1 /******************************************************************** 2 * COPYRIGHT: 3 * Copyright (c) 1997-2010, International Business Machines Corporation and 4 * others. All Rights Reserved. 5 ********************************************************************/ 6 7 #include "unicode/utypes.h" 8 9 #if !UCONFIG_NO_NORMALIZATION 10 11 #include "unicode/uchar.h" 12 #include "unicode/errorcode.h" 13 #include "unicode/normlzr.h" 14 #include "unicode/uniset.h" 15 #include "unicode/usetiter.h" 16 #include "unicode/schriter.h" 17 #include "cstring.h" 18 #include "normalizer2impl.h" 19 #include "tstnorm.h" 20 21 #define LENGTHOF(array) ((int32_t)(sizeof(array)/sizeof((array)[0]))) 22 #define ARRAY_LENGTH(array) LENGTHOF(array) 23 24 #define CASE(id,test) case id: \ 25 name = #test; \ 26 if (exec) { \ 27 logln(#test "---"); \ 28 logln((UnicodeString)""); \ 29 test(); \ 30 } \ 31 break 32 33 static UErrorCode status = U_ZERO_ERROR; 34 35 void BasicNormalizerTest::runIndexedTest(int32_t index, UBool exec, 36 const char* &name, char* /*par*/) { 37 switch (index) { 38 CASE(0,TestDecomp); 39 CASE(1,TestCompatDecomp); 40 CASE(2,TestCanonCompose); 41 CASE(3,TestCompatCompose); 42 CASE(4,TestPrevious); 43 CASE(5,TestHangulDecomp); 44 CASE(6,TestHangulCompose); 45 CASE(7,TestTibetan); 46 CASE(8,TestCompositionExclusion); 47 CASE(9,TestZeroIndex); 48 CASE(10,TestVerisign); 49 CASE(11,TestPreviousNext); 50 CASE(12,TestNormalizerAPI); 51 CASE(13,TestConcatenate); 52 CASE(14,FindFoldFCDExceptions); 53 CASE(15,TestCompare); 54 CASE(16,TestSkippable); 55 #if !UCONFIG_NO_FILE_IO && !UCONFIG_NO_LEGACY_CONVERSION 56 CASE(17,TestCustomComp); 57 CASE(18,TestCustomFCC); 58 #endif 59 CASE(19,TestFilteredNormalizer2Coverage); 60 default: name = ""; break; 61 } 62 } 63 64 /** 65 * Convert Java-style strings with \u Unicode escapes into UnicodeString objects 66 */ 67 static UnicodeString str(const char *input) 68 { 69 UnicodeString str(input, ""); // Invariant conversion 70 return str.unescape(); 71 } 72 73 74 BasicNormalizerTest::BasicNormalizerTest() 75 { 76 // canonTest 77 // Input Decomposed Composed 78 79 canonTests[0][0] = str("cat"); canonTests[0][1] = str("cat"); canonTests[0][2] = str("cat"); 80 81 canonTests[1][0] = str("\\u00e0ardvark"); canonTests[1][1] = str("a\\u0300ardvark"); canonTests[1][2] = str("\\u00e0ardvark"); 82 83 canonTests[2][0] = str("\\u1e0a"); canonTests[2][1] = str("D\\u0307"); canonTests[2][2] = str("\\u1e0a"); // D-dot_above 84 85 canonTests[3][0] = str("D\\u0307"); canonTests[3][1] = str("D\\u0307"); canonTests[3][2] = str("\\u1e0a"); // D dot_above 86 87 canonTests[4][0] = str("\\u1e0c\\u0307"); canonTests[4][1] = str("D\\u0323\\u0307"); canonTests[4][2] = str("\\u1e0c\\u0307"); // D-dot_below dot_above 88 89 canonTests[5][0] = str("\\u1e0a\\u0323"); canonTests[5][1] = str("D\\u0323\\u0307"); canonTests[5][2] = str("\\u1e0c\\u0307"); // D-dot_above dot_below 90 91 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 92 93 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 94 95 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 96 97 canonTests[9][0] = str("\\u1E14"); canonTests[9][1] = str("E\\u0304\\u0300"); canonTests[9][2] = str("\\u1E14"); // E-macron-grave 98 99 canonTests[10][0] = str("\\u0112\\u0300"); canonTests[10][1] = str("E\\u0304\\u0300"); canonTests[10][2] = str("\\u1E14"); // E-macron + grave 100 101 canonTests[11][0] = str("\\u00c8\\u0304"); canonTests[11][1] = str("E\\u0300\\u0304"); canonTests[11][2] = str("\\u00c8\\u0304"); // E-grave + macron 102 103 canonTests[12][0] = str("\\u212b"); canonTests[12][1] = str("A\\u030a"); canonTests[12][2] = str("\\u00c5"); // angstrom_sign 104 105 canonTests[13][0] = str("\\u00c5"); canonTests[13][1] = str("A\\u030a"); canonTests[13][2] = str("\\u00c5"); // A-ring 106 107 canonTests[14][0] = str("\\u00C4ffin"); canonTests[14][1] = str("A\\u0308ffin"); canonTests[14][2] = str("\\u00C4ffin"); 108 109 canonTests[15][0] = str("\\u00C4\\uFB03n"); canonTests[15][1] = str("A\\u0308\\uFB03n"); canonTests[15][2] = str("\\u00C4\\uFB03n"); 110 111 canonTests[16][0] = str("Henry IV"); canonTests[16][1] = str("Henry IV"); canonTests[16][2] = str("Henry IV"); 112 113 canonTests[17][0] = str("Henry \\u2163"); canonTests[17][1] = str("Henry \\u2163"); canonTests[17][2] = str("Henry \\u2163"); 114 115 canonTests[18][0] = str("\\u30AC"); canonTests[18][1] = str("\\u30AB\\u3099"); canonTests[18][2] = str("\\u30AC"); // ga (Katakana) 116 117 canonTests[19][0] = str("\\u30AB\\u3099"); canonTests[19][1] = str("\\u30AB\\u3099"); canonTests[19][2] = str("\\u30AC"); // ka + ten 118 119 canonTests[20][0] = str("\\uFF76\\uFF9E"); canonTests[20][1] = str("\\uFF76\\uFF9E"); canonTests[20][2] = str("\\uFF76\\uFF9E"); // hw_ka + hw_ten 120 121 canonTests[21][0] = str("\\u30AB\\uFF9E"); canonTests[21][1] = str("\\u30AB\\uFF9E"); canonTests[21][2] = str("\\u30AB\\uFF9E"); // ka + hw_ten 122 123 canonTests[22][0] = str("\\uFF76\\u3099"); canonTests[22][1] = str("\\uFF76\\u3099"); canonTests[22][2] = str("\\uFF76\\u3099"); // hw_ka + ten 124 125 canonTests[23][0] = str("A\\u0300\\u0316"); canonTests[23][1] = str("A\\u0316\\u0300"); canonTests[23][2] = str("\\u00C0\\u0316"); 126 127 /* compatTest */ 128 // Input Decomposed Composed 129 compatTests[0][0] = str("cat"); compatTests[0][1] = str("cat"); compatTests[0][2] = str("cat") ; 130 131 compatTests[1][0] = str("\\uFB4f"); compatTests[1][1] = str("\\u05D0\\u05DC"); compatTests[1][2] = str("\\u05D0\\u05DC"); // Alef-Lamed vs. Alef, Lamed 132 133 compatTests[2][0] = str("\\u00C4ffin"); compatTests[2][1] = str("A\\u0308ffin"); compatTests[2][2] = str("\\u00C4ffin") ; 134 135 compatTests[3][0] = str("\\u00C4\\uFB03n"); compatTests[3][1] = str("A\\u0308ffin"); compatTests[3][2] = str("\\u00C4ffin") ; // ffi ligature -> f + f + i 136 137 compatTests[4][0] = str("Henry IV"); compatTests[4][1] = str("Henry IV"); compatTests[4][2] = str("Henry IV") ; 138 139 compatTests[5][0] = str("Henry \\u2163"); compatTests[5][1] = str("Henry IV"); compatTests[5][2] = str("Henry IV") ; 140 141 compatTests[6][0] = str("\\u30AC"); compatTests[6][1] = str("\\u30AB\\u3099"); compatTests[6][2] = str("\\u30AC") ; // ga (Katakana) 142 143 compatTests[7][0] = str("\\u30AB\\u3099"); compatTests[7][1] = str("\\u30AB\\u3099"); compatTests[7][2] = str("\\u30AC") ; // ka + ten 144 145 compatTests[8][0] = str("\\uFF76\\u3099"); compatTests[8][1] = str("\\u30AB\\u3099"); compatTests[8][2] = str("\\u30AC") ; // hw_ka + ten 146 147 /* These two are broken in Unicode 2.1.2 but fixed in 2.1.5 and later */ 148 compatTests[9][0] = str("\\uFF76\\uFF9E"); compatTests[9][1] = str("\\u30AB\\u3099"); compatTests[9][2] = str("\\u30AC") ; // hw_ka + hw_ten 149 150 compatTests[10][0] = str("\\u30AB\\uFF9E"); compatTests[10][1] = str("\\u30AB\\u3099"); compatTests[10][2] = str("\\u30AC") ; // ka + hw_ten 151 152 /* Hangul Canonical */ 153 // Input Decomposed Composed 154 hangulCanon[0][0] = str("\\ud4db"); hangulCanon[0][1] = str("\\u1111\\u1171\\u11b6"); hangulCanon[0][2] = str("\\ud4db") ; 155 156 hangulCanon[1][0] = str("\\u1111\\u1171\\u11b6"), hangulCanon[1][1] = str("\\u1111\\u1171\\u11b6"), hangulCanon[1][2] = str("\\ud4db"); 157 } 158 159 BasicNormalizerTest::~BasicNormalizerTest() 160 { 161 } 162 163 void BasicNormalizerTest::TestPrevious() 164 { 165 Normalizer* norm = new Normalizer("", UNORM_NFD); 166 167 logln("testing decomp..."); 168 uint32_t i; 169 for (i = 0; i < ARRAY_LENGTH(canonTests); i++) { 170 backAndForth(norm, canonTests[i][0]); 171 } 172 173 logln("testing compose..."); 174 norm->setMode(UNORM_NFC); 175 for (i = 0; i < ARRAY_LENGTH(canonTests); i++) { 176 backAndForth(norm, canonTests[i][0]); 177 } 178 179 delete norm; 180 } 181 182 void BasicNormalizerTest::TestDecomp() 183 { 184 Normalizer* norm = new Normalizer("", UNORM_NFD); 185 iterateTest(norm, canonTests, ARRAY_LENGTH(canonTests), 1); 186 staticTest(UNORM_NFD, 0, canonTests, ARRAY_LENGTH(canonTests), 1); 187 delete norm; 188 } 189 190 void BasicNormalizerTest::TestCompatDecomp() 191 { 192 Normalizer* norm = new Normalizer("", UNORM_NFKD); 193 iterateTest(norm, compatTests, ARRAY_LENGTH(compatTests), 1); 194 195 staticTest(UNORM_NFKD, 0, 196 compatTests, ARRAY_LENGTH(compatTests), 1); 197 delete norm; 198 } 199 200 void BasicNormalizerTest::TestCanonCompose() 201 { 202 Normalizer* norm = new Normalizer("", UNORM_NFC); 203 iterateTest(norm, canonTests, ARRAY_LENGTH(canonTests), 2); 204 205 staticTest(UNORM_NFC, 0, canonTests, 206 ARRAY_LENGTH(canonTests), 2); 207 delete norm; 208 } 209 210 void BasicNormalizerTest::TestCompatCompose() 211 { 212 Normalizer* norm = new Normalizer("", UNORM_NFKC); 213 iterateTest(norm, compatTests, ARRAY_LENGTH(compatTests), 2); 214 215 staticTest(UNORM_NFKC, 0, 216 compatTests, ARRAY_LENGTH(compatTests), 2); 217 delete norm; 218 } 219 220 221 //------------------------------------------------------------------------------- 222 223 void BasicNormalizerTest::TestHangulCompose() 224 { 225 // Make sure that the static composition methods work 226 logln("Canonical composition..."); 227 staticTest(UNORM_NFC, 0, hangulCanon, ARRAY_LENGTH(hangulCanon), 2); 228 logln("Compatibility composition..."); 229 230 // Now try iterative composition.... 