1// © 2016 and later: Unicode, Inc. and others. 2// License & terms of use: http://www.unicode.org/copyright.html 3/******************************************************************** 4 * COPYRIGHT: 5 * Copyright (c) 1997-2016, International Business Machines Corporation and 6 * others. All Rights Reserved. 7 ********************************************************************/ 8 9#include "unicode/utypes.h" 10 11#if !UCONFIG_NO_NORMALIZATION 12 13#include "unicode/uchar.h" 14#include "unicode/errorcode.h" 15#include "unicode/normlzr.h" 16#include "unicode/stringoptions.h" 17#include "unicode/uniset.h" 18#include "unicode/usetiter.h" 19#include "unicode/schriter.h" 20#include "unicode/utf16.h" 21#include "cmemory.h" 22#include "cstring.h" 23#include "normalizer2impl.h" 24#include "testutil.h" 25#include "tstnorm.h" 26 27#define ARRAY_LENGTH(array) UPRV_LENGTHOF(array) 28 29void BasicNormalizerTest::runIndexedTest(int32_t index, UBool exec, 30 const char* &name, char* /*par*/) { 31 if(exec) { 32 logln("TestSuite BasicNormalizerTest: "); 33 } 34 TESTCASE_AUTO_BEGIN; 35 TESTCASE_AUTO(TestDecomp); 36 TESTCASE_AUTO(TestCompatDecomp); 37 TESTCASE_AUTO(TestCanonCompose); 38 TESTCASE_AUTO(TestCompatCompose); 39 TESTCASE_AUTO(TestPrevious); 40 TESTCASE_AUTO(TestHangulDecomp); 41 TESTCASE_AUTO(TestHangulCompose); 42 TESTCASE_AUTO(TestTibetan); 43 TESTCASE_AUTO(TestCompositionExclusion); 44 TESTCASE_AUTO(TestZeroIndex); 45 TESTCASE_AUTO(TestVerisign); 46 TESTCASE_AUTO(TestPreviousNext); 47 TESTCASE_AUTO(TestNormalizerAPI); 48 TESTCASE_AUTO(TestConcatenate); 49 TESTCASE_AUTO(FindFoldFCDExceptions); 50 TESTCASE_AUTO(TestCompare); 51 TESTCASE_AUTO(TestSkippable); 52#if !UCONFIG_NO_FILE_IO && !UCONFIG_NO_LEGACY_CONVERSION 53 TESTCASE_AUTO(TestCustomComp); 54 TESTCASE_AUTO(TestCustomFCC); 55#endif 56 TESTCASE_AUTO(TestFilteredNormalizer2Coverage); 57 TESTCASE_AUTO(TestNormalizeUTF8WithEdits); 58 TESTCASE_AUTO(TestLowMappingToEmpty_D); 59 TESTCASE_AUTO(TestLowMappingToEmpty_FCD); 60 TESTCASE_AUTO(TestNormalizeIllFormedText); 61 TESTCASE_AUTO(TestComposeJamoTBase); 62 TESTCASE_AUTO(TestComposeBoundaryAfter); 63 TESTCASE_AUTO_END; 64} 65 66/** 67 * Convert Java-style strings with \u Unicode escapes into UnicodeString objects 68 */ 69static UnicodeString str(const char *input) 70{ 71 UnicodeString str(input, ""); // Invariant conversion 72 return str.unescape(); 73} 74 75 76BasicNormalizerTest::BasicNormalizerTest() 77{ 78 // canonTest 79 // Input Decomposed Composed 80 81 canonTests[0][0] = str("cat"); canonTests[0][1] = str("cat"); canonTests[0][2] = str("cat"); 82 83 canonTests[1][0] = str("\\u00e0ardvark"); canonTests[1][1] = str("a\\u0300ardvark"); canonTests[1][2] = str("\\u00e0ardvark"); 84 85 canonTests[2][0] = str("\\u1e0a"); canonTests[2][1] = str("D\\u0307"); canonTests[2][2] = str("\\u1e0a"); // D-dot_above 86 87 canonTests[3][0] = str("D\\u0307"); canonTests[3][1] = str("D\\u0307"); canonTests[3][2] = str("\\u1e0a"); // D dot_above 88 89 canonTests[4][0] = str("\\u1e0c\\u0307"); canonTests[4][1] = str("D\\u0323\\u0307"); canonTests[4][2] = str("\\u1e0c\\u0307"); // D-dot_below dot_above 90 91 canonTests[5][0] = str("\\u1e0a\\u0323"); canonTests[5][1] = str("D\\u0323\\u0307"); canonTests[5][2] = str("\\u1e0c\\u0307"); // D-dot_above dot_below 92 93 canonTests[6][0] = str("D\\u0307\\u0323"); canonTests[6][1] = str("D\\u0323\\u0307"); canonTests[6][2] = str("\\u1e0c\\u0307"); // D dot_below dot_above 94 95 canonTests[7][0] = str("\\u1e10\\u0307\\u0323"); canonTests[7][1] = str("D\\u0327\\u0323\\u0307"); canonTests[7][2] = str("\\u1e10\\u0323\\u0307"); // D dot_below cedilla dot_above 96 97 canonTests[8][0] = str("D\\u0307\\u0328\\u0323"); canonTests[8][1] = str("D\\u0328\\u0323\\u0307"); canonTests[8][2] = str("\\u1e0c\\u0328\\u0307"); // D dot_above ogonek dot_below 98 99 canonTests[9][0] = str("\\u1E14"); canonTests[9][1] = str("E\\u0304\\u0300"); canonTests[9][2] = str("\\u1E14"); // E-macron-grave 100 101 canonTests[10][0] = str("\\u0112\\u0300"); canonTests[10][1] = str("E\\u0304\\u0300"); canonTests[10][2] = str("\\u1E14"); // E-macron + grave 102 103 canonTests[11][0] = str("\\u00c8\\u0304"); canonTests[11][1] = str("E\\u0300\\u0304"); canonTests[11][2] = str("\\u00c8\\u0304"); // E-grave + macron 104 105 canonTests[12][0] = str("\\u212b"); canonTests[12][1] = str("A\\u030a"); canonTests[12][2] = str("\\u00c5"); // angstrom_sign 106 107 canonTests[13][0] = str("\\u00c5"); canonTests[13][1] = str("A\\u030a"); canonTests[13][2] = str("\\u00c5"); // A-ring 108 109 canonTests[14][0] = str("\\u00C4ffin"); canonTests[14][1] = str("A\\u0308ffin"); canonTests[14][2] = str("\\u00C4ffin"); 110 111 canonTests[15][0] = str("\\u00C4\\uFB03n"); canonTests[15][1] = str("A\\u0308\\uFB03n"); canonTests[15][2] = str("\\u00C4\\uFB03n"); 112 113 canonTests[16][0] = str("Henry IV"); canonTests[16][1] = str("Henry IV"); canonTests[16][2] = str("Henry IV"); 114 115 canonTests[17][0] = str("Henry \\u2163"); canonTests[17][1] = str("Henry \\u2163"); canonTests[17][2] = str("Henry \\u2163"); 116 117 canonTests[18][0] = str("\\u30AC"); canonTests[18][1] = str("\\u30AB\\u3099"); canonTests[18][2] = str("\\u30AC"); // ga (Katakana) 118 119 canonTests[19][0] = str("\\u30AB\\u3099"); canonTests[19][1] = str("\\u30AB\\u3099"); canonTests[19][2] = str("\\u30AC"); // ka + ten 120 121 canonTests[20][0] = str("\\uFF76\\uFF9E"); canonTests[20][1] = str("\\uFF76\\uFF9E"); canonTests[20][2] = str("\\uFF76\\uFF9E"); // hw_ka + hw_ten 122 123 canonTests[21][0] = str("\\u30AB\\uFF9E"); canonTests[21][1] = str("\\u30AB\\uFF9E"); canonTests[21][2] = str("\\u30AB\\uFF9E"); // ka + hw_ten 124 125 canonTests[22][0] = str("\\uFF76\\u3099"); canonTests[22][1] = str("\\uFF76\\u3099"); canonTests[22][2] = str("\\uFF76\\u3099"); // hw_ka + ten 126 127 canonTests[23][0] = str("A\\u0300\\u0316"); canonTests[23][1] = str("A\\u0316\\u0300"); canonTests[23][2] = str("\\u00C0\\u0316"); 128 129 /* compatTest */ 130 // Input Decomposed Composed 131 compatTests[0][0] = str("cat"); compatTests[0][1] = str("cat"); compatTests[0][2] = str("cat") ; 132 133 compatTests[1][0] = str("\\uFB4f"); compatTests[1][1] = str("\\u05D0\\u05DC"); compatTests[1][2] = str("\\u05D0\\u05DC"); // Alef-Lamed vs. Alef, Lamed 134 135 compatTests[2][0] = str("\\u00C4ffin"); compatTests[2][1] = str("A\\u0308ffin"); compatTests[2][2] = str("\\u00C4ffin") ; 136 137 compatTests[3][0] = str("\\u00C4\\uFB03n"); compatTests[3][1] = str("A\\u0308ffin"); compatTests[3][2] = str("\\u00C4ffin") ; // ffi ligature -> f + f + i 138 139 compatTests[4][0] = str("Henry IV"); compatTests[4][1] = str("Henry IV"); compatTests[4][2] = str("Henry IV") ; 140 141 compatTests[5][0] = str("Henry \\u2163"); compatTests[5][1] = str("Henry IV"); compatTests[5][2] = str("Henry IV") ; 142 143 compatTests[6][0] = str("\\u30AC"); compatTests[6][1] = str("\\u30AB\\u3099"); compatTests[6][2] = str("\\u30AC") ; // ga (Katakana) 144 145 compatTests[7][0] = str("\\u30AB\\u3099"); compatTests[7][1] = str("\\u30AB\\u3099"); compatTests[7][2] = str("\\u30AC") ; // ka + ten 146 147 compatTests[8][0] = str("\\uFF76\\u3099"); compatTests[8][1] = str("\\u30AB\\u3099"); compatTests[8][2] = str("\\u30AC") ; // hw_ka + ten 148 149 /* These two are broken in Unicode 2.1.2 but fixed in 2.1.5 and later */ 150 compatTests[9][0] = str("\\uFF76\\uFF9E"); compatTests[9][1] = str("\\u30AB\\u3099"); compatTests[9][2] = str("\\u30AC") ; // hw_ka + hw_ten 151 152 compatTests[10][0] = str("\\u30AB\\uFF9E"); compatTests[10][1] = str("\\u30AB\\u3099"); compatTests[10][2] = str("\\u30AC") ; // ka + hw_ten 153 154 /* Hangul Canonical */ 155 // Input Decomposed Composed 156 hangulCanon[0][0] = str("\\ud4db"); hangulCanon[0][1] = str("\\u1111\\u1171\\u11b6"); hangulCanon[0][2] = str("\\ud4db") ; 157 158 hangulCanon[1][0] = str("\\u1111\\u1171\\u11b6"), hangulCanon[1][1] = str("\\u1111\\u1171\\u11b6"), hangulCanon[1][2] = str("\\ud4db"); 159} 160 161BasicNormalizerTest::~BasicNormalizerTest() 162{ 163} 164 165void BasicNormalizerTest::TestPrevious() 166{ 167 Normalizer* norm = new Normalizer("", UNORM_NFD); 168 169 logln("testing decomp..."); 170 uint32_t i; 171 for (i = 0; i < ARRAY_LENGTH(canonTests); i++) { 172 backAndForth(norm, canonTests[i][0]); 173 } 174 175 logln("testing compose..."); 176 