositeAndFwd>>1 : U_SENTINEL(-1); | ||||||||
| 1444 | #endif | ||||||||
| 1445 | } | ||||||||
| 1446 | |||||||||
| 1447 | // Very similar to composeQuickCheck(): Make the same changes in both places if relevant. | ||||||||
| 1448 | // doCompose: normalize | ||||||||
| 1449 | // !doCompose: isNormalized (buffer must be empty and initialized) | ||||||||
| 1450 | UBool | ||||||||
| 1451 | Normalizer2Impl::compose(const char16_t *src, const char16_t *limit, | ||||||||
| 1452 | UBool onlyContiguous, | ||||||||
| 1453 | UBool doCompose, | ||||||||
| 1454 | ReorderingBuffer &buffer, | ||||||||
| 1455 | UErrorCode &errorCode) const { | ||||||||
| 1456 | const char16_t *prevBoundary=src; | ||||||||
| 1457 | UChar32 minNoMaybeCP=minCompNoMaybeCP; | ||||||||
| 1458 | if(limit==nullptr) { | ||||||||
| |||||||||
| 1459 | src=copyLowPrefixFromNulTerminated(src, minNoMaybeCP, | ||||||||
| 1460 | doCompose ? &buffer : nullptr, | ||||||||
| 1461 | errorCode); | ||||||||
| 1462 | if(U_FAILURE(errorCode)) { | ||||||||
| 1463 | return false; | ||||||||
| 1464 | } | ||||||||
| 1465 | limit=u_strchru_strchr_77(src, 0); | ||||||||
| 1466 | if (prevBoundary != src) { | ||||||||
| 1467 | if (hasCompBoundaryAfter(*(src-1), onlyContiguous)) { | ||||||||
| 1468 | prevBoundary = src; | ||||||||
| 1469 | } else { | ||||||||
| 1470 | buffer.removeSuffix(1); | ||||||||
| 1471 | prevBoundary = --src; | ||||||||
| 1472 | } | ||||||||
| 1473 | } | ||||||||
| 1474 | } | ||||||||
| 1475 | |||||||||
| 1476 | for (;;) { | ||||||||
| 1477 | // Fast path: Scan over a sequence of characters below the minimum "no or maybe" code point, | ||||||||
| 1478 | // or with (compYes && ccc==0) properties. | ||||||||
| 1479 | const char16_t *prevSrc; | ||||||||
| 1480 | UChar32 c = 0; | ||||||||
| 1481 | uint16_t norm16 = 0; | ||||||||
| 1482 | for (;;) { | ||||||||
| 1483 | if (src == limit) { | ||||||||
| 1484 | if (prevBoundary != limit && doCompose) { | ||||||||
| 1485 | buffer.appendZeroCC(prevBoundary, limit, errorCode); | ||||||||
| 1486 | } | ||||||||
| 1487 | return true; | ||||||||
| 1488 | } | ||||||||
| 1489 | if( (c=*src)<minNoMaybeCP || | ||||||||
| 1490 | isCompYesAndZeroCC(norm16=UCPTRIE_FAST_BMP_GET(normTrie, UCPTRIE_16, c)((normTrie)->data.ptr16[((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT] + ((c) & UCPTRIE_FAST_DATA_MASK ))])) | ||||||||
| 1491 | ) { | ||||||||
| 1492 | ++src; | ||||||||
| 1493 | } else { | ||||||||
| 1494 | prevSrc = src++; | ||||||||
| 1495 | if(!U16_IS_LEAD(c)(((c)&0xfffffc00)==0xd800)) { | ||||||||
| 1496 | break; | ||||||||
| 1497 | } else { | ||||||||
| 1498 | char16_t c2; | ||||||||
| 1499 | if(src!=limit && U16_IS_TRAIL(c2=*src)(((c2=*src)&0xfffffc00)==0xdc00)) { | ||||||||
| 1500 | ++src; | ||||||||
| 1501 | c=U16_GET_SUPPLEMENTARY(c, c2)(((UChar32)(c)<<10UL)+(UChar32)(c2)-((0xd800<<10UL )+0xdc00-0x10000)); | ||||||||
| 1502 | norm16=UCPTRIE_FAST_SUPP_GET(normTrie, UCPTRIE_16, c)((normTrie)->data.ptr16[((c) >= (normTrie)->highStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77(normTrie, c))]); | ||||||||
| 1503 | if(!isCompYesAndZeroCC(norm16)) { | ||||||||
| 1504 | break; | ||||||||
| 1505 | } | ||||||||
| 1506 | } | ||||||||
| 1507 | } | ||||||||
| 1508 | } | ||||||||
| 1509 | } | ||||||||
| 1510 | // isCompYesAndZeroCC(norm16) is false, that is, norm16>=minNoNo. | ||||||||
| 1511 | // The current character is either a "noNo" (has a mapping) | ||||||||
| 1512 | // or a "maybeYes" / "maybeNo" (combines backward) | ||||||||
| 1513 | // or a "yesYes" with ccc!=0. | ||||||||
| 1514 | // It is not a Hangul syllable or Jamo L because those have "yes" properties. | ||||||||
| 1515 | |||||||||
| 1516 | // Medium-fast path: Handle cases that do not require full decomposition and recomposition. | ||||||||
| 1517 | if (norm16 < minMaybeNo) { // minNoNo <= norm16 < minMaybeNo | ||||||||
| 1518 | if (!doCompose) { | ||||||||
| 1519 | return false; | ||||||||
| 1520 | } | ||||||||
| 1521 | // Fast path for mapping a character that is immediately surrounded by boundaries. | ||||||||
| 1522 | // In this case, we need not decompose around the current character. | ||||||||
| 1523 | if (isDecompNoAlgorithmic(norm16)) { | ||||||||
| 1524 | // Maps to a single isCompYesAndZeroCC character | ||||||||
| 1525 | // which also implies hasCompBoundaryBefore. | ||||||||
| 1526 | if (norm16HasCompBoundaryAfter(norm16, onlyContiguous) || | ||||||||
| 1527 | hasCompBoundaryBefore(src, limit)) { | ||||||||
| 1528 | if (prevBoundary != prevSrc && !buffer.appendZeroCC(prevBoundary, prevSrc, errorCode)) { | ||||||||
| 1529 | break; | ||||||||
| 1530 | } | ||||||||
| 1531 | if(!buffer.append(mapAlgorithmic(c, norm16), 0, errorCode)) { | ||||||||
| 1532 | break; | ||||||||
| 1533 | } | ||||||||
| 1534 | prevBoundary = src; | ||||||||
| 1535 | continue; | ||||||||
| 1536 | } | ||||||||
| 1537 | } else if (norm16 < minNoNoCompBoundaryBefore) { | ||||||||
| 1538 | // The mapping is comp-normalized which also implies hasCompBoundaryBefore. | ||||||||
| 1539 | if (norm16HasCompBoundaryAfter(norm16, onlyContiguous) || | ||||||||
| 1540 | hasCompBoundaryBefore(src, limit)) { | ||||||||
| 1541 | if (prevBoundary != prevSrc && !buffer.appendZeroCC(prevBoundary, prevSrc, errorCode)) { | ||||||||
| 1542 | break; | ||||||||
| 1543 | } | ||||||||
| 1544 | const char16_t *mapping = reinterpret_cast<const char16_t *>(getDataForYesOrNo(norm16)); | ||||||||
| 1545 | int32_t length = *mapping++ & MAPPING_LENGTH_MASK; | ||||||||
| 1546 | if(!buffer.appendZeroCC(mapping, mapping + length, errorCode)) { | ||||||||
| 1547 | break; | ||||||||
| 1548 | } | ||||||||
| 1549 | prevBoundary = src; | ||||||||
| 1550 | continue; | ||||||||
| 1551 | } | ||||||||
| 1552 | } else if (norm16 >= minNoNoEmpty) { | ||||||||
| 1553 | // The current character maps to nothing. | ||||||||
| 1554 | // Simply omit it from the output if there is a boundary before _or_ after it. | ||||||||
| 1555 | // The character itself implies no boundaries. | ||||||||
| 1556 | if (hasCompBoundaryBefore(src, limit) || | ||||||||
| 1557 | hasCompBoundaryAfter(prevBoundary, prevSrc, onlyContiguous)) { | ||||||||
| 1558 | if (prevBoundary != prevSrc && !buffer.appendZeroCC(prevBoundary, prevSrc, errorCode)) { | ||||||||
| 1559 | break; | ||||||||
| 1560 | } | ||||||||
| 1561 | prevBoundary = src; | ||||||||
| 1562 | continue; | ||||||||
| 1563 | } | ||||||||
| 1564 | } | ||||||||
| 1565 | // Other "noNo" type, or need to examine more text around this character: | ||||||||
| 1566 | // Fall through to the slow path. | ||||||||
| 1567 | } else if (isJamoVT(norm16) && prevBoundary != prevSrc) { | ||||||||
| 1568 | char16_t prev=*(prevSrc-1); | ||||||||
| 1569 | if(c<Hangul::JAMO_T_BASE) { | ||||||||
| 1570 | // The current character is a Jamo Vowel, | ||||||||
| 1571 | // compose with previous Jamo L and following Jamo T. | ||||||||
| 1572 | char16_t l = static_cast<char16_t>(prev - Hangul::JAMO_L_BASE); | ||||||||
| 1573 | if(l<Hangul::JAMO_L_COUNT) { | ||||||||
| 1574 | if (!doCompose) { | ||||||||
| 1575 | return false; | ||||||||
| 1576 | } | ||||||||
| 1577 | int32_t t; | ||||||||
| 1578 | if (src != limit && | ||||||||
| 1579 | 0 < (t = (static_cast<int32_t>(*src) - Hangul::JAMO_T_BASE)) && | ||||||||
| 1580 | t < Hangul::JAMO_T_COUNT) { | ||||||||
| 1581 | // The next character is a Jamo T. | ||||||||
| 1582 | ++src; | ||||||||
| 1583 | } else if (hasCompBoundaryBefore(src, limit)) { | ||||||||
| 1584 | // No Jamo T follows, not even via decomposition. | ||||||||
| 1585 | t = 0; | ||||||||
| 1586 | } else { | ||||||||
| 1587 | t = -1; | ||||||||
| 1588 | } | ||||||||
| 1589 | if (t >= 0) { | ||||||||
| 1590 | UChar32 syllable = Hangul::HANGUL_BASE + | ||||||||
| 1591 | (l*Hangul::JAMO_V_COUNT + (c-Hangul::JAMO_V_BASE)) * | ||||||||
| 1592 | Hangul::JAMO_T_COUNT + t; | ||||||||
| 1593 | --prevSrc; // Replace the Jamo L as well. | ||||||||
| 1594 | if (prevBoundary != prevSrc && !buffer.appendZeroCC(prevBoundary, prevSrc, errorCode)) { | ||||||||
| 1595 | break; | ||||||||
| 1596 | } | ||||||||
| 1597 | if (!buffer.appendBMP(static_cast<char16_t>(syllable), 0, errorCode)) { | ||||||||
| 1598 | break; | ||||||||
| 1599 | } | ||||||||
| 1600 | prevBoundary = src; | ||||||||
| 1601 | continue; | ||||||||
| 1602 | } | ||||||||
| 1603 | // If we see L+V+x where x!=T then we drop to the slow path, | ||||||||
| 1604 | // decompose and recompose. | ||||||||
| 1605 | // This is to deal with NFKC finding normal L and V but a | ||||||||
| 1606 | // compatibility variant of a T. | ||||||||
| 1607 | // We need to either fully compose that combination here | ||||||||
| 1608 | // (which would complicate the code and may not work with strange custom data) | ||||||||
| 1609 | // or use the slow path. | ||||||||
| 1610 | } | ||||||||
| 1611 | } else if (Hangul::isHangulLV(prev)) { | ||||||||
| 1612 | // The current character is a Jamo Trailing consonant, | ||||||||
| 1613 | // compose with previous Hangul LV that does not contain a Jamo T. | ||||||||
| 1614 | if (!doCompose) { | ||||||||
| 1615 | return false; | ||||||||
| 1616 | } | ||||||||
| 1617 | UChar32 syllable = prev + c - Hangul::JAMO_T_BASE; | ||||||||
| 1618 | --prevSrc; // Replace the Hangul LV as well. | ||||||||
| 1619 | if (prevBoundary != prevSrc && !buffer.appendZeroCC(prevBoundary, prevSrc, errorCode)) { | ||||||||
| 1620 | break; | ||||||||
| 1621 | } | ||||||||
| 1622 | if (!buffer.appendBMP(static_cast<char16_t>(syllable), 0, errorCode)) { | ||||||||
| 1623 | break; | ||||||||
| 1624 | } | ||||||||
| 1625 | prevBoundary = src; | ||||||||
| 1626 | continue; | ||||||||
| 1627 | } | ||||||||
| 1628 | // No matching context, or may need to decompose surrounding text first: | ||||||||
| 1629 | // Fall through to the slow path. | ||||||||
| 1630 | } else if (norm16 > JAMO_VT) { // norm16 >= MIN_YES_YES_WITH_CC | ||||||||
| 1631 | // One or more combining marks that do not combine-back: | ||||||||
| 1632 | // Check for canonical order, copy unchanged if ok and | ||||||||
| 1633 | // if followed by a character with a boundary-before. | ||||||||
| 1634 | uint8_t cc = getCCFromNormalYesOrMaybe(norm16); // cc!=0 | ||||||||
| 1635 | if (onlyContiguous /* FCC */ && getPreviousTrailCC(prevBoundary, prevSrc) > cc) { | ||||||||
| 1636 | // Fails FCD test, need to decompose and contiguously recompose. | ||||||||
| 1637 | if (!doCompose) { | ||||||||
| 1638 | return false; | ||||||||
| 1639 | } | ||||||||
| 1640 | } else { | ||||||||
| 1641 | // If !onlyContiguous (not FCC), then we ignore the tccc of | ||||||||
| 1642 | // the previous character which passed the quick check "yes && ccc==0" test. | ||||||||
| 1643 | const char16_t *nextSrc; | ||||||||
| 1644 | uint16_t n16; | ||||||||
| 1645 | for (;;) { | ||||||||
| 1646 | if (src == limit) { | ||||||||
| 1647 | if (doCompose) { | ||||||||
| 1648 | buffer.appendZeroCC(prevBoundary, limit, errorCode); | ||||||||
| 1649 | } | ||||||||
| 1650 | return true; | ||||||||
| 1651 | } | ||||||||
| 1652 | uint8_t prevCC = cc; | ||||||||
| 1653 | nextSrc = src; | ||||||||
| 1654 | UCPTRIE_FAST_U16_NEXT(normTrie, UCPTRIE_16, nextSrc, limit, c, n16)do { (c) = *(nextSrc)++; int32_t __index; if (!(((c)&0xfffff800 )==0xd800)) { __index = ((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT] + ((c) & UCPTRIE_FAST_DATA_MASK)); } else { uint16_t __c2; if ((((c)&0x400)==0) && (nextSrc ) != (limit) && (((__c2 = *(nextSrc))&0xfffffc00) ==0xdc00)) { ++(nextSrc); (c) = (((UChar32)((c))<<10UL) +(UChar32)(__c2)-((0xd800<<10UL)+0xdc00-0x10000)); __index = ((c) >= (normTrie)->highStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77 (normTrie, c)); } else { __index = (normTrie)->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET; } } (n16) = ((normTrie) ->data.ptr16[__index]); } while (false); | ||||||||
| 1655 | if (n16 >= MIN_YES_YES_WITH_CC) { | ||||||||
| 1656 | cc = getCCFromNormalYesOrMaybe(n16); | ||||||||
| 1657 | if (prevCC > cc) { | ||||||||
| 1658 | if (!doCompose) { | ||||||||
| 1659 | return false; | ||||||||
| 1660 | } | ||||||||
| 1661 | break; | ||||||||
| 1662 | } | ||||||||
| 1663 | } else { | ||||||||
| 1664 | break; | ||||||||
| 1665 | } | ||||||||
| 1666 | src = nextSrc; | ||||||||
| 1667 | } | ||||||||
| 1668 | // src is after the last in-order combining mark. | ||||||||
| 1669 | // If there is a boundary here, then we continue with no change. | ||||||||
| 1670 | if (norm16HasCompBoundaryBefore(n16)) { | ||||||||
| 1671 | if (isCompYesAndZeroCC(n16)) { | ||||||||
| 1672 | src = nextSrc; | ||||||||
| 1673 | } | ||||||||
| 1674 | continue; | ||||||||
| 1675 | } | ||||||||
| 1676 | // Use the slow path. There is no boundary in [prevSrc, src[. | ||||||||
| 1677 | } | ||||||||
| 1678 | } | ||||||||
| 1679 | |||||||||
| 1680 | // Slow path: Find the nearest boundaries around the current character, | ||||||||
| 1681 | // decompose and recompose. | ||||||||
| 1682 | if (prevBoundary
| ||||||||
| 1683 | const char16_t *p = prevSrc; | ||||||||
| 1684 | UCPTRIE_FAST_U16_PREV(normTrie, UCPTRIE_16, prevBoundary, p, c, norm16)do { (c) = *--(p); int32_t __index; if (!(((c)&0xfffff800 )==0xd800)) { __index = ((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT] + ((c) & UCPTRIE_FAST_DATA_MASK)); } else { uint16_t __c2; if ((((c)&0x400)!=0) && (p ) != (prevBoundary) && (((__c2 = *((p) - 1))&0xfffffc00 )==0xd800)) { --(p); (c) = (((UChar32)(__c2)<<10UL)+(UChar32 )((c))-((0xd800<<10UL)+0xdc00-0x10000)); __index = ((c) >= (normTrie)->highStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77 (normTrie, c)); } else { __index = (normTrie)->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET; } } (norm16) = ((normTrie )->data.ptr16[__index]); } while (false); | ||||||||
| 1685 | if (!norm16HasCompBoundaryAfter(norm16, onlyContiguous)) { | ||||||||
| 1686 | prevSrc = p; | ||||||||
| 1687 | } | ||||||||
| 1688 | } | ||||||||
| 1689 | if (doCompose && prevBoundary != prevSrc && !buffer.appendZeroCC(prevBoundary, prevSrc, errorCode)) { | ||||||||
| 1690 | break; | ||||||||
| 1691 | } | ||||||||
| 1692 | int32_t recomposeStartIndex=buffer.length(); | ||||||||
| 1693 | // We know there is not a boundary here. | ||||||||
| 1694 | decomposeShort(prevSrc, src, false /* !stopAtCompBoundary */, onlyContiguous, | ||||||||
| 1695 | buffer, errorCode); | ||||||||
| 1696 | // Decompose until the next boundary. | ||||||||
| 1697 | src = decomposeShort(src, limit, true /* stopAtCompBoundary */, onlyContiguous, | ||||||||
| 1698 | buffer, errorCode); | ||||||||
| 1699 | if (U_FAILURE(errorCode)) { | ||||||||
| 1700 | break; | ||||||||
| 1701 | } | ||||||||
| 1702 | if ((src - prevSrc) > INT32_MAX(2147483647)) { // guard before buffer.equals() | ||||||||
| 1703 | errorCode = U_INDEX_OUTOFBOUNDS_ERROR; | ||||||||
| 1704 | return true; | ||||||||
| 1705 | } | ||||||||
| 1706 | recompose(buffer, recomposeStartIndex, onlyContiguous); | ||||||||
| 1707 | if(!doCompose) { | ||||||||
| 1708 | if(!buffer.equals(prevSrc, src)) { | ||||||||
| 1709 | return false; | ||||||||
| 1710 | } | ||||||||
| 1711 | buffer.remove(); | ||||||||
| 1712 | } | ||||||||
| 1713 | prevBoundary=src; | ||||||||
| 1714 | } | ||||||||
| 1715 | return true; | ||||||||
| 1716 | } | ||||||||
| 1717 | |||||||||
| 1718 | // Very similar to compose(): Make the same changes in both places if relevant. | ||||||||
| 1719 | // pQCResult==nullptr: spanQuickCheckYes | ||||||||
| 1720 | // pQCResult!=nullptr: quickCheck (*pQCResult must be UNORM_YES) | ||||||||
| 1721 | const char16_t * | ||||||||
| 1722 | Normalizer2Impl::composeQuickCheck(const char16_t *src, const char16_t *limit, | ||||||||
| 1723 | UBool onlyContiguous, | ||||||||
| 1724 | UNormalizationCheckResult *pQCResult) const { | ||||||||
| 1725 | const char16_t *prevBoundary=src; | ||||||||
| 1726 | UChar32 minNoMaybeCP=minCompNoMaybeCP; | ||||||||
| 1727 | if(limit==nullptr) { | ||||||||
| 1728 | UErrorCode errorCode=U_ZERO_ERROR; | ||||||||
| 1729 | src=copyLowPrefixFromNulTerminated(src, minNoMaybeCP, nullptr, errorCode); | ||||||||
| 1730 | limit=u_strchru_strchr_77(src, 0); | ||||||||
| 1731 | if (prevBoundary != src) { | ||||||||
| 1732 | if (hasCompBoundaryAfter(*(src-1), onlyContiguous)) { | ||||||||
| 1733 | prevBoundary = src; | ||||||||
| 1734 | } else { | ||||||||
| 1735 | prevBoundary = --src; | ||||||||
| 1736 | } | ||||||||
| 1737 | } | ||||||||
| 1738 | } | ||||||||
| 1739 | |||||||||
| 1740 | for(;;) { | ||||||||
| 1741 | // Fast path: Scan over a sequence of characters below the minimum "no or maybe" code point, | ||||||||
| 1742 | // or with (compYes && ccc==0) properties. | ||||||||
| 1743 | const char16_t *prevSrc; | ||||||||
| 1744 | UChar32 c = 0; | ||||||||
| 1745 | uint16_t norm16 = 0; | ||||||||
| 1746 | for (;;) { | ||||||||
| 1747 | if(src==limit) { | ||||||||
| 1748 | return src; | ||||||||
| 1749 | } | ||||||||
| 1750 | if( (c=*src)<minNoMaybeCP || | ||||||||
| 1751 | isCompYesAndZeroCC(norm16=UCPTRIE_FAST_BMP_GET(normTrie, UCPTRIE_16, c)((normTrie)->data.ptr16[((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT] + ((c) & UCPTRIE_FAST_DATA_MASK ))])) | ||||||||
| 1752 | ) { | ||||||||
| 1753 | ++src; | ||||||||
| 1754 | } else { | ||||||||
| 1755 | prevSrc = src++; | ||||||||
| 1756 | if(!U16_IS_LEAD(c)(((c)&0xfffffc00)==0xd800)) { | ||||||||
| 1757 | break; | ||||||||
| 1758 | } else { | ||||||||