231 logln("Static composition..."); 232 Normalizer* norm = new Normalizer("", UNORM_NFC); 233 iterateTest(norm, hangulCanon, ARRAY_LENGTH(hangulCanon), 2); 234 norm->setMode(UNORM_NFKC); 235 236 // And finally, make sure you can do it in reverse too 237 logln("Reverse iteration..."); 238 norm->setMode(UNORM_NFC); 239 for (uint32_t i = 0; i < ARRAY_LENGTH(hangulCanon); i++) { 240 backAndForth(norm, hangulCanon[i][0]); 241 } 242 delete norm; 243 } 244 245 void BasicNormalizerTest::TestHangulDecomp() 246 { 247 // Make sure that the static decomposition methods work 248 logln("Canonical decomposition..."); 249 staticTest(UNORM_NFD, 0, hangulCanon, ARRAY_LENGTH(hangulCanon), 1); 250 logln("Compatibility decomposition..."); 251 252 // Now the iterative decomposition methods... 253 logln("Iterative decomposition..."); 254 Normalizer* norm = new Normalizer("", UNORM_NFD); 255 iterateTest(norm, hangulCanon, ARRAY_LENGTH(hangulCanon), 1); 256 norm->setMode(UNORM_NFKD); 257 258 // And finally, make sure you can do it in reverse too 259 logln("Reverse iteration..."); 260 norm->setMode(UNORM_NFD); 261 for (uint32_t i = 0; i < ARRAY_LENGTH(hangulCanon); i++) { 262 backAndForth(norm, hangulCanon[i][0]); 263 } 264 delete norm; 265 } 266 267 /** 268 * The Tibetan vowel sign AA, 0f71, was messed up prior to Unicode version 2.1.9. 269 */ 270 void BasicNormalizerTest::TestTibetan(void) { 271 UnicodeString decomp[1][3]; 272 decomp[0][0] = str("\\u0f77"); 273 decomp[0][1] = str("\\u0f77"); 274 decomp[0][2] = str("\\u0fb2\\u0f71\\u0f80"); 275 276 UnicodeString compose[1][3]; 277 compose[0][0] = str("\\u0fb2\\u0f71\\u0f80"); 278 compose[0][1] = str("\\u0fb2\\u0f71\\u0f80"); 279 compose[0][2] = str("\\u0fb2\\u0f71\\u0f80"); 280 281 staticTest(UNORM_NFD, 0, decomp, ARRAY_LENGTH(decomp), 1); 282 staticTest(UNORM_NFKD, 0, decomp, ARRAY_LENGTH(decomp), 2); 283 staticTest(UNORM_NFC, 0, compose, ARRAY_LENGTH(compose), 1); 284 staticTest(UNORM_NFKC, 0, compose, ARRAY_LENGTH(compose), 2); 285 } 286 287 /** 288 * Make sure characters in the CompositionExclusion.txt list do not get 289 * composed to. 290 */ 291 void BasicNormalizerTest::TestCompositionExclusion(void) { 292 // This list is generated from CompositionExclusion.txt. 293 // Update whenever the normalizer tables are updated. Note 294 // that we test all characters listed, even those that can be 295 // derived from the Unicode DB and are therefore commented 296 // out. 297 // ### TODO read composition exclusion from source/data/unidata file 298 // and test against that 299 UnicodeString EXCLUDED = str( 300 "\\u0340\\u0341\\u0343\\u0344\\u0374\\u037E\\u0387\\u0958" 301 "\\u0959\\u095A\\u095B\\u095C\\u095D\\u095E\\u095F\\u09DC" 302 "\\u09DD\\u09DF\\u0A33\\u0A36\\u0A59\\u0A5A\\u0A5B\\u0A5E" 303 "\\u0B5C\\u0B5D\\u0F43\\u0F4D\\u0F52\\u0F57\\u0F5C\\u0F69" 304 "\\u0F73\\u0F75\\u0F76\\u0F78\\u0F81\\u0F93\\u0F9D\\u0FA2" 305 "\\u0FA7\\u0FAC\\u0FB9\\u1F71\\u1F73\\u1F75\\u1F77\\u1F79" 306 "\\u1F7B\\u1F7D\\u1FBB\\u1FBE\\u1FC9\\u1FCB\\u1FD3\\u1FDB" 307 "\\u1FE3\\u1FEB\\u1FEE\\u1FEF\\u1FF9\\u1FFB\\u1FFD\\u2000" 308 "\\u2001\\u2126\\u212A\\u212B\\u2329\\u232A\\uF900\\uFA10" 309 "\\uFA12\\uFA15\\uFA20\\uFA22\\uFA25\\uFA26\\uFA2A\\uFB1F" 310 "\\uFB2A\\uFB2B\\uFB2C\\uFB2D\\uFB2E\\uFB2F\\uFB30\\uFB31" 311 "\\uFB32\\uFB33\\uFB34\\uFB35\\uFB36\\uFB38\\uFB39\\uFB3A" 312 "\\uFB3B\\uFB3C\\uFB3E\\uFB40\\uFB41\\uFB43\\uFB44\\uFB46" 313 "\\uFB47\\uFB48\\uFB49\\uFB4A\\uFB4B\\uFB4C\\uFB4D\\uFB4E" 314 ); 315 for (int32_t i=0; i<EXCLUDED.length(); ++i) { 316 UnicodeString a(EXCLUDED.charAt(i)); 317 UnicodeString b; 318 UnicodeString c; 319 Normalizer::normalize(a, UNORM_NFKD, 0, b, status); 320 Normalizer::normalize(b, UNORM_NFC, 0, c, status); 321 if (c == a) { 322 errln("FAIL: " + hex(a) + " x DECOMP_COMPAT => " + 323 hex(b) + " x COMPOSE => " + 324 hex(c)); 325 } else if (verbose) { 326 logln("Ok: " + hex(a) + " x DECOMP_COMPAT => " + 327 hex(b) + " x COMPOSE => " + 328 hex(c)); 329 } 330 } 331 } 332 333 /** 334 * Test for a problem that showed up just before ICU 1.6 release 335 * having to do with combining characters with an index of zero. 336 * Such characters do not participate in any canonical 337 * decompositions. However, having an index of zero means that 338 * they all share one typeMask[] entry, that is, they all have to 339 * map to the same canonical class, which is not the case, in 340 * reality. 