norm->setMode(UNORM_NFC); 177 for (i = 0; i < ARRAY_LENGTH(canonTests); i++) { 178 backAndForth(norm, canonTests[i][0]); 179 } 180 181 delete norm; 182} 183 184void BasicNormalizerTest::TestDecomp() 185{ 186 Normalizer* norm = new Normalizer("", UNORM_NFD); 187 iterateTest(norm, canonTests, ARRAY_LENGTH(canonTests), 1); 188 staticTest(UNORM_NFD, 0, canonTests, ARRAY_LENGTH(canonTests), 1); 189 delete norm; 190} 191 192void BasicNormalizerTest::TestCompatDecomp() 193{ 194 Normalizer* norm = new Normalizer("", UNORM_NFKD); 195 iterateTest(norm, compatTests, ARRAY_LENGTH(compatTests), 1); 196 197 staticTest(UNORM_NFKD, 0, 198 compatTests, ARRAY_LENGTH(compatTests), 1); 199 delete norm; 200} 201 202void BasicNormalizerTest::TestCanonCompose() 203{ 204 Normalizer* norm = new Normalizer("", UNORM_NFC); 205 iterateTest(norm, canonTests, ARRAY_LENGTH(canonTests), 2); 206 207 staticTest(UNORM_NFC, 0, canonTests, 208 ARRAY_LENGTH(canonTests), 2); 209 delete norm; 210} 211 212void BasicNormalizerTest::TestCompatCompose() 213{ 214 Normalizer* norm = new Normalizer("", UNORM_NFKC); 215 iterateTest(norm, compatTests, ARRAY_LENGTH(compatTests), 2); 216 217 staticTest(UNORM_NFKC, 0, 218 compatTests, ARRAY_LENGTH(compatTests), 2); 219 delete norm; 220} 221 222 223//------------------------------------------------------------------------------- 224 225void BasicNormalizerTest::TestHangulCompose() 226{ 227 // Make sure that the static composition methods work 228 logln("Canonical composition..."); 229 staticTest(UNORM_NFC, 0, hangulCanon, ARRAY_LENGTH(hangulCanon), 2); 230 logln("Compatibility composition..."); 231 232 // Now try iterative composition.... 233 logln("Static composition..."); 234 Normalizer* norm = new Normalizer("", UNORM_NFC); 235 iterateTest(norm, hangulCanon, ARRAY_LENGTH(hangulCanon), 2); 236 norm->setMode(UNORM_NFKC); 237 238 // And finally, make sure you can do it in reverse too 239 logln("Reverse iteration..."); 240 norm->setMode(UNORM_NFC); 241 for (uint32_t i = 0; i < ARRAY_LENGTH(hangulCanon); i++) { 242 backAndForth(norm, hangulCanon[i][0]); 243 } 244 delete norm; 245} 246 247void BasicNormalizerTest::TestHangulDecomp() 248{ 249 // Make sure that the static decomposition methods work 250 logln("Canonical decomposition..."); 251 staticTest(UNORM_NFD, 0, hangulCanon, ARRAY_LENGTH(hangulCanon), 1); 252 logln("Compatibility decomposition..."); 253 254 // Now the iterative decomposition methods... 255 logln("Iterative decomposition..."); 256 Normalizer* norm = new Normalizer("", UNORM_NFD); 257 iterateTest(norm, hangulCanon, ARRAY_LENGTH(hangulCanon), 1); 258 norm->setMode(UNORM_NFKD); 259 260 // And finally, make sure you can do it in reverse too 261 logln("Reverse iteration..."); 262 norm->setMode(UNORM_NFD); 263 for (uint32_t i = 0; i < ARRAY_LENGTH(hangulCanon); i++) { 264 backAndForth(norm, hangulCanon[i][0]); 265 } 266 delete norm; 267} 268 269/** 270 * The Tibetan vowel sign AA, 0f71, was messed up prior to Unicode version 2.1.9. 271 */ 272void BasicNormalizerTest::TestTibetan(void) { 273 UnicodeString decomp[1][3]; 274 decomp[0][0] = str("\\u0f77"); 275 decomp[0][1] = str("\\u0f77"); 276 decomp[0][2] = str("\\u0fb2\\u0f71\\u0f80"); 277 278 UnicodeString compose[1][3]; 279 compose[0][0] = str("\\u0fb2\\u0f71\\u0f80"); 280 compose[0][1] = str("\\u0fb2\\u0f71\\u0f80"); 281 compose[0][2] = str("\\u0fb2\\u0f71\\u0f80"); 282 283 staticTest(UNORM_NFD, 0, decomp, ARRAY_LENGTH(decomp), 1); 284 staticTest(UNORM_NFKD, 0, decomp, ARRAY_LENGTH(decomp), 2); 285 staticTest(UNORM_NFC, 0, compose, ARRAY_LENGTH(compose), 1); 286 staticTest(UNORM_NFKC, 0, compose, ARRAY_LENGTH(compose), 2); 287} 288 289/** 290 * Make sure characters in the CompositionExclusion.txt list do not get 291 * composed to. 292 */ 293void BasicNormalizerTest::TestCompositionExclusion(void) { 294 // This list is generated from CompositionExclusion.txt. 295 // Update whenever the normalizer tables are updated. Note 296 // that we test all characters listed, even those that can be 297 // derived from the Unicode DB and are therefore commented 298 // out. 299 // ### TODO read composition exclusion from source/data/unidata file 300 // and test against that 301 UnicodeString EXCLUDED = str( 302 "\\u0340\\u0341\\u0343\\u0344\\u0374\\u037E\\u0387\\u0958" 303 "\\u0959\\u095A\\u095B\\u095C\\u095D\\u095E\\u095F\\u09DC" 304 "\\u09DD\\u09DF\\u0A33\\u0A36\\u0A59\\u0A5A\\u0A5B\\u0A5E" 305 "\\u0B5C\\u0B5D\\u0F43\\u0F4D\\u0F52\\u0F57\\u0F5C\\u0F69" 306 "\\u0F73\\u0F75\\u0F76\\u0F78\\u0F81\\u0F93\\u0F9D\\u0FA2" 307 "\\u0FA7\\u0FAC\\u0FB9\\u1F71\\u1F73\\u1F75\\u1F77\\u1F79" 308 "\\u1F7B\\u1F7D\\u1FBB\\u1FBE\\u1FC9\\u1FCB\\u1FD3\\u1FDB" 309 "\\u1FE3\\u1FEB\\u1FEE\\u1FEF\\u1FF9\\u1FFB\\u1FFD\\u2000" 310 "\\u2001\\u2126\\u212A\\u212B\\u2329\\u232A\\uF900\\uFA10" 311 "\\uFA12\\uFA15\\uFA20\\uFA22\\uFA25\\uFA26\\uFA2A\\uFB1F" 312 "\\uFB2A\\uFB2B\\uFB2C\\uFB2D\\uFB2E\\uFB2F\\uFB30\\uFB31" 313 "\\uFB32\\uFB33\\uFB34\\uFB35\\uFB36\\uFB38\\uFB39\\uFB3A" 314 "\\uFB3B\\uFB3C\\uFB3E\\uFB40\\uFB41\\uFB43\\uFB44\\uFB46" 315 "\\uFB47\\uFB48\\uFB49\\uFB4A\\uFB4B\\uFB4C\\uFB4D\\uFB4E" 316 ); 317 UErrorCode status = U_ZERO_ERROR; 318 for (int32_t i=0; i<EXCLUDED.length(); ++i) { 319 UnicodeString a(EXCLUDED.charAt(i)); 320 UnicodeString b; 321 UnicodeString c; 322 Normalizer::normalize(a, UNORM_NFKD, 0, b, status); 323 Normalizer::normalize(b, UNORM_NFC, 0, c, status); 324 if (c == a) { 325 errln("FAIL: " + hex(a) + " x DECOMP_COMPAT => " + 326 hex(b) + " x COMPOSE => " + 327 hex(c)); 328 } else if (verbose) { 329 logln("Ok: " + hex(a) + " x DECOMP_COMPAT => " + 330 hex(b) + " x COMPOSE => " + 331 hex(c)); 332 } 333 } 334} 335 336/** 337 * Test for a problem that showed up just before ICU 1.6 release 338 * having to do with combining characters with an index of zero. 339 * Such characters do not participate in any canonical 340 * decompositions. However, having an index of zero means that 341 * they all share one typeMask[] entry, that is, they all have to 342 * map to the same canonical class, which is not the case, in 343 * reality. 344 */ 345void BasicNormalizerTest::TestZeroIndex(void) { 346 const char* DATA[] = { 347 // Expect col1 x COMPOSE_COMPAT => col2 348 // Expect col2 x DECOMP => col3 349 "A\\u0316\\u0300", "\\u00C0\\u0316", "A\\u0316\\u0300", 350 "A\\u0300\\u0316", "\\u00C0\\u0316", "A\\u0316\\u0300", 351 "A\\u0327\\u0300", "\\u00C0\\u0327", "A\\u0327\\u0300", 352 "c\\u0321\\u0327", "c\\u0321\\u0327", "c\\u0321\\u0327", 353 "c\\u0327\\u0321", "\\u00E7\\u0321", "c\\u0327\\u0321", 354 }; 355 int32_t DATA_length = UPRV_LENGTHOF(DATA); 356 357 for (int32_t i=0; i<DATA_length; i+=3) { 358 UErrorCode status = U_ZERO_ERROR; 359 UnicodeString a(DATA[i], ""); 360 a = a.unescape(); 361 UnicodeString b; 362 Normalizer::normalize(a, UNORM_NFKC, 0, b, status); 363 if (U_FAILURE(status)) { 364 dataerrln("Error calling normalize UNORM_NFKC: %s", u_errorName(status)); 365 } else { 366 UnicodeString exp(DATA[i+1], ""); 367 exp = exp.unescape(); 368 if (b == exp) { 369 logln((UnicodeString)"Ok: " + hex(a) + " x COMPOSE_COMPAT => " + hex(b)); 370 } else { 371 errln((UnicodeString)"FAIL: " + hex(a) + " x COMPOSE_COMPAT => " + hex(b) + 372 ", expect " + hex(exp)); 373 } 374 } 375 Normalizer::normalize(b, UNORM_NFD, 0, a, status); 376 if (U_FAILURE(status)) { 377 dataerrln("Error calling normalize UNORM_NFD: %s", u_errorName(status)); 378 } else { 379 UnicodeString exp = UnicodeString(DATA[i+2], "").unescape(); 380 if (a == exp) { 381 logln((UnicodeString)"Ok: " + hex(b) + " x DECOMP => " + hex(a)); 382 } else { 383 errln((UnicodeString)"FAIL: " + hex(b) + " x DECOMP => " + hex(a) + 384 ", expect " + hex(exp)); 385 } 386 } 387 } 388} 389 390/** 391 * Run a few specific cases that are failing for Verisign. 