| 1759 | char16_t c2; | ||||||||
| 1760 | if(src!=limit && U16_IS_TRAIL(c2=*src)(((c2=*src)&0xfffffc00)==0xdc00)) { | ||||||||
| 1761 | ++src; | ||||||||
| 1762 | c=U16_GET_SUPPLEMENTARY(c, c2)(((UChar32)(c)<<10UL)+(UChar32)(c2)-((0xd800<<10UL )+0xdc00-0x10000)); | ||||||||
| 1763 | norm16=UCPTRIE_FAST_SUPP_GET(normTrie, UCPTRIE_16, c)((normTrie)->data.ptr16[((c) >= (normTrie)->highStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77(normTrie, c))]); | ||||||||
| 1764 | if(!isCompYesAndZeroCC(norm16)) { | ||||||||
| 1765 | break; | ||||||||
| 1766 | } | ||||||||
| 1767 | } | ||||||||
| 1768 | } | ||||||||
| 1769 | } | ||||||||
| 1770 | } | ||||||||
| 1771 | // isCompYesAndZeroCC(norm16) is false, that is, norm16>=minNoNo. | ||||||||
| 1772 | // The current character is either a "noNo" (has a mapping) | ||||||||
| 1773 | // or a "maybeYes" / "maybeNo" (combines backward) | ||||||||
| 1774 | // or a "yesYes" with ccc!=0. | ||||||||
| 1775 | // It is not a Hangul syllable or Jamo L because those have "yes" properties. | ||||||||
| 1776 | |||||||||
| 1777 | uint16_t prevNorm16 = INERT; | ||||||||
| 1778 | if (prevBoundary != prevSrc) { | ||||||||
| 1779 | if (norm16HasCompBoundaryBefore(norm16)) { | ||||||||
| 1780 | prevBoundary = prevSrc; | ||||||||
| 1781 | } else { | ||||||||
| 1782 | const char16_t *p = prevSrc; | ||||||||
| 1783 | uint16_t n16; | ||||||||
| 1784 | UCPTRIE_FAST_U16_PREV(normTrie, UCPTRIE_16, prevBoundary, p, c, n16)do { (c) = *--(p); int32_t __index; if (!(((c)&0xfffff800 )==0xd800)) { __index = ((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT] + ((c) & UCPTRIE_FAST_DATA_MASK)); } else { uint16_t __c2; if ((((c)&0x400)!=0) && (p ) != (prevBoundary) && (((__c2 = *((p) - 1))&0xfffffc00 )==0xd800)) { --(p); (c) = (((UChar32)(__c2)<<10UL)+(UChar32 )((c))-((0xd800<<10UL)+0xdc00-0x10000)); __index = ((c) >= (normTrie)->highStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77 (normTrie, c)); } else { __index = (normTrie)->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET; } } (n16) = ((normTrie) ->data.ptr16[__index]); } while (false); | ||||||||
| 1785 | if (norm16HasCompBoundaryAfter(n16, onlyContiguous)) { | ||||||||
| 1786 | prevBoundary = prevSrc; | ||||||||
| 1787 | } else { | ||||||||
| 1788 | prevBoundary = p; | ||||||||
| 1789 | prevNorm16 = n16; | ||||||||
| 1790 | } | ||||||||
| 1791 | } | ||||||||
| 1792 | } | ||||||||
| 1793 | |||||||||
| 1794 | if (norm16 >= minMaybeNo) { | ||||||||
| 1795 | uint16_t fcd16 = getFCD16FromMaybeOrNonZeroCC(norm16); | ||||||||
| 1796 | uint8_t cc = fcd16 >> 8; | ||||||||
| 1797 | if (onlyContiguous /* FCC */ && cc != 0 && | ||||||||
| 1798 | getTrailCCFromCompYesAndZeroCC(prevNorm16) > cc) { | ||||||||
| 1799 | // The [prevBoundary..prevSrc[ character | ||||||||
| 1800 | // passed the quick check "yes && ccc==0" test | ||||||||
| 1801 | // but is out of canonical order with the current combining mark. | ||||||||
| 1802 | } else { | ||||||||
| 1803 | // If !onlyContiguous (not FCC), then we ignore the tccc of | ||||||||
| 1804 | // the previous character which passed the quick check "yes && ccc==0" test. | ||||||||
| 1805 | const char16_t *nextSrc; | ||||||||
| 1806 | for (;;) { | ||||||||
| 1807 | if (norm16 < MIN_YES_YES_WITH_CC) { | ||||||||
| 1808 | if (pQCResult != nullptr) { | ||||||||
| 1809 | *pQCResult = UNORM_MAYBE; | ||||||||
| 1810 | } else { | ||||||||
| 1811 | return prevBoundary; | ||||||||
| 1812 | } | ||||||||
| 1813 | } | ||||||||
| 1814 | if (src == limit) { | ||||||||
| 1815 | return src; | ||||||||
| 1816 | } | ||||||||
| 1817 | uint8_t prevCC = fcd16; | ||||||||
| 1818 | nextSrc = src; | ||||||||
| 1819 | UCPTRIE_FAST_U16_NEXT(normTrie, UCPTRIE_16, nextSrc, limit, c, norm16)do { (c) = *(nextSrc)++; int32_t __index; if (!(((c)&0xfffff800 )==0xd800)) { __index = ((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT] + ((c) & UCPTRIE_FAST_DATA_MASK)); } else { uint16_t __c2; if ((((c)&0x400)==0) && (nextSrc ) != (limit) && (((__c2 = *(nextSrc))&0xfffffc00) ==0xdc00)) { ++(nextSrc); (c) = (((UChar32)((c))<<10UL) +(UChar32)(__c2)-((0xd800<<10UL)+0xdc00-0x10000)); __index = ((c) >= (normTrie)->highStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77 (normTrie, c)); } else { __index = (normTrie)->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET; } } (norm16) = ((normTrie )->data.ptr16[__index]); } while (false); | ||||||||
| 1820 | if (norm16 >= minMaybeNo) { | ||||||||
| 1821 | fcd16 = getFCD16FromMaybeOrNonZeroCC(norm16); | ||||||||
| 1822 | cc = fcd16 >> 8; | ||||||||
| 1823 | if (!(prevCC <= cc || cc == 0)) { | ||||||||
| 1824 | break; | ||||||||
| 1825 | } | ||||||||
| 1826 | } else { | ||||||||
| 1827 | break; | ||||||||
| 1828 | } | ||||||||
| 1829 | src = nextSrc; | ||||||||
| 1830 | } | ||||||||
| 1831 | // src is after the last in-order combining mark. | ||||||||
| 1832 | if (isCompYesAndZeroCC(norm16)) { | ||||||||
| 1833 | prevBoundary = src; | ||||||||
| 1834 | src = nextSrc; | ||||||||
| 1835 | continue; | ||||||||
| 1836 | } | ||||||||
| 1837 | } | ||||||||
| 1838 | } | ||||||||
| 1839 | if(pQCResult!=nullptr) { | ||||||||
| 1840 | *pQCResult=UNORM_NO; | ||||||||
| 1841 | } | ||||||||
| 1842 | return prevBoundary; | ||||||||
| 1843 | } | ||||||||
| 1844 | } | ||||||||
| 1845 | |||||||||
| 1846 | void Normalizer2Impl::composeAndAppend(const char16_t *src, const char16_t *limit, | ||||||||
| 1847 | UBool doCompose, | ||||||||
| 1848 | UBool onlyContiguous, | ||||||||
| 1849 | UnicodeString &safeMiddle, | ||||||||
| 1850 | ReorderingBuffer &buffer, | ||||||||
| 1851 | UErrorCode &errorCode) const { | ||||||||
| 1852 | if(!buffer.isEmpty()) { | ||||||||
| 1853 | const char16_t *firstStarterInSrc=findNextCompBoundary(src, limit, onlyContiguous); | ||||||||
| 1854 | if(src!=firstStarterInSrc) { | ||||||||
| 1855 | const char16_t *lastStarterInDest=findPreviousCompBoundary(buffer.getStart(), | ||||||||
| 1856 | buffer.getLimit(), onlyContiguous); | ||||||||
| 1857 | int32_t destSuffixLength = static_cast<int32_t>(buffer.getLimit() - lastStarterInDest); | ||||||||
| 1858 | UnicodeString middle(lastStarterInDest, destSuffixLength); | ||||||||
| 1859 | buffer.removeSuffix(destSuffixLength); | ||||||||
| 1860 | safeMiddle=middle; | ||||||||
| 1861 | middle.append(src, static_cast<int32_t>(firstStarterInSrc - src)); | ||||||||
| 1862 | const char16_t *middleStart=middle.getBuffer(); | ||||||||
| 1863 | compose(middleStart, middleStart+middle.length(), onlyContiguous, | ||||||||
| 1864 | true, buffer, errorCode); | ||||||||
| 1865 | if(U_FAILURE(errorCode)) { | ||||||||
| 1866 | return; | ||||||||
| 1867 | } | ||||||||
| 1868 | src=firstStarterInSrc; | ||||||||
| 1869 | } | ||||||||
| 1870 | } | ||||||||
| 1871 | if(doCompose) { | ||||||||
| 1872 | compose(src, limit, onlyContiguous, true, buffer, errorCode); | ||||||||
| 1873 | } else { | ||||||||
| 1874 | if(limit==nullptr) { // appendZeroCC() needs limit!=nullptr | ||||||||
| 1875 | limit=u_strchru_strchr_77(src, 0); | ||||||||
| 1876 | } | ||||||||
| 1877 | buffer.appendZeroCC(src, limit, errorCode); | ||||||||
| 1878 | } | ||||||||
| 1879 | } | ||||||||
| 1880 | |||||||||
| 1881 | UBool | ||||||||
| 1882 | Normalizer2Impl::composeUTF8(uint32_t options, UBool onlyContiguous, | ||||||||
| 1883 | const uint8_t *src, const uint8_t *limit, | ||||||||
| 1884 | ByteSink *sink, Edits *edits, UErrorCode &errorCode) const { | ||||||||
| 1885 | U_ASSERT(limit != nullptr)(static_cast <bool> (limit != nullptr) ? void (0) : __assert_fail ("limit != nullptr", __builtin_FILE (), __builtin_LINE (), __extension__ __PRETTY_FUNCTION__)); | ||||||||
| 1886 | UnicodeString s16; | ||||||||
| 1887 | uint8_t minNoMaybeLead = leadByteForCP(minCompNoMaybeCP); | ||||||||
| 1888 | const uint8_t *prevBoundary = src; | ||||||||
| 1889 | |||||||||
| 1890 | for (;;) { | ||||||||
| 1891 | // Fast path: Scan over a sequence of characters below the minimum "no or maybe" code point, | ||||||||
| 1892 | // or with (compYes && ccc==0) properties. | ||||||||
| 1893 | const uint8_t *prevSrc; | ||||||||
| 1894 | uint16_t norm16 = 0; | ||||||||
| 1895 | for (;;) { | ||||||||
| 1896 | if (src == limit) { | ||||||||
| 1897 | if (prevBoundary != limit && sink != nullptr) { | ||||||||
| 1898 | ByteSinkUtil::appendUnchanged(prevBoundary, limit, | ||||||||
| 1899 | *sink, options, edits, errorCode); | ||||||||
| 1900 | } | ||||||||
| 1901 | return true; | ||||||||
| 1902 | } | ||||||||
| 1903 | if (*src < minNoMaybeLead) { | ||||||||
| 1904 | ++src; | ||||||||
| 1905 | } else { | ||||||||
| 1906 | prevSrc = src; | ||||||||
| 1907 | UCPTRIE_FAST_U8_NEXT(normTrie, UCPTRIE_16, src, limit, norm16)do { int32_t __lead = (uint8_t)*(src)++; if (!(((__lead)& 0x80)==0)) { uint8_t __t1, __t2, __t3; if ((src) != (limit) && (__lead >= 0xe0 ? __lead < 0xf0 ? "\x20\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x10\x30\x30" [__lead &= 0xf] & (1 << ((__t1 = *(src)) >> 5)) && ++(src) != (limit) && (__t2 = *(src) - 0x80) <= 0x3f && (__lead = ((int32_t)(normTrie)-> index[(__lead << 6) + (__t1 & 0x3f)]) + __t2, 1) : ( __lead -= 0xf0) <= 4 && "\x00\x00\x00\x00\x00\x00\x00\x00\x1E\x0F\x0F\x0F\x00\x00\x00\x00" [(__t1 = *(src)) >> 4] & (1 << __lead) && (__lead = (__lead << 6) | (__t1 & 0x3f), ++(src) != (limit)) && (__t2 = *(src) - 0x80) <= 0x3f && ++(src) != (limit) && (__t3 = *(src) - 0x80) <= 0x3f && (__lead = __lead >= (normTrie)->shifted12HighStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallU8Index_77((normTrie), __lead, __t2, __t3 ), 1) : __lead >= 0xc2 && (__t1 = *(src) - 0x80) <= 0x3f && (__lead = (int32_t)(normTrie)->index[__lead & 0x1f] + __t1, 1))) { ++(src); } else { __lead = (normTrie )->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET; } } ( norm16) = ((normTrie)->data.ptr16[__lead]); } while (false ); | ||||||||
| 1908 | if (!isCompYesAndZeroCC(norm16)) { | ||||||||
| 1909 | break; | ||||||||
| 1910 | } | ||||||||
| 1911 | } | ||||||||
| 1912 | } | ||||||||
| 1913 | // isCompYesAndZeroCC(norm16) is false, that is, norm16>=minNoNo. | ||||||||
| 1914 | // The current character is either a "noNo" (has a mapping) | ||||||||
| 1915 | // or a "maybeYes" / "maybeNo" (combines backward) | ||||||||
| 1916 | // or a "yesYes" with ccc!=0. | ||||||||
| 1917 | // It is not a Hangul syllable or Jamo L because those have "yes" properties. | ||||||||
| 1918 | |||||||||
| 1919 | // Medium-fast path: Handle cases that do not require full decomposition and recomposition. | ||||||||
| 1920 | if (norm16 < minMaybeNo) { // minNoNo <= norm16 < minMaybeNo | ||||||||
| 1921 | if (sink == nullptr) { | ||||||||
| 1922 | return false; | ||||||||
| 1923 | } | ||||||||
| 1924 | // Fast path for mapping a character that is immediately surrounded by boundaries. | ||||||||
| 1925 | // In this case, we need not decompose around the current character. | ||||||||
| 1926 | if (isDecompNoAlgorithmic(norm16)) { | ||||||||
| 1927 | // Maps to a single isCompYesAndZeroCC character | ||||||||
| 1928 | // which also implies hasCompBoundaryBefore. | ||||||||
| 1929 | if (norm16HasCompBoundaryAfter(norm16, onlyContiguous) || | ||||||||
| 1930 | hasCompBoundaryBefore(src, limit)) { | ||||||||
| 1931 | if (prevBoundary != prevSrc && | ||||||||
| 1932 | !ByteSinkUtil::appendUnchanged(prevBoundary, prevSrc, | ||||||||
| 1933 | *sink, options, edits, errorCode)) { | ||||||||
| 1934 | break; | ||||||||
| 1935 | } | ||||||||
| 1936 | appendCodePointDelta(prevSrc, src, getAlgorithmicDelta(norm16), *sink, edits); | ||||||||
| 1937 | prevBoundary = src; | ||||||||
| 1938 | continue; | ||||||||
| 1939 | } | ||||||||
| 1940 | } else if (norm16 < minNoNoCompBoundaryBefore) { | ||||||||
| 1941 | // The mapping is comp-normalized which also implies hasCompBoundaryBefore. | ||||||||
| 1942 | if (norm16HasCompBoundaryAfter(norm16, onlyContiguous) || | ||||||||
| 1943 | hasCompBoundaryBefore(src, limit)) { | ||||||||
| 1944 | if (prevBoundary != prevSrc && | ||||||||
| 1945 | !ByteSinkUtil::appendUnchanged(prevBoundary, prevSrc, | ||||||||
| 1946 | *sink, options, edits, errorCode)) { | ||||||||
| 1947 | break; | ||||||||
| 1948 | } | ||||||||
| 1949 | const uint16_t *mapping = getDataForYesOrNo(norm16); | ||||||||
| 1950 | int32_t length = *mapping++ & MAPPING_LENGTH_MASK; | ||||||||
| 1951 | if (!ByteSinkUtil::appendChange(prevSrc, src, reinterpret_cast<const char16_t*>(mapping), length, | ||||||||
| 1952 | *sink, edits, errorCode)) { | ||||||||
| 1953 | break; | ||||||||
| 1954 | } | ||||||||
| 1955 | prevBoundary = src; | ||||||||
| 1956 | continue; | ||||||||
| 1957 | } | ||||||||
| 1958 | } else if (norm16 >= minNoNoEmpty) { | ||||||||
| 1959 | // The current character maps to nothing. | ||||||||
| 1960 | // Simply omit it from the output if there is a boundary before _or_ after it. | ||||||||
| 1961 | // The character itself implies no boundaries. | ||||||||
| 1962 | if (hasCompBoundaryBefore(src, limit) || | ||||||||
| 1963 | hasCompBoundaryAfter(prevBoundary, prevSrc, onlyContiguous)) { | ||||||||
| 1964 | if (prevBoundary != prevSrc && | ||||||||
| 1965 | !ByteSinkUtil::appendUnchanged(prevBoundary, prevSrc, | ||||||||
| 1966 | *sink, options, edits, errorCode)) { | ||||||||
| 1967 | break; | ||||||||
| 1968 | } | ||||||||
| 1969 | if (edits != nullptr) { | ||||||||
| 1970 | edits->addReplace(static_cast<int32_t>(src - prevSrc), 0); | ||||||||
| 1971 | } | ||||||||
| 1972 | prevBoundary = src; | ||||||||
| 1973 | continue; | ||||||||
| 1974 | } | ||||||||
| 1975 | } | ||||||||
| 1976 | // Other "noNo" type, or need to examine more text around this character: | ||||||||
| 1977 | // Fall through to the slow path. | ||||||||
| 1978 | } else if (isJamoVT(norm16)) { | ||||||||
| 1979 | // Jamo L: E1 84 80..92 | ||||||||
| 1980 | // Jamo V: E1 85 A1..B5 | ||||||||
| 1981 | // Jamo T: E1 86 A8..E1 87 82 | ||||||||
| 1982 | U_ASSERT((src - prevSrc) == 3 && *prevSrc == 0xe1)(static_cast <bool> ((src - prevSrc) == 3 && *prevSrc == 0xe1) ? void (0) : __assert_fail ("(src - prevSrc) == 3 && *prevSrc == 0xe1" , __builtin_FILE (), __builtin_LINE (), __extension__ __PRETTY_FUNCTION__ )); | ||||||||
| 1983 | UChar32 prev = previousHangulOrJamo(prevBoundary, prevSrc); | ||||||||
| 1984 | if (prevSrc[1] == 0x85) { | ||||||||
| 1985 | // The current character is a Jamo Vowel, | ||||||||
| 1986 | // compose with previous Jamo L and following Jamo T. | ||||||||
| 1987 | UChar32 l = prev - Hangul::JAMO_L_BASE; | ||||||||
| 1988 | if (static_cast<uint32_t>(l) < Hangul::JAMO_L_COUNT) { | ||||||||
| 1989 | if (sink == nullptr) { | ||||||||
| 1990 | return false; | ||||||||
| 1991 | } | ||||||||
| 1992 | int32_t t = getJamoTMinusBase(src, limit); | ||||||||
| 1993 | if (t >= 0) { | ||||||||
| 1994 | // The next character is a Jamo T. | ||||||||
| 1995 | src += 3; | ||||||||
| 1996 | } else if (hasCompBoundaryBefore(src, limit)) { | ||||||||
| 1997 | // No Jamo T follows, not even via decomposition. | ||||||||
| 1998 | t = 0; | ||||||||
| 1999 | } | ||||||||
| 2000 | if (t >= 0) { | ||||||||
| 2001 | UChar32 syllable = Hangul::HANGUL_BASE + | ||||||||
| 2002 | (l*Hangul::JAMO_V_COUNT + (prevSrc[2]-0xa1)) * | ||||||||
| 2003 | Hangul::JAMO_T_COUNT + t; | ||||||||
| 2004 | prevSrc -= 3; // Replace the Jamo L as well. | ||||||||
| 2005 | if (prevBoundary != prevSrc && | ||||||||
| 2006 | !ByteSinkUtil::appendUnchanged(prevBoundary, prevSrc, | ||||||||
| 2007 | *sink, options, edits, errorCode)) { | ||||||||
| 2008 | break; | ||||||||
| 2009 | } | ||||||||
| 2010 | ByteSinkUtil::appendCodePoint(prevSrc, src, syllable, *sink, edits); | ||||||||
| 2011 | prevBoundary = src; | ||||||||
| 2012 | continue; | ||||||||
| 2013 | } | ||||||||
| 2014 | // If we see L+V+x where x!=T then we drop to the slow path, | ||||||||
| 2015 | // decompose and recompose. | ||||||||
| 2016 | // This is to deal with NFKC finding normal L and V but a | ||||||||
| 2017 | // compatibility variant of a T. | ||||||||
| 2018 | // We need to either fully compose that combination here | ||||||||
| 2019 | // (which would complicate the code and may not work with strange custom data) | ||||||||
| 2020 | // or use the slow path. | ||||||||
| 2021 | } | ||||||||
| 2022 | } else if (Hangul::isHangulLV(prev)) { | ||||||||
| 2023 | // The current character is a Jamo Trailing consonant, | ||||||||
| 2024 | // compose with previous Hangul LV that does not contain a Jamo T. | ||||||||
| 2025 | if (sink == nullptr) { | ||||||||
| 2026 | return false; | ||||||||
| 2027 | } | ||||||||
| 2028 | UChar32 syllable = prev + getJamoTMinusBase(prevSrc, src); | ||||||||
| 2029 | prevSrc -= 3; // Replace the Hangul LV as well. | ||||||||
| 2030 | if (prevBoundary != prevSrc && | ||||||||
| 2031 | !ByteSinkUtil::appendUnchanged(prevBoundary, prevSrc, | ||||||||
| 2032 | *sink, options, edits, errorCode)) { | ||||||||
| 2033 | break; | ||||||||
| 2034 | } | ||||||||
| 2035 | ByteSinkUtil::appendCodePoint(prevSrc, src, syllable, *sink, edits); | ||||||||
| 2036 | prevBoundary = src; | ||||||||
| 2037 | continue; | ||||||||
| 2038 | } | ||||||||
| 2039 | // No matching context, or may need to decompose surrounding text first: | ||||||||
| 2040 | // Fall through to the slow path. | ||||||||
| 2041 | } else if (norm16 > JAMO_VT) { // norm16 >= MIN_YES_YES_WITH_CC | ||||||||
| 2042 | // One or more combining marks that do not combine-back: | ||||||||