341 */ 342 void BasicNormalizerTest::TestZeroIndex(void) { 343 const char* DATA[] = { 344 // Expect col1 x COMPOSE_COMPAT => col2 345 // Expect col2 x DECOMP => col3 346 "A\\u0316\\u0300", "\\u00C0\\u0316", "A\\u0316\\u0300", 347 "A\\u0300\\u0316", "\\u00C0\\u0316", "A\\u0316\\u0300", 348 "A\\u0327\\u0300", "\\u00C0\\u0327", "A\\u0327\\u0300", 349 "c\\u0321\\u0327", "c\\u0321\\u0327", "c\\u0321\\u0327", 350 "c\\u0327\\u0321", "\\u00E7\\u0321", "c\\u0327\\u0321", 351 }; 352 int32_t DATA_length = (int32_t)(sizeof(DATA) / sizeof(DATA[0])); 353 354 for (int32_t i=0; i<DATA_length; i+=3) { 355 UErrorCode status = U_ZERO_ERROR; 356 UnicodeString a(DATA[i], ""); 357 a = a.unescape(); 358 UnicodeString b; 359 Normalizer::normalize(a, UNORM_NFKC, 0, b, status); 360 if (U_FAILURE(status)) { 361 dataerrln("Error calling normalize UNORM_NFKC: %s", u_errorName(status)); 362 } else { 363 UnicodeString exp(DATA[i+1], ""); 364 exp = exp.unescape(); 365 if (b == exp) { 366 logln((UnicodeString)"Ok: " + hex(a) + " x COMPOSE_COMPAT => " + hex(b)); 367 } else { 368 errln((UnicodeString)"FAIL: " + hex(a) + " x COMPOSE_COMPAT => " + hex(b) + 369 ", expect " + hex(exp)); 370 } 371 } 372 Normalizer::normalize(b, UNORM_NFD, 0, a, status); 373 if (U_FAILURE(status)) { 374 dataerrln("Error calling normalize UNORM_NFD: %s", u_errorName(status)); 375 } else { 376 UnicodeString exp = UnicodeString(DATA[i+2], "").unescape(); 377 if (a == exp) { 378 logln((UnicodeString)"Ok: " + hex(b) + " x DECOMP => " + hex(a)); 379 } else { 380 errln((UnicodeString)"FAIL: " + hex(b) + " x DECOMP => " + hex(a) + 381 ", expect " + hex(exp)); 382 } 383 } 384 } 385 } 386 387 /** 388 * Run a few specific cases that are failing for Verisign. 389 */ 390 void BasicNormalizerTest::TestVerisign(void) { 391 /* 392 > Their input: 393 > 05B8 05B9 05B1 0591 05C3 05B0 05AC 059F 394 > Their output (supposedly from ICU): 395 > 05B8 05B1 05B9 0591 05C3 05B0 05AC 059F 396 > My output from charlint: 397 > 05B1 05B8 05B9 0591 05C3 05B0 05AC 059F 398 399 05B8 05B9 05B1 0591 05C3 05B0 05AC 059F => 05B1 05B8 05B9 0591 05C3 05B0 400 05AC 059F 401 402 U+05B8 18 E HEBREW POINT QAMATS 403 U+05B9 19 F HEBREW POINT HOLAM 404 U+05B1 11 HEBREW POINT HATAF SEGOL 405 U+0591 220 HEBREW ACCENT ETNAHTA 406 U+05C3 0 HEBREW PUNCTUATION SOF PASUQ 407 U+05B0 10 HEBREW POINT SHEVA 408 U+05AC 230 HEBREW ACCENT ILUY 409 U+059F 230 HEBREW ACCENT QARNEY PARA 410 411 U+05B1 11 HEBREW POINT HATAF SEGOL 412 U+05B8 18 HEBREW POINT QAMATS 413 U+05B9 19 HEBREW POINT HOLAM 414 U+0591 220 HEBREW ACCENT ETNAHTA 415 U+05C3 0 HEBREW PUNCTUATION SOF PASUQ 416 U+05B0 10 HEBREW POINT SHEVA 417 U+05AC 230 HEBREW ACCENT ILUY 418 U+059F 230 HEBREW ACCENT QARNEY PARA 419 420 Wrong result: 421 U+05B8 18 HEBREW POINT QAMATS 422 U+05B1 11 HEBREW POINT HATAF SEGOL 423 U+05B9 19 HEBREW POINT HOLAM 424 U+0591 220 HEBREW ACCENT ETNAHTA 425 U+05C3 0 HEBREW PUNCTUATION SOF PASUQ 426 U+05B0 10 HEBREW POINT SHEVA 427 U+05AC 230 HEBREW ACCENT ILUY 428 U+059F 230 HEBREW ACCENT QARNEY PARA 429 430 431 > Their input: 432 >0592 05B7 05BC 05A5 05B0 05C0 05C4 05AD 433 >Their output (supposedly from ICU): 434 >0592 05B0 05B7 05BC 05A5 05C0 05AD 05C4 435 >My output from charlint: 436 >05B0 05B7 05BC 05A5 0592 05C0 05AD 05C4 437 438 0592 05B7 05BC 05A5 05B0 05C0 05C4 05AD => 05B0 05B7 05BC 05A5 0592 05C0 439 05AD 05C4 440 441 U+0592 230 HEBREW ACCENT SEGOL 442 U+05B7 17 HEBREW POINT PATAH 443 U+05BC 21 HEBREW POINT DAGESH OR MAPIQ 444 U+05A5 220 HEBREW ACCENT MERKHA 445 U+05B0 10 HEBREW POINT SHEVA 446 U+05C0 0 HEBREW PUNCTUATION PASEQ 447 U+05C4 230 HEBREW MARK UPPER DOT 448 U+05AD 222 HEBREW ACCENT DEHI 449 450 U+05B0 10 HEBREW POINT SHEVA 451 U+05B7 17 HEBREW POINT PATAH 452 U+05BC 21 HEBREW POINT DAGESH OR MAPIQ 453 U+05A5 220 HEBREW ACCENT MERKHA 454 U+0592 230 HEBREW ACCENT SEGOL 455 U+05C0 0 HEBREW PUNCTUATION PASEQ 456 U+05AD 222 HEBREW ACCENT DEHI 457 U+05C4 230 HEBREW MARK UPPER DOT 458 459 Wrong result: 460 U+0592 230 HEBREW ACCENT SEGOL 461 U+05B0 10 HEBREW POINT SHEVA 462 U+05B7 17 HEBREW POINT PATAH 463 U+05BC 21 HEBREW POINT DAGESH OR MAPIQ 464 U+05A5 220 HEBREW ACCENT MERKHA 465 U+05C0 0 HEBREW PUNCTUATION PASEQ 466 U+05AD 222 HEBREW ACCENT DEHI 467 U+05C4 230 HEBREW MARK UPPER DOT 468 */ 469 UnicodeString data[2][3]; 470 data[0][0] = str("\\u05B8\\u05B9\\u05B1\\u0591\\u05C3\\u05B0\\u05AC\\u059F"); 471 data[0][1] = str("\\u05B1\\u05B8\\u05B9\\u0591\\u05C3\\u05B0\\u05AC\\u059F"); 472 data[0][2] = str(""); 473 data[1][0] = str("\\u0592\\u05B7\\u05BC\\u05A5\\u05B0\\u05C0\\u05C4\\u05AD"); 