392 */ 393void BasicNormalizerTest::TestVerisign(void) { 394 /* 395 > Their input: 396 > 05B8 05B9 05B1 0591 05C3 05B0 05AC 059F 397 > Their output (supposedly from ICU): 398 > 05B8 05B1 05B9 0591 05C3 05B0 05AC 059F 399 > My output from charlint: 400 > 05B1 05B8 05B9 0591 05C3 05B0 05AC 059F 401 402 05B8 05B9 05B1 0591 05C3 05B0 05AC 059F => 05B1 05B8 05B9 0591 05C3 05B0 403 05AC 059F 404 405 U+05B8 18 E HEBREW POINT QAMATS 406 U+05B9 19 F HEBREW POINT HOLAM 407 U+05B1 11 HEBREW POINT HATAF SEGOL 408 U+0591 220 HEBREW ACCENT ETNAHTA 409 U+05C3 0 HEBREW PUNCTUATION SOF PASUQ 410 U+05B0 10 HEBREW POINT SHEVA 411 U+05AC 230 HEBREW ACCENT ILUY 412 U+059F 230 HEBREW ACCENT QARNEY PARA 413 414 U+05B1 11 HEBREW POINT HATAF SEGOL 415 U+05B8 18 HEBREW POINT QAMATS 416 U+05B9 19 HEBREW POINT HOLAM 417 U+0591 220 HEBREW ACCENT ETNAHTA 418 U+05C3 0 HEBREW PUNCTUATION SOF PASUQ 419 U+05B0 10 HEBREW POINT SHEVA 420 U+05AC 230 HEBREW ACCENT ILUY 421 U+059F 230 HEBREW ACCENT QARNEY PARA 422 423 Wrong result: 424 U+05B8 18 HEBREW POINT QAMATS 425 U+05B1 11 HEBREW POINT HATAF SEGOL 426 U+05B9 19 HEBREW POINT HOLAM 427 U+0591 220 HEBREW ACCENT ETNAHTA 428 U+05C3 0 HEBREW PUNCTUATION SOF PASUQ 429 U+05B0 10 HEBREW POINT SHEVA 430 U+05AC 230 HEBREW ACCENT ILUY 431 U+059F 230 HEBREW ACCENT QARNEY PARA 432 433 434 > Their input: 435 >0592 05B7 05BC 05A5 05B0 05C0 05C4 05AD 436 >Their output (supposedly from ICU): 437 >0592 05B0 05B7 05BC 05A5 05C0 05AD 05C4 438 >My output from charlint: 439 >05B0 05B7 05BC 05A5 0592 05C0 05AD 05C4 440 441 0592 05B7 05BC 05A5 05B0 05C0 05C4 05AD => 05B0 05B7 05BC 05A5 0592 05C0 442 05AD 05C4 443 444 U+0592 230 HEBREW ACCENT SEGOL 445 U+05B7 17 HEBREW POINT PATAH 446 U+05BC 21 HEBREW POINT DAGESH OR MAPIQ 447 U+05A5 220 HEBREW ACCENT MERKHA 448 U+05B0 10 HEBREW POINT SHEVA 449 U+05C0 0 HEBREW PUNCTUATION PASEQ 450 U+05C4 230 HEBREW MARK UPPER DOT 451 U+05AD 222 HEBREW ACCENT DEHI 452 453 U+05B0 10 HEBREW POINT SHEVA 454 U+05B7 17 HEBREW POINT PATAH 455 U+05BC 21 HEBREW POINT DAGESH OR MAPIQ 456 U+05A5 220 HEBREW ACCENT MERKHA 457 U+0592 230 HEBREW ACCENT SEGOL 458 U+05C0 0 HEBREW PUNCTUATION PASEQ 459 U+05AD 222 HEBREW ACCENT DEHI 460 U+05C4 230 HEBREW MARK UPPER DOT 461 462 Wrong result: 463 U+0592 230 HEBREW ACCENT SEGOL 464 U+05B0 10 HEBREW POINT SHEVA 465 U+05B7 17 HEBREW POINT PATAH 466 U+05BC 21 HEBREW POINT DAGESH OR MAPIQ 467 U+05A5 220 HEBREW ACCENT MERKHA 468 U+05C0 0 HEBREW PUNCTUATION PASEQ 469 U+05AD 222 HEBREW ACCENT DEHI 470 U+05C4 230 HEBREW MARK UPPER DOT 471 */ 472 UnicodeString data[2][3]; 473 data[0][0] = str("\\u05B8\\u05B9\\u05B1\\u0591\\u05C3\\u05B0\\u05AC\\u059F"); 474 data[0][1] = str("\\u05B1\\u05B8\\u05B9\\u0591\\u05C3\\u05B0\\u05AC\\u059F"); 475 data[0][2] = str(""); 476 data[1][0] = str("\\u0592\\u05B7\\u05BC\\u05A5\\u05B0\\u05C0\\u05C4\\u05AD"); 477 data[1][1] = str("\\u05B0\\u05B7\\u05BC\\u05A5\\u0592\\u05C0\\u05AD\\u05C4"); 478 data[1][2] = str(""); 479 480 staticTest(UNORM_NFD, 0, data, ARRAY_LENGTH(data), 1); 481 staticTest(UNORM_NFC, 0, data, ARRAY_LENGTH(data), 1); 482} 483 484//------------------------------------------------------------------------ 485// Internal utilities 486// 487 488UnicodeString BasicNormalizerTest::hex(UChar ch) { 489 UnicodeString result; 490 return appendHex(ch, 4, result); 491} 492 493UnicodeString BasicNormalizerTest::hex(const UnicodeString& s) { 494 UnicodeString result; 495 for (int i = 0; i < s.length(); ++i) { 496 if (i != 0) result += (UChar)0x2c/*,*/; 497 appendHex(s[i], 4, result); 498 } 499 return result; 500} 501 502 503inline static void insert(UnicodeString& dest, int pos, UChar32 ch) 504{ 505 dest.replace(pos, 0, ch); 506} 507 508void BasicNormalizerTest::backAndForth(Normalizer* iter, const UnicodeString& input) 509{ 510 UChar32 ch; 511 UErrorCode status = U_ZERO_ERROR; 512 iter->setText(input, status); 513 514 // Run through the iterator forwards and stick it into a StringBuffer 515 UnicodeString forward; 516 for (ch = iter->first(); ch != iter->DONE; ch = iter->next()) { 517 forward += ch; 518 } 519 520 // Now do it backwards 521 UnicodeString reverse; 522 for (ch = iter->last(); ch != iter->DONE; ch = iter->previous()) { 523 insert(reverse, 0, ch); 524 } 525 526 if (forward != reverse) { 527 errln("Forward/reverse mismatch for input " + hex(input) 528 + ", forward: " + hex(forward) + ", backward: " + hex(reverse)); 529 } 530} 531 532void BasicNormalizerTest::staticTest(UNormalizationMode mode, int options, 533 UnicodeString tests[][3], int length, 534 int outCol) 535{ 536 UErrorCode status = U_ZERO_ERROR; 537 for (int i = 0; i < length; i++) 538 { 539 UnicodeString& input = tests[i][0]; 540 UnicodeString& expect = tests[i][outCol]; 541 542 logln("Normalizing '" + input + "' (" + hex(input) + ")" ); 543 544 UnicodeString output; 545 Normalizer::normalize(input, mode, options, output, status); 546 547 if (output != expect) { 548 dataerrln(UnicodeString("ERROR: case ") + i + " normalized " + hex(input) + "\n" 549 + " expected " + hex(expect) + "\n" 550 + " static got " + hex(output) ); 551 } 552 } 553} 554 555void BasicNormalizerTest::iterateTest(Normalizer* iter, 556 UnicodeString tests[][3], int length, 557 int outCol) 558{ 559 UErrorCode status = U_ZERO_ERROR; 560 for (int i = 0; i < length; i++) 561 { 562 UnicodeString& input = tests[i][0]; 563 UnicodeString& expect = tests[i][outCol]; 564 565 logln("Normalizing '" + input + "' (" + hex(input) + ")" ); 566 567 iter->setText(input, status); 568 assertEqual(input, expect, iter, UnicodeString("ERROR: case ") + i + " "); 569 } 570} 571 572void BasicNormalizerTest::assertEqual(const UnicodeString& input, 573 const UnicodeString& expected, 574 Normalizer* iter, 575 const UnicodeString& errPrefix) 576{ 577 UnicodeString result; 578 579 for (UChar32 ch = iter->first(); ch != iter->DONE; ch = iter->next()) { 580 result += ch; 581 } 582 if (result != expected) { 583 dataerrln(errPrefix + "normalized " + hex(input) + "\n" 584 + " expected " + hex(expected) + "\n" 585 + " iterate got " + hex(result) ); 586 } 587} 588 589// helper class for TestPreviousNext() 590// simple UTF-32 character iterator 591class UChar32Iterator { 592public: 593 UChar32Iterator(const UChar32 *text, int32_t len, int32_t index) : 594 s(text), length(len), i(index) {} 595 596 UChar32 current() { 597 if(i<length) { 598 return s[i]; 599 } else { 600 return 0xffff; 601 } 602 } 603 604 UChar32 next() { 605 if(i<length) { 606 return s[i++]; 607 } else { 608 return 0xffff; 609 } 610 } 611 612 UChar32 previous() { 613 if(i>0) { 614 return s[--i]; 615 } else { 616 return 0xffff; 617 } 618 } 619 620 int32_t getIndex() { 621 return i; 622 } 623private: 624 const UChar32 *s; 625 int32_t length, i; 626}; 627 628void 629BasicNormalizerTest::TestPreviousNext(const UChar *src, int32_t srcLength, 630 const UChar32 *expect, int32_t expectLength, 631 const int32_t *expectIndex, // its length=expectLength+1 632 int32_t srcMiddle, int32_t expectMiddle, 633 const char *moves, 634 UNormalizationMode mode, 635 const char *name) { 636 // iterators 637 Normalizer iter(src, srcLength, mode); 638 639 // test getStaticClassID and getDynamicClassID 640 if(iter.getDynamicClassID() != Normalizer::getStaticClassID()) { 641 errln("getStaticClassID != getDynamicClassID for Normalizer."); 642 } 643 644 UChar32Iterator iter32(expect, expectLength, expectMiddle); 645 646 UChar32 c1, c2; 647 char m; 648 649 // initially set the indexes into the middle of the strings 650 iter.setIndexOnly(srcMiddle); 651 652 // move around and compare the iteration code points with 653 // the expected ones 654 const char *move=moves; 655 while((m=*move++)!=0) { 656 if(m=='-') { 657 c1=iter.previous(); 658 c2=iter32.previous(); 659 } else if(m=='0') { 660 c1=iter.current(); 661 c2=iter32.current(); 662 } else /* m=='+' */ { 663 c1=iter.next(); 664 c2=iter32.next(); 665 } 666 667 // compare results 668 if(c1!=c2) { 669 // copy the moves until the current (m) move, and terminate 670 char history[64]; 671 uprv_strcpy(history, moves); 672 history[move-moves]=0; 673 dataerrln("error: mismatch in Normalizer iteration (%s) at %s: " 674 "got c1=U+%04lx != expected c2=U+%04lx", 675 name, history, c1, c2); 676 break; 677 } 678 679 // compare indexes 680 if(iter.getIndex()!