| 2043 | // Check for canonical order, copy unchanged if ok and | ||||||||
| 2044 | // if followed by a character with a boundary-before. | ||||||||
| 2045 | uint8_t cc = getCCFromNormalYesOrMaybe(norm16); // cc!=0 | ||||||||
| 2046 | if (onlyContiguous /* FCC */ && getPreviousTrailCC(prevBoundary, prevSrc) > cc) { | ||||||||
| 2047 | // Fails FCD test, need to decompose and contiguously recompose. | ||||||||
| 2048 | if (sink == nullptr) { | ||||||||
| 2049 | return false; | ||||||||
| 2050 | } | ||||||||
| 2051 | } else { | ||||||||
| 2052 | // If !onlyContiguous (not FCC), then we ignore the tccc of | ||||||||
| 2053 | // the previous character which passed the quick check "yes && ccc==0" test. | ||||||||
| 2054 | const uint8_t *nextSrc; | ||||||||
| 2055 | uint16_t n16; | ||||||||
| 2056 | for (;;) { | ||||||||
| 2057 | if (src == limit) { | ||||||||
| 2058 | if (sink != nullptr) { | ||||||||
| 2059 | ByteSinkUtil::appendUnchanged(prevBoundary, limit, | ||||||||
| 2060 | *sink, options, edits, errorCode); | ||||||||
| 2061 | } | ||||||||
| 2062 | return true; | ||||||||
| 2063 | } | ||||||||
| 2064 | uint8_t prevCC = cc; | ||||||||
| 2065 | nextSrc = src; | ||||||||
| 2066 | UCPTRIE_FAST_U8_NEXT(normTrie, UCPTRIE_16, nextSrc, limit, n16)do { int32_t __lead = (uint8_t)*(nextSrc)++; if (!(((__lead)& 0x80)==0)) { uint8_t __t1, __t2, __t3; if ((nextSrc) != (limit ) && (__lead >= 0xe0 ? __lead < 0xf0 ? "\x20\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x10\x30\x30" [__lead &= 0xf] & (1 << ((__t1 = *(nextSrc)) >> 5)) && ++(nextSrc) != (limit) && (__t2 = *(nextSrc ) - 0x80) <= 0x3f && (__lead = ((int32_t)(normTrie )->index[(__lead << 6) + (__t1 & 0x3f)]) + __t2, 1) : (__lead -= 0xf0) <= 4 && "\x00\x00\x00\x00\x00\x00\x00\x00\x1E\x0F\x0F\x0F\x00\x00\x00\x00" [(__t1 = *(nextSrc)) >> 4] & (1 << __lead) && (__lead = (__lead << 6) | (__t1 & 0x3f), ++(nextSrc ) != (limit)) && (__t2 = *(nextSrc) - 0x80) <= 0x3f && ++(nextSrc) != (limit) && (__t3 = *(nextSrc ) - 0x80) <= 0x3f && (__lead = __lead >= (normTrie )->shifted12HighStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallU8Index_77((normTrie), __lead, __t2, __t3 ), 1) : __lead >= 0xc2 && (__t1 = *(nextSrc) - 0x80 ) <= 0x3f && (__lead = (int32_t)(normTrie)->index [__lead & 0x1f] + __t1, 1))) { ++(nextSrc); } else { __lead = (normTrie)->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET ; } } (n16) = ((normTrie)->data.ptr16[__lead]); } while (false ); | ||||||||
| 2067 | if (n16 >= MIN_YES_YES_WITH_CC) { | ||||||||
| 2068 | cc = getCCFromNormalYesOrMaybe(n16); | ||||||||
| 2069 | if (prevCC > cc) { | ||||||||
| 2070 | if (sink == nullptr) { | ||||||||
| 2071 | return false; | ||||||||
| 2072 | } | ||||||||
| 2073 | break; | ||||||||
| 2074 | } | ||||||||
| 2075 | } else { | ||||||||
| 2076 | break; | ||||||||
| 2077 | } | ||||||||
| 2078 | src = nextSrc; | ||||||||
| 2079 | } | ||||||||
| 2080 | // src is after the last in-order combining mark. | ||||||||
| 2081 | // If there is a boundary here, then we continue with no change. | ||||||||
| 2082 | if (norm16HasCompBoundaryBefore(n16)) { | ||||||||
| 2083 | if (isCompYesAndZeroCC(n16)) { | ||||||||
| 2084 | src = nextSrc; | ||||||||
| 2085 | } | ||||||||
| 2086 | continue; | ||||||||
| 2087 | } | ||||||||
| 2088 | // Use the slow path. There is no boundary in [prevSrc, src[. | ||||||||
| 2089 | } | ||||||||
| 2090 | } | ||||||||
| 2091 | |||||||||
| 2092 | // Slow path: Find the nearest boundaries around the current character, | ||||||||
| 2093 | // decompose and recompose. | ||||||||
| 2094 | if (prevBoundary != prevSrc && !norm16HasCompBoundaryBefore(norm16)) { | ||||||||
| 2095 | const uint8_t *p = prevSrc; | ||||||||
| 2096 | UCPTRIE_FAST_U8_PREV(normTrie, UCPTRIE_16, prevBoundary, p, norm16)do { int32_t __index = (uint8_t)*--(p); if (!(((__index)& 0x80)==0)) { __index = ucptrie_internalU8PrevIndex_77((normTrie ), __index, (const uint8_t *)(prevBoundary), (const uint8_t * )(p)); (p) -= __index & 7; __index >>= 3; } (norm16 ) = ((normTrie)->data.ptr16[__index]); } while (false); | ||||||||
| 2097 | if (!norm16HasCompBoundaryAfter(norm16, onlyContiguous)) { | ||||||||
| 2098 | prevSrc = p; | ||||||||
| 2099 | } | ||||||||
| 2100 | } | ||||||||
| 2101 | ReorderingBuffer buffer(*this, s16, errorCode); | ||||||||
| 2102 | if (U_FAILURE(errorCode)) { | ||||||||
| 2103 | break; | ||||||||
| 2104 | } | ||||||||
| 2105 | // We know there is not a boundary here. | ||||||||
| 2106 | decomposeShort(prevSrc, src, STOP_AT_LIMIT, onlyContiguous, | ||||||||
| 2107 | buffer, errorCode); | ||||||||
| 2108 | // Decompose until the next boundary. | ||||||||
| 2109 | src = decomposeShort(src, limit, STOP_AT_COMP_BOUNDARY, onlyContiguous, | ||||||||
| 2110 | buffer, errorCode); | ||||||||
| 2111 | if (U_FAILURE(errorCode)) { | ||||||||
| 2112 | break; | ||||||||
| 2113 | } | ||||||||
| 2114 | if ((src - prevSrc) > INT32_MAX(2147483647)) { // guard before buffer.equals() | ||||||||
| 2115 | errorCode = U_INDEX_OUTOFBOUNDS_ERROR; | ||||||||
| 2116 | return true; | ||||||||
| 2117 | } | ||||||||
| 2118 | recompose(buffer, 0, onlyContiguous); | ||||||||
| 2119 | if (!buffer.equals(prevSrc, src)) { | ||||||||
| 2120 | if (sink == nullptr) { | ||||||||
| 2121 | return false; | ||||||||
| 2122 | } | ||||||||
| 2123 | if (prevBoundary != prevSrc && | ||||||||
| 2124 | !ByteSinkUtil::appendUnchanged(prevBoundary, prevSrc, | ||||||||
| 2125 | *sink, options, edits, errorCode)) { | ||||||||
| 2126 | break; | ||||||||
| 2127 | } | ||||||||
| 2128 | if (!ByteSinkUtil::appendChange(prevSrc, src, buffer.getStart(), buffer.length(), | ||||||||
| 2129 | *sink, edits, errorCode)) { | ||||||||
| 2130 | break; | ||||||||
| 2131 | } | ||||||||
| 2132 | prevBoundary = src; | ||||||||
| 2133 | } | ||||||||
| 2134 | } | ||||||||
| 2135 | return true; | ||||||||
| 2136 | } | ||||||||
| 2137 | |||||||||
| 2138 | UBool Normalizer2Impl::hasCompBoundaryBefore(const char16_t *src, const char16_t *limit) const { | ||||||||
| 2139 | if (src == limit || *src < minCompNoMaybeCP) { | ||||||||
| 2140 | return true; | ||||||||
| 2141 | } | ||||||||
| 2142 | UChar32 c; | ||||||||
| 2143 | uint16_t norm16; | ||||||||
| 2144 | UCPTRIE_FAST_U16_NEXT(normTrie, UCPTRIE_16, src, limit, c, norm16)do { (c) = *(src)++; int32_t __index; if (!(((c)&0xfffff800 )==0xd800)) { __index = ((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT] + ((c) & UCPTRIE_FAST_DATA_MASK)); } else { uint16_t __c2; if ((((c)&0x400)==0) && (src ) != (limit) && (((__c2 = *(src))&0xfffffc00)==0xdc00 )) { ++(src); (c) = (((UChar32)((c))<<10UL)+(UChar32)(__c2 )-((0xd800<<10UL)+0xdc00-0x10000)); __index = ((c) >= (normTrie)->highStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77(normTrie, c)); } else { __index = (normTrie)->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET ; } } (norm16) = ((normTrie)->data.ptr16[__index]); } while (false); | ||||||||
| 2145 | return norm16HasCompBoundaryBefore(norm16); | ||||||||
| 2146 | } | ||||||||
| 2147 | |||||||||
| 2148 | UBool Normalizer2Impl::hasCompBoundaryBefore(const uint8_t *src, const uint8_t *limit) const { | ||||||||
| 2149 | if (src == limit) { | ||||||||
| 2150 | return true; | ||||||||
| 2151 | } | ||||||||
| 2152 | uint16_t norm16; | ||||||||
| 2153 | UCPTRIE_FAST_U8_NEXT(normTrie, UCPTRIE_16, src, limit, norm16)do { int32_t __lead = (uint8_t)*(src)++; if (!(((__lead)& 0x80)==0)) { uint8_t __t1, __t2, __t3; if ((src) != (limit) && (__lead >= 0xe0 ? __lead < 0xf0 ? "\x20\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x10\x30\x30" [__lead &= 0xf] & (1 << ((__t1 = *(src)) >> 5)) && ++(src) != (limit) && (__t2 = *(src) - 0x80) <= 0x3f && (__lead = ((int32_t)(normTrie)-> index[(__lead << 6) + (__t1 & 0x3f)]) + __t2, 1) : ( __lead -= 0xf0) <= 4 && "\x00\x00\x00\x00\x00\x00\x00\x00\x1E\x0F\x0F\x0F\x00\x00\x00\x00" [(__t1 = *(src)) >> 4] & (1 << __lead) && (__lead = (__lead << 6) | (__t1 & 0x3f), ++(src) != (limit)) && (__t2 = *(src) - 0x80) <= 0x3f && ++(src) != (limit) && (__t3 = *(src) - 0x80) <= 0x3f && (__lead = __lead >= (normTrie)->shifted12HighStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallU8Index_77((normTrie), __lead, __t2, __t3 ), 1) : __lead >= 0xc2 && (__t1 = *(src) - 0x80) <= 0x3f && (__lead = (int32_t)(normTrie)->index[__lead & 0x1f] + __t1, 1))) { ++(src); } else { __lead = (normTrie )->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET; } } ( norm16) = ((normTrie)->data.ptr16[__lead]); } while (false ); | ||||||||
| 2154 | return norm16HasCompBoundaryBefore(norm16); | ||||||||
| 2155 | } | ||||||||
| 2156 | |||||||||
| 2157 | UBool Normalizer2Impl::hasCompBoundaryAfter(const char16_t *start, const char16_t *p, | ||||||||
| 2158 | UBool onlyContiguous) const { | ||||||||
| 2159 | if (start == p) { | ||||||||
| 2160 | return true; | ||||||||
| 2161 | } | ||||||||
| 2162 | UChar32 c; | ||||||||
| 2163 | uint16_t norm16; | ||||||||
| 2164 | UCPTRIE_FAST_U16_PREV(normTrie, UCPTRIE_16, start, p, c, norm16)do { (c) = *--(p); int32_t __index; if (!(((c)&0xfffff800 )==0xd800)) { __index = ((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT] + ((c) & UCPTRIE_FAST_DATA_MASK)); } else { uint16_t __c2; if ((((c)&0x400)!=0) && (p ) != (start) && (((__c2 = *((p) - 1))&0xfffffc00) ==0xd800)) { --(p); (c) = (((UChar32)(__c2)<<10UL)+(UChar32 )((c))-((0xd800<<10UL)+0xdc00-0x10000)); __index = ((c) >= (normTrie)->highStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77 (normTrie, c)); } else { __index = (normTrie)->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET; } } (norm16) = ((normTrie )->data.ptr16[__index]); } while (false); | ||||||||
| 2165 | return norm16HasCompBoundaryAfter(norm16, onlyContiguous); | ||||||||
| 2166 | } | ||||||||
| 2167 | |||||||||
| 2168 | UBool Normalizer2Impl::hasCompBoundaryAfter(const uint8_t *start, const uint8_t *p, | ||||||||
| 2169 | UBool onlyContiguous) const { | ||||||||
| 2170 | if (start == p) { | ||||||||
| 2171 | return true; | ||||||||
| 2172 | } | ||||||||
| 2173 | uint16_t norm16; | ||||||||
| 2174 | UCPTRIE_FAST_U8_PREV(normTrie, UCPTRIE_16, start, p, norm16)do { int32_t __index = (uint8_t)*--(p); if (!(((__index)& 0x80)==0)) { __index = ucptrie_internalU8PrevIndex_77((normTrie ), __index, (const uint8_t *)(start), (const uint8_t *)(p)); ( p) -= __index & 7; __index >>= 3; } (norm16) = ((normTrie )->data.ptr16[__index]); } while (false); | ||||||||
| 2175 | return norm16HasCompBoundaryAfter(norm16, onlyContiguous); | ||||||||
| 2176 | } | ||||||||
| 2177 | |||||||||
| 2178 | const char16_t *Normalizer2Impl::findPreviousCompBoundary(const char16_t *start, const char16_t *p, | ||||||||
| 2179 | UBool onlyContiguous) const { | ||||||||
| 2180 | while (p != start) { | ||||||||
| 2181 | const char16_t *codePointLimit = p; | ||||||||
| 2182 | UChar32 c; | ||||||||
| 2183 | uint16_t norm16; | ||||||||
| 2184 | UCPTRIE_FAST_U16_PREV(normTrie, UCPTRIE_16, start, p, c, norm16)do { (c) = *--(p); int32_t __index; if (!(((c)&0xfffff800 )==0xd800)) { __index = ((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT] + ((c) & UCPTRIE_FAST_DATA_MASK)); } else { uint16_t __c2; if ((((c)&0x400)!=0) && (p ) != (start) && (((__c2 = *((p) - 1))&0xfffffc00) ==0xd800)) { --(p); (c) = (((UChar32)(__c2)<<10UL)+(UChar32 )((c))-((0xd800<<10UL)+0xdc00-0x10000)); __index = ((c) >= (normTrie)->highStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77 (normTrie, c)); } else { __index = (normTrie)->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET; } } (norm16) = ((normTrie )->data.ptr16[__index]); } while (false); | ||||||||
| 2185 | if (norm16HasCompBoundaryAfter(norm16, onlyContiguous)) { | ||||||||
| 2186 | return codePointLimit; | ||||||||
| 2187 | } | ||||||||
| 2188 | if (hasCompBoundaryBefore(c, norm16)) { | ||||||||
| 2189 | return p; | ||||||||
| 2190 | } | ||||||||
| 2191 | } | ||||||||
| 2192 | return p; | ||||||||
| 2193 | } | ||||||||
| 2194 | |||||||||
| 2195 | const char16_t *Normalizer2Impl::findNextCompBoundary(const char16_t *p, const char16_t *limit, | ||||||||
| 2196 | UBool onlyContiguous) const { | ||||||||
| 2197 | while (p != limit) { | ||||||||
| 2198 | const char16_t *codePointStart = p; | ||||||||
| 2199 | UChar32 c; | ||||||||
| 2200 | uint16_t norm16; | ||||||||
| 2201 | UCPTRIE_FAST_U16_NEXT(normTrie, UCPTRIE_16, p, limit, c, norm16)do { (c) = *(p)++; int32_t __index; if (!(((c)&0xfffff800 )==0xd800)) { __index = ((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT] + ((c) & UCPTRIE_FAST_DATA_MASK)); } else { uint16_t __c2; if ((((c)&0x400)==0) && (p ) != (limit) && (((__c2 = *(p))&0xfffffc00)==0xdc00 )) { ++(p); (c) = (((UChar32)((c))<<10UL)+(UChar32)(__c2 )-((0xd800<<10UL)+0xdc00-0x10000)); __index = ((c) >= (normTrie)->highStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77(normTrie, c)); } else { __index = (normTrie)->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET ; } } (norm16) = ((normTrie)->data.ptr16[__index]); } while (false); | ||||||||
| 2202 | if (hasCompBoundaryBefore(c, norm16)) { | ||||||||
| 2203 | return codePointStart; | ||||||||
| 2204 | } | ||||||||
| 2205 | if (norm16HasCompBoundaryAfter(norm16, onlyContiguous)) { | ||||||||
| 2206 | return p; | ||||||||
| 2207 | } | ||||||||
| 2208 | } | ||||||||
| 2209 | return p; | ||||||||
| 2210 | } | ||||||||
| 2211 | |||||||||
| 2212 | uint8_t Normalizer2Impl::getPreviousTrailCC(const char16_t *start, const char16_t *p) const { | ||||||||
| 2213 | if (start == p) { | ||||||||
| 2214 | return 0; | ||||||||
| 2215 | } | ||||||||
| 2216 | int32_t i = static_cast<int32_t>(p - start); | ||||||||
| 2217 | UChar32 c; | ||||||||
| 2218 | U16_PREV(start, 0, i, c)do { (c)=(start)[--(i)]; if((((c)&0xfffffc00)==0xdc00)) { uint16_t __c2; if((i)>(0) && (((__c2=(start)[(i)- 1])&0xfffffc00)==0xd800)) { --(i); (c)=(((UChar32)(__c2)<< 10UL)+(UChar32)((c))-((0xd800<<10UL)+0xdc00-0x10000)); } } } while (false); | ||||||||
| 2219 | return static_cast<uint8_t>(getFCD16(c)); | ||||||||
| 2220 | } | ||||||||
| 2221 | |||||||||
| 2222 | uint8_t Normalizer2Impl::getPreviousTrailCC(const uint8_t *start, const uint8_t *p) const { | ||||||||
| 2223 | if (start == p) { | ||||||||
| 2224 | return 0; | ||||||||
| 2225 | } | ||||||||
| 2226 | int32_t i = static_cast<int32_t>(p - start); | ||||||||
| 2227 | UChar32 c; | ||||||||
| 2228 | U8_PREV(start, 0, i, c)do { (c)=(uint8_t)(start)[--(i)]; if(!(((c)&0x80)==0)) { ( c)=utf8_prevCharSafeBody_77((const uint8_t *)start, 0, &( i), c, -1); } } while (false); | ||||||||
| 2229 | return static_cast<uint8_t>(getFCD16(c)); | ||||||||
| 2230 | } | ||||||||
| 2231 | |||||||||
| 2232 | // Note: normalizer2impl.cpp r30982 (2011-nov-27) | ||||||||
| 2233 | // still had getFCDTrie() which built and cached an FCD trie. | ||||||||
| 2234 | // That provided faster access to FCD data than getFCD16FromNormData() | ||||||||
| 2235 | // but required synchronization and consumed some 10kB of heap memory | ||||||||
| 2236 | // in any process that uses FCD (e.g., via collation). | ||||||||
| 2237 | // minDecompNoCP etc. and smallFCD[] are intended to help with any loss of performance, | ||||||||
| 2238 | // at least for ASCII & CJK. | ||||||||
| 2239 | |||||||||
| 2240 | // Ticket 20907 - The optimizer in MSVC/Visual Studio versions below 16.4 has trouble with this | ||||||||
| 2241 | // function on Windows ARM64. As a work-around, we disable optimizations for this function. | ||||||||
| 2242 | // This work-around could/should be removed once the following versions of Visual Studio are no | ||||||||
| 2243 | // longer supported: All versions of VS2017, and versions of VS2019 below 16.4. | ||||||||
| 2244 | #if (defined(_MSC_VER) && (defined(_M_ARM64)) && (_MSC_VER < 1924)) | ||||||||
| 2245 | #pragma optimize( "", off ) | ||||||||
| 2246 | #endif | ||||||||
| 2247 | // Gets the FCD value from the regular normalization data. | ||||||||
| 2248 | uint16_t Normalizer2Impl::getFCD16FromNormData(UChar32 c) const { | ||||||||
| 2249 | uint16_t norm16=getNorm16(c); | ||||||||
| 2250 | if (norm16 >= limitNoNo) { | ||||||||
| 2251 | if(norm16>=MIN_NORMAL_MAYBE_YES) { | ||||||||
| 2252 | // combining mark | ||||||||
| 2253 | norm16=getCCFromNormalYesOrMaybe(norm16); | ||||||||
| 2254 | return norm16|(norm16<<8); | ||||||||
| 2255 | } else if(norm16>=minMaybeYes) { | ||||||||
| 2256 | return 0; | ||||||||
| 2257 | } else if(norm16<minMaybeNo) { // isDecompNoAlgorithmic(norm16) | ||||||||
| 2258 | uint16_t deltaTrailCC = norm16 & DELTA_TCCC_MASK; | ||||||||
| 2259 | if (deltaTrailCC <= DELTA_TCCC_1) { | ||||||||
| 2260 | return deltaTrailCC >> OFFSET_SHIFT; | ||||||||
| 2261 | } | ||||||||
| 2262 | // Maps to an isCompYesAndZeroCC. | ||||||||
| 2263 | c=mapAlgorithmic(c, norm16); | ||||||||
| 2264 | norm16=getRawNorm16(c); | ||||||||
| 2265 | } | ||||||||
| 2266 | } | ||||||||
| 2267 | if(norm16<=minYesNo || isHangulLVT(norm16)) { | ||||||||
| 2268 | // no decomposition or Hangul syllable, all zeros | ||||||||
| 2269 | return 0; | ||||||||
| 2270 | } | ||||||||
| 2271 | // c decomposes, get everything from the variable-length extra data | ||||||||
| 2272 | const uint16_t *mapping=getData(norm16); | ||||||||
| 2273 | uint16_t firstUnit=*mapping; | ||||||||
| 2274 | norm16=firstUnit>>8; // tccc | ||||||||
| 2275 | if(firstUnit&MAPPING_HAS_CCC_LCCC_WORD) { | ||||||||