474 data[1][1] = str("\\u05B0\\u05B7\\u05BC\\u05A5\\u0592\\u05C0\\u05AD\\u05C4"); 475 data[1][2] = str(""); 476 477 staticTest(UNORM_NFD, 0, data, ARRAY_LENGTH(data), 1); 478 staticTest(UNORM_NFC, 0, data, ARRAY_LENGTH(data), 1); 479 } 480 481 //------------------------------------------------------------------------ 482 // Internal utilities 483 // 484 485 UnicodeString BasicNormalizerTest::hex(UChar ch) { 486 UnicodeString result; 487 return appendHex(ch, 4, result); 488 } 489 490 UnicodeString BasicNormalizerTest::hex(const UnicodeString& s) { 491 UnicodeString result; 492 for (int i = 0; i < s.length(); ++i) { 493 if (i != 0) result += (UChar)0x2c/*,*/; 494 appendHex(s[i], 4, result); 495 } 496 return result; 497 } 498 499 500 inline static void insert(UnicodeString& dest, int pos, UChar32 ch) 501 { 502 dest.replace(pos, 0, ch); 503 } 504 505 void BasicNormalizerTest::backAndForth(Normalizer* iter, const UnicodeString& input) 506 { 507 UChar32 ch; 508 iter->setText(input, status); 509 510 // Run through the iterator forwards and stick it into a StringBuffer 511 UnicodeString forward; 512 for (ch = iter->first(); ch != iter->DONE; ch = iter->next()) { 513 forward += ch; 514 } 515 516 // Now do it backwards 517 UnicodeString reverse; 518 for (ch = iter->last(); ch != iter->DONE; ch = iter->previous()) { 519 insert(reverse, 0, ch); 520 } 521 522 if (forward != reverse) { 523 errln("Forward/reverse mismatch for input " + hex(input) 524 + ", forward: " + hex(forward) + ", backward: " + hex(reverse)); 525 } 526 } 527 528 void BasicNormalizerTest::staticTest(UNormalizationMode mode, int options, 529 UnicodeString tests[][3], int length, 530 int outCol) 531 { 532 for (int i = 0; i < length; i++) 533 { 534 UnicodeString& input = tests[i][0]; 535 UnicodeString& expect = tests[i][outCol]; 536 537 logln("Normalizing '" + input + "' (" + hex(input) + ")" ); 538 539 UnicodeString output; 540 Normalizer::normalize(input, mode, options, output, status); 541 542 if (output != expect) { 543 dataerrln(UnicodeString("ERROR: case ") + i + " normalized " + hex(input) + "\n" 544 + " expected " + hex(expect) + "\n" 545 + " static got " + hex(output) ); 546 } 547 } 548 } 549 550 void BasicNormalizerTest::iterateTest(Normalizer* iter, 551 UnicodeString tests[][3], int length, 552 int outCol) 553 { 554 for (int i = 0; i < length; i++) 555 { 556 UnicodeString& input = tests[i][0]; 557 UnicodeString& expect = tests[i][outCol]; 558 559 logln("Normalizing '" + input + "' (" + hex(input) + ")" ); 560 561 iter->setText(input, status); 562 assertEqual(input, expect, iter, UnicodeString("ERROR: case ") + i + " "); 563 } 564 } 565 566 void BasicNormalizerTest::assertEqual(const UnicodeString& input, 567 const UnicodeString& expected, 568 Normalizer* iter, 569 const UnicodeString& errPrefix) 570 { 571 UnicodeString result; 572 573 for (UChar32 ch = iter->first(); ch != iter->DONE; ch = iter->next()) { 574 result += ch; 575 } 576 if (result != expected) { 577 dataerrln(errPrefix + "normalized " + hex(input) + "\n" 578 + " expected " + hex(expected) + "\n" 579 + " iterate got " + hex(result) ); 580 } 581 } 582 583 // helper class for TestPreviousNext() 584 // simple UTF-32 character iterator 585 class UChar32Iterator { 586 public: 587 UChar32Iterator(const UChar32 *text, int32_t len, int32_t index) : 588 s(text), length(len), i(index) {} 589 590 UChar32 current() { 591 if(i<length) { 592 return s[i]; 593 } else { 594 return 0xffff; 595 } 596 } 597 598 UChar32 next() { 599 if(i<length) { 600 return s[i++]; 601 } else { 602 return 0xffff; 603 } 604 } 605 606 UChar32 previous() { 607 if(i>0) { 608 return s[--i]; 609 } else { 610 return 0xffff; 611 } 612 } 613 614 int32_t getIndex() { 615 return i; 616 } 617 private: 618 const UChar32 *s; 619 int32_t length, i; 620 }; 621 622 void 623 BasicNormalizerTest::TestPreviousNext(const UChar *src, int32_t srcLength, 624 const UChar32 *expect, int32_t expectLength, 625 const int32_t *expectIndex, // its length=expectLength+1 626 int32_t srcMiddle, int32_t expectMiddle, 627 const char *moves, 628 UNormalizationMode mode, 629 const char *name) { 630 // iterators 631 Normalizer iter(src, srcLength, mode); 632 633 // test getStaticClassID and getDynamicClassID 634 if(iter.getDynamicClassID() != Normalizer::getStaticClassID()) { 635 errln("getStaticClassID != getDynamicClassID for Normalizer."); 636 } 637 638 UChar32Iterator iter32(expect, expectLength, expectMiddle); 639 640 UChar32 c1, c2; 641 char m; 642 643 // initially set the indexes into the middle of the strings 644 iter.setIndexOnly(srcMiddle); 645 646 // move around and compare the iteration code points with 647 // the expected ones 648 const char *move=moves; 649 while((m=*move++)!