=expectIndex[iter32.getIndex()]) { 681 // copy the moves until the current (m) move, and terminate 682 char history[64]; 683 uprv_strcpy(history, moves); 684 history[move-moves]=0; 685 errln("error: index mismatch in Normalizer iteration (%s) at %s: " 686 "Normalizer index %ld expected %ld\n", 687 name, history, iter.getIndex(), expectIndex[iter32.getIndex()]); 688 break; 689 } 690 } 691} 692 693void 694BasicNormalizerTest::TestPreviousNext() { 695 // src and expect strings 696 static const UChar src[]={ 697 U16_LEAD(0x2f999), U16_TRAIL(0x2f999), 698 U16_LEAD(0x1d15f), U16_TRAIL(0x1d15f), 699 0xc4, 700 0x1ed0 701 }; 702 static const UChar32 expect[]={ 703 0x831d, 704 0x1d158, 0x1d165, 705 0x41, 0x308, 706 0x4f, 0x302, 0x301 707 }; 708 709 // expected src indexes corresponding to expect indexes 710 static const int32_t expectIndex[]={ 711 0, 712 2, 2, 713 4, 4, 714 5, 5, 5, 715 6 // behind last character 716 }; 717 718 // src and expect strings for regression test for j2911 719 static const UChar src_j2911[]={ 720 U16_LEAD(0x2f999), U16_TRAIL(0x2f999), 721 0xdd00, 0xd900, // unpaired surrogates - regression test for j2911 722 0xc4, 723 0x4f, 0x302, 0x301 724 }; 725 static const UChar32 expect_j2911[]={ 726 0x831d, 727 0xdd00, 0xd900, // unpaired surrogates - regression test for j2911 728 0xc4, 729 0x1ed0 730 }; 731 732 // expected src indexes corresponding to expect indexes 733 static const int32_t expectIndex_j2911[]={ 734 0, 735 2, 3, 736 4, 737 5, 738 8 // behind last character 739 }; 740 741 // initial indexes into the src and expect strings 742 // for both sets of test data 743 enum { 744 SRC_MIDDLE=4, 745 EXPECT_MIDDLE=3, 746 SRC_MIDDLE_2=2, 747 EXPECT_MIDDLE_2=1 748 }; 749 750 // movement vector 751 // - for previous(), 0 for current(), + for next() 752 // for both sets of test data 753 static const char *const moves="0+0+0--0-0-+++0--+++++++0--------"; 754 755 TestPreviousNext(src, UPRV_LENGTHOF(src), 756 expect, UPRV_LENGTHOF(expect), 757 expectIndex, 758 SRC_MIDDLE, EXPECT_MIDDLE, 759 moves, UNORM_NFD, "basic"); 760 761 TestPreviousNext(src_j2911, UPRV_LENGTHOF(src_j2911), 762 expect_j2911, UPRV_LENGTHOF(expect_j2911), 763 expectIndex_j2911, 764 SRC_MIDDLE, EXPECT_MIDDLE, 765 moves, UNORM_NFKC, "j2911"); 766 767 // try again from different "middle" indexes 768 TestPreviousNext(src, UPRV_LENGTHOF(src), 769 expect, UPRV_LENGTHOF(expect), 770 expectIndex, 771 SRC_MIDDLE_2, EXPECT_MIDDLE_2, 772 moves, UNORM_NFD, "basic_2"); 773 774 TestPreviousNext(src_j2911, UPRV_LENGTHOF(src_j2911), 775 expect_j2911, UPRV_LENGTHOF(expect_j2911), 776 expectIndex_j2911, 777 SRC_MIDDLE_2, EXPECT_MIDDLE_2, 778 moves, UNORM_NFKC, "j2911_2"); 779} 780 781void BasicNormalizerTest::TestConcatenate() { 782 static const char *const 783 cases[][4]={ 784 /* mode, left, right, result */ 785 { 786 "C", 787 "re", 788 "\\u0301sum\\u00e9", 789 "r\\u00e9sum\\u00e9" 790 }, 791 { 792 "C", 793 "a\\u1100", 794 "\\u1161bcdefghijk", 795 "a\\uac00bcdefghijk" 796 }, 797 /* ### TODO: add more interesting cases */ 798 { 799 "D", 800 "\\u03B1\\u0345", 801 "\\u0C4D\\U000110BA\\U0001D169", 802 "\\u03B1\\U0001D169\\U000110BA\\u0C4D\\u0345" 803 } 804 }; 805 806 UnicodeString left, right, expect, result, r; 807 UErrorCode errorCode; 808 UNormalizationMode mode; 809 int32_t i; 810 811 /* test concatenation */ 812 for(i=0; i<UPRV_LENGTHOF(cases); ++i) { 813 switch(*cases[i][0]) { 814 case 'C': mode=UNORM_NFC; break; 815 case 'D': mode=UNORM_NFD; break; 816 case 'c': mode=UNORM_NFKC; break; 817 case 'd': mode=UNORM_NFKD; break; 818 default: mode=UNORM_NONE; break; 819 } 820 821 left=UnicodeString(cases[i][1], "").unescape(); 822 right=UnicodeString(cases[i][2], "").unescape(); 823 expect=UnicodeString(cases[i][3], "").unescape(); 824 825 //result=r=UnicodeString(); 826 errorCode=U_ZERO_ERROR; 827 828 r=Normalizer::concatenate(left, right, result, mode, 0, errorCode); 829 if(U_FAILURE(errorCode) || /*result!=r ||*/ result!=expect) { 830 dataerrln("error in Normalizer::concatenate(), cases[] fails with "+ 831 UnicodeString(u_errorName(errorCode))+", result==expect: expected: "+ 832 hex(expect)+" =========> got: " + hex(result)); 833 } 834 } 835 836 /* test error cases */ 837 838 /* left.getBuffer()==result.getBuffer() */ 839 result=r=expect=UnicodeString("zz", ""); 840 errorCode=U_UNEXPECTED_TOKEN; 841 r=Normalizer::concatenate(left, right, result, mode, 0, errorCode); 842 if(errorCode!=U_UNEXPECTED_TOKEN || result!=r || !result.isBogus()) { 843 errln("error in Normalizer::concatenate(), violates UErrorCode protocol"); 844 } 845 846 left.setToBogus(); 847 errorCode=U_ZERO_ERROR; 848 r=Normalizer::concatenate(left, right, result, mode, 0, errorCode); 849 if(errorCode!=U_ILLEGAL_ARGUMENT_ERROR || result!=r || !result.isBogus()) { 850 errln("error in Normalizer::concatenate(), does not detect left.isBogus()"); 851 } 852} 853 854// reference implementation of Normalizer::compare 855static int32_t 856ref_norm_compare(const UnicodeString &s1, const UnicodeString &s2, uint32_t options, UErrorCode &errorCode) { 857 UnicodeString r1, r2, t1, t2; 858 int32_t normOptions=(int32_t)(options>>UNORM_COMPARE_NORM_OPTIONS_SHIFT); 859 860 if(options&U_COMPARE_IGNORE_CASE) { 861 Normalizer::decompose(s1, FALSE, normOptions, r1, errorCode); 862 Normalizer::decompose(s2, FALSE, normOptions, r2, errorCode); 863 864 r1.foldCase(options); 865 r2.foldCase(options); 866 } else { 867 r1=s1; 868 r2=s2; 869 } 870 871 Normalizer::decompose(r1, FALSE, normOptions, t1, errorCode); 872 Normalizer::decompose(r2, FALSE, normOptions, t2, errorCode); 873 874 if(options&U_COMPARE_CODE_POINT_ORDER) { 875 return t1.compareCodePointOrder(t2); 876 } else { 877 return t1.compare(t2); 878 } 879} 880 881// test wrapper for Normalizer::compare, sets UNORM_INPUT_IS_FCD appropriately 882static int32_t 883_norm_compare(const UnicodeString &s1, const UnicodeString &s2, uint32_t options, UErrorCode &errorCode) { 884 int32_t normOptions=(int32_t)(options>>UNORM_COMPARE_NORM_OPTIONS_SHIFT); 885 886 if( UNORM_YES==Normalizer::quickCheck(s1, UNORM_FCD, normOptions, errorCode) && 887 UNORM_YES==Normalizer::quickCheck(s2, UNORM_FCD, normOptions, errorCode)) { 888 options|=UNORM_INPUT_IS_FCD; 889 } 890 891 return Normalizer::compare(s1, s2, options, errorCode); 892} 893 894// reference implementation of UnicodeString::caseCompare 895static int32_t 896ref_case_compare(const UnicodeString &s1, const UnicodeString &s2, uint32_t options) { 897 UnicodeString t1, t2; 898 899 t1=s1; 900 t2=s2; 901 902 t1.foldCase(options); 903 t2.foldCase(options); 904 905 if(options&U_COMPARE_CODE_POINT_ORDER) { 906 return t1.compareCodePointOrder(t2); 907 } else { 908 return t1.compare(t2); 909 } 910} 911 912// reduce an integer to -1/0/1 913static inline int32_t 914_sign(int32_t value) { 915 if(value==0) { 916 return 0; 917 } else { 918 return (value>>31)|1; 919 } 920} 921 922static const char * 923_signString(int32_t value) { 924 if(value<0) { 925 return "<0"; 926 } else if(value==0) { 927 return "=0"; 928 } else /* value>0 */ { 929 return ">0"; 930 } 931} 932 933void 934BasicNormalizerTest::TestCompare() { 935 // test Normalizer::compare and unorm_compare (thinly wrapped by the former) 936 // by comparing it with its semantic equivalent 937 // since we trust the pieces, this is sufficient 938 939 // test each string with itself and each other 940 // each time with all options 941 static const char *const 942 strings[]={ 943 // some cases from NormalizationTest.txt 944 // 0..3 945 "D\\u031B\\u0307\\u0323", 946 "\\u1E0C\\u031B\\u0307", 947 "D\\u031B\\u0323\\u0307", 948 "d\\u031B\\u0323\\u0307", 949 950 // 4..6 951 "\\u00E4", 952 "a\\u0308", 953 "A\\u0308", 954 955 // Angstrom sign = A ring 956 // 7..10 957 "\\u212B", 958 "\\u00C5", 959 "A\\u030A", 960 "a\\u030A", 961 962 // 11.14 963 "a\\u059A\\u0316\\u302A\\u032Fb", 964 "a\\u302A\\u0316\\u032F\\u059Ab", 965 "a\\u302A\\u0316\\u032F\\u059Ab", 966 "A\\u059A\\u0316\\u302A\\u032Fb", 967 968 // from ICU case folding tests 969 // 15..20 970 "A\\u00df\\u00b5\\ufb03\\U0001040c\\u0131", 971 "ass\\u03bcffi\\U00010434i", 972 "\\u0061\\u0042\\u0131\\u03a3\\u00df\\ufb03\\ud93f\\udfff", 973 "\\u0041\\u0062\\u0069\\u03c3\\u0073\\u0053\\u0046\\u0066\\u0049\\ud93f\\udfff", 974 "\\u0041\\u0062\\u0131\\u03c3\\u0053\\u0073\\u0066\\u0046\\u0069\\ud93f\\udfff", 975 "\\u0041\\u0062\\u0069\\u03c3\\u0073\\u0053\\u0046\\u0066\\u0049\\ud93f\\udffd", 976 977 // U+d800 U+10001 see implementation comment in unorm_cmpEquivFold 978 // vs. U+10000 at bottom - code point order 979 // 21..22 980 "\\ud800\\ud800\\udc01", 981 "\\ud800\\udc00", 982 983 // other code point order tests from ustrtest.cpp 984 // 23..31 985 "\\u20ac\\ud801", 986 "\\u20ac\\ud800\\udc00", 987 "\\ud800", 988 "\\ud800\\uff61", 989 "\\udfff", 990 "\\uff61\\udfff", 991 "\\uff61\\ud800\\udc02", 992 "\\ud800\\udc02", 993 "\\ud84d\\udc56", 994 995 // long strings, see cnormtst.c/TestNormCoverage() 