| 2276 | norm16|=*(mapping-1)&0xff00; // lccc | ||||||||
| 2277 | } | ||||||||
| 2278 | return norm16; | ||||||||
| 2279 | } | ||||||||
| 2280 | #if (defined(_MSC_VER) && (defined(_M_ARM64)) && (_MSC_VER < 1924)) | ||||||||
| 2281 | #pragma optimize( "", on ) | ||||||||
| 2282 | #endif | ||||||||
| 2283 | |||||||||
| 2284 | uint16_t Normalizer2Impl::getFCD16FromMaybeOrNonZeroCC(uint16_t norm16) const { | ||||||||
| 2285 | U_ASSERT(norm16 >= minMaybeNo)(static_cast <bool> (norm16 >= minMaybeNo) ? void (0 ) : __assert_fail ("norm16 >= minMaybeNo", __builtin_FILE ( ), __builtin_LINE (), __extension__ __PRETTY_FUNCTION__)); | ||||||||
| 2286 | if (norm16 >= MIN_NORMAL_MAYBE_YES) { | ||||||||
| 2287 | // combining mark | ||||||||
| 2288 | norm16 = getCCFromNormalYesOrMaybe(norm16); | ||||||||
| 2289 | return norm16 | (norm16<<8); | ||||||||
| 2290 | } else if (norm16 >= minMaybeYes) { | ||||||||
| 2291 | return 0; | ||||||||
| 2292 | } | ||||||||
| 2293 | // c decomposes, get everything from the variable-length extra data | ||||||||
| 2294 | const uint16_t *mapping = getDataForMaybe(norm16); | ||||||||
| 2295 | uint16_t firstUnit = *mapping; | ||||||||
| 2296 | // maybeNo has lccc = 0 | ||||||||
| 2297 | U_ASSERT((firstUnit & MAPPING_HAS_CCC_LCCC_WORD) == 0 || (*(mapping - 1) & 0xff00) == 0)(static_cast <bool> ((firstUnit & MAPPING_HAS_CCC_LCCC_WORD ) == 0 || (*(mapping - 1) & 0xff00) == 0) ? void (0) : __assert_fail ("(firstUnit & MAPPING_HAS_CCC_LCCC_WORD) == 0 || (*(mapping - 1) & 0xff00) == 0" , __builtin_FILE (), __builtin_LINE (), __extension__ __PRETTY_FUNCTION__ )); | ||||||||
| 2298 | return firstUnit >> 8; // tccc | ||||||||
| 2299 | } | ||||||||
| 2300 | |||||||||
| 2301 | // Dual functionality: | ||||||||
| 2302 | // buffer!=nullptr: normalize | ||||||||
| 2303 | // buffer==nullptr: isNormalized/quickCheck/spanQuickCheckYes | ||||||||
| 2304 | const char16_t * | ||||||||
| 2305 | Normalizer2Impl::makeFCD(const char16_t *src, const char16_t *limit, | ||||||||
| 2306 | ReorderingBuffer *buffer, | ||||||||
| 2307 | UErrorCode &errorCode) const { | ||||||||
| 2308 | // Tracks the last FCD-safe boundary, before lccc=0 or after properly-ordered tccc<=1. | ||||||||
| 2309 | // Similar to the prevBoundary in the compose() implementation. | ||||||||
| 2310 | const char16_t *prevBoundary=src; | ||||||||
| 2311 | int32_t prevFCD16=0; | ||||||||
| 2312 | if(limit==nullptr) { | ||||||||
| 2313 | src=copyLowPrefixFromNulTerminated(src, minLcccCP, buffer, errorCode); | ||||||||
| 2314 | if(U_FAILURE(errorCode)) { | ||||||||
| 2315 | return src; | ||||||||
| 2316 | } | ||||||||
| 2317 | if(prevBoundary<src) { | ||||||||
| 2318 | prevBoundary=src; | ||||||||
| 2319 | // We know that the previous character's lccc==0. | ||||||||
| 2320 | // Fetching the fcd16 value was deferred for this below-U+0300 code point. | ||||||||
| 2321 | prevFCD16=getFCD16(*(src-1)); | ||||||||
| 2322 | if(prevFCD16>1) { | ||||||||
| 2323 | --prevBoundary; | ||||||||
| 2324 | } | ||||||||
| 2325 | } | ||||||||
| 2326 | limit=u_strchru_strchr_77(src, 0); | ||||||||
| 2327 | } | ||||||||
| 2328 | |||||||||
| 2329 | // Note: In this function we use buffer->appendZeroCC() because we track | ||||||||
| 2330 | // the lead and trail combining classes here, rather than leaving it to | ||||||||
| 2331 | // the ReorderingBuffer. | ||||||||
| 2332 | // The exception is the call to decomposeShort() which uses the buffer | ||||||||
| 2333 | // in the normal way. | ||||||||
| 2334 | |||||||||
| 2335 | const char16_t *prevSrc; | ||||||||
| 2336 | UChar32 c=0; | ||||||||
| 2337 | uint16_t fcd16=0; | ||||||||
| 2338 | |||||||||
| 2339 | for(;;) { | ||||||||
| 2340 | // count code units with lccc==0 | ||||||||
| 2341 | for(prevSrc=src; src!=limit;) { | ||||||||
| 2342 | if((c=*src)<minLcccCP) { | ||||||||
| 2343 | prevFCD16=~c; | ||||||||
| 2344 | ++src; | ||||||||
| 2345 | } else if(!singleLeadMightHaveNonZeroFCD16(c)) { | ||||||||
| 2346 | prevFCD16=0; | ||||||||
| 2347 | ++src; | ||||||||
| 2348 | } else { | ||||||||
| 2349 | if(U16_IS_LEAD(c)(((c)&0xfffffc00)==0xd800)) { | ||||||||
| 2350 | char16_t c2; | ||||||||
| 2351 | if((src+1)!=limit && U16_IS_TRAIL(c2=src[1])(((c2=src[1])&0xfffffc00)==0xdc00)) { | ||||||||
| 2352 | c=U16_GET_SUPPLEMENTARY(c, c2)(((UChar32)(c)<<10UL)+(UChar32)(c2)-((0xd800<<10UL )+0xdc00-0x10000)); | ||||||||
| 2353 | } | ||||||||
| 2354 | } | ||||||||
| 2355 | if((fcd16=getFCD16FromNormData(c))<=0xff) { | ||||||||
| 2356 | prevFCD16=fcd16; | ||||||||
| 2357 | src+=U16_LENGTH(c)((uint32_t)(c)<=0xffff ? 1 : 2); | ||||||||
| 2358 | } else { | ||||||||
| 2359 | break; | ||||||||
| 2360 | } | ||||||||
| 2361 | } | ||||||||
| 2362 | } | ||||||||
| 2363 | // copy these code units all at once | ||||||||
| 2364 | if(src!=prevSrc) { | ||||||||
| 2365 | if(buffer!=nullptr && !buffer->appendZeroCC(prevSrc, src, errorCode)) { | ||||||||
| 2366 | break; | ||||||||
| 2367 | } | ||||||||
| 2368 | if(src==limit) { | ||||||||
| 2369 | break; | ||||||||
| 2370 | } | ||||||||
| 2371 | prevBoundary=src; | ||||||||
| 2372 | // We know that the previous character's lccc==0. | ||||||||
| 2373 | if(prevFCD16<0) { | ||||||||
| 2374 | // Fetching the fcd16 value was deferred for this below-minLcccCP code point. | ||||||||
| 2375 | UChar32 prev=~prevFCD16; | ||||||||
| 2376 | if(prev<minDecompNoCP) { | ||||||||
| 2377 | prevFCD16=0; | ||||||||
| 2378 | } else { | ||||||||
| 2379 | prevFCD16=getFCD16FromNormData(prev); | ||||||||
| 2380 | if(prevFCD16>1) { | ||||||||
| 2381 | --prevBoundary; | ||||||||
| 2382 | } | ||||||||
| 2383 | } | ||||||||
| 2384 | } else { | ||||||||
| 2385 | const char16_t *p=src-1; | ||||||||
| 2386 | if(U16_IS_TRAIL(*p)(((*p)&0xfffffc00)==0xdc00) && prevSrc<p && U16_IS_LEAD(*(p-1))(((*(p-1))&0xfffffc00)==0xd800)) { | ||||||||
| 2387 | --p; | ||||||||
| 2388 | // Need to fetch the previous character's FCD value because | ||||||||
| 2389 | // prevFCD16 was just for the trail surrogate code point. | ||||||||
| 2390 | prevFCD16=getFCD16FromNormData(U16_GET_SUPPLEMENTARY(p[0], p[1])(((UChar32)(p[0])<<10UL)+(UChar32)(p[1])-((0xd800<< 10UL)+0xdc00-0x10000))); | ||||||||
| 2391 | // Still known to have lccc==0 because its lead surrogate unit had lccc==0. | ||||||||
| 2392 | } | ||||||||
| 2393 | if(prevFCD16>1) { | ||||||||
| 2394 | prevBoundary=p; | ||||||||
| 2395 | } | ||||||||
| 2396 | } | ||||||||
| 2397 | // The start of the current character (c). | ||||||||
| 2398 | prevSrc=src; | ||||||||
| 2399 | } else if(src==limit) { | ||||||||
| 2400 | break; | ||||||||
| 2401 | } | ||||||||
| 2402 | |||||||||
| 2403 | src+=U16_LENGTH(c)((uint32_t)(c)<=0xffff ? 1 : 2); | ||||||||
| 2404 | // The current character (c) at [prevSrc..src[ has a non-zero lead combining class. | ||||||||
| 2405 | // Check for proper order, and decompose locally if necessary. | ||||||||
| 2406 | if((prevFCD16&0xff)<=(fcd16>>8)) { | ||||||||
| 2407 | // proper order: prev tccc <= current lccc | ||||||||
| 2408 | if((fcd16&0xff)<=1) { | ||||||||
| 2409 | prevBoundary=src; | ||||||||
| 2410 | } | ||||||||
| 2411 | if(buffer!=nullptr && !buffer->appendZeroCC(c, errorCode)) { | ||||||||
| 2412 | break; | ||||||||
| 2413 | } | ||||||||
| 2414 | prevFCD16=fcd16; | ||||||||
| 2415 | continue; | ||||||||
| 2416 | } else if(buffer==nullptr) { | ||||||||
| 2417 | return prevBoundary; // quick check "no" | ||||||||
| 2418 | } else { | ||||||||
| 2419 | /* | ||||||||
| 2420 | * Back out the part of the source that we copied or appended | ||||||||
| 2421 | * already but is now going to be decomposed. | ||||||||
| 2422 | * prevSrc is set to after what was copied/appended. | ||||||||
| 2423 | */ | ||||||||
| 2424 | buffer->removeSuffix(static_cast<int32_t>(prevSrc - prevBoundary)); | ||||||||
| 2425 | /* | ||||||||
| 2426 | * Find the part of the source that needs to be decomposed, | ||||||||
| 2427 | * up to the next safe boundary. | ||||||||
| 2428 | */ | ||||||||
| 2429 | src=findNextFCDBoundary(src, limit); | ||||||||
| 2430 | /* | ||||||||
| 2431 | * The source text does not fulfill the conditions for FCD. | ||||||||
| 2432 | * Decompose and reorder a limited piece of the text. | ||||||||
| 2433 | */ | ||||||||
| 2434 | decomposeShort(prevBoundary, src, false, false, *buffer, errorCode); | ||||||||
| 2435 | if (U_FAILURE(errorCode)) { | ||||||||
| 2436 | break; | ||||||||
| 2437 | } | ||||||||
| 2438 | prevBoundary=src; | ||||||||
| 2439 | prevFCD16=0; | ||||||||
| 2440 | } | ||||||||
| 2441 | } | ||||||||
| 2442 | return src; | ||||||||
| 2443 | } | ||||||||
| 2444 | |||||||||
| 2445 | void Normalizer2Impl::makeFCDAndAppend(const char16_t *src, const char16_t *limit, | ||||||||
| 2446 | UBool doMakeFCD, | ||||||||
| 2447 | UnicodeString &safeMiddle, | ||||||||
| 2448 | ReorderingBuffer &buffer, | ||||||||
| 2449 | UErrorCode &errorCode) const { | ||||||||
| 2450 | if(!buffer.isEmpty()) { | ||||||||
| 2451 | const char16_t *firstBoundaryInSrc=findNextFCDBoundary(src, limit); | ||||||||
| 2452 | if(src!=firstBoundaryInSrc) { | ||||||||
| 2453 | const char16_t *lastBoundaryInDest=findPreviousFCDBoundary(buffer.getStart(), | ||||||||
| 2454 | buffer.getLimit()); | ||||||||
| 2455 | int32_t destSuffixLength = static_cast<int32_t>(buffer.getLimit() - lastBoundaryInDest); | ||||||||
| 2456 | UnicodeString middle(lastBoundaryInDest, destSuffixLength); | ||||||||
| 2457 | buffer.removeSuffix(destSuffixLength); | ||||||||
| 2458 | safeMiddle=middle; | ||||||||
| 2459 | middle.append(src, static_cast<int32_t>(firstBoundaryInSrc - src)); | ||||||||
| 2460 | const char16_t *middleStart=middle.getBuffer(); | ||||||||
| 2461 | makeFCD(middleStart, middleStart+middle.length(), &buffer, errorCode); | ||||||||
| 2462 | if(U_FAILURE(errorCode)) { | ||||||||
| 2463 | return; | ||||||||
| 2464 | } | ||||||||
| 2465 | src=firstBoundaryInSrc; | ||||||||
| 2466 | } | ||||||||
| 2467 | } | ||||||||
| 2468 | if(doMakeFCD) { | ||||||||
| 2469 | makeFCD(src, limit, &buffer, errorCode); | ||||||||
| 2470 | } else { | ||||||||
| 2471 | if(limit==nullptr) { // appendZeroCC() needs limit!=nullptr | ||||||||
| 2472 | limit=u_strchru_strchr_77(src, 0); | ||||||||
| 2473 | } | ||||||||
| 2474 | buffer.appendZeroCC(src, limit, errorCode); | ||||||||
| 2475 | } | ||||||||
| 2476 | } | ||||||||
| 2477 | |||||||||
| 2478 | const char16_t *Normalizer2Impl::findPreviousFCDBoundary(const char16_t *start, const char16_t *p) const { | ||||||||
| 2479 | while(start<p) { | ||||||||
| 2480 | const char16_t *codePointLimit = p; | ||||||||
| 2481 | UChar32 c; | ||||||||
| 2482 | uint16_t norm16; | ||||||||
| 2483 | UCPTRIE_FAST_U16_PREV(normTrie, UCPTRIE_16, start, p, c, norm16)do { (c) = *--(p); int32_t __index; if (!(((c)&0xfffff800 )==0xd800)) { __index = ((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT] + ((c) & UCPTRIE_FAST_DATA_MASK)); } else { uint16_t __c2; if ((((c)&0x400)!=0) && (p ) != (start) && (((__c2 = *((p) - 1))&0xfffffc00) ==0xd800)) { --(p); (c) = (((UChar32)(__c2)<<10UL)+(UChar32 )((c))-((0xd800<<10UL)+0xdc00-0x10000)); __index = ((c) >= (normTrie)->highStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77 (normTrie, c)); } else { __index = (normTrie)->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET; } } (norm16) = ((normTrie )->data.ptr16[__index]); } while (false); | ||||||||
| 2484 | if (c < minDecompNoCP || norm16HasDecompBoundaryAfter(norm16)) { | ||||||||
| 2485 | return codePointLimit; | ||||||||
| 2486 | } | ||||||||
| 2487 | if (norm16HasDecompBoundaryBefore(norm16)) { | ||||||||
| 2488 | return p; | ||||||||
| 2489 | } | ||||||||
| 2490 | } | ||||||||
| 2491 | return p; | ||||||||
| 2492 | } | ||||||||
| 2493 | |||||||||
| 2494 | const char16_t *Normalizer2Impl::findNextFCDBoundary(const char16_t *p, const char16_t *limit) const { | ||||||||
| 2495 | while(p<limit) { | ||||||||
| 2496 | const char16_t *codePointStart=p; | ||||||||
| 2497 | UChar32 c; | ||||||||
| 2498 | uint16_t norm16; | ||||||||
| 2499 | UCPTRIE_FAST_U16_NEXT(normTrie, UCPTRIE_16, p, limit, c, norm16)do { (c) = *(p)++; int32_t __index; if (!(((c)&0xfffff800 )==0xd800)) { __index = ((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT] + ((c) & UCPTRIE_FAST_DATA_MASK)); } else { uint16_t __c2; if ((((c)&0x400)==0) && (p ) != (limit) && (((__c2 = *(p))&0xfffffc00)==0xdc00 )) { ++(p); (c) = (((UChar32)((c))<<10UL)+(UChar32)(__c2 )-((0xd800<<10UL)+0xdc00-0x10000)); __index = ((c) >= (normTrie)->highStart ? (normTrie)->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77(normTrie, c)); } else { __index = (normTrie)->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET ; } } (norm16) = ((normTrie)->data.ptr16[__index]); } while (false); | ||||||||
| 2500 | if (c < minLcccCP || norm16HasDecompBoundaryBefore(norm16)) { | ||||||||
| 2501 | return codePointStart; | ||||||||
| 2502 | } | ||||||||
| 2503 | if (norm16HasDecompBoundaryAfter(norm16)) { | ||||||||
| 2504 | return p; | ||||||||
| 2505 | } | ||||||||
| 2506 | } | ||||||||
| 2507 | return p; | ||||||||
| 2508 | } | ||||||||
| 2509 | |||||||||
| 2510 | // CanonicalIterator data -------------------------------------------------- *** | ||||||||
| 2511 | |||||||||
| 2512 | CanonIterData::CanonIterData(UErrorCode &errorCode) : | ||||||||
| 2513 | mutableTrie(umutablecptrie_openumutablecptrie_open_77(0, 0, &errorCode)), trie(nullptr), | ||||||||
| 2514 | canonStartSets(uprv_deleteUObjectuprv_deleteUObject_77, nullptr, errorCode) {} | ||||||||
| 2515 | |||||||||
| 2516 | CanonIterData::~CanonIterData() { | ||||||||
| 2517 | umutablecptrie_closeumutablecptrie_close_77(mutableTrie); | ||||||||
| 2518 | ucptrie_closeucptrie_close_77(trie); | ||||||||
| 2519 | } | ||||||||
| 2520 | |||||||||
| 2521 | void CanonIterData::addToStartSet(UChar32 origin, UChar32 decompLead, UErrorCode &errorCode) { | ||||||||
| 2522 | uint32_t canonValue = umutablecptrie_getumutablecptrie_get_77(mutableTrie, decompLead); | ||||||||
| 2523 | if((canonValue&(CANON_HAS_SET0x200000|CANON_VALUE_MASK0x1fffff))==0 && origin!=0) { | ||||||||
| 2524 | // origin is the first character whose decomposition starts with | ||||||||
| 2525 | // the character for which we are setting the value. | ||||||||
| 2526 | umutablecptrie_setumutablecptrie_set_77(mutableTrie, decompLead, canonValue|origin, &errorCode); | ||||||||
| 2527 | } else { | ||||||||
| 2528 | // origin is not the first character, or it is U+0000. | ||||||||
| 2529 | UnicodeSet *set; | ||||||||
| 2530 | if((canonValue&CANON_HAS_SET0x200000)==0) { | ||||||||
| 2531 | LocalPointer<UnicodeSet> lpSet(new UnicodeSet, errorCode); | ||||||||
| 2532 | set=lpSet.getAlias(); | ||||||||
| 2533 | if(U_FAILURE(errorCode)) { | ||||||||
| 2534 | return; | ||||||||
| 2535 | } | ||||||||
| 2536 | UChar32 firstOrigin = static_cast<UChar32>(canonValue & CANON_VALUE_MASK0x1fffff); | ||||||||
| 2537 | canonValue = (canonValue & ~CANON_VALUE_MASK0x1fffff) | CANON_HAS_SET0x200000 | static_cast<uint32_t>(canonStartSets.size()); | ||||||||
| 2538 | umutablecptrie_setumutablecptrie_set_77(mutableTrie, decompLead, canonValue, &errorCode); | ||||||||
| 2539 | canonStartSets.adoptElement(lpSet.orphan(), errorCode); | ||||||||
| 2540 | if (U_FAILURE(errorCode)) { | ||||||||
| 2541 | return; | ||||||||
| 2542 | } | ||||||||
| 2543 | if(firstOrigin!=0) { | ||||||||
| 2544 | set->add(firstOrigin); | ||||||||
| 2545 | } | ||||||||
| 2546 | } else { | ||||||||
| 2547 | set = static_cast<UnicodeSet*>(canonStartSets[static_cast<int32_t>(canonValue & CANON_VALUE_MASK0x1fffff)]); | ||||||||
| 2548 | } | ||||||||
| 2549 | set->add(origin); | ||||||||
| 2550 | } | ||||||||
| 2551 | } | ||||||||
| 2552 | |||||||||
| 2553 | // C++ class for friend access to private Normalizer2Impl members. | ||||||||
| 2554 | class InitCanonIterData { | ||||||||
| 2555 | public: | ||||||||
| 2556 | static void doInit(Normalizer2Impl *impl, UErrorCode &errorCode); | ||||||||
| 2557 | }; | ||||||||
| 2558 | |||||||||
| 2559 | U_CDECL_BEGINextern "C" { | ||||||||
| 2560 | |||||||||
| 2561 | // UInitOnce instantiation function for CanonIterData | ||||||||
| 2562 | static void U_CALLCONV | ||||||||
| 2563 | initCanonIterData(Normalizer2Impl *impl, UErrorCode &errorCode) { | ||||||||
| 2564 | InitCanonIterData::doInit(impl, errorCode); | ||||||||
| 2565 | } | ||||||||
| 2566 | |||||||||
| 2567 | U_CDECL_END} | ||||||||
| 2568 | |||||||||
| 2569 | void InitCanonIterData::doInit(Normalizer2Impl *impl, UErrorCode &errorCode) { | ||||||||
| 2570 | U_ASSERT(impl->fCanonIterData == nullptr)(static_cast <bool> (impl->fCanonIterData == nullptr ) ? void (0) : __assert_fail ("impl->fCanonIterData == nullptr" , __builtin_FILE (), __builtin_LINE (), __extension__ __PRETTY_FUNCTION__ )); | ||||||||
| 2571 | impl->fCanonIterData = new CanonIterData(errorCode); | ||||||||
| 2572 | if (impl->fCanonIterData == nullptr) { | ||||||||
| 2573 | errorCode=U_MEMORY_ALLOCATION_ERROR; | ||||||||
| 2574 | } | ||||||||
| 2575 | if (U_SUCCESS(errorCode)) { | ||||||||
| 2576 | UChar32 start = 0, end; | ||||||||
| 2577 | uint32_t value; | ||||||||
| 2578 | while ((end = ucptrie_getRangeucptrie_getRange_77(impl->normTrie, start, | ||||||||
| 2579 | UCPMAP_RANGE_FIXED_LEAD_SURROGATES, Normalizer2Impl::INERT, | ||||||||
| 2580 | nullptr, nullptr, &value)) >= 0) { | ||||||||
| 2581 | // Call Normalizer2Impl::makeCanonIterDataFromNorm16() for a range of same-norm16 characters. | ||||||||