=0) { 650 if(m=='-') { 651 c1=iter.previous(); 652 c2=iter32.previous(); 653 } else if(m=='0') { 654 c1=iter.current(); 655 c2=iter32.current(); 656 } else /* m=='+' */ { 657 c1=iter.next(); 658 c2=iter32.next(); 659 } 660 661 // compare results 662 if(c1!=c2) { 663 // copy the moves until the current (m) move, and terminate 664 char history[64]; 665 uprv_strcpy(history, moves); 666 history[move-moves]=0; 667 dataerrln("error: mismatch in Normalizer iteration (%s) at %s: " 668 "got c1=U+%04lx != expected c2=U+%04lx", 669 name, history, c1, c2); 670 break; 671 } 672 673 // compare indexes 674 if(iter.getIndex()!=expectIndex[iter32.getIndex()]) { 675 // copy the moves until the current (m) move, and terminate 676 char history[64]; 677 uprv_strcpy(history, moves); 678 history[move-moves]=0; 679 errln("error: index mismatch in Normalizer iteration (%s) at %s: " 680 "Normalizer index %ld expected %ld\n", 681 name, history, iter.getIndex(), expectIndex[iter32.getIndex()]); 682 break; 683 } 684 } 685 } 686 687 void 688 BasicNormalizerTest::TestPreviousNext() { 689 // src and expect strings 690 static const UChar src[]={ 691 UTF16_LEAD(0x2f999), UTF16_TRAIL(0x2f999), 692 UTF16_LEAD(0x1d15f), UTF16_TRAIL(0x1d15f), 693 0xc4, 694 0x1ed0 695 }; 696 static const UChar32 expect[]={ 697 0x831d, 698 0x1d158, 0x1d165, 699 0x41, 0x308, 700 0x4f, 0x302, 0x301 701 }; 702 703 // expected src indexes corresponding to expect indexes 704 static const int32_t expectIndex[]={ 705 0, 706 2, 2, 707 4, 4, 708 5, 5, 5, 709 6 // behind last character 710 }; 711 712 // src and expect strings for regression test for j2911 713 static const UChar src_j2911[]={ 714 UTF16_LEAD(0x2f999), UTF16_TRAIL(0x2f999), 715 0xdd00, 0xd900, // unpaired surrogates - regression test for j2911 716 0xc4, 717 0x4f, 0x302, 0x301 718 }; 719 static const UChar32 expect_j2911[]={ 720 0x831d, 721 0xdd00, 0xd900, // unpaired surrogates - regression test for j2911 722 0xc4, 723 0x1ed0 724 }; 725 726 // expected src indexes corresponding to expect indexes 727 static const int32_t expectIndex_j2911[]={ 728 0, 729 2, 3, 730 4, 731 5, 732 8 // behind last character 733 }; 734 735 // initial indexes into the src and expect strings 736 // for both sets of test data 737 enum { 738 SRC_MIDDLE=4, 739 EXPECT_MIDDLE=3, 740 SRC_MIDDLE_2=2, 741 EXPECT_MIDDLE_2=1 742 }; 743 744 // movement vector 745 // - for previous(), 0 for current(), + for next() 746 // for both sets of test data 747 static const char *const moves="0+0+0--0-0-+++0--+++++++0--------"; 748 749 TestPreviousNext(src, LENGTHOF(src), 750 expect, LENGTHOF(expect), 751 expectIndex, 752 SRC_MIDDLE, EXPECT_MIDDLE, 753 moves, UNORM_NFD, "basic"); 754 755 TestPreviousNext(src_j2911, LENGTHOF(src_j2911), 756 expect_j2911, LENGTHOF(expect_j2911), 757 expectIndex_j2911, 758 SRC_MIDDLE, EXPECT_MIDDLE, 759 moves, UNORM_NFKC, "j2911"); 760 761 // try again from different "middle" indexes 762 TestPreviousNext(src, LENGTHOF(src), 763 expect, LENGTHOF(expect), 764 expectIndex, 765 SRC_MIDDLE_2, EXPECT_MIDDLE_2, 766 moves, UNORM_NFD, "basic_2"); 767 768 TestPreviousNext(src_j2911, LENGTHOF(src_j2911), 769 expect_j2911, LENGTHOF(expect_j2911), 770 expectIndex_j2911, 771 SRC_MIDDLE_2, EXPECT_MIDDLE_2, 772 moves, UNORM_NFKC, "j2911_2"); 773 } 774 775 void BasicNormalizerTest::TestConcatenate() { 776 static const char *const 777 cases[][4]={ 778 /* mode, left, right, result */ 779 { 780 "C", 781 "re", 782 "\\u0301sum\\u00e9", 783 "r\\u00e9sum\\u00e9" 784 }, 785 { 786 "C", 787 "a\\u1100", 788 "\\u1161bcdefghijk", 789 "a\\uac00bcdefghijk" 790 }, 791 /* ### TODO: add more interesting cases */ 792 { 793 "D", 794 "\\u03B1\\u0345", 795 "\\u0C4D\\U000110BA\\U0001D169", 796 "\\u03B1\\U0001D169\\U000110BA\\u0C4D\\u0345" 797 } 798 }; 799 800 UnicodeString left, right, expect, result, r; 801 UErrorCode errorCode; 802 UNormalizationMode mode; 803 int32_t i; 804 805 /* test concatenation */ 806 for(i=0; i<(int32_t)(sizeof(cases)/sizeof(cases[0])); ++i) { 807 switch(*cases[i][0]) { 808 case 'C': mode=UNORM_NFC; break; 809 case 'D': mode=UNORM_NFD; break; 810 case 'c': mode=UNORM_NFKC; break; 811 case 'd': mode=UNORM_NFKD; break; 812 default: mode=UNORM_NONE; break; 813 } 814 815 left=UnicodeString(cases[i][1], "").unescape(); 816 right=UnicodeString(cases[i][2], "").unescape(); 817 expect=UnicodeString(cases[i][3], "").unescape(); 818 819 //result=r=UnicodeString(); 820 errorCode=U_ZERO_ERROR; 821 822 r=Normalizer::concatenate(left, right, result, mode, 0, errorCode); 823 if(U_FAILURE(errorCode) || /*result!