996 // equivalent if case-insensitive 997 // 32..33 998 "\\uAD8B\\uAD8B\\uAD8B\\uAD8B" 999 "\\U0001d15e\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 1000 "\\U0001d15e\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 1001 "\\U0001d15e\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 1002 "\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 1003 "\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 1004 "aaaaaaaaaaaaaaaaaazzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz" 1005 "bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb" 1006 "ccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc" 1007 "ddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddd" 1008 "\\uAD8B\\uAD8B\\uAD8B\\uAD8B" 1009 "d\\u031B\\u0307\\u0323", 1010 1011 "\\u1100\\u116f\\u11aa\\uAD8B\\uAD8B\\u1100\\u116f\\u11aa" 1012 "\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 1013 "\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 1014 "\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 1015 "\\U0001d15e\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 1016 "\\U0001d15e\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 1017 "aaaaaaaaaaAAAAAAAAZZZZZZZZZZZZZZZZzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz" 1018 "bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb" 1019 "ccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc" 1020 "ddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddd" 1021 "\\u1100\\u116f\\u11aa\\uAD8B\\uAD8B\\u1100\\u116f\\u11aa" 1022 "\\u1E0C\\u031B\\u0307", 1023 1024 // some strings that may make a difference whether the compare function 1025 // case-folds or decomposes first 1026 // 34..41 1027 "\\u0360\\u0345\\u0334", 1028 "\\u0360\\u03b9\\u0334", 1029 1030 "\\u0360\\u1f80\\u0334", 1031 "\\u0360\\u03b1\\u0313\\u03b9\\u0334", 1032 1033 "\\u0360\\u1ffc\\u0334", 1034 "\\u0360\\u03c9\\u03b9\\u0334", 1035 1036 "a\\u0360\\u0345\\u0360\\u0345b", 1037 "a\\u0345\\u0360\\u0345\\u0360b", 1038 1039 // interesting cases for canonical caseless match with turkic i handling 1040 // 42..43 1041 "\\u00cc", 1042 "\\u0069\\u0300", 1043 1044 // strings with post-Unicode 3.2 normalization or normalization corrections 1045 // 44..45 1046 "\\u00e4\\u193b\\U0002f868", 1047 "\\u0061\\u193b\\u0308\\u36fc", 1048 1049 // empty string 1050 // 46 1051 "" 1052 }; 1053 1054 UnicodeString s[100]; // at least as many items as in strings[] ! 1055 1056 // all combinations of options 1057 // UNORM_INPUT_IS_FCD is set automatically if both input strings fulfill FCD conditions 1058 // set UNORM_UNICODE_3_2 in one additional combination 1059 static const struct { 1060 uint32_t options; 1061 const char *name; 1062 } opt[]={ 1063 { 0, "default" }, 1064 { U_COMPARE_CODE_POINT_ORDER, "c.p. order" }, 1065 { U_COMPARE_IGNORE_CASE, "ignore case" }, 1066 { U_COMPARE_CODE_POINT_ORDER|U_COMPARE_IGNORE_CASE, "c.p. order & ignore case" }, 1067 { U_COMPARE_IGNORE_CASE|U_FOLD_CASE_EXCLUDE_SPECIAL_I, "ignore case & special i" }, 1068 { U_COMPARE_CODE_POINT_ORDER|U_COMPARE_IGNORE_CASE|U_FOLD_CASE_EXCLUDE_SPECIAL_I, "c.p. order & ignore case & special i" }, 1069 { UNORM_UNICODE_3_2<<UNORM_COMPARE_NORM_OPTIONS_SHIFT, "Unicode 3.2" } 1070 }; 1071 1072 int32_t i, j, k, count=UPRV_LENGTHOF(strings); 1073 int32_t result, refResult; 1074 1075 UErrorCode errorCode; 1076 1077 // create the UnicodeStrings 1078 for(i=0; i<count; ++i) { 1079 s[i]=UnicodeString(strings[i], "").unescape(); 1080 } 1081 1082 // test them each with each other 1083 for(i=0; i<count; ++i) { 1084 for(j=i; j<count; ++j) { 1085 for(k=0; k<UPRV_LENGTHOF(opt); ++k) { 1086 // test Normalizer::compare 1087 errorCode=U_ZERO_ERROR; 1088 result=_norm_compare(s[i], s[j], opt[k].options, errorCode); 1089 refResult=ref_norm_compare(s[i], s[j], opt[k].options, errorCode); 1090 if(_sign(result)!=_sign(refResult)) { 1091 errln("Normalizer::compare(%d, %d, %s)%s should be %s %s", 1092 i, j, opt[k].name, _signString(result), _signString(refResult), 1093 U_SUCCESS(errorCode) ? "" : u_errorName(errorCode)); 1094 } 1095 1096 // test UnicodeString::caseCompare - same internal implementation function 1097 if(opt[k].options&U_COMPARE_IGNORE_CASE) { 1098 errorCode=U_ZERO_ERROR; 1099 result=s[i].caseCompare(s[j], opt[k].options); 1100 refResult=ref_case_compare(s[i], s[j], opt[k].options); 1101 if(_sign(result)!=_sign(refResult)) { 1102 errln("UniStr::caseCompare(%d, %d, %s)%s should be %s %s", 1103 i, j, opt[k].name, _signString(result), _signString(refResult), 1104 U_SUCCESS(errorCode) ? "" : u_errorName(errorCode)); 1105 } 1106 } 1107 } 1108 } 1109 } 1110 1111 // test cases with i and I to make sure Turkic works 1112 static const UChar iI[]={ 0x49, 0x69, 0x130, 0x131 }; 1113 UnicodeSet iSet, set; 1114 1115 UnicodeString s1, s2; 1116 1117 const Normalizer2Impl *nfcImpl=Normalizer2Factory::getNFCImpl(errorCode); 1118 if(U_FAILURE(errorCode) || !nfcImpl->ensureCanonIterData(errorCode)) { 1119 dataerrln("Normalizer2Factory::getNFCImpl().ensureCanonIterData() failed: %s", 1120 u_errorName(errorCode)); 1121 return; 1122 } 1123 1124 // collect all sets into one for contiguous output 1125 for(i=0; i<UPRV_LENGTHOF(iI); ++i) { 1126 if(nfcImpl->getCanonStartSet(iI[i], iSet)) { 1127 set.addAll(iSet); 1128 } 1129 } 1130 1131 // test all of these precomposed characters 1132 const Normalizer2 *nfcNorm2=Normalizer2::getNFCInstance(errorCode); 1133 UnicodeSetIterator it(set); 1134 while(it.next() && !it.isString()) { 1135 UChar32 c=it.getCodepoint(); 1136 if(!nfcNorm2->getDecomposition(c, s2)) { 1137 dataerrln("NFC.getDecomposition(i-composite U+%04lx) failed", (long)c); 1138 return; 1139 } 1140 1141 s1.setTo(c); 1142 for(k=0; k<UPRV_LENGTHOF(opt); ++k) { 1143 // test Normalizer::compare 1144 errorCode=U_ZERO_ERROR; 1145 result=_norm_compare(s1, s2, opt[k].options, errorCode); 1146 refResult=ref_norm_compare(s1, s2, opt[k].options, errorCode); 1147 if(_sign(result)!=_sign(refResult)) { 1148 errln("Normalizer::compare(U+%04x with its NFD, %s)%s should be %s %s", 1149 c, opt[k].name, _signString(result), _signString(refResult), 1150 U_SUCCESS(errorCode) ? "" : u_errorName(errorCode)); 1151 } 1152 1153 // test UnicodeString::caseCompare - same internal implementation function 1154 if(opt[k].options&U_COMPARE_IGNORE_CASE) { 1155 errorCode=U_ZERO_ERROR; 1156 result=s1.caseCompare(s2, opt[k].options); 1157 refResult=ref_case_compare(s1, s2, opt[k].options); 1158 if(_sign(result)!=_sign(refResult)) { 1159 errln("UniStr::caseCompare(U+%04x with its NFD, %s)%s should be %s %s", 1160 c, opt[k].name, _signString(result), _signString(refResult), 1161 U_SUCCESS(errorCode) ? "" : u_errorName(errorCode)); 1162 } 1163 } 1164 } 1165 } 1166 1167 // test getDecomposition() for some characters that do not decompose 1168 if( nfcNorm2->getDecomposition(0x20, s2) || 1169 nfcNorm2->getDecomposition(0x4e00, s2) || 1170 nfcNorm2->getDecomposition(0x20002, s2) 1171 ) { 1172 errln("NFC.getDecomposition() returns TRUE for characters which do not have decompositions"); 1173 } 1174 1175 // test getRawDecomposition() for some characters that do not decompose 1176 if( nfcNorm2->getRawDecomposition(0x20, s2) || 1177 nfcNorm2->getRawDecomposition(0x4e00, s2) || 1178 nfcNorm2->getRawDecomposition(0x20002, s2) 1179 ) { 1180 errln("NFC.getRawDecomposition() returns TRUE for characters which do not have decompositions"); 1181 } 1182 1183 // test composePair() for some pairs of characters that do not compose 1184 if( nfcNorm2->composePair(0x20, 0x301)>=0 || 1185 nfcNorm2->composePair(0x61, 0x305)>=0 || 1186 nfcNorm2->composePair(0x1100, 0x1160)>=0 || 1187 nfcNorm2->composePair(0xac00, 0x11a7)>=0 1188 ) { 1189 errln("NFC.composePair() incorrectly composes some pairs of characters"); 1190 } 1191 1192 // test FilteredNormalizer2::getDecomposition() 1193 UnicodeSet filter(UNICODE_STRING_SIMPLE("[^\\u00a0-\\u00ff]"), errorCode); 1194 FilteredNormalizer2 fn2(*nfcNorm2, filter); 1195 if( fn2.getDecomposition(0xe4, s1) || !fn2.getDecomposition(0x100, s2) || 1196 s2.length()!=2 || s2[0]!=0x41 || s2[1]!=0x304 1197 ) { 1198 errln("FilteredNormalizer2(NFC, ^A0-FF).getDecomposition() failed"); 1199 } 1200 1201 // test FilteredNormalizer2::getRawDecomposition() 1202 if( fn2.getRawDecomposition(0xe4, s1) || !fn2.getRawDecomposition(0x100, s2) || 1203 s2.length()!