| 2582 | if (value != Normalizer2Impl::INERT) { | ||||||||
| 2583 | impl->makeCanonIterDataFromNorm16(start, end, value, *impl->fCanonIterData, errorCode); | ||||||||
| 2584 | } | ||||||||
| 2585 | start = end + 1; | ||||||||
| 2586 | } | ||||||||
| 2587 | #ifdef UCPTRIE_DEBUG | ||||||||
| 2588 | umutablecptrie_setName(impl->fCanonIterData->mutableTrie, "CanonIterData"); | ||||||||
| 2589 | #endif | ||||||||
| 2590 | impl->fCanonIterData->trie = umutablecptrie_buildImmutableumutablecptrie_buildImmutable_77( | ||||||||
| 2591 | impl->fCanonIterData->mutableTrie, UCPTRIE_TYPE_SMALL, UCPTRIE_VALUE_BITS_32, &errorCode); | ||||||||
| 2592 | umutablecptrie_closeumutablecptrie_close_77(impl->fCanonIterData->mutableTrie); | ||||||||
| 2593 | impl->fCanonIterData->mutableTrie = nullptr; | ||||||||
| 2594 | } | ||||||||
| 2595 | if (U_FAILURE(errorCode)) { | ||||||||
| 2596 | delete impl->fCanonIterData; | ||||||||
| 2597 | impl->fCanonIterData = nullptr; | ||||||||
| 2598 | } | ||||||||
| 2599 | } | ||||||||
| 2600 | |||||||||
| 2601 | void Normalizer2Impl::makeCanonIterDataFromNorm16(UChar32 start, UChar32 end, const uint16_t norm16, | ||||||||
| 2602 | CanonIterData &newData, | ||||||||
| 2603 | UErrorCode &errorCode) const { | ||||||||
| 2604 | if(isInert(norm16) || | ||||||||
| 2605 | (minYesNo<=norm16 && norm16<minNoNo) || | ||||||||
| 2606 | (minMaybeNo<=norm16 && norm16<minMaybeYes)) { | ||||||||
| 2607 | // Inert, or 2-way mapping (including Hangul syllable). | ||||||||
| 2608 | // We do not write a canonStartSet for any yesNo/maybeNo character. | ||||||||
| 2609 | // Composites from 2-way mappings are added at runtime from the | ||||||||
| 2610 | // starter's compositions list, and the other characters in | ||||||||
| 2611 | // 2-way mappings get CANON_NOT_SEGMENT_STARTER set because they are | ||||||||
| 2612 | // "maybe" characters. | ||||||||
| 2613 | return; | ||||||||
| 2614 | } | ||||||||
| 2615 | for(UChar32 c=start; c<=end; ++c) { | ||||||||
| 2616 | uint32_t oldValue = umutablecptrie_getumutablecptrie_get_77(newData.mutableTrie, c); | ||||||||
| 2617 | uint32_t newValue=oldValue; | ||||||||
| 2618 | if(isMaybeYesOrNonZeroCC(norm16)) { | ||||||||
| 2619 | // not a segment starter if it occurs in a decomposition or has cc!=0 | ||||||||
| 2620 | newValue|=CANON_NOT_SEGMENT_STARTER0x80000000; | ||||||||
| 2621 | if(norm16<MIN_NORMAL_MAYBE_YES) { | ||||||||
| 2622 | newValue|=CANON_HAS_COMPOSITIONS0x40000000; | ||||||||
| 2623 | } | ||||||||
| 2624 | } else if(norm16<minYesNo) { | ||||||||
| 2625 | newValue|=CANON_HAS_COMPOSITIONS0x40000000; | ||||||||
| 2626 | } else { | ||||||||
| 2627 | // c has a one-way decomposition | ||||||||
| 2628 | UChar32 c2=c; | ||||||||
| 2629 | // Do not modify the whole-range norm16 value. | ||||||||
| 2630 | uint16_t norm16_2=norm16; | ||||||||
| 2631 | if (isDecompNoAlgorithmic(norm16_2)) { | ||||||||
| 2632 | // Maps to an isCompYesAndZeroCC. | ||||||||
| 2633 | c2 = mapAlgorithmic(c2, norm16_2); | ||||||||
| 2634 | norm16_2 = getRawNorm16(c2); | ||||||||
| 2635 | // No compatibility mappings for the CanonicalIterator. | ||||||||
| 2636 | U_ASSERT(!(isHangulLV(norm16_2) || isHangulLVT(norm16_2)))(static_cast <bool> (!(isHangulLV(norm16_2) || isHangulLVT (norm16_2))) ? void (0) : __assert_fail ("!(isHangulLV(norm16_2) || isHangulLVT(norm16_2))" , __builtin_FILE (), __builtin_LINE (), __extension__ __PRETTY_FUNCTION__ )); | ||||||||
| 2637 | } | ||||||||
| 2638 | if (norm16_2 > minYesNo) { | ||||||||
| 2639 | // c decomposes, get everything from the variable-length extra data | ||||||||
| 2640 | const uint16_t *mapping=getDataForYesOrNo(norm16_2); | ||||||||
| 2641 | uint16_t firstUnit=*mapping; | ||||||||
| 2642 | int32_t length=firstUnit&MAPPING_LENGTH_MASK; | ||||||||
| 2643 | if((firstUnit&MAPPING_HAS_CCC_LCCC_WORD)!=0) { | ||||||||
| 2644 | if(c==c2 && (*(mapping-1)&0xff)!=0) { | ||||||||
| 2645 | newValue|=CANON_NOT_SEGMENT_STARTER0x80000000; // original c has cc!=0 | ||||||||
| 2646 | } | ||||||||
| 2647 | } | ||||||||
| 2648 | // Skip empty mappings (no characters in the decomposition). | ||||||||
| 2649 | if(length!=0) { | ||||||||
| 2650 | ++mapping; // skip over the firstUnit | ||||||||
| 2651 | // add c to first code point's start set | ||||||||
| 2652 | int32_t i=0; | ||||||||
| 2653 | U16_NEXT_UNSAFE(mapping, i, c2)do { (c2)=(mapping)[(i)++]; if((((c2)&0xfffffc00)==0xd800 )) { (c2)=(((UChar32)((c2))<<10UL)+(UChar32)((mapping)[ (i)++])-((0xd800<<10UL)+0xdc00-0x10000)); } } while (false ); | ||||||||
| 2654 | newData.addToStartSet(c, c2, errorCode); | ||||||||
| 2655 | // Set CANON_NOT_SEGMENT_STARTER for each remaining code point of a | ||||||||
| 2656 | // one-way mapping. A 2-way mapping is possible here after | ||||||||
| 2657 | // intermediate algorithmic mapping. | ||||||||
| 2658 | if(norm16_2>=minNoNo) { | ||||||||
| 2659 | while(i<length) { | ||||||||
| 2660 | U16_NEXT_UNSAFE(mapping, i, c2)do { (c2)=(mapping)[(i)++]; if((((c2)&0xfffffc00)==0xd800 )) { (c2)=(((UChar32)((c2))<<10UL)+(UChar32)((mapping)[ (i)++])-((0xd800<<10UL)+0xdc00-0x10000)); } } while (false ); | ||||||||
| 2661 | uint32_t c2Value = umutablecptrie_getumutablecptrie_get_77(newData.mutableTrie, c2); | ||||||||
| 2662 | if((c2Value&CANON_NOT_SEGMENT_STARTER0x80000000)==0) { | ||||||||
| 2663 | umutablecptrie_setumutablecptrie_set_77(newData.mutableTrie, c2, | ||||||||
| 2664 | c2Value|CANON_NOT_SEGMENT_STARTER0x80000000, &errorCode); | ||||||||
| 2665 | } | ||||||||
| 2666 | } | ||||||||
| 2667 | } | ||||||||
| 2668 | } | ||||||||
| 2669 | } else { | ||||||||
| 2670 | // c decomposed to c2 algorithmically; c has cc==0 | ||||||||
| 2671 | newData.addToStartSet(c, c2, errorCode); | ||||||||
| 2672 | } | ||||||||
| 2673 | } | ||||||||
| 2674 | if(newValue!=oldValue) { | ||||||||
| 2675 | umutablecptrie_setumutablecptrie_set_77(newData.mutableTrie, c, newValue, &errorCode); | ||||||||
| 2676 | } | ||||||||
| 2677 | } | ||||||||
| 2678 | } | ||||||||
| 2679 | |||||||||
| 2680 | UBool Normalizer2Impl::ensureCanonIterData(UErrorCode &errorCode) const { | ||||||||
| 2681 | // Logically const: Synchronized instantiation. | ||||||||
| 2682 | Normalizer2Impl *me=const_cast<Normalizer2Impl *>(this); | ||||||||
| 2683 | umtx_initOnce(me->fCanonIterDataInitOnce, &initCanonIterData, me, errorCode); | ||||||||
| 2684 | return U_SUCCESS(errorCode); | ||||||||
| 2685 | } | ||||||||
| 2686 | |||||||||
| 2687 | int32_t Normalizer2Impl::getCanonValue(UChar32 c) const { | ||||||||
| 2688 | return static_cast<int32_t>(ucptrie_getucptrie_get_77(fCanonIterData->trie, c)); | ||||||||
| 2689 | } | ||||||||
| 2690 | |||||||||
| 2691 | const UnicodeSet &Normalizer2Impl::getCanonStartSet(int32_t n) const { | ||||||||
| 2692 | return *static_cast<const UnicodeSet*>(fCanonIterData->canonStartSets[n]); | ||||||||
| 2693 | } | ||||||||
| 2694 | |||||||||
| 2695 | UBool Normalizer2Impl::isCanonSegmentStarter(UChar32 c) const { | ||||||||
| 2696 | return getCanonValue(c)>=0; | ||||||||
| 2697 | } | ||||||||
| 2698 | |||||||||
| 2699 | UBool Normalizer2Impl::getCanonStartSet(UChar32 c, UnicodeSet &set) const { | ||||||||
| 2700 | int32_t canonValue=getCanonValue(c)&~CANON_NOT_SEGMENT_STARTER0x80000000; | ||||||||
| 2701 | if(canonValue==0) { | ||||||||
| 2702 | return false; | ||||||||
| 2703 | } | ||||||||
| 2704 | set.clear(); | ||||||||
| 2705 | int32_t value=canonValue&CANON_VALUE_MASK0x1fffff; | ||||||||
| 2706 | if((canonValue&CANON_HAS_SET0x200000)!=0) { | ||||||||
| 2707 | set.addAll(getCanonStartSet(value)); | ||||||||
| 2708 | } else if(value!=0) { | ||||||||
| 2709 | set.add(value); | ||||||||
| 2710 | } | ||||||||
| 2711 | if((canonValue&CANON_HAS_COMPOSITIONS0x40000000)!=0) { | ||||||||
| 2712 | uint16_t norm16=getRawNorm16(c); | ||||||||
| 2713 | if(norm16==JAMO_L) { | ||||||||
| 2714 | UChar32 syllable= | ||||||||
| 2715 | static_cast<UChar32>(Hangul::HANGUL_BASE + (c - Hangul::JAMO_L_BASE) * Hangul::JAMO_VT_COUNT); | ||||||||
| 2716 | set.add(syllable, syllable+Hangul::JAMO_VT_COUNT-1); | ||||||||
| 2717 | } else { | ||||||||
| 2718 | addComposites(getCompositionsList(norm16), set); | ||||||||
| 2719 | } | ||||||||
| 2720 | } | ||||||||
| 2721 | return true; | ||||||||
| 2722 | } | ||||||||
| 2723 | |||||||||
| 2724 | U_NAMESPACE_END} | ||||||||
| 2725 | |||||||||
| 2726 | // Normalizer2 data swapping ----------------------------------------------- *** | ||||||||
| 2727 | |||||||||
| 2728 | U_NAMESPACE_USEusing namespace icu_77; | ||||||||
| 2729 | |||||||||
| 2730 | U_CAPIextern "C" int32_t U_EXPORT2 | ||||||||
| 2731 | unorm2_swapunorm2_swap_77(const UDataSwapper *ds, | ||||||||
| 2732 | const void *inData, int32_t length, void *outData, | ||||||||
| 2733 | UErrorCode *pErrorCode) { | ||||||||
| 2734 | const UDataInfo *pInfo; | ||||||||
| 2735 | int32_t headerSize; | ||||||||
| 2736 | |||||||||
| 2737 | const uint8_t *inBytes; | ||||||||
| 2738 | uint8_t *outBytes; | ||||||||
| 2739 | |||||||||
| 2740 | const int32_t *inIndexes; | ||||||||
| 2741 | int32_t indexes[Normalizer2Impl::IX_TOTAL_SIZE+1]; | ||||||||
| 2742 | |||||||||
| 2743 | int32_t i, offset, nextOffset, size; | ||||||||
| 2744 | |||||||||
| 2745 | /* udata_swapDataHeader checks the arguments */ | ||||||||
| 2746 | headerSize=udata_swapDataHeaderudata_swapDataHeader_77(ds, inData, length, outData, pErrorCode); | ||||||||
| 2747 | if(pErrorCode==nullptr || U_FAILURE(*pErrorCode)) { | ||||||||
| 2748 | return 0; | ||||||||
| 2749 | } | ||||||||
| 2750 | |||||||||
| 2751 | /* check data format and format version */ | ||||||||
| 2752 | pInfo=(const UDataInfo *)((const char *)inData+4); | ||||||||
| 2753 | uint8_t formatVersion0=pInfo->formatVersion[0]; | ||||||||
| 2754 | if(!( | ||||||||
| 2755 | pInfo->dataFormat[0]==0x4e && /* dataFormat="Nrm2" */ | ||||||||
| 2756 | pInfo->dataFormat[1]==0x72 && | ||||||||
| 2757 | pInfo->dataFormat[2]==0x6d && | ||||||||
| 2758 | pInfo->dataFormat[3]==0x32 && | ||||||||
| 2759 | (1<=formatVersion0 && formatVersion0<=5) | ||||||||
| 2760 | )) { | ||||||||
| 2761 | udata_printErrorudata_printError_77(ds, "unorm2_swap(): data format %02x.%02x.%02x.%02x (format version %02x) is not recognized as Normalizer2 data\n", | ||||||||
| 2762 | pInfo->dataFormat[0], pInfo->dataFormat[1], | ||||||||
| 2763 | pInfo->dataFormat[2], pInfo->dataFormat[3], | ||||||||
| 2764 | pInfo->formatVersion[0]); | ||||||||
| 2765 | *pErrorCode=U_UNSUPPORTED_ERROR; | ||||||||
| 2766 | return 0; | ||||||||
| 2767 | } | ||||||||
| 2768 | |||||||||
| 2769 | inBytes=(const uint8_t *)inData+headerSize; | ||||||||
| 2770 | outBytes=(outData == nullptr) ? nullptr : (uint8_t *)outData+headerSize; | ||||||||
| 2771 | |||||||||
| 2772 | inIndexes=(const int32_t *)inBytes; | ||||||||
| 2773 | int32_t minIndexesLength; | ||||||||
| 2774 | if(formatVersion0==1) { | ||||||||
| 2775 | minIndexesLength=Normalizer2Impl::IX_MIN_MAYBE_YES+1; | ||||||||
| 2776 | } else if(formatVersion0==2) { | ||||||||
| 2777 | minIndexesLength=Normalizer2Impl::IX_MIN_YES_NO_MAPPINGS_ONLY+1; | ||||||||
| 2778 | } else if(formatVersion0<=4) { | ||||||||
| 2779 | minIndexesLength=Normalizer2Impl::IX_MIN_LCCC_CP+1; | ||||||||
| 2780 | } else { | ||||||||
| 2781 | minIndexesLength=Normalizer2Impl::IX_MIN_MAYBE_NO_COMBINES_FWD+1; | ||||||||
| 2782 | } | ||||||||
| 2783 | |||||||||
| 2784 | if(length>=0) { | ||||||||
| 2785 | length-=headerSize; | ||||||||
| 2786 | if(length<minIndexesLength*4) { | ||||||||
| 2787 | udata_printErrorudata_printError_77(ds, "unorm2_swap(): too few bytes (%d after header) for Normalizer2 data\n", | ||||||||
| 2788 | length); | ||||||||
| 2789 | *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; | ||||||||
| 2790 | return 0; | ||||||||
| 2791 | } | ||||||||
| 2792 | } | ||||||||
| 2793 | |||||||||
| 2794 | /* read the first few indexes */ | ||||||||
| 2795 | for(i=0; i<UPRV_LENGTHOF(indexes)(int32_t)(sizeof(indexes)/sizeof((indexes)[0])); ++i) { | ||||||||
| 2796 | indexes[i]=udata_readInt32udata_readInt32_77(ds, inIndexes[i]); | ||||||||
| 2797 | } | ||||||||
| 2798 | |||||||||
| 2799 | /* get the total length of the data */ | ||||||||
| 2800 | size=indexes[Normalizer2Impl::IX_TOTAL_SIZE]; | ||||||||
| 2801 | |||||||||
| 2802 | if(length>=0) { | ||||||||
| 2803 | if(length<size) { | ||||||||
| 2804 | udata_printErrorudata_printError_77(ds, "unorm2_swap(): too few bytes (%d after header) for all of Normalizer2 data\n", | ||||||||
| 2805 | length); | ||||||||
| 2806 | *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; | ||||||||
| 2807 | return 0; | ||||||||
| 2808 | } | ||||||||
| 2809 | |||||||||
| 2810 | /* copy the data for inaccessible bytes */ | ||||||||
| 2811 | if(inBytes!=outBytes) { | ||||||||
| 2812 | uprv_memcpy(outBytes, inBytes, size)do { clang diagnostic push
clang diagnostic ignored "-Waddress" (static_cast <bool> (outBytes != __null) ? void (0) : __assert_fail ("outBytes != __null", __builtin_FILE (), __builtin_LINE (), __extension__ __PRETTY_FUNCTION__)); (static_cast <bool > (inBytes != __null) ? void (0) : __assert_fail ("inBytes != __null" , __builtin_FILE (), __builtin_LINE (), __extension__ __PRETTY_FUNCTION__ )); clang diagnostic pop :: memcpy(outBytes, inBytes, size); } while (false); | ||||||||
| 2813 | } | ||||||||
| 2814 | |||||||||
| 2815 | offset=0; | ||||||||
| 2816 | |||||||||
| 2817 | /* swap the int32_t indexes[] */ | ||||||||
| 2818 | nextOffset=indexes[Normalizer2Impl::IX_NORM_TRIE_OFFSET]; | ||||||||
| 2819 | ds->swapArray32(ds, inBytes, nextOffset-offset, outBytes, pErrorCode); | ||||||||
| 2820 | offset=nextOffset; | ||||||||
| 2821 | |||||||||
| 2822 | /* swap the trie */ | ||||||||
| 2823 | nextOffset=indexes[Normalizer2Impl::IX_EXTRA_DATA_OFFSET]; | ||||||||
| 2824 | utrie_swapAnyVersionutrie_swapAnyVersion_77(ds, inBytes+offset, nextOffset-offset, outBytes+offset, pErrorCode); | ||||||||
| 2825 | offset=nextOffset; | ||||||||
| 2826 | |||||||||
| 2827 | /* swap the uint16_t extraData[] */ | ||||||||
| 2828 | nextOffset=indexes[Normalizer2Impl::IX_SMALL_FCD_OFFSET]; | ||||||||
| 2829 | ds->swapArray16(ds, inBytes+offset, nextOffset-offset, outBytes+offset, pErrorCode); | ||||||||
| 2830 | offset=nextOffset; | ||||||||
| 2831 | |||||||||
| 2832 | /* no need to swap the uint8_t smallFCD[] (new in formatVersion 2) */ | ||||||||
| 2833 | nextOffset=indexes[Normalizer2Impl::IX_SMALL_FCD_OFFSET+1]; | ||||||||
| 2834 | offset=nextOffset; | ||||||||
| 2835 | |||||||||
| 2836 | U_ASSERT(offset==size)(static_cast <bool> (offset==size) ? void (0) : __assert_fail ("offset==size", __builtin_FILE (), __builtin_LINE (), __extension__ __PRETTY_FUNCTION__)); | ||||||||
| 2837 | } | ||||||||
| 2838 | |||||||||
| 2839 | return headerSize+size; | ||||||||
| 2840 | } | ||||||||
| 2841 | |||||||||
| 2842 | #endif // !UCONFIG_NO_NORMALIZATION |
| 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) 2009-2014, International Business Machines | ||||
| 7 | * Corporation and others. All Rights Reserved. | ||||
| 8 | * | ||||
| 9 | ******************************************************************************* | ||||
| 10 | * file name: normalizer2impl.h | ||||
| 11 | * encoding: UTF-8 | ||||
| 12 | * tab size: 8 (not used) | ||||
| 13 | * indentation:4 | ||||
| 14 | * | ||||
| 15 | * created on: 2009nov22 | ||||
| 16 | * created by: Markus W. Scherer | ||||
| 17 | */ | ||||
| 18 | |||||
| 19 | #ifndef __NORMALIZER2IMPL_H__ | ||||
| 20 | #define __NORMALIZER2IMPL_H__ | ||||
| 21 | |||||
| 22 | #include "unicode/utypes.h" | ||||
| 23 | |||||
| 24 | #if !UCONFIG_NO_NORMALIZATION0 | ||||
| 25 | |||||
| 26 | #include "unicode/normalizer2.h" | ||||
| 27 | #include "unicode/ucptrie.h" | ||||
| 28 | #include "unicode/unistr.h" | ||||
| 29 | #include "unicode/unorm.h" | ||||
| 30 | #include "unicode/utf.h" | ||||
| 31 | #include "unicode/utf16.h" | ||||
| 32 | #include "mutex.h" | ||||
| 33 | #include "udataswp.h" | ||||
| 34 | #include "uset_imp.h" | ||||
| 35 | |||||
| 36 | // When the nfc.nrm data is *not* hardcoded into the common library | ||||
| 37 | // (with this constant set to 0), | ||||
| 38 | // then it needs to be built into the data package: | ||||
| 39 | // Add nfc.nrm to icu4c/source/data/Makefile.in DAT_FILES_SHORT | ||||
| 40 | #define NORM2_HARDCODE_NFC_DATA1 1 | ||||
| 41 | |||||
| 42 | U_NAMESPACE_BEGINnamespace icu_77 { | ||||
| 43 | |||||
| 44 | struct CanonIterData; | ||||
| 45 | |||||
| 46 | class ByteSink; | ||||
| 47 | class Edits; | ||||
| 48 | class InitCanonIterData; | ||||
| 49 | class LcccContext; | ||||
| 50 | |||||
| 51 | class U_COMMON_API Hangul { | ||||
| 52 | public: | ||||
| 53 | /* Korean Hangul and Jamo constants */ | ||||
| 54 | enum { | ||||
| 55 | JAMO_L_BASE=0x1100, /* "lead" jamo */ | ||||
| 56 | JAMO_L_END=0x1112, | ||||
| 57 | JAMO_V_BASE=0x1161, /* "vowel" jamo */ | ||||
| 58 | JAMO_V_END=0x1175, | ||||
| 59 | JAMO_T_BASE=0x11a7, /* "trail" jamo */ | ||||
| 60 | JAMO_T_END=0x11c2, | ||||
| 61 | |||||
| 62 | HANGUL_BASE=0xac00, | ||||
| 63 | HANGUL_END=0xd7a3, | ||||
| 64 | |||||
| 65 | JAMO_L_COUNT=19, | ||||
| 66 | JAMO_V_COUNT=21, | ||||
| 67 | JAMO_T_COUNT=28, | ||||
| 68 | |||||
| 69 | JAMO_VT_COUNT=JAMO_V_COUNT*JAMO_T_COUNT, | ||||
| 70 | |||||
| 71 | HANGUL_COUNT=JAMO_L_COUNT*JAMO_V_COUNT*JAMO_T_COUNT, | ||||
| 72 | HANGUL_LIMIT=HANGUL_BASE+HANGUL_COUNT | ||||