=r ||*/ result!=expect) { 824 dataerrln("error in Normalizer::concatenate(), cases[] fails with "+ 825 UnicodeString(u_errorName(errorCode))+", result==expect: expected: "+ 826 hex(expect)+" =========> got: " + hex(result)); 827 } 828 } 829 830 /* test error cases */ 831 832 /* left.getBuffer()==result.getBuffer() */ 833 result=r=expect=UnicodeString("zz", ""); 834 errorCode=U_UNEXPECTED_TOKEN; 835 r=Normalizer::concatenate(left, right, result, mode, 0, errorCode); 836 if(errorCode!=U_UNEXPECTED_TOKEN || result!=r || !result.isBogus()) { 837 errln("error in Normalizer::concatenate(), violates UErrorCode protocol"); 838 } 839 840 left.setToBogus(); 841 errorCode=U_ZERO_ERROR; 842 r=Normalizer::concatenate(left, right, result, mode, 0, errorCode); 843 if(errorCode!=U_ILLEGAL_ARGUMENT_ERROR || result!=r || !result.isBogus()) { 844 errln("error in Normalizer::concatenate(), does not detect left.isBogus()"); 845 } 846 } 847 848 // reference implementation of Normalizer::compare 849 static int32_t 850 ref_norm_compare(const UnicodeString &s1, const UnicodeString &s2, uint32_t options, UErrorCode &errorCode) { 851 UnicodeString r1, r2, t1, t2; 852 int32_t normOptions=(int32_t)(options>>UNORM_COMPARE_NORM_OPTIONS_SHIFT); 853 854 if(options&U_COMPARE_IGNORE_CASE) { 855 Normalizer::decompose(s1, FALSE, normOptions, r1, errorCode); 856 Normalizer::decompose(s2, FALSE, normOptions, r2, errorCode); 857 858 r1.foldCase(options); 859 r2.foldCase(options); 860 } else { 861 r1=s1; 862 r2=s2; 863 } 864 865 Normalizer::decompose(r1, FALSE, normOptions, t1, errorCode); 866 Normalizer::decompose(r2, FALSE, normOptions, t2, errorCode); 867 868 if(options&U_COMPARE_CODE_POINT_ORDER) { 869 return t1.compareCodePointOrder(t2); 870 } else { 871 return t1.compare(t2); 872 } 873 } 874 875 // test wrapper for Normalizer::compare, sets UNORM_INPUT_IS_FCD appropriately 876 static int32_t 877 _norm_compare(const UnicodeString &s1, const UnicodeString &s2, uint32_t options, UErrorCode &errorCode) { 878 int32_t normOptions=(int32_t)(options>>UNORM_COMPARE_NORM_OPTIONS_SHIFT); 879 880 if( UNORM_YES==Normalizer::quickCheck(s1, UNORM_FCD, normOptions, errorCode) && 881 UNORM_YES==Normalizer::quickCheck(s2, UNORM_FCD, normOptions, errorCode)) { 882 options|=UNORM_INPUT_IS_FCD; 883 } 884 885 return Normalizer::compare(s1, s2, options, errorCode); 886 } 887 888 // reference implementation of UnicodeString::caseCompare 889 static int32_t 890 ref_case_compare(const UnicodeString &s1, const UnicodeString &s2, uint32_t options) { 891 UnicodeString t1, t2; 892 893 t1=s1; 894 t2=s2; 895 896 t1.foldCase(options); 897 t2.foldCase(options); 898 899 if(options&U_COMPARE_CODE_POINT_ORDER) { 900 return t1.compareCodePointOrder(t2); 901 } else { 902 return t1.compare(t2); 903 } 904 } 905 906 // reduce an integer to -1/0/1 907 static inline int32_t 908 _sign(int32_t value) { 909 if(value==0) { 910 return 0; 911 } else { 912 return (value>>31)|1; 913 } 914 } 915 916 static const char * 917 _signString(int32_t value) { 918 if(value<0) { 919 return "<0"; 920 } else if(value==0) { 921 return "=0"; 922 } else /* value>0 */ { 923 return ">0"; 924 } 925 } 926 927 void 928 BasicNormalizerTest::TestCompare() { 929 // test Normalizer::compare and unorm_compare (thinly wrapped by the former) 930 // by comparing it with its semantic equivalent 931 // since we trust the pieces, this is sufficient 932 933 // test each string with itself and each other 934 // each time with all options 935 static const char *const 936 strings[]={ 937 // some cases from NormalizationTest.txt 938 // 0..3 939 "D\\u031B\\u0307\\u0323", 940 "\\u1E0C\\u031B\\u0307", 941 "D\\u031B\\u0323\\u0307", 942 "d\\u031B\\u0323\\u0307", 943 944 // 4..6 945 "\\u00E4", 946 "a\\u0308", 947 "A\\u0308", 948 949 // Angstrom sign = A ring 950 // 7..10 951 "\\u212B", 952 "\\u00C5", 953 "A\\u030A", 954 "a\\u030A", 955 956 // 11.14 957 "a\\u059A\\u0316\\u302A\\u032Fb", 958 "a\\u302A\\u0316\\u032F\\u059Ab", 959 "a\\u302A\\u0316\\u032F\\u059Ab", 960 "A\\u059A\\u0316\\u302A\\u032Fb", 961 962 // from ICU case folding tests 963 // 15..20 964 "A\\u00df\\u00b5\\ufb03\\U0001040c\\u0131", 965 "ass\\u03bcffi\\U00010434i", 966 "\\u0061\\u0042\\u0131\\u03a3\\u00df\\ufb03\\ud93f\\udfff", 967 "\\u0041\\u0062\\u0069\\u03c3\\u0073\\u0053\\u0046\\u0066\\u0049\\ud93f\\udfff", 968 "\\u0041\\u0062\\u0131\\u03c3\\u0053\\u0073\\u0066\\u0046\\u0069\\ud93f\\udfff", 