=2 || s2[0]!=0x41 || s2[1]!=0x304 1204 ) { 1205 errln("FilteredNormalizer2(NFC, ^A0-FF).getRawDecomposition() failed"); 1206 } 1207 1208 // test FilteredNormalizer2::composePair() 1209 if( 0x100!=fn2.composePair(0x41, 0x304) || 1210 fn2.composePair(0xc7, 0x301)>=0 // unfiltered result: U+1E08 1211 ) { 1212 errln("FilteredNormalizer2(NFC, ^A0-FF).composePair() failed"); 1213 } 1214} 1215 1216// verify that case-folding does not un-FCD strings 1217int32_t 1218BasicNormalizerTest::countFoldFCDExceptions(uint32_t foldingOptions) { 1219 UnicodeString s, fold, d; 1220 UChar32 c; 1221 int32_t count; 1222 uint8_t cc, trailCC, foldCC, foldTrailCC; 1223 UNormalizationCheckResult qcResult; 1224 int8_t category; 1225 UBool isNFD; 1226 UErrorCode errorCode; 1227 1228 logln("Test if case folding may un-FCD a string (folding options %04lx)", foldingOptions); 1229 1230 count=0; 1231 for(c=0; c<=0x10ffff; ++c) { 1232 errorCode = U_ZERO_ERROR; 1233 category=u_charType(c); 1234 if(category==U_UNASSIGNED) { 1235 continue; // skip unassigned code points 1236 } 1237 if(c==0xac00) { 1238 c=0xd7a3; // skip Hangul - no case folding there 1239 continue; 1240 } 1241 // skip Han blocks - no case folding there either 1242 if(c==0x3400) { 1243 c=0x4db5; 1244 continue; 1245 } 1246 if(c==0x4e00) { 1247 c=0x9fa5; 1248 continue; 1249 } 1250 if(c==0x20000) { 1251 c=0x2a6d6; 1252 continue; 1253 } 1254 1255 s.setTo(c); 1256 1257 // get leading and trailing cc for c 1258 Normalizer::decompose(s, FALSE, 0, d, errorCode); 1259 isNFD= s==d; 1260 cc=u_getCombiningClass(d.char32At(0)); 1261 trailCC=u_getCombiningClass(d.char32At(d.length()-1)); 1262 1263 // get leading and trailing cc for the case-folding of c 1264 s.foldCase(foldingOptions); 1265 Normalizer::decompose(s, FALSE, 0, d, errorCode); 1266 foldCC=u_getCombiningClass(d.char32At(0)); 1267 foldTrailCC=u_getCombiningClass(d.char32At(d.length()-1)); 1268 1269 qcResult=Normalizer::quickCheck(s, UNORM_FCD, errorCode); 1270 1271 if (U_FAILURE(errorCode)) { 1272 ++count; 1273 dataerrln("U+%04lx: Failed with error %s", u_errorName(errorCode)); 1274 } 1275 1276 // bad: 1277 // - character maps to empty string: adjacent characters may then need reordering 1278 // - folding has different leading/trailing cc's, and they don't become just 0 1279 // - folding itself is not FCD 1280 if( qcResult!=UNORM_YES || 1281 s.isEmpty() || 1282 (cc!=foldCC && foldCC!=0) || (trailCC!=foldTrailCC && foldTrailCC!=0) 1283 ) { 1284 ++count; 1285 dataerrln("U+%04lx: case-folding may un-FCD a string (folding options %04lx)", c, foldingOptions); 1286 dataerrln(" cc %02x trailCC %02x foldCC(U+%04lx) %02x foldTrailCC(U+%04lx) %02x quickCheck(folded)=%d", cc, trailCC, d.char32At(0), foldCC, d.char32At(d.length()-1), foldTrailCC, qcResult); 1287 continue; 1288 } 1289 1290 // also bad: 1291 // if a code point is in NFD but its case folding is not, then 1292 // unorm_compare will also fail 1293 if(isNFD && UNORM_YES!=Normalizer::quickCheck(s, UNORM_NFD, errorCode)) { 1294 ++count; 1295 errln("U+%04lx: case-folding un-NFDs this character (folding options %04lx)", c, foldingOptions); 1296 } 1297 } 1298 1299 logln("There are %ld code points for which case-folding may un-FCD a string (folding options %04lx)", count, foldingOptions); 1300 return count; 1301} 1302 1303void 1304BasicNormalizerTest::FindFoldFCDExceptions() { 1305 int32_t count; 1306 1307 count=countFoldFCDExceptions(0); 1308 count+=countFoldFCDExceptions(U_FOLD_CASE_EXCLUDE_SPECIAL_I); 1309 if(count>0) { 1310 /* 1311 * If case-folding un-FCDs any strings, then unorm_compare() must be 1312 * re-implemented. 1313 * It currently assumes that one can check for FCD then case-fold 1314 * and then still have FCD strings for raw decomposition without reordering. 1315 */ 1316 dataerrln("error: There are %ld code points for which case-folding may un-FCD a string for all folding options.\n" 1317 "See comment in BasicNormalizerTest::FindFoldFCDExceptions()!", count); 1318 } 1319} 1320 1321static void 1322initExpectedSkippables(UnicodeSet skipSets[UNORM_MODE_COUNT], UErrorCode &errorCode) { 1323 skipSets[UNORM_NFD].applyPattern( 1324 UNICODE_STRING_SIMPLE("[[:NFD_QC=Yes:]&[:ccc=0:]]"), errorCode); 1325 skipSets[UNORM_NFC].applyPattern( 1326 UNICODE_STRING_SIMPLE("[[:NFC_QC=Yes:]&[:ccc=0:]-[:HST=LV:]]"), errorCode); 1327 skipSets[UNORM_NFKD].applyPattern( 1328 UNICODE_STRING_SIMPLE("[[:NFKD_QC=Yes:]&[:ccc=0:]]"), errorCode); 1329 skipSets[UNORM_NFKC].applyPattern( 1330 UNICODE_STRING_SIMPLE("[[:NFKC_QC=Yes:]&[:ccc=0:]-[:HST=LV:]]"), errorCode); 1331 1332 // Remove from the NFC and NFKC sets all those characters that change 1333 // when a back-combining character is added. 1334 // First, get all of the back-combining characters and their combining classes. 1335 UnicodeSet combineBack("[:NFC_QC=Maybe:]", errorCode); 1336 int32_t numCombineBack=combineBack.size(); 1337 int32_t *combineBackCharsAndCc=new int32_t[numCombineBack*2]; 1338 UnicodeSetIterator iter(combineBack); 1339 for(int32_t i=0; i<numCombineBack; ++i) { 1340 iter.next(); 1341 UChar32 c=iter.getCodepoint(); 1342 combineBackCharsAndCc[2*i]=c; 1343 combineBackCharsAndCc[2*i+1]=u_getCombiningClass(c); 1344 } 1345 1346 // We need not look at control codes, Han characters nor Hangul LVT syllables because they 1347 // do not combine forward. LV syllables are already removed. 1348 UnicodeSet notInteresting("[[:C:][:Unified_Ideograph:][:HST=LVT:]]", errorCode); 1349 LocalPointer<UnicodeSet> unsure(&((UnicodeSet *)(skipSets[UNORM_NFC].clone()))->removeAll(notInteresting)); 1350 // System.out.format("unsure.size()=%d\n", unsure.size()); 1351 1352 // For each character about which we are unsure, see if it changes when we add 1353 // one of the back-combining characters. 1354 const Normalizer2 *norm2=Normalizer2::getNFCInstance(errorCode); 1355 UnicodeString s; 1356 iter.reset(*unsure); 1357 while(iter.next()) { 1358 UChar32 c=iter.getCodepoint(); 1359 s.setTo(c); 1360 int32_t cLength=s.length(); 1361 int32_t tccc=u_getIntPropertyValue(c, UCHAR_TRAIL_CANONICAL_COMBINING_CLASS); 1362 for(int32_t i=0; i<numCombineBack; ++i) { 1363 // If c's decomposition ends with a character with non-zero combining class, then 1364 // c can only change if it combines with a character with a non-zero combining class. 1365 int32_t cc2=combineBackCharsAndCc[2*i+1]; 1366 if(tccc==0 || cc2!=0) { 1367 UChar32 c2=combineBackCharsAndCc[2*i]; 1368 s.append(c2); 1369 if(!norm2->isNormalized(s, errorCode)) { 1370 // System.out.format("remove U+%04x (tccc=%d) + U+%04x (cc=%d)\n", c, tccc, c2, cc2); 1371 skipSets[UNORM_NFC].remove(c); 1372 skipSets[UNORM_NFKC].remove(c); 1373 break; 1374 } 1375 s.truncate(cLength); 1376 } 1377 } 1378 } 1379 delete [] combineBackCharsAndCc; 1380} 1381 1382static const char *const kModeStrings[UNORM_MODE_COUNT] = { 1383 "?", "none", "D", "KD", "C", "KC", "FCD" 1384}; 1385 1386void 1387BasicNormalizerTest::TestSkippable() { 1388 UnicodeSet diff, skipSets[UNORM_MODE_COUNT], expectSets[UNORM_MODE_COUNT]; 1389 UnicodeString s, pattern; 1390 1391 /* build NF*Skippable sets from runtime data */ 1392 IcuTestErrorCode errorCode(*this, "TestSkippable"); 1393 skipSets[UNORM_NFD].applyPattern(UNICODE_STRING_SIMPLE("[:NFD_Inert:]"), errorCode); 1394 skipSets[UNORM_NFKD].applyPattern(UNICODE_STRING_SIMPLE("[:NFKD_Inert:]"), errorCode); 1395 skipSets[UNORM_NFC].applyPattern(UNICODE_STRING_SIMPLE("[:NFC_Inert:]"), errorCode); 1396 skipSets[UNORM_NFKC].applyPattern(UNICODE_STRING_SIMPLE("[:NFKC_Inert:]"), errorCode); 1397 if(errorCode.logDataIfFailureAndReset("UnicodeSet(NF..._Inert) failed")) { 1398 return; 1399 } 1400 1401 /* get expected sets from hardcoded patterns */ 1402 initExpectedSkippables(expectSets, errorCode); 1403 errorCode.assertSuccess(); 1404 1405 for(int32_t i=UNORM_NONE; i<UNORM_MODE_COUNT; ++i) { 1406 if(skipSets[i]!=expectSets[i]) { 1407 const char *ms=kModeStrings[i]; 1408 errln("error: TestSkippable skipSets[%s]!=expectedSets[%s]\n", ms, ms); 1409 // Note: This used to depend on hardcoded UnicodeSet patterns generated by 1410 // Mark's unicodetools.com.ibm.text.UCD.NFSkippable, by 1411 // running com.ibm.text.UCD.Main with the option NFSkippable. 1412 // Since ICU 4.6/Unicode 6, we are generating the 1413 // expectSets ourselves in initSkippables(). 