| 73 | }; | ||||
| 74 | |||||
| 75 | static inline UBool isHangul(UChar32 c) { | ||||
| 76 | return HANGUL_BASE<=c && c<HANGUL_LIMIT; | ||||
| 77 | } | ||||
| 78 | static inline UBool | ||||
| 79 | isHangulLV(UChar32 c) { | ||||
| 80 | c-=HANGUL_BASE; | ||||
| 81 | return 0<=c && c<HANGUL_COUNT && c%JAMO_T_COUNT==0; | ||||
| 82 | } | ||||
| 83 | static inline UBool isJamoL(UChar32 c) { | ||||
| 84 | return static_cast<uint32_t>(c - JAMO_L_BASE) < JAMO_L_COUNT; | ||||
| 85 | } | ||||
| 86 | static inline UBool isJamoV(UChar32 c) { | ||||
| 87 | return static_cast<uint32_t>(c - JAMO_V_BASE) < JAMO_V_COUNT; | ||||
| 88 | } | ||||
| 89 | static inline UBool isJamoT(UChar32 c) { | ||||
| 90 | int32_t t=c-JAMO_T_BASE; | ||||
| 91 | return 0<t && t<JAMO_T_COUNT; // not JAMO_T_BASE itself | ||||
| 92 | } | ||||
| 93 | static UBool isJamo(UChar32 c) { | ||||
| 94 | return JAMO_L_BASE<=c && c<=JAMO_T_END && | ||||
| 95 | (c<=JAMO_L_END || (JAMO_V_BASE<=c && c<=JAMO_V_END) || JAMO_T_BASE<c); | ||||
| 96 | } | ||||
| 97 | |||||
| 98 | /** | ||||
| 99 | * Decomposes c, which must be a Hangul syllable, into buffer | ||||
| 100 | * and returns the length of the decomposition (2 or 3). | ||||
| 101 | */ | ||||
| 102 | static inline int32_t decompose(UChar32 c, char16_t buffer[3]) { | ||||
| 103 | c-=HANGUL_BASE; | ||||
| 104 | UChar32 c2=c%JAMO_T_COUNT; | ||||
| 105 | c/=JAMO_T_COUNT; | ||||
| 106 | buffer[0] = static_cast<char16_t>(JAMO_L_BASE + c / JAMO_V_COUNT); | ||||
| 107 | buffer[1] = static_cast<char16_t>(JAMO_V_BASE + c % JAMO_V_COUNT); | ||||
| 108 | if(c2==0) { | ||||
| 109 | return 2; | ||||
| 110 | } else { | ||||
| 111 | buffer[2] = static_cast<char16_t>(JAMO_T_BASE + c2); | ||||
| 112 | return 3; | ||||
| 113 | } | ||||
| 114 | } | ||||
| 115 | |||||
| 116 | /** | ||||
| 117 | * Decomposes c, which must be a Hangul syllable, into buffer. | ||||
| 118 | * This is the raw, not recursive, decomposition. Its length is always 2. | ||||
| 119 | */ | ||||
| 120 | static inline void getRawDecomposition(UChar32 c, char16_t buffer[2]) { | ||||
| 121 | UChar32 orig=c; | ||||
| 122 | c-=HANGUL_BASE; | ||||
| 123 | UChar32 c2=c%JAMO_T_COUNT; | ||||
| 124 | if(c2==0) { | ||||
| 125 | c/=JAMO_T_COUNT; | ||||
| 126 | buffer[0] = static_cast<char16_t>(JAMO_L_BASE + c / JAMO_V_COUNT); | ||||
| 127 | buffer[1] = static_cast<char16_t>(JAMO_V_BASE + c % JAMO_V_COUNT); | ||||
| 128 | } else { | ||||
| 129 | buffer[0] = static_cast<char16_t>(orig - c2); // LV syllable | ||||
| 130 | buffer[1] = static_cast<char16_t>(JAMO_T_BASE + c2); | ||||
| 131 | } | ||||
| 132 | } | ||||
| 133 | private: | ||||
| 134 | Hangul() = delete; // no instantiation | ||||
| 135 | }; | ||||
| 136 | |||||
| 137 | class Normalizer2Impl; | ||||
| 138 | |||||
| 139 | class U_COMMON_API ReorderingBuffer : public UMemory { | ||||
| 140 | public: | ||||
| 141 | /** Constructs only; init() should be called. */ | ||||
| 142 | ReorderingBuffer(const Normalizer2Impl &ni, UnicodeString &dest) : | ||||
| 143 | impl(ni), str(dest), | ||||
| 144 | start(nullptr), reorderStart(nullptr), limit(nullptr), | ||||
| 145 | remainingCapacity(0), lastCC(0) {} | ||||
| 146 | /** Constructs, removes the string contents, and initializes for a small initial capacity. */ | ||||
| 147 | ReorderingBuffer(const Normalizer2Impl &ni, UnicodeString &dest, UErrorCode &errorCode); | ||||
| 148 | ~ReorderingBuffer() { | ||||
| 149 | if (start != nullptr) { | ||||
| 150 | str.releaseBuffer(static_cast<int32_t>(limit - start)); | ||||
| 151 | } | ||||
| 152 | } | ||||
| 153 | UBool init(int32_t destCapacity, UErrorCode &errorCode); | ||||
| 154 | |||||
| 155 | UBool isEmpty() const { return start==limit; } | ||||
| 156 | int32_t length() const { return static_cast<int32_t>(limit - start); } | ||||
| 157 | char16_t *getStart() { return start; } | ||||
| 158 | char16_t *getLimit() { return limit; } | ||||
| 159 | uint8_t getLastCC() const { return lastCC; } | ||||
| 160 | |||||
| 161 | UBool equals(const char16_t *start, const char16_t *limit) const; | ||||
| 162 | UBool equals(const uint8_t *otherStart, const uint8_t *otherLimit) const; | ||||
| 163 | |||||
| 164 | UBool append(UChar32 c, uint8_t cc, UErrorCode &errorCode) { | ||||
| 165 | return (c<=0xffff) ? | ||||
| 166 | appendBMP(static_cast<char16_t>(c), cc, errorCode) : | ||||
| 167 | appendSupplementary(c, cc, errorCode); | ||||
| 168 | } | ||||
| 169 | UBool append(const char16_t *s, int32_t length, UBool isNFD, | ||||
| 170 | uint8_t leadCC, uint8_t trailCC, | ||||
| 171 | UErrorCode &errorCode); | ||||
| 172 | UBool appendBMP(char16_t c, uint8_t cc, UErrorCode &errorCode) { | ||||
| 173 | if(remainingCapacity==0 && !resize(1, errorCode)) { | ||||
| 174 | return false; | ||||
| 175 | } | ||||
| 176 | if(lastCC<=cc || cc==0) { | ||||
| 177 | *limit++=c; | ||||
| 178 | lastCC=cc; | ||||
| 179 | if(cc<=1) { | ||||
| 180 | reorderStart=limit; | ||||
| 181 | } | ||||
| 182 | } else { | ||||
| 183 | insert(c, cc); | ||||
| 184 | } | ||||
| 185 | --remainingCapacity; | ||||
| 186 | return true; | ||||
| 187 | } | ||||
| 188 | UBool appendZeroCC(UChar32 c, UErrorCode &errorCode); | ||||
| 189 | UBool appendZeroCC(const char16_t *s, const char16_t *sLimit, UErrorCode &errorCode); | ||||
| 190 | void remove(); | ||||
| 191 | void removeSuffix(int32_t suffixLength); | ||||
| 192 | void setReorderingLimit(char16_t *newLimit) { | ||||
| 193 | remainingCapacity += static_cast<int32_t>(limit - newLimit); | ||||
| 194 | reorderStart=limit=newLimit; | ||||
| 195 | lastCC=0; | ||||
| 196 | } | ||||
| 197 | void copyReorderableSuffixTo(UnicodeString &s) const { | ||||
| 198 | s.setTo(ConstChar16Ptr(reorderStart), static_cast<int32_t>(limit - reorderStart)); | ||||
| 199 | } | ||||
| 200 | private: | ||||
| 201 | /* | ||||
| 202 | * TODO: Revisit whether it makes sense to track reorderStart. | ||||
| 203 | * It is set to after the last known character with cc<=1, | ||||
| 204 | * which stops previousCC() before it reads that character and looks up its cc. | ||||
| 205 | * previousCC() is normally only called from insert(). | ||||
| 206 | * In other words, reorderStart speeds up the insertion of a combining mark | ||||
| 207 | * into a multi-combining mark sequence where it does not belong at the end. | ||||
| 208 | * This might not be worth the trouble. | ||||
| 209 | * On the other hand, it's not a huge amount of trouble. | ||||
| 210 | * | ||||
| 211 | * We probably need it for UNORM_SIMPLE_APPEND. | ||||
| 212 | */ | ||||
| 213 | |||||
| 214 | UBool appendSupplementary(UChar32 c, uint8_t cc, UErrorCode &errorCode); | ||||
| 215 | void insert(UChar32 c, uint8_t cc); | ||||
| 216 | static void writeCodePoint(char16_t *p, UChar32 c) { | ||||
| 217 | if(c<=0xffff) { | ||||
| 218 | *p = static_cast<char16_t>(c); | ||||
| 219 | } else { | ||||
| 220 | p[0]=U16_LEAD(c)(UChar)(((c)>>10)+0xd7c0); | ||||
| 221 | p[1]=U16_TRAIL(c)(UChar)(((c)&0x3ff)|0xdc00); | ||||
| 222 | } | ||||
| 223 | } | ||||
| 224 | UBool resize(int32_t appendLength, UErrorCode &errorCode); | ||||
| 225 | |||||
| 226 | const Normalizer2Impl &impl; | ||||
| 227 | UnicodeString &str; | ||||
| 228 | char16_t *start, *reorderStart, *limit; | ||||
| 229 | int32_t remainingCapacity; | ||||
| 230 | uint8_t lastCC; | ||||
| 231 | |||||
| 232 | // private backward iterator | ||||
| 233 | void setIterator() { codePointStart=limit; } | ||||
| 234 | void skipPrevious(); // Requires start<codePointStart. | ||||
| 235 | uint8_t previousCC(); // Returns 0 if there is no previous character. | ||||
| 236 | |||||
| 237 | char16_t *codePointStart, *codePointLimit; | ||||
| 238 | }; | ||||
| 239 | |||||
| 240 | /** | ||||
| 241 | * Low-level implementation of the Unicode Normalization Algorithm. | ||||
| 242 | * For the data structure and details see the documentation at the end of | ||||
| 243 | * this normalizer2impl.h and in the design doc at | ||||
| 244 | * https://unicode-org.github.io/icu/design/normalization/custom.html | ||||
| 245 | */ | ||||
| 246 | class U_COMMON_API Normalizer2Impl : public UObject { | ||||
| 247 | public: | ||||
| 248 | Normalizer2Impl() : normTrie(nullptr), fCanonIterData(nullptr) {} | ||||
| 249 | virtual ~Normalizer2Impl(); | ||||
| 250 | |||||
| 251 | void init(const int32_t *inIndexes, const UCPTrie *inTrie, | ||||
| 252 | const uint16_t *inExtraData, const uint8_t *inSmallFCD); | ||||
| 253 | |||||
| 254 | void addLcccChars(UnicodeSet &set) const; | ||||
| 255 | void addPropertyStarts(const USetAdder *sa, UErrorCode &errorCode) const; | ||||
| 256 | void addCanonIterPropertyStarts(const USetAdder *sa, UErrorCode &errorCode) const; | ||||
| 257 | |||||
| 258 | // low-level properties ------------------------------------------------ *** | ||||
| 259 | |||||
| 260 | UBool ensureCanonIterData(UErrorCode &errorCode) const; | ||||
| 261 | |||||
| 262 | // The trie stores values for lead surrogate code *units*. | ||||
| 263 | // Surrogate code *points* are inert. | ||||
| 264 | uint16_t getNorm16(UChar32 c) const { | ||||
| 265 | return U_IS_LEAD(c)(((c)&0xfffffc00)==0xd800) ? | ||||
| 266 | static_cast<uint16_t>(INERT) : | ||||
| 267 | UCPTRIE_FAST_GET(normTrie, UCPTRIE_16, c)((normTrie)->data.ptr16[((uint32_t)(c) <= (uint32_t)(0xffff ) ? ((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT ] + ((c) & UCPTRIE_FAST_DATA_MASK)) : (uint32_t)(c) <= 0x10ffff ? ((c) >= (normTrie)->highStart ? (normTrie)-> dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77 (normTrie, c)) : (normTrie)->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET )]); | ||||
| 268 | } | ||||
| 269 | uint16_t getRawNorm16(UChar32 c) const { return UCPTRIE_FAST_GET(normTrie, UCPTRIE_16, c)((normTrie)->data.ptr16[((uint32_t)(c) <= (uint32_t)(0xffff ) ? ((int32_t)(normTrie)->index[(c) >> UCPTRIE_FAST_SHIFT ] + ((c) & UCPTRIE_FAST_DATA_MASK)) : (uint32_t)(c) <= 0x10ffff ? ((c) >= (normTrie)->highStart ? (normTrie)-> dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET : ucptrie_internalSmallIndex_77 (normTrie, c)) : (normTrie)->dataLength - UCPTRIE_ERROR_VALUE_NEG_DATA_OFFSET )]); } | ||||
| 270 | |||||
| 271 | UNormalizationCheckResult getCompQuickCheck(uint16_t norm16) const { | ||||
| 272 | if(norm16<minNoNo || MIN_YES_YES_WITH_CC<=norm16) { | ||||
| 273 | return UNORM_YES; | ||||
| 274 | } else if(minMaybeNo<=norm16) { | ||||
| 275 | return UNORM_MAYBE; | ||||
| 276 | } else { | ||||
| 277 | return UNORM_NO; | ||||
| 278 | } | ||||
| 279 | } | ||||
| 280 | UBool isAlgorithmicNoNo(uint16_t norm16) const { return limitNoNo<=norm16 && norm16<minMaybeNo; } | ||||
| 281 | UBool isCompNo(uint16_t norm16) const { return minNoNo<=norm16 && norm16<minMaybeNo; } | ||||
| 282 | UBool isDecompYes(uint16_t norm16) const { return norm16<minYesNo || minMaybeYes<=norm16; } | ||||
| 283 | |||||
| 284 | uint8_t getCC(uint16_t norm16) const { | ||||
| 285 | if(norm16>=MIN_NORMAL_MAYBE_YES) { | ||||
| 286 | return getCCFromNormalYesOrMaybe(norm16); | ||||
| 287 | } | ||||
| 288 | if(norm16<minNoNo || limitNoNo<=norm16) { | ||||
| 289 | return 0; | ||||
| 290 | } | ||||
| 291 | return getCCFromNoNo(norm16); | ||||
| 292 | } | ||||
| 293 | static uint8_t getCCFromNormalYesOrMaybe(uint16_t norm16) { | ||||
| 294 | return static_cast<uint8_t>(norm16 >> OFFSET_SHIFT); | ||||
| 295 | } | ||||
| 296 | static uint8_t getCCFromYesOrMaybeYes(uint16_t norm16) { | ||||
| 297 | return norm16>=MIN_NORMAL_MAYBE_YES ? getCCFromNormalYesOrMaybe(norm16) : 0; | ||||
| 298 | } | ||||
| 299 | uint8_t getCCFromYesOrMaybeYesCP(UChar32 c) const { | ||||
| 300 | if (c < minCompNoMaybeCP) { return 0; } | ||||
| 301 | return getCCFromYesOrMaybeYes(getNorm16(c)); | ||||
| 302 | } | ||||
| 303 | |||||
| 304 | /** | ||||
| 305 | * Returns the FCD data for code point c. | ||||
| 306 | * @param c A Unicode code point. | ||||
| 307 | * @return The lccc(c) in bits 15..8 and tccc(c) in bits 7..0. | ||||
| 308 | */ | ||||
| 309 | uint16_t getFCD16(UChar32 c) const { | ||||
| 310 | if(c<minDecompNoCP) { | ||||
| 311 | return 0; | ||||
| 312 | } else if(c<=0xffff) { | ||||
| 313 | if(!singleLeadMightHaveNonZeroFCD16(c)) { return 0; } | ||||
| 314 | } | ||||
| 315 | return getFCD16FromNormData(c); | ||||
| 316 | } | ||||
| 317 | /** | ||||
| 318 | * Returns the FCD data for the next code point (post-increment). | ||||
| 319 | * Might skip only a lead surrogate rather than the whole surrogate pair if none of | ||||
| 320 | * the supplementary code points associated with the lead surrogate have non-zero FCD data. | ||||
| 321 | * @param s A valid pointer into a string. Requires s!=limit. | ||||
| 322 | * @param limit The end of the string, or NULL. | ||||
| 323 | * @return The lccc(c) in bits 15..8 and tccc(c) in bits 7..0. | ||||
| 324 | */ | ||||
| 325 | uint16_t nextFCD16(const char16_t *&s, const char16_t *limit) const { | ||||
| 326 | UChar32 c=*s++; | ||||
| 327 | if(c<minDecompNoCP || !singleLeadMightHaveNonZeroFCD16(c)) { | ||||
| 328 | return 0; | ||||
| 329 | } | ||||
| 330 | char16_t c2; | ||||
| 331 | if(U16_IS_LEAD(c)(((c)&0xfffffc00)==0xd800) && s!=limit && U16_IS_TRAIL(c2=*s)(((c2=*s)&0xfffffc00)==0xdc00)) { | ||||
| 332 | c=U16_GET_SUPPLEMENTARY(c, c2)(((UChar32)(c)<<10UL)+(UChar32)(c2)-((0xd800<<10UL )+0xdc00-0x10000)); | ||||
| 333 | ++s; | ||||
| 334 | } | ||||
| 335 | return getFCD16FromNormData(c); | ||||
| 336 | } | ||||
| 337 | /** | ||||
| 338 | * Returns the FCD data for the previous code point (pre-decrement). | ||||
| 339 | * @param start The start of the string. | ||||
| 340 | * @param s A valid pointer into a string. Requires start<s. | ||||
| 341 | * @return The lccc(c) in bits 15..8 and tccc(c) in bits 7..0. | ||||
| 342 | */ | ||||
| 343 | uint16_t previousFCD16(const char16_t *start, const char16_t *&s) const { | ||||
| 344 | UChar32 c=*--s; | ||||
| 345 | if(c<minDecompNoCP) { | ||||
| 346 | return 0; | ||||
| 347 | } | ||||
| 348 | if(!U16_IS_TRAIL(c)(((c)&0xfffffc00)==0xdc00)) { | ||||
| 349 | if(!singleLeadMightHaveNonZeroFCD16(c)) { | ||||
| 350 | return 0; | ||||
| 351 | } | ||||
| 352 | } else { | ||||
| 353 | char16_t c2; | ||||
| 354 | if(start<s && U16_IS_LEAD(c2=*(s-1))(((c2=*(s-1))&0xfffffc00)==0xd800)) { | ||||
| 355 | c=U16_GET_SUPPLEMENTARY(c2, c)(((UChar32)(c2)<<10UL)+(UChar32)(c)-((0xd800<<10UL )+0xdc00-0x10000)); | ||||
| 356 | --s; | ||||
| 357 | } | ||||
| 358 | } | ||||
| 359 | return getFCD16FromNormData(c); | ||||
| 360 | } | ||||
| 361 | |||||
| 362 | /** Returns true if the single-or-lead code unit c might have non-zero FCD data. */ | ||||
| 363 | UBool singleLeadMightHaveNonZeroFCD16(UChar32 lead) const { | ||||
| 364 | // 0<=lead<=0xffff | ||||
| 365 | uint8_t bits=smallFCD[lead>>8]; | ||||
| 366 | if(bits==0) { return false; } | ||||
| 367 | return (bits >> ((lead >> 5) & 7)) & 1; | ||||
| 368 | } | ||||
| 369 | /** Returns the FCD value from the regular normalization data. */ | ||||
| 370 | uint16_t getFCD16FromNormData(UChar32 c) const; | ||||
| 371 | |||||
| 372 | uint16_t getFCD16FromMaybeOrNonZeroCC(uint16_t norm16) const; | ||||
| 373 | |||||
| 374 | /** | ||||
| 375 | * Gets the decomposition for one code point. | ||||
| 376 | * @param c code point | ||||
| 377 | * @param buffer out-only buffer for algorithmic decompositions | ||||
| 378 | * @param length out-only, takes the length of the decomposition, if any | ||||
| 379 | * @return pointer to the decomposition, or NULL if none | ||||
| 380 | */ | ||||
| 381 | const char16_t *getDecomposition(UChar32 c, char16_t buffer[4], int32_t &length) const; | ||||
| 382 | |||||
| 383 | /** | ||||
| 384 | * Gets the raw decomposition for one code point. | ||||
| 385 | * @param c code point | ||||
| 386 | * @param buffer out-only buffer for algorithmic decompositions | ||||
| 387 | * @param length out-only, takes the length of the decomposition, if any | ||||
| 388 | * @return pointer to the decomposition, or NULL if none | ||||
| 389 | */ | ||||
| 390 | const char16_t *getRawDecomposition(UChar32 c, char16_t buffer[30], int32_t &length) const; | ||||
| 391 | |||||
| 392 | UChar32 composePair(UChar32 a, UChar32 b) const; | ||||
| 393 | |||||
| 394 | UBool isCanonSegmentStarter(UChar32 c) const; | ||||
| 395 | UBool getCanonStartSet(UChar32 c, UnicodeSet &set) const; | ||||
| 396 | |||||
| 397 | enum { | ||||
| 398 | // Fixed norm16 values. | ||||
| 399 | MIN_YES_YES_WITH_CC=0xfe02, | ||||
| 400 | JAMO_VT=0xfe00, | ||||
| 401 | MIN_NORMAL_MAYBE_YES=0xfc00, | ||||
| 402 | JAMO_L=2, // offset=1 hasCompBoundaryAfter=false | ||||
| 403 | INERT=1, // offset=0 hasCompBoundaryAfter=true | ||||
| 404 | |||||
| 405 | // norm16 bit 0 is comp-boundary-after. | ||||
| 406 | HAS_COMP_BOUNDARY_AFTER=1, | ||||
| 407 | OFFSET_SHIFT=1, | ||||
| 408 | |||||
| 409 | // For algorithmic one-way mappings, norm16 bits 2..1 indicate the | ||||
| 410 | // tccc (0, 1, >1) for quick FCC boundary-after tests. | ||||
| 411 | DELTA_TCCC_0=0, | ||||
| 412 | DELTA_TCCC_1=2, | ||||
| 413 | DELTA_TCCC_GT_1=4, | ||||
| 414 | DELTA_TCCC_MASK=6, | ||||
| 415 | DELTA_SHIFT=3, | ||||
| 416 | |||||
| 417 | MAX_DELTA=0x40 | ||||
| 418 | }; | ||||
| 419 | |||||
| 420 | enum { | ||||
| 421 | // Byte offsets from the start of the data, after the generic header. | ||||
| 422 | IX_NORM_TRIE_OFFSET, | ||||
| 423 | IX_EXTRA_DATA_OFFSET, | ||||
| 424 | IX_SMALL_FCD_OFFSET, | ||||
| 425 | IX_RESERVED3_OFFSET, | ||||