1414 1415 s=UNICODE_STRING_SIMPLE("skip-expect="); 1416 (diff=skipSets[i]).removeAll(expectSets[i]).toPattern(pattern, TRUE); 1417 s.append(pattern); 1418 1419 pattern.remove(); 1420 s.append(UNICODE_STRING_SIMPLE("\n\nexpect-skip=")); 1421 (diff=expectSets[i]).removeAll(skipSets[i]).toPattern(pattern, TRUE); 1422 s.append(pattern); 1423 s.append(UNICODE_STRING_SIMPLE("\n\n")); 1424 1425 errln(s); 1426 } 1427 } 1428} 1429 1430struct StringPair { const char *input, *expected; }; 1431 1432void 1433BasicNormalizerTest::TestCustomComp() { 1434 static const StringPair pairs[]={ 1435 { "\\uD801\\uE000\\uDFFE", "" }, 1436 { "\\uD800\\uD801\\uE000\\uDFFE\\uDFFF", "\\uD7FF\\uFFFF" }, 1437 { "\\uD800\\uD801\\uDFFE\\uDFFF", "\\uD7FF\\U000107FE\\uFFFF" }, 1438 { "\\uE001\\U000110B9\\u0345\\u0308\\u0327", "\\uE002\\U000110B9\\u0327\\u0345" }, 1439 { "\\uE010\\U000F0011\\uE012", "\\uE011\\uE012" }, 1440 { "\\uE010\\U000F0011\\U000F0011\\uE012", "\\uE011\\U000F0010" }, 1441 { "\\uE111\\u1161\\uE112\\u1162", "\\uAE4C\\u1102\\u0062\\u1162" }, 1442 { "\\uFFF3\\uFFF7\\U00010036\\U00010077", "\\U00010037\\U00010037\\uFFF6\\U00010037" } 1443 }; 1444 IcuTestErrorCode errorCode(*this, "BasicNormalizerTest/TestCustomComp"); 1445 const Normalizer2 *customNorm2= 1446 Normalizer2::getInstance(loadTestData(errorCode), "testnorm", 1447 UNORM2_COMPOSE, errorCode); 1448 if(errorCode.logDataIfFailureAndReset("unable to load testdata/testnorm.nrm")) { 1449 return; 1450 } 1451 for(int32_t i=0; i<UPRV_LENGTHOF(pairs); ++i) { 1452 const StringPair &pair=pairs[i]; 1453 UnicodeString input=UnicodeString(pair.input, -1, US_INV).unescape(); 1454 UnicodeString expected=UnicodeString(pair.expected, -1, US_INV).unescape(); 1455 UnicodeString result=customNorm2->normalize(input, errorCode); 1456 if(result!=expected) { 1457 errln("custom compose Normalizer2 did not normalize input %d as expected", i); 1458 } 1459 } 1460} 1461 1462void 1463BasicNormalizerTest::TestCustomFCC() { 1464 static const StringPair pairs[]={ 1465 { "\\uD801\\uE000\\uDFFE", "" }, 1466 { "\\uD800\\uD801\\uE000\\uDFFE\\uDFFF", "\\uD7FF\\uFFFF" }, 1467 { "\\uD800\\uD801\\uDFFE\\uDFFF", "\\uD7FF\\U000107FE\\uFFFF" }, 1468 // The following expected result is different from CustomComp 1469 // because of only-contiguous composition. 1470 { "\\uE001\\U000110B9\\u0345\\u0308\\u0327", "\\uE001\\U000110B9\\u0327\\u0308\\u0345" }, 1471 { "\\uE010\\U000F0011\\uE012", "\\uE011\\uE012" }, 1472 { "\\uE010\\U000F0011\\U000F0011\\uE012", "\\uE011\\U000F0010" }, 1473 { "\\uE111\\u1161\\uE112\\u1162", "\\uAE4C\\u1102\\u0062\\u1162" }, 1474 { "\\uFFF3\\uFFF7\\U00010036\\U00010077", "\\U00010037\\U00010037\\uFFF6\\U00010037" } 1475 }; 1476 IcuTestErrorCode errorCode(*this, "BasicNormalizerTest/TestCustomFCC"); 1477 const Normalizer2 *customNorm2= 1478 Normalizer2::getInstance(loadTestData(errorCode), "testnorm", 1479 UNORM2_COMPOSE_CONTIGUOUS, errorCode); 1480 if(errorCode.logDataIfFailureAndReset("unable to load testdata/testnorm.nrm")) { 1481 return; 1482 } 1483 for(int32_t i=0; i<UPRV_LENGTHOF(pairs); ++i) { 1484 const StringPair &pair=pairs[i]; 1485 UnicodeString input=UnicodeString(pair.input, -1, US_INV).unescape(); 1486 UnicodeString expected=UnicodeString(pair.expected, -1, US_INV).unescape(); 1487 UnicodeString result=customNorm2->normalize(input, errorCode); 1488 if(result!=expected) { 1489 errln("custom FCC Normalizer2 did not normalize input %d as expected", i); 1490 } 1491 } 1492} 1493 1494/* Improve code coverage of Normalizer2 */ 1495void 1496BasicNormalizerTest::TestFilteredNormalizer2Coverage() { 1497 UErrorCode errorCode = U_ZERO_ERROR; 1498 const Normalizer2 *nfcNorm2=Normalizer2::getNFCInstance(errorCode); 1499 if (U_FAILURE(errorCode)) { 1500 dataerrln("Normalizer2::getNFCInstance() call failed - %s", u_errorName(errorCode)); 1501 return; 1502 } 1503 UnicodeSet filter(UNICODE_STRING_SIMPLE("[^\\u00a0-\\u00ff\\u0310-\\u031f]"), errorCode); 1504 FilteredNormalizer2 fn2(*nfcNorm2, filter); 1505 1506 UChar32 char32 = 0x0054; 1507 1508 if (fn2.isInert(char32)) { 1509 errln("FilteredNormalizer2.isInert() failed."); 1510 } 1511 1512 if (fn2.hasBoundaryAfter(char32)) { 1513 errln("FilteredNormalizer2.hasBoundaryAfter() failed."); 1514 } 1515 1516 UChar32 c; 1517 for(c=0; c<=0x3ff; ++c) { 1518 uint8_t expectedCC= filter.contains(c) ? nfcNorm2->getCombiningClass(c) : 0; 1519 uint8_t cc=fn2.getCombiningClass(c); 1520 if(cc!=expectedCC) { 1521 errln( 1522 UnicodeString("FilteredNormalizer2(NFC, ^A0-FF,310-31F).getCombiningClass(U+")+ 1523 hex(c)+ 1524 ")==filtered NFC.getCC()"); 1525 } 1526 } 1527 1528 UnicodeString newString1 = UNICODE_STRING_SIMPLE("[^\\u0100-\\u01ff]"); 1529 UnicodeString newString2 = UNICODE_STRING_SIMPLE("[^\\u0200-\\u02ff]"); 1530 fn2.append(newString1, newString2, errorCode); 1531 if (U_FAILURE(errorCode)) { 1532 errln("FilteredNormalizer2.append() failed."); 1533 } 1534} 1535 1536void 1537BasicNormalizerTest::TestNormalizeUTF8WithEdits() { 1538 IcuTestErrorCode errorCode(*this, "TestNormalizeUTF8WithEdits"); 1539 const Normalizer2 *nfkc_cf=Normalizer2::getNFKCCasefoldInstance(errorCode); 1540 if(errorCode.logDataIfFailureAndReset("Normalizer2::getNFKCCasefoldInstance() call failed")) { 1541 return; 1542 } 1543 static const char *const src = 1544 u8" AÄA\u0308A\u0308\u00ad\u0323Ä\u0323,\u00ad\u1100\u1161가\u11A8가\u3133 "; 1545 std::string expected = u8" aääạ\u0308ạ\u0308,가각갃 "; 1546 std::string result; 1547 StringByteSink<std::string> sink(&result, expected.length()); 1548 Edits edits; 1549 nfkc_cf->normalizeUTF8(0, src, sink, &edits, errorCode); 1550 assertSuccess("normalizeUTF8 with Edits", errorCode.get()); 1551 assertEquals("normalizeUTF8 with Edits", expected.c_str(), result.c_str()); 1552 static const EditChange expectedChanges[] = { 1553 { FALSE, 2, 2 }, // 2 spaces 1554 { TRUE, 1, 1 }, // A→a 1555 { TRUE, 2, 2 }, // Ä→ä 1556 { TRUE, 3, 2 }, // A\u0308→ä 1557 { TRUE, 7, 5 }, // A\u0308\u00ad\u0323→ạ\u0308 removes the soft hyphen 1558 { TRUE, 4, 5 }, // Ä\u0323→ ạ\u0308 1559 { FALSE, 1, 1 }, // comma 1560 { TRUE, 2, 0 }, // U+00AD soft hyphen maps to empty 1561 { TRUE, 6, 3 }, // \u1100\u1161→ 가 1562 { TRUE, 6, 3 }, // 가\u11A8→ 각 1563 { TRUE, 6, 3 }, // 가\u3133→ 갃 1564 { FALSE, 2, 2 } // 2 spaces 1565 }; 1566 assertTrue("normalizeUTF8 with Edits hasChanges", edits.hasChanges()); 1567 assertEquals("normalizeUTF8 with Edits numberOfChanges", 9, edits.numberOfChanges()); 1568 TestUtility::checkEditsIter(*this, u"normalizeUTF8 with Edits", 1569 edits.getFineIterator(), edits.getFineIterator(), 1570 expectedChanges, UPRV_LENGTHOF(expectedChanges), 1571 TRUE, errorCode); 1572 1573 assertFalse("isNormalizedUTF8(source)", nfkc_cf->isNormalizedUTF8(src, errorCode)); 1574 assertTrue("isNormalizedUTF8(normalized)", nfkc_cf->isNormalizedUTF8(result, errorCode)); 1575 1576 // Omit unchanged text. 1577 expected = u8"aääạ\u0308ạ\u0308가각갃"; 1578 result.clear(); 1579 edits.reset(); 1580 nfkc_cf->normalizeUTF8(U_OMIT_UNCHANGED_TEXT, src, sink, &edits, errorCode); 1581 assertSuccess("normalizeUTF8 omit unchanged", errorCode.get()); 1582 assertEquals("normalizeUTF8 omit unchanged", expected.c_str(), result.c_str()); 1583 assertTrue("normalizeUTF8 omit unchanged hasChanges", edits.hasChanges()); 1584 assertEquals("normalizeUTF8 omit unchanged numberOfChanges", 9, edits.numberOfChanges()); 1585 TestUtility::checkEditsIter(*this, u"normalizeUTF8 omit unchanged", 1586 edits.getFineIterator(), edits.getFineIterator(), 1587 expectedChanges, UPRV_LENGTHOF(expectedChanges), 1588 TRUE, errorCode); 1589 1590 // With filter: The normalization code does not see the "A" substrings. 