| 426 | IX_RESERVED4_OFFSET, | ||||
| 427 | IX_RESERVED5_OFFSET, | ||||
| 428 | IX_RESERVED6_OFFSET, | ||||
| 429 | IX_TOTAL_SIZE, | ||||
| 430 | |||||
| 431 | // Code point thresholds for quick check codes. | ||||
| 432 | IX_MIN_DECOMP_NO_CP, | ||||
| 433 | IX_MIN_COMP_NO_MAYBE_CP, | ||||
| 434 | |||||
| 435 | // Norm16 value thresholds for quick check combinations and types of extra data. | ||||
| 436 | |||||
| 437 | /** Mappings & compositions in [minYesNo..minYesNoMappingsOnly[. */ | ||||
| 438 | IX_MIN_YES_NO, | ||||
| 439 | /** Mappings are comp-normalized. */ | ||||
| 440 | IX_MIN_NO_NO, | ||||
| 441 | IX_LIMIT_NO_NO, | ||||
| 442 | IX_MIN_MAYBE_YES, | ||||
| 443 | |||||
| 444 | /** Mappings only in [minYesNoMappingsOnly..minNoNo[. */ | ||||
| 445 | IX_MIN_YES_NO_MAPPINGS_ONLY, | ||||
| 446 | /** Mappings are not comp-normalized but have a comp boundary before. */ | ||||
| 447 | IX_MIN_NO_NO_COMP_BOUNDARY_BEFORE, | ||||
| 448 | /** Mappings do not have a comp boundary before. */ | ||||
| 449 | IX_MIN_NO_NO_COMP_NO_MAYBE_CC, | ||||
| 450 | /** Mappings to the empty string. */ | ||||
| 451 | IX_MIN_NO_NO_EMPTY, | ||||
| 452 | |||||
| 453 | IX_MIN_LCCC_CP, | ||||
| 454 | IX_RESERVED19, | ||||
| 455 | |||||
| 456 | /** Two-way mappings; each starts with a character that combines backward. */ | ||||
| 457 | IX_MIN_MAYBE_NO, // 20 | ||||
| 458 | /** Two-way mappings & compositions. */ | ||||
| 459 | IX_MIN_MAYBE_NO_COMBINES_FWD, | ||||
| 460 | |||||
| 461 | IX_COUNT // 22 | ||||
| 462 | }; | ||||
| 463 | |||||
| 464 | enum { | ||||
| 465 | MAPPING_HAS_CCC_LCCC_WORD=0x80, | ||||
| 466 | MAPPING_HAS_RAW_MAPPING=0x40, | ||||
| 467 | // unused bit 0x20, | ||||
| 468 | MAPPING_LENGTH_MASK=0x1f | ||||
| 469 | }; | ||||
| 470 | |||||
| 471 | enum { | ||||
| 472 | COMP_1_LAST_TUPLE=0x8000, | ||||
| 473 | COMP_1_TRIPLE=1, | ||||
| 474 | COMP_1_TRAIL_LIMIT=0x3400, | ||||
| 475 | COMP_1_TRAIL_MASK=0x7ffe, | ||||
| 476 | COMP_1_TRAIL_SHIFT=9, // 10-1 for the "triple" bit | ||||
| 477 | COMP_2_TRAIL_SHIFT=6, | ||||
| 478 | COMP_2_TRAIL_MASK=0xffc0 | ||||
| 479 | }; | ||||
| 480 | |||||
| 481 | // higher-level functionality ------------------------------------------ *** | ||||
| 482 | |||||
| 483 | // NFD without an NFD Normalizer2 instance. | ||||
| 484 | UnicodeString &decompose(const UnicodeString &src, UnicodeString &dest, | ||||
| 485 | UErrorCode &errorCode) const; | ||||
| 486 | /** | ||||
| 487 | * Decomposes [src, limit[ and writes the result to dest. | ||||
| 488 | * limit can be NULL if src is NUL-terminated. | ||||
| 489 | * destLengthEstimate is the initial dest buffer capacity and can be -1. | ||||
| 490 | */ | ||||
| 491 | void decompose(const char16_t *src, const char16_t *limit, | ||||
| 492 | UnicodeString &dest, int32_t destLengthEstimate, | ||||
| 493 | UErrorCode &errorCode) const; | ||||
| 494 | |||||
| 495 | const char16_t *decompose(const char16_t *src, const char16_t *limit, | ||||
| 496 | ReorderingBuffer *buffer, UErrorCode &errorCode) const; | ||||
| 497 | void decomposeAndAppend(const char16_t *src, const char16_t *limit, | ||||
| 498 | UBool doDecompose, | ||||
| 499 | UnicodeString &safeMiddle, | ||||
| 500 | ReorderingBuffer &buffer, | ||||
| 501 | UErrorCode &errorCode) const; | ||||
| 502 | |||||
| 503 | /** sink==nullptr: isNormalized()/spanQuickCheckYes() */ | ||||
| 504 | const uint8_t *decomposeUTF8(uint32_t options, | ||||
| 505 | const uint8_t *src, const uint8_t *limit, | ||||
| 506 | ByteSink *sink, Edits *edits, UErrorCode &errorCode) const; | ||||
| 507 | |||||
| 508 | UBool compose(const char16_t *src, const char16_t *limit, | ||||
| 509 | UBool onlyContiguous, | ||||
| 510 | UBool doCompose, | ||||
| 511 | ReorderingBuffer &buffer, | ||||
| 512 | UErrorCode &errorCode) const; | ||||
| 513 | const char16_t *composeQuickCheck(const char16_t *src, const char16_t *limit, | ||||
| 514 | UBool onlyContiguous, | ||||
| 515 | UNormalizationCheckResult *pQCResult) const; | ||||
| 516 | void composeAndAppend(const char16_t *src, const char16_t *limit, | ||||
| 517 | UBool doCompose, | ||||
| 518 | UBool onlyContiguous, | ||||
| 519 | UnicodeString &safeMiddle, | ||||
| 520 | ReorderingBuffer &buffer, | ||||
| 521 | UErrorCode &errorCode) const; | ||||
| 522 | |||||
| 523 | /** sink==nullptr: isNormalized() */ | ||||
| 524 | UBool composeUTF8(uint32_t options, UBool onlyContiguous, | ||||
| 525 | const uint8_t *src, const uint8_t *limit, | ||||
| 526 | ByteSink *sink, icu::Edits *edits, UErrorCode &errorCode) const; | ||||
| 527 | |||||
| 528 | const char16_t *makeFCD(const char16_t *src, const char16_t *limit, | ||||
| 529 | ReorderingBuffer *buffer, UErrorCode &errorCode) const; | ||||
| 530 | void makeFCDAndAppend(const char16_t *src, const char16_t *limit, | ||||
| 531 | UBool doMakeFCD, | ||||
| 532 | UnicodeString &safeMiddle, | ||||
| 533 | ReorderingBuffer &buffer, | ||||
| 534 | UErrorCode &errorCode) const; | ||||
| 535 | |||||
| 536 | UBool hasDecompBoundaryBefore(UChar32 c) const; | ||||
| 537 | UBool norm16HasDecompBoundaryBefore(uint16_t norm16) const; | ||||
| 538 | UBool hasDecompBoundaryAfter(UChar32 c) const; | ||||
| 539 | UBool norm16HasDecompBoundaryAfter(uint16_t norm16) const; | ||||
| 540 | UBool isDecompInert(UChar32 c) const { return isDecompYesAndZeroCC(getNorm16(c)); } | ||||
| 541 | |||||
| 542 | UBool hasCompBoundaryBefore(UChar32 c) const { | ||||
| 543 | return c<minCompNoMaybeCP || norm16HasCompBoundaryBefore(getNorm16(c)); | ||||
| 544 | } | ||||
| 545 | UBool hasCompBoundaryAfter(UChar32 c, UBool onlyContiguous) const { | ||||
| 546 | return norm16HasCompBoundaryAfter(getNorm16(c), onlyContiguous); | ||||
| 547 | } | ||||
| 548 | UBool isCompInert(UChar32 c, UBool onlyContiguous) const { | ||||
| 549 | uint16_t norm16=getNorm16(c); | ||||
| 550 | return isCompYesAndZeroCC(norm16) && | ||||
| 551 | (norm16 & HAS_COMP_BOUNDARY_AFTER) != 0 && | ||||
| 552 | (!onlyContiguous || isInert(norm16) || *getDataForYesOrNo(norm16) <= 0x1ff); | ||||
| 553 | // The last check fetches the mapping's first unit and checks tccc<=1. | ||||
| 554 | } | ||||
| 555 | |||||
| 556 | UBool hasFCDBoundaryBefore(UChar32 c) const { return hasDecompBoundaryBefore(c); } | ||||
| 557 | UBool hasFCDBoundaryAfter(UChar32 c) const { return hasDecompBoundaryAfter(c); } | ||||
| 558 | UBool isFCDInert(UChar32 c) const { return getFCD16(c)<=1; } | ||||
| 559 | private: | ||||
| 560 | friend class InitCanonIterData; | ||||
| 561 | friend class LcccContext; | ||||
| 562 | |||||
| 563 | UBool isMaybe(uint16_t norm16) const { return minMaybeNo<=norm16 && norm16<=JAMO_VT; } | ||||
| 564 | UBool isMaybeYesOrNonZeroCC(uint16_t norm16) const { return norm16>=minMaybeYes; } | ||||
| 565 | static UBool isInert(uint16_t norm16) { return norm16==INERT; } | ||||
| 566 | static UBool isJamoL(uint16_t norm16) { return norm16==JAMO_L; } | ||||
| 567 | static UBool isJamoVT(uint16_t norm16) { return norm16==JAMO_VT; } | ||||
| 568 | uint16_t hangulLVT() const { return minYesNoMappingsOnly|HAS_COMP_BOUNDARY_AFTER; } | ||||
| 569 | UBool isHangulLV(uint16_t norm16) const { return norm16==minYesNo; } | ||||
| 570 | UBool isHangulLVT(uint16_t norm16) const { | ||||
| 571 | return norm16==hangulLVT(); | ||||
| 572 | } | ||||
| 573 | UBool isCompYesAndZeroCC(uint16_t norm16) const { return norm16<minNoNo; } | ||||
| 574 | // UBool isCompYes(uint16_t norm16) const { | ||||
| 575 | // return norm16>=MIN_YES_YES_WITH_CC || norm16<minNoNo; | ||||
| 576 | // } | ||||
| 577 | // UBool isCompYesOrMaybe(uint16_t norm16) const { | ||||
| 578 | // return norm16<minNoNo || minMaybeNo<=norm16; | ||||
| 579 | // } | ||||
| 580 | // UBool hasZeroCCFromDecompYes(uint16_t norm16) const { | ||||
| 581 | // return norm16<=MIN_NORMAL_MAYBE_YES || norm16==JAMO_VT; | ||||
| 582 | // } | ||||
| 583 | UBool isDecompYesAndZeroCC(uint16_t norm16) const { | ||||
| 584 | return norm16<minYesNo || | ||||
| 585 | norm16==JAMO_VT || | ||||
| 586 | (minMaybeYes<=norm16 && norm16<=MIN_NORMAL_MAYBE_YES); | ||||
| 587 | } | ||||
| 588 | /** | ||||
| 589 | * A little faster and simpler than isDecompYesAndZeroCC() but does not include | ||||
| 590 | * the MaybeYes which combine-forward and have ccc=0. | ||||
| 591 | */ | ||||
| 592 | UBool isMostDecompYesAndZeroCC(uint16_t norm16) const { | ||||
| 593 | return norm16<minYesNo || norm16==MIN_NORMAL_MAYBE_YES || norm16==JAMO_VT; | ||||
| 594 | } | ||||
| 595 | /** Since formatVersion 5: same as isAlgorithmicNoNo() */ | ||||
| 596 | UBool isDecompNoAlgorithmic(uint16_t norm16) const { return limitNoNo<=norm16 && norm16<minMaybeNo; } | ||||
| 597 | |||||
| 598 | // For use with isCompYes(). | ||||
| 599 | // Perhaps the compiler can combine the two tests for MIN_YES_YES_WITH_CC. | ||||
| 600 | // static uint8_t getCCFromYes(uint16_t norm16) { | ||||
| 601 | // return norm16>=MIN_YES_YES_WITH_CC ? getCCFromNormalYesOrMaybe(norm16) : 0; | ||||
| 602 | // } | ||||
| 603 | uint8_t getCCFromNoNo(uint16_t norm16) const { | ||||
| 604 | const uint16_t *mapping=getDataForYesOrNo(norm16); | ||||
| 605 | if(*mapping&MAPPING_HAS_CCC_LCCC_WORD) { | ||||
| 606 | return static_cast<uint8_t>(*(mapping - 1)); | ||||
| 607 | } else { | ||||
| 608 | return 0; | ||||
| 609 | } | ||||
| 610 | } | ||||
| 611 | // requires that the [cpStart..cpLimit[ character passes isCompYesAndZeroCC() | ||||
| 612 | uint8_t getTrailCCFromCompYesAndZeroCC(uint16_t norm16) const { | ||||
| 613 | if(norm16<=minYesNo) { | ||||
| 614 | return 0; // yesYes and Hangul LV have ccc=tccc=0 | ||||
| 615 | } else { | ||||
| 616 | // For Hangul LVT we harmlessly fetch a firstUnit with tccc=0 here. | ||||
| 617 | return static_cast<uint8_t>(*getDataForYesOrNo(norm16) >> 8); // tccc from yesNo | ||||
| 618 | } | ||||
| 619 | } | ||||
| 620 | uint8_t getPreviousTrailCC(const char16_t *start, const char16_t *p) const; | ||||
| 621 | uint8_t getPreviousTrailCC(const uint8_t *start, const uint8_t *p) const; | ||||
| 622 | |||||
| 623 | // Requires algorithmic-NoNo. | ||||
| 624 | UChar32 mapAlgorithmic(UChar32 c, uint16_t norm16) const { | ||||
| 625 | return c+(norm16>>DELTA_SHIFT)-centerNoNoDelta; | ||||
| 626 | } | ||||
| 627 | UChar32 getAlgorithmicDelta(uint16_t norm16) const { | ||||
| 628 | return (norm16>>DELTA_SHIFT)-centerNoNoDelta; | ||||
| 629 | } | ||||
| 630 | |||||
| 631 | const uint16_t *getDataForYesOrNo(uint16_t norm16) const { | ||||
| 632 | return extraData+(norm16>>OFFSET_SHIFT); | ||||
| 633 | } | ||||
| 634 | const uint16_t *getDataForMaybe(uint16_t norm16) const { | ||||
| 635 | return extraData+((norm16-minMaybeNo+limitNoNo)>>OFFSET_SHIFT); | ||||
| 636 | } | ||||
| 637 | const uint16_t *getData(uint16_t norm16) const { | ||||
| 638 | if(norm16>=minMaybeNo) { | ||||
| 639 | norm16=norm16-minMaybeNo+limitNoNo; | ||||
| 640 | } | ||||
| 641 | return extraData+(norm16>>OFFSET_SHIFT); | ||||
| |||||
| 642 | } | ||||
| 643 | const uint16_t *getCompositionsListForDecompYes(uint16_t norm16) const { | ||||
| 644 | if(norm16<JAMO_L || MIN_NORMAL_MAYBE_YES<=norm16
| ||||
| 645 | return nullptr; | ||||
| 646 | } else { | ||||
| 647 | // if yesYes: if Jamo L: harmless empty list | ||||
| 648 | return getData(norm16); | ||||
| 649 | } | ||||
| 650 | } | ||||
| 651 | const uint16_t *getCompositionsListForComposite(uint16_t norm16) const { | ||||
| 652 | // A composite has both mapping & compositions list. | ||||
| 653 | const uint16_t *list=getData(norm16); | ||||
| 654 | return list+ // mapping pointer | ||||
| 655 | 1+ // +1 to skip the first unit with the mapping length | ||||
| 656 | (*list&MAPPING_LENGTH_MASK); // + mapping length | ||||
| 657 | } | ||||
| 658 | /** | ||||
| 659 | * @param c code point must have compositions | ||||
| 660 | * @return compositions list pointer | ||||
| 661 | */ | ||||
| 662 | const uint16_t *getCompositionsList(uint16_t norm16) const { | ||||
| 663 | return isDecompYes(norm16) ? | ||||
| 664 | getCompositionsListForDecompYes(norm16) : | ||||
| 665 | getCompositionsListForComposite(norm16); | ||||
| 666 | } | ||||
| 667 | |||||
| 668 | const char16_t *copyLowPrefixFromNulTerminated(const char16_t *src, | ||||
| 669 | UChar32 minNeedDataCP, | ||||
| 670 | ReorderingBuffer *buffer, | ||||
| 671 | UErrorCode &errorCode) const; | ||||
| 672 | |||||
| 673 | enum StopAt { STOP_AT_LIMIT, STOP_AT_DECOMP_BOUNDARY, STOP_AT_COMP_BOUNDARY }; | ||||
| 674 | |||||
| 675 | const char16_t *decomposeShort(const char16_t *src, const char16_t *limit, | ||||
| 676 | UBool stopAtCompBoundary, UBool onlyContiguous, | ||||
| 677 | ReorderingBuffer &buffer, UErrorCode &errorCode) const; | ||||
| 678 | UBool decompose(UChar32 c, uint16_t norm16, | ||||
| 679 | ReorderingBuffer &buffer, UErrorCode &errorCode) const; | ||||
| 680 | |||||
| 681 | const uint8_t *decomposeShort(const uint8_t *src, const uint8_t *limit, | ||||
| 682 | StopAt stopAt, UBool onlyContiguous, | ||||
| 683 | ReorderingBuffer &buffer, UErrorCode &errorCode) const; | ||||
| 684 | |||||
| 685 | static int32_t combine(const uint16_t *list, UChar32 trail); | ||||
| 686 | void addComposites(const uint16_t *list, UnicodeSet &set) const; | ||||
| 687 | void recompose(ReorderingBuffer &buffer, int32_t recomposeStartIndex, | ||||
| 688 | UBool onlyContiguous) const; | ||||
| 689 | |||||
| 690 | UBool hasCompBoundaryBefore(UChar32 c, uint16_t norm16) const { | ||||
| 691 | return c<minCompNoMaybeCP || norm16HasCompBoundaryBefore(norm16); | ||||
| 692 | } | ||||
| 693 | UBool norm16HasCompBoundaryBefore(uint16_t norm16) const { | ||||
| 694 | return norm16 < minNoNoCompNoMaybeCC || isAlgorithmicNoNo(norm16); | ||||
| 695 | } | ||||
| 696 | UBool hasCompBoundaryBefore(const char16_t *src, const char16_t *limit) const; | ||||
| 697 | UBool hasCompBoundaryBefore(const uint8_t *src, const uint8_t *limit) const; | ||||
| 698 | UBool hasCompBoundaryAfter(const char16_t *start, const char16_t *p, | ||||
| 699 | UBool onlyContiguous) const; | ||||
| 700 | UBool hasCompBoundaryAfter(const uint8_t *start, const uint8_t *p, | ||||
| 701 | UBool onlyContiguous) const; | ||||
| 702 | UBool norm16HasCompBoundaryAfter(uint16_t norm16, UBool onlyContiguous) const { | ||||
| 703 | return (norm16 & HAS_COMP_BOUNDARY_AFTER) != 0 && | ||||
| 704 | (!onlyContiguous || isTrailCC01ForCompBoundaryAfter(norm16)); | ||||
| 705 | } | ||||
| 706 | /** For FCC: Given norm16 HAS_COMP_BOUNDARY_AFTER, does it have tccc<=1? */ | ||||
| 707 | UBool isTrailCC01ForCompBoundaryAfter(uint16_t norm16) const { | ||||
| 708 | return isInert(norm16) || (isDecompNoAlgorithmic(norm16) ? | ||||
| 709 | (norm16 & DELTA_TCCC_MASK) <= DELTA_TCCC_1 : *getDataForYesOrNo(norm16) <= 0x1ff); | ||||
| 710 | } | ||||
| 711 | |||||
| 712 | const char16_t *findPreviousCompBoundary(const char16_t *start, const char16_t *p, | ||||
| 713 | UBool onlyContiguous) const; | ||||
| 714 | const char16_t *findNextCompBoundary(const char16_t *p, const char16_t *limit, | ||||
| 715 | UBool onlyContiguous) const; | ||||
| 716 | |||||
| 717 | const char16_t *findPreviousFCDBoundary(const char16_t *start, const char16_t *p) const; | ||||
| 718 | const char16_t *findNextFCDBoundary(const char16_t *p, const char16_t *limit) const; | ||||
| 719 | |||||
| 720 | void makeCanonIterDataFromNorm16(UChar32 start, UChar32 end, const uint16_t norm16, | ||||
| 721 | CanonIterData &newData, UErrorCode &errorCode) const; | ||||
| 722 | |||||
| 723 | int32_t getCanonValue(UChar32 c) const; | ||||
| 724 | const UnicodeSet &getCanonStartSet(int32_t n) const; | ||||
| 725 | |||||
| 726 | // UVersionInfo dataVersion; | ||||
| 727 | |||||
| 728 | // BMP code point thresholds for quick check loops looking at single UTF-16 code units. | ||||
| 729 | char16_t minDecompNoCP; | ||||
| 730 | char16_t minCompNoMaybeCP; | ||||
| 731 | char16_t minLcccCP; | ||||
| 732 | |||||
| 733 | // Norm16 value thresholds for quick check combinations and types of extra data. | ||||
| 734 | uint16_t minYesNo; | ||||
| 735 | uint16_t minYesNoMappingsOnly; | ||||
| 736 | uint16_t minNoNo; | ||||
| 737 | uint16_t minNoNoCompBoundaryBefore; | ||||
| 738 | uint16_t minNoNoCompNoMaybeCC; | ||||
| 739 | uint16_t minNoNoEmpty; | ||||
| 740 | uint16_t limitNoNo; | ||||
| 741 | uint16_t centerNoNoDelta; | ||||
| 742 | uint16_t minMaybeNo; | ||||
| 743 | uint16_t minMaybeNoCombinesFwd; | ||||
| 744 | uint16_t minMaybeYes; | ||||
| 745 | |||||
| 746 | const UCPTrie *normTrie; | ||||
| 747 | const uint16_t *extraData; // mappings and/or compositions | ||||
| 748 | const uint8_t *smallFCD; // [0x100] one bit per 32 BMP code points, set if any FCD!=0 | ||||
| 749 | |||||
| 750 | UInitOnce fCanonIterDataInitOnce {}; | ||||
| 751 | CanonIterData *fCanonIterData; | ||||
| 752 | }; | ||||
| 753 | |||||
| 754 | // bits in canonIterData | ||||
| 755 | #define CANON_NOT_SEGMENT_STARTER0x80000000 0x80000000 | ||||
| 756 | #define CANON_HAS_COMPOSITIONS0x40000000 0x40000000 | ||||
| 757 | #define CANON_HAS_SET0x200000 0x200000 | ||||
| 758 | #define CANON_VALUE_MASK0x1fffff 0x1fffff | ||||
| 759 | |||||
| 760 | /** | ||||
| 761 | * ICU-internal shortcut for quick access to standard Unicode normalization. | ||||
| 762 | */ | ||||
| 763 | class U_COMMON_API Normalizer2Factory { | ||||
| 764 | public: | ||||
| 765 | static const Normalizer2 *getFCDInstance(UErrorCode &errorCode); | ||||
| 766 | static const Normalizer2 *getFCCInstance(UErrorCode &errorCode); | ||||
| 767 | static const Normalizer2 *getNoopInstance(UErrorCode &errorCode); | ||||
| 768 | |||||
| 769 | static const Normalizer2 *getInstance(UNormalizationMode mode, UErrorCode &errorCode); | ||||
| 770 | |||||
| 771 | static const Normalizer2Impl *getNFCImpl(UErrorCode &errorCode); | ||||
| 772 | static const Normalizer2Impl *getNFKCImpl(UErrorCode &errorCode); | ||||
| 773 | static const Normalizer2Impl *getNFKC_CFImpl(UErrorCode &errorCode); | ||||
| 774 | |||||
| 775 | // Get the Impl instance of the Normalizer2. | ||||
| 776 | // Must be used only when it is known that norm2 is a Normalizer2WithImpl instance. | ||||
| 777 | static const Normalizer2Impl *getImpl(const Normalizer2 *norm2); | ||||
| 778 | private: | ||||
| 779 | Normalizer2Factory() = delete; // No instantiation. | ||||
| 780 | }; | ||||
| 781 | |||||
| 782 | U_NAMESPACE_END} | ||||
| 783 | |||||
| 784 | U_CAPIextern "C" int32_t U_EXPORT2 | ||||
| 785 | unorm2_swapunorm2_swap_77(const UDataSwapper *ds, | ||||
| 786 | const void *inData, int32_t length, void *outData, | ||||
| 787 | UErrorCode *pErrorCode); | ||||
| 788 | |||||
| 789 | /** | ||||
| 790 | * Get the NF*_QC property for a code point, for u_getIntPropertyValue(). | ||||
| 791 | * @internal | ||||
| 792 | */ | ||||
| 793 | U_CFUNCextern "C" UNormalizationCheckResult | ||||
| 794 | unorm_getQuickCheckunorm_getQuickCheck_77(UChar32 c, UNormalizationMode mode); | ||||
| 795 | |||||
| 796 | /** | ||||
| 797 | * Gets the 16-bit FCD value (lead & trail CCs) for a code point, for u_getIntPropertyValue(). | ||||
| 798 | * @internal | ||||
| 799 | */ | ||||
| 800 | U_CFUNCextern "C" uint16_t | ||||
| 801 | unorm_getFCD16unorm_getFCD16_77(UChar32 c); | ||||
| 802 | |||||
| 803 | /** | ||||
| 804 | * Format of Normalizer2 .nrm data files. | ||||
| 805 | * Format version 5.0. | ||||
| 806 | * | ||||
| 807 | * Normalizer2 .nrm data files provide data for the Unicode Normalization algorithms. | ||||
| 808 | * ICU ships with data files for standard Unicode Normalization Forms | ||||
| 809 | * NFC and NFD (nfc.nrm), NFKC and NFKD (nfkc.nrm), | ||||
| 810 | * NFKC_Casefold (nfkc_cf.nrm) and NFKC_Simple_Casefold (nfkc_scf.nrm). | ||||
| 811 | * Custom (application-specific) data can be built into additional .nrm files | ||||
| 812 | * with the gennorm2 build tool. | ||||
| 813 | * ICU ships with one such file, uts46.nrm, for the implementation of UTS #46. | ||||
| 814 | * | ||||
| 815 | * Normalizer2.getInstance() causes a .nrm file to be loaded, unless it has been | ||||
| 816 | * cached already. Internally, Normalizer2Impl.load() reads the .nrm file. | ||||
| 817 | * | ||||
| 818 | * A .nrm file begins with a standard ICU data file header | ||||
| 819 | * (DataHeader, see ucmndata.h and unicode/udata.h). | ||||
| 820 | * The UDataInfo.dataVersion field usually contains the Unicode version | ||||
| 821 | * for which the data was generated. | ||||
| 822 | * | ||||
| 823 | * After the header, the file contains the following parts. | ||||
| 824 | * Constants are defined as enum values of the Normalizer2Impl class. | ||||
| 825 | * | ||||
| 826 | * Many details of the data structures are described in the design doc | ||||
| 827 | * which is at https://unicode-org.github.io/icu/design/normalization/custom.html | ||||
| 828 | * | ||||
| 829 | * int32_t indexes[indexesLength]; -- indexesLength=indexes[IX_NORM_TRIE_OFFSET]/4; | ||||
| 830 | * | ||||
| 831 | * The first eight indexes are byte offsets in ascending order. | ||||
| 832 | * Each byte offset marks the start of the next part in the data file, | ||||
| 833 | * and the end of the previous one. | ||||
| 834 | * When two consecutive byte offsets are the same, then the corresponding part is empty. | ||||
| 835 | * Byte offsets are offsets from after the header, | ||||
| 836 | * that is, from the beginning of the indexes[]. | ||||
| 837 | * Each part starts at an offset with proper alignment for its data. | ||||
| 838 | * If necessary, the previous part may include padding bytes to achieve this alignment. | ||||
| 839 | * | ||||
| 840 | * minDecompNoCP=indexes[IX_MIN_DECOMP_NO_CP] is the lowest code point | ||||
| 841 | * with a decomposition mapping, that is, with NF*D_QC=No. | ||||
| 842 | * minCompNoMaybeCP=indexes[IX_MIN_COMP_NO_MAYBE_CP] is the lowest code point | ||||
| 843 | * with NF*C_QC=No (has a one-way mapping) or Maybe (combines backward). | ||||
| 844 | * minLcccCP=indexes[IX_MIN_LCCC_CP] (index 18, new in formatVersion 3) | ||||
| 845 | * is the lowest code point with lccc!=0. | ||||
| 846 | * | ||||
| 847 | * The next eight indexes are thresholds of 16-bit trie values for ranges of | ||||
| 848 | * values indicating multiple normalization properties. | ||||
| 849 | * Format version 5 adds the two minMaybeNo* threshold indexes. | ||||
| 850 | * The thresholds are listed here in threshold order, | ||||
| 851 | * not in the order they are stored in the indexes. | ||||
| 852 | * minYesNo=indexes[IX_MIN_YES_NO]; | ||||
| 853 | * minYesNoMappingsOnly=indexes[IX_MIN_YES_NO_MAPPINGS_ONLY]; | ||||
| 854 | * minNoNo=indexes[IX_MIN_NO_NO]; | ||||
| 855 | * minNoNoCompBoundaryBefore=indexes[IX_MIN_NO_NO_COMP_BOUNDARY_BEFORE]; | ||||
| 856 | * minNoNoCompNoMaybeCC=indexes[IX_MIN_NO_NO_COMP_NO_MAYBE_CC]; | ||||
| 857 | * minNoNoEmpty=indexes[IX_MIN_NO_NO_EMPTY]; | ||||
| 858 | * limitNoNo=indexes[IX_LIMIT_NO_NO]; | ||||
| 859 | * minMaybeNo=indexes[IX_MIN_MAYBE_NO]; | ||||
| 860 | * minMaybeNoCombinesFwd=indexes[IX_MIN_MAYBE_NO_COMBINES_FWD]; | ||||
| 861 | * minMaybeYes=indexes[IX_MIN_MAYBE_YES]; | ||||
| 862 | * See the normTrie description below and the design doc for details. | ||||
| 863 | * | ||||
| 864 | * UCPTrie normTrie; -- see ucptrie_impl.h and ucptrie.h, same as Java CodePointTrie | ||||
| 865 | * | ||||
| 866 | * The trie holds the main normalization data. Each code point is mapped to a 16-bit value. | ||||
| 867 | * Rather than using independent bits in the value (which would require more than 16 bits), | ||||
| 868 | * information is extracted primarily via range checks. | ||||
| 869 | * Except, format version 3+ uses bit 0 for hasCompBoundaryAfter(). | ||||
| 870 | * For example, a 16-bit value norm16 in the range minYesNo<=norm16<minNoNo | ||||
| 871 | * means that the character has NF*C_QC=Yes and NF*D_QC=No properties, | ||||
| 872 | * which means it has a two-way (round-trip) decomposition mapping. | ||||
| 873 | * Values in the ranges 2<=norm16<limitNoNo and minMaybeNo<=norm16<minMaybeYes | ||||
| 874 | * are also directly indexes into the extraData | ||||
| 875 | * pointing to mappings, compositions lists, or both. | ||||
| 876 | * Value norm16==INERT (0 in versions 1 & 2, 1 in version 3+) | ||||
| 877 | * means that the character is normalization-inert, that is, | ||||
| 878 | * it does not have a mapping, does not participate in composition, has a zero | ||||
| 879 | * canonical combining class, and forms a boundary where text before it and after it | ||||
| 880 | * can be normalized independently. | ||||
| 881 | * For details about how multiple properties are encoded in 16-bit values | ||||
| 882 | * see the design doc. | ||||
| 883 | * Note that the encoding cannot express all combinations of the properties involved; | ||||
| 884 | * it only supports those combinations that are allowed by | ||||
| 885 | * the Unicode Normalization algorithms. Details are in the design doc as well. | ||||
| 886 | * The gennorm2 tool only builds .nrm files for data that conforms to the limitations. | ||||
| 887 | * | ||||
| 888 | * The trie has a value for each lead surrogate code unit representing the "worst case" | ||||
| 889 | * properties of the 1024 supplementary characters whose UTF-16 form starts with | ||||
| 890 | * the lead surrogate. If all of the 1024 supplementary characters are normalization-inert, | ||||
| 891 | * then their lead surrogate code unit has the trie value INERT. | ||||
| 892 | * When the lead surrogate unit's value exceeds the quick check minimum during processing, | ||||
| 893 | * the properties for the full supplementary code point need to be looked up. | ||||
| 894 | * | ||||
| 895 | * uint16_t extraData[]; | ||||
| 896 | * | ||||
| 897 | * The extraData array contains many per-character data sections. | ||||
| 898 | * Each section contains mappings and/or composition lists. | ||||
| 899 | * The norm16 value of each character that has such data is directly an index to | ||||
| 900 | * a section of the extraData array. | ||||
| 901 | * | ||||
| 902 | * In version 3+, the norm16 values must be shifted right by OFFSET_SHIFT | ||||
| 903 | * for accessing extraData. | ||||
| 904 | * | ||||
| 905 | * The data structures for compositions lists and mappings are described in the design doc. | ||||
| 906 | * | ||||
| 907 | * In version 4 and below, the composition lists for MaybeYes characters were stored before | ||||
| 908 | * the data for other characters. | ||||
| 909 | * This sub-array had a length of MIN_NORMAL_MAYBE_YES-minMaybeYes. | ||||
| 910 | * In version 3 & 4, the difference must be shifted right by OFFSET_SHIFT. | ||||
| 911 | * | ||||
| 912 | * In version 5, the data for MaybeNo and MaybeYes characters is stored after | ||||
| 913 | * the data for other characters. | ||||
| 914 | * | ||||
| 915 | * If there are no MaybeNo and no MaybeYes characters, | ||||
| 916 | * then minMaybeYes==minMaybeNo==MIN_NORMAL_MAYBE_YES. | ||||
| 917 | * If there are such characters, then minMaybeNo is subtracted from their norm16 values | ||||
| 918 | * to get the index into the extraData. | ||||
| 919 | * In version 4 and below, the data index for Yes* and No* characters needs to be | ||||
| 920 | * offset by the length of the MaybeYes data. | ||||
| 921 | * In version 5, the data index for Maybe* characters needs to be offset by limitNoNo. | ||||
| 922 | * | ||||
| 923 | * Version 5 is the first to support MaybeNo characters, and | ||||
| 924 | * adds the minMaybeNo and minMaybeNoCombinesFwd thresholds and | ||||
| 925 | * the corresponding sections of the extraData. | ||||
| 926 | * | ||||
| 927 | * uint8_t smallFCD[0x100]; -- new in format version 2 | ||||
| 928 | * | ||||
| 929 | * This is a bit set to help speed up FCD value lookups in the absence of a full | ||||
| 930 | * UTrie2 or other large data structure with the full FCD value mapping. | ||||
| 931 | * | ||||
| 932 | * Each smallFCD bit is set if any of the corresponding 32 BMP code points | ||||
| 933 | * has a non-zero FCD value (lccc!=0 or tccc!=0). | ||||
| 934 | * Bit 0 of smallFCD[0] is for U+0000..U+001F. Bit 7 of smallFCD[0xff] is for U+FFE0..U+FFFF. | ||||
| 935 | * A bit for 32 lead surrogates is set if any of the 32k corresponding | ||||
| 936 | * _supplementary_ code points has a non-zero FCD value. | ||||
| 937 | * | ||||
| 938 | * This bit set is most useful for the large blocks of CJK characters with FCD=0. | ||||
| 939 | * | ||||
| 940 | * Changes from format version 1 to format version 2 --------------------------- | ||||
| 941 | * | ||||
| 942 | * - Addition of data for raw (not recursively decomposed) mappings. | ||||
| 943 | * + The MAPPING_NO_COMP_BOUNDARY_AFTER bit in the extraData is now also set when | ||||
| 944 | * the mapping is to an empty string or when the character combines-forward. | ||||
| 945 | * This subsumes the one actual use of the MAPPING_PLUS_COMPOSITION_LIST bit which | ||||
| 946 | * is then repurposed for the MAPPING_HAS_RAW_MAPPING bit. | ||||
| 947 | * + For details see the design doc. | ||||
| 948 | * - Addition of indexes[IX_MIN_YES_NO_MAPPINGS_ONLY] and separation of the yesNo extraData into | ||||
| 949 | * distinct ranges (combines-forward vs. not) | ||||
| 950 | * so that a range check can be used to find out if there is a compositions list. | ||||
| 951 | * This is fully equivalent with formatVersion 1's MAPPING_PLUS_COMPOSITION_LIST flag. | ||||
| 952 | * It is needed for the new (in ICU 49) composePair(), not for other normalization. | ||||
| 953 | * - Addition of the smallFCD[] bit set. | ||||
| 954 | * | ||||
| 955 | * Changes from format version 2 to format version 3 (ICU 60) ------------------ | ||||
| 956 | * | ||||
| 957 | * - norm16 bit 0 indicates hasCompBoundaryAfter(), | ||||
| 958 | * except that for contiguous composition (FCC) the tccc must be checked as well. | ||||
| 959 | * Data indexes and ccc values are shifted left by one (OFFSET_SHIFT). | ||||
| 960 | * Thresholds like minNoNo are tested before shifting. | ||||
| 961 | * | ||||
| 962 | * - Algorithmic mapping deltas are shifted left by two more bits (total DELTA_SHIFT), | ||||
| 963 | * to make room for two bits (three values) indicating whether the tccc is 0, 1, or greater. | ||||
| 964 | * See DELTA_TCCC_MASK etc. | ||||
| 965 | * This helps with fetching tccc/FCD values and FCC hasCompBoundaryAfter(). | ||||
| 966 | * minMaybeNo is 8-aligned so that the DELTA_TCCC_MASK bits can be tested directly. | ||||
| 967 | * | ||||
| 968 | * - Algorithmic mappings are only used for mapping to "comp yes and ccc=0" characters, | ||||
| 969 | * and ASCII characters are mapped algorithmically only to other ASCII characters. | ||||
| 970 | * This helps with hasCompBoundaryBefore() and compose() fast paths. | ||||
| 971 | * It is never necessary any more to loop for algorithmic mappings. | ||||
| 972 | * | ||||
| 973 | * - Addition of indexes[IX_MIN_NO_NO_COMP_BOUNDARY_BEFORE], | ||||
| 974 | * indexes[IX_MIN_NO_NO_COMP_NO_MAYBE_CC], and indexes[IX_MIN_NO_NO_EMPTY], | ||||
| 975 | * and separation of the noNo extraData into distinct ranges. | ||||
| 976 | * With this, the noNo norm16 value indicates whether the mapping is | ||||
| 977 | * compose-normalized, not normalized but hasCompBoundaryBefore(), | ||||
| 978 | * not even that, or maps to an empty string. | ||||
| 979 | * hasCompBoundaryBefore() can be determined solely from the norm16 value. | ||||
| 980 | * | ||||
| 981 | * - The norm16 value for Hangul LVT is now different from that for Hangul LV, | ||||
| 982 | * so that hasCompBoundaryAfter() need not check for the syllable type. | ||||
| 983 | * For Hangul LV, minYesNo continues to be used (no comp-boundary-after). | ||||
| 984 | * For Hangul LVT, minYesNoMappingsOnly|HAS_COMP_BOUNDARY_AFTER is used. | ||||
| 985 | * The extraData units at these indexes are set to firstUnit=2 and firstUnit=3, respectively, | ||||
| 986 | * to simplify some code. | ||||
| 987 | * | ||||
| 988 | * - The extraData firstUnit bit 5 is no longer necessary | ||||
| 989 | * (norm16 bit 0 used instead of firstUnit MAPPING_NO_COMP_BOUNDARY_AFTER), | ||||
| 990 | * is reserved again, and always set to 0. | ||||
| 991 | * | ||||
| 992 | * - Addition of indexes[IX_MIN_LCCC_CP], the first code point where lccc!=0. | ||||
| 993 | * This used to be hardcoded to U+0300, but in data like NFKC_Casefold it is lower: | ||||
| 994 | * U+00AD Soft Hyphen maps to an empty string, | ||||
| 995 | * which is artificially assigned "worst case" values lccc=1 and tccc=255. | ||||
| 996 | * | ||||
| 997 | * - A mapping to an empty string has explicit lccc=1 and tccc=255 values. | ||||
| 998 | * | ||||
| 999 | * Changes from format version 3 to format version 4 (ICU 63) ------------------ | ||||
| 1000 | * | ||||
| 1001 | * Switched from UTrie2 to UCPTrie/CodePointTrie. | ||||
| 1002 | * | ||||
| 1003 | * The new trie no longer stores different values for surrogate code *units* vs. | ||||
| 1004 | * surrogate code *points*. | ||||
| 1005 | * Lead surrogates still have values for optimized UTF-16 string processing. | ||||
| 1006 | * When looking up code point properties, the code now checks for lead surrogates and | ||||
| 1007 | * treats them as inert. | ||||
| 1008 | * | ||||
| 1009 | * gennorm2 now has to reject mappings for surrogate code points. | ||||
| 1010 | * UTS #46 maps unpaired surrogates to U+FFFD in code rather than via its | ||||
| 1011 | * custom normalization data file. | ||||
| 1012 | * | ||||
| 1013 | * Changes from format version 4 to format version 5 (ICU 76) ------------------ | ||||
| 1014 | * | ||||
| 1015 | * Unicode 16 adds the first MaybeYes characters which combine both backward and forward, | ||||
| 1016 | * taking this formerly theoretical data structure into reality. | ||||
| 1017 | * | ||||
| 1018 | * Unicode 16 also adds the first characters that have two-way mappings whose first characters | ||||
| 1019 | * combine backward. In order for normalization and the quick check to work properly, | ||||
| 1020 | * these composite characters also must be marked as NFC_QC=Maybe, | ||||
| 1021 | * corresponding to "combines back", although the composites themselves do not combine backward. | ||||
| 1022 | * Format version 5 adds two new ranges between "algorithmic NoNo" and MaybeYes, | ||||
| 1023 | * with thresholds minMaybeNo and minMaybeNoCombinesFwd, | ||||
| 1024 | * and indexes[IX_MIN_MAYBE_NO] and indexes[IX_MIN_MAYBE_NO_COMBINES_FWD], | ||||
| 1025 | * and corresponding mappings and composition lists in the extraData. | ||||
| 1026 | * | ||||
| 1027 | * Format version 5 moves the data for Maybe* characters from the start of the extraData array | ||||
| 1028 | * to its end. | ||||
| 1029 | */ | ||||
| 1030 | |||||
| 1031 | #endif /* !UCONFIG_NO_NORMALIZATION */ | ||||
| 1032 | #endif /* __NORMALIZER2IMPL_H__ */ |