1591 UnicodeSet filter(u"[^A]", errorCode); 1592 FilteredNormalizer2 fn2(*nfkc_cf, filter); 1593 expected = u8" AäA\u0308A\u0323\u0308ạ\u0308,가각갃 "; 1594 result.clear(); 1595 edits.reset(); 1596 fn2.normalizeUTF8(0, src, sink, &edits, errorCode); 1597 assertSuccess("filtered normalizeUTF8", errorCode.get()); 1598 assertEquals("filtered normalizeUTF8", expected.c_str(), result.c_str()); 1599 static const EditChange filteredChanges[] = { 1600 { FALSE, 3, 3 }, // 2 spaces + A 1601 { TRUE, 2, 2 }, // Ä→ä 1602 { FALSE, 4, 4 }, // A\u0308A 1603 { TRUE, 6, 4 }, // \u0308\u00ad\u0323→\u0323\u0308 removes the soft hyphen 1604 { TRUE, 4, 5 }, // Ä\u0323→ ạ\u0308 1605 { FALSE, 1, 1 }, // comma 1606 { TRUE, 2, 0 }, // U+00AD soft hyphen maps to empty 1607 { TRUE, 6, 3 }, // \u1100\u1161→ 가 1608 { TRUE, 6, 3 }, // 가\u11A8→ 각 1609 { TRUE, 6, 3 }, // 가\u3133→ 갃 1610 { FALSE, 2, 2 } // 2 spaces 1611 }; 1612 assertTrue("filtered normalizeUTF8 hasChanges", edits.hasChanges()); 1613 assertEquals("filtered normalizeUTF8 numberOfChanges", 7, edits.numberOfChanges()); 1614 TestUtility::checkEditsIter(*this, u"filtered normalizeUTF8", 1615 edits.getFineIterator(), edits.getFineIterator(), 1616 filteredChanges, UPRV_LENGTHOF(filteredChanges), 1617 TRUE, errorCode); 1618 1619 assertFalse("filtered isNormalizedUTF8(source)", fn2.isNormalizedUTF8(src, errorCode)); 1620 assertTrue("filtered isNormalizedUTF8(normalized)", fn2.isNormalizedUTF8(result, errorCode)); 1621 1622 // Omit unchanged text. 1623 // Note that the result is not normalized because the inner normalizer 1624 // does not see text across filter spans. 1625 expected = u8"ä\u0323\u0308ạ\u0308가각갃"; 1626 result.clear(); 1627 edits.reset(); 1628 fn2.normalizeUTF8(U_OMIT_UNCHANGED_TEXT, src, sink, &edits, errorCode); 1629 assertSuccess("filtered normalizeUTF8 omit unchanged", errorCode.get()); 1630 assertEquals("filtered normalizeUTF8 omit unchanged", expected.c_str(), result.c_str()); 1631 assertTrue("filtered normalizeUTF8 omit unchanged hasChanges", edits.hasChanges()); 1632 assertEquals("filtered normalizeUTF8 omit unchanged numberOfChanges", 7, edits.numberOfChanges()); 1633 TestUtility::checkEditsIter(*this, u"filtered normalizeUTF8 omit unchanged", 1634 edits.getFineIterator(), edits.getFineIterator(), 1635 filteredChanges, UPRV_LENGTHOF(filteredChanges), 1636 TRUE, errorCode); 1637} 1638 1639void 1640BasicNormalizerTest::TestLowMappingToEmpty_D() { 1641 IcuTestErrorCode errorCode(*this, "TestLowMappingToEmpty_D"); 1642 const Normalizer2 *n2 = Normalizer2::getInstance( 1643 nullptr, "nfkc_cf", UNORM2_DECOMPOSE, errorCode); 1644 if (errorCode.logDataIfFailureAndReset("Normalizer2::getInstance() call failed")) { 1645 return; 1646 } 1647 checkLowMappingToEmpty(*n2); 1648 1649 UnicodeString sh(u'\u00AD'); 1650 assertFalse("soft hyphen is not normalized", n2->isNormalized(sh, errorCode)); 1651 UnicodeString result = n2->normalize(sh, errorCode); 1652 assertTrue("soft hyphen normalizes to empty", result.isEmpty()); 1653 assertEquals("soft hyphen QC=No", UNORM_NO, n2->quickCheck(sh, errorCode)); 1654 assertEquals("soft hyphen spanQuickCheckYes", 0, n2->spanQuickCheckYes(sh, errorCode)); 1655 1656 UnicodeString s(u"\u00ADÄ\u00AD\u0323"); 1657 result = n2->normalize(s, errorCode); 1658 assertEquals("normalize string with soft hyphens", u"a\u0323\u0308", result); 1659} 1660 1661void 1662BasicNormalizerTest::TestLowMappingToEmpty_FCD() { 1663 IcuTestErrorCode errorCode(*this, "TestLowMappingToEmpty_FCD"); 1664 const Normalizer2 *n2 = Normalizer2::getInstance( 1665 nullptr, "nfkc_cf", UNORM2_FCD, errorCode); 1666 if (errorCode.logDataIfFailureAndReset("Normalizer2::getInstance() call failed")) { 1667 return; 1668 } 1669 checkLowMappingToEmpty(*n2); 1670 1671 UnicodeString sh(u'\u00AD'); 1672 assertTrue("soft hyphen is FCD", n2->isNormalized(sh, errorCode)); 1673 1674 UnicodeString s(u"\u00ADÄ\u00AD\u0323"); 1675 UnicodeString result = n2->normalize(s, errorCode); 1676 assertEquals("normalize string with soft hyphens", u"\u00ADa\u0323\u0308", result); 1677} 1678 1679void 1680BasicNormalizerTest::checkLowMappingToEmpty(const Normalizer2 &n2) { 1681 UnicodeString mapping; 1682 assertTrue("getDecomposition(soft hyphen)", n2.getDecomposition(0xad, mapping)); 1683 assertTrue("soft hyphen maps to empty", mapping.isEmpty()); 1684 assertFalse("soft hyphen has no boundary before", n2.hasBoundaryBefore(0xad)); 1685 assertFalse("soft hyphen has no boundary after", n2.hasBoundaryAfter(0xad)); 1686 assertFalse("soft hyphen is not inert", n2.isInert(0xad)); 1687} 1688 1689void 1690BasicNormalizerTest::TestNormalizeIllFormedText() { 1691 IcuTestErrorCode errorCode(*this, "TestNormalizeIllFormedText"); 1692 const Normalizer2 *nfkc_cf = Normalizer2::getNFKCCasefoldInstance(errorCode); 1693 if(errorCode.logDataIfFailureAndReset("Normalizer2::getNFKCCasefoldInstance() call failed")) { 1694 return; 1695 } 1696 // Normalization behavior for ill-formed text is not defined. 1697 // ICU currently treats ill-formed sequences as normalization-inert 1698 // and copies them unchanged. 1699 UnicodeString src(u" A"); 1700 src.append((char16_t)0xD800).append(u"ÄA\u0308").append((char16_t)0xD900). 1701 append(u"A\u0308\u00ad\u0323").append((char16_t)0xDBFF). 1702 append(u"Ä\u0323,\u00ad").append((char16_t)0xDC00). 1703 append(u"\u1100\u1161가\u11A8가\u3133 ").append((char16_t)0xDFFF); 1704 UnicodeString expected(u" a"); 1705 expected.append((char16_t)0xD800).append(u"ää").append((char16_t)0xD900). 1706 append(u"ạ\u0308").append((char16_t)0xDBFF). 1707 append(u"ạ\u0308,").append((char16_t)0xDC00). 1708 append(u"가각갃 ").append((char16_t)0xDFFF); 1709 UnicodeString result = nfkc_cf->normalize(src, errorCode); 1710 assertSuccess("normalize", errorCode.get()); 1711 assertEquals("normalize", expected, result); 1712 1713 std::string src8(u8" A"); 1714 src8.append("\x80").append(u8"ÄA\u0308").append("\xC0\x80"). 1715 append(u8"A\u0308\u00ad\u0323").append("\xED\xA0\x80"). 1716 append(u8"Ä\u0323,\u00ad").append("\xF4\x90\x80\x80"). 1717 append(u8"\u1100\u1161가\u11A8가\u3133 ").append("\xF0"); 1718 std::string expected8(u8" a"); 1719 expected8.append("\x80").append(u8"ää").append("\xC0\x80"). 1720 append(u8"ạ\u0308").append("\xED\xA0\x80"). 1721 append(u8"ạ\u0308,").append("\xF4\x90\x80\x80"). 1722 append(u8"가각갃 ").append("\xF0"); 1723 std::string result8; 1724 StringByteSink<std::string> sink(&result8); 1725 nfkc_cf->normalizeUTF8(0, src8, sink, nullptr, errorCode); 1726 assertSuccess("normalizeUTF8", errorCode.get()); 1727 assertEquals("normalizeUTF8", expected8.c_str(), result8.c_str()); 1728} 1729 1730void 1731BasicNormalizerTest::TestComposeJamoTBase() { 1732 // Algorithmic composition of Hangul syllables must not combine with JAMO_T_BASE = U+11A7 1733 // which is not a conjoining Jamo Trailing consonant. 1734 IcuTestErrorCode errorCode(*this, "TestComposeJamoTBase"); 1735 const Normalizer2 *nfkc = Normalizer2::getNFKCInstance(errorCode); 1736 if(errorCode.logDataIfFailureAndReset("Normalizer2::getNFKCInstance() call failed")) { 1737 return; 1738 } 1739 UnicodeString s(u"\u1100\u1161\u11A7\u1100\u314F\u11A7가\u11A7"); 1740 UnicodeString expected(u"가\u11A7가\u11A7가\u11A7"); 1741 UnicodeString result = nfkc->normalize(s, errorCode); 1742 assertSuccess("normalize(LV+11A7)", errorCode.get()); 1743 assertEquals("normalize(LV+11A7)", expected, result); 1744 assertFalse("isNormalized(LV+11A7)", nfkc->isNormalized(s, errorCode)); 1745 assertTrue("isNormalized(normalized)", nfkc->isNormalized(result, errorCode)); 1746 1747 std::string s8(u8"\u1100\u1161\u11A7\u1100\u314F\u11A7가\u11A7"); 1748 std::string expected8(u8"가\u11A7가\u11A7가\u11A7"); 1749 std::string result8; 1750 StringByteSink<std::string> sink(&result8, expected8.length()); 1751 nfkc->normalizeUTF8(0, s8, sink, nullptr, errorCode); 1752 assertSuccess("normalizeUTF8(LV+11A7)", errorCode.get()); 1753 assertEquals("normalizeUTF8(LV+11A7)", expected8.c_str(), result8.c_str()); 1754 assertFalse("isNormalizedUTF8(LV+11A7)", nfkc->isNormalizedUTF8(s8, errorCode)); 1755 assertTrue("isNormalizedUTF8(normalized)", nfkc->isNormalizedUTF8(result8, errorCode)); 1756} 1757 1758void 1759BasicNormalizerTest::TestComposeBoundaryAfter() { 1760 IcuTestErrorCode errorCode(*this, "TestComposeBoundaryAfter"); 1761 const Normalizer2 *nfkc = Normalizer2::getNFKCInstance(errorCode); 1762 if(errorCode.logDataIfFailureAndReset("Normalizer2::getNFKCInstance() call failed")) { 1763 return; 1764 } 1765 // U+02DA and U+FB2C do not have compose-boundaries-after. 1766 UnicodeString s(u"\u02DA\u0339 \uFB2C\u05B6"); 1767 UnicodeString expected(u" \u0339\u030A \u05E9\u05B6\u05BC\u05C1"); 1768 UnicodeString result = nfkc->normalize(s, errorCode); 1769 assertSuccess("nfkc", errorCode.get()); 1770 assertEquals("nfkc", expected, result); 1771 assertFalse("U+02DA boundary-after", nfkc->hasBoundaryAfter(0x2DA)); 1772 assertFalse("U+FB2C boundary-after", nfkc->hasBoundaryAfter(0xFB2C)); 1773} 1774 1775#endif /* #if !UCONFIG_NO_NORMALIZATION */ 1776