969 "\\u0041\\u0062\\u0069\\u03c3\\u0073\\u0053\\u0046\\u0066\\u0049\\ud93f\\udffd", 970 971 // U+d800 U+10001 see implementation comment in unorm_cmpEquivFold 972 // vs. U+10000 at bottom - code point order 973 // 21..22 974 "\\ud800\\ud800\\udc01", 975 "\\ud800\\udc00", 976 977 // other code point order tests from ustrtest.cpp 978 // 23..31 979 "\\u20ac\\ud801", 980 "\\u20ac\\ud800\\udc00", 981 "\\ud800", 982 "\\ud800\\uff61", 983 "\\udfff", 984 "\\uff61\\udfff", 985 "\\uff61\\ud800\\udc02", 986 "\\ud800\\udc02", 987 "\\ud84d\\udc56", 988 989 // long strings, see cnormtst.c/TestNormCoverage() 990 // equivalent if case-insensitive 991 // 32..33 992 "\\uAD8B\\uAD8B\\uAD8B\\uAD8B" 993 "\\U0001d15e\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 994 "\\U0001d15e\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 995 "\\U0001d15e\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 996 "\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 997 "\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 998 "aaaaaaaaaaaaaaaaaazzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz" 999 "bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb" 1000 "ccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc" 1001 "ddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddd" 1002 "\\uAD8B\\uAD8B\\uAD8B\\uAD8B" 1003 "d\\u031B\\u0307\\u0323", 1004 1005 "\\u1100\\u116f\\u11aa\\uAD8B\\uAD8B\\u1100\\u116f\\u11aa" 1006 "\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 1007 "\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 1008 "\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 1009 "\\U0001d15e\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 1010 "\\U0001d15e\\U0001d157\\U0001d165\\U0001d15e\\U0001d15e\\U0001d15e\\U0001d15e" 1011 "aaaaaaaaaaAAAAAAAAZZZZZZZZZZZZZZZZzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz" 1012 "bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb" 1013 "ccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc" 1014 "ddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddd" 1015 "\\u1100\\u116f\\u11aa\\uAD8B\\uAD8B\\u1100\\u116f\\u11aa" 1016 "\\u1E0C\\u031B\\u0307", 1017 1018 // some strings that may make a difference whether the compare function 1019 // case-folds or decomposes first 1020 // 34..41 1021 "\\u0360\\u0345\\u0334", 1022 "\\u0360\\u03b9\\u0334", 1023 1024 "\\u0360\\u1f80\\u0334", 1025 "\\u0360\\u03b1\\u0313\\u03b9\\u0334", 1026 1027 "\\u0360\\u1ffc\\u0334", 1028 "\\u0360\\u03c9\\u03b9\\u0334", 1029 1030 "a\\u0360\\u0345\\u0360\\u0345b", 1031 "a\\u0345\\u0360\\u0345\\u0360b", 1032 1033 // interesting cases for canonical caseless match with turkic i handling 1034 // 42..43 1035 "\\u00cc", 1036 "\\u0069\\u0300", 1037 1038 // strings with post-Unicode 3.2 normalization or normalization corrections 1039 // 44..45 1040 "\\u00e4\\u193b\\U0002f868", 1041 "\\u0061\\u193b\\u0308\\u36fc", 1042 1043 // empty string 1044 // 46 1045 "" 1046 }; 1047 1048 UnicodeString s[100]; // at least as many items as in strings[] ! 1049 1050 // all combinations of options 1051 // UNORM_INPUT_IS_FCD is set automatically if both input strings fulfill FCD conditions 1052 // set UNORM_UNICODE_3_2 in one additional combination 1053 static const struct { 1054 uint32_t options; 1055 const char *name; 1056 } opt[]={ 1057 { 0, "default" }, 1058 { U_COMPARE_CODE_POINT_ORDER, "c.p. order" }, 1059 { U_COMPARE_IGNORE_CASE, "ignore case" }, 1060 { U_COMPARE_CODE_POINT_ORDER|U_COMPARE_IGNORE_CASE, "c.p. order & ignore case" }, 1061 { U_COMPARE_IGNORE_CASE|U_FOLD_CASE_EXCLUDE_SPECIAL_I, "ignore case & special i" }, 1062 { U_COMPARE_CODE_POINT_ORDER|U_COMPARE_IGNORE_CASE|U_FOLD_CASE_EXCLUDE_SPECIAL_I, "c.p. order & ignore case & special i" }, 1063 { UNORM_UNICODE_3_2<<UNORM_COMPARE_NORM_OPTIONS_SHIFT, "Unicode 3.2" } 1064 }; 1065 1066 int32_t i, j, k, count=LENGTHOF(strings); 1067 int32_t result, refResult; 1068 1069 UErrorCode errorCode; 1070 1071 // create the UnicodeStrings 1072 for(i=0; i<count; ++i) { 1073 s[i]=UnicodeString(strings[i], "").unescape(); 1074 } 1075 1076 // test them each with each other 1077 for(i=0; i<count; ++i) { 1078 for(j=i; j<count; ++j) { 1079 for(k=0; k<LENGTHOF(opt); ++k) { 1080 // test Normalizer::compare 1081 errorCode=U_ZERO_ERROR; 1082 result=_norm_compare(s[i], s[j], opt[k].options, errorCode); 1083 refResult=ref_norm_compare(s[i], s[j], opt[k].options, errorCode); 1084 if(_sign(result)!=_sign(refResult)) { 1085 errln("Normalizer::compare(%d, %d, %s)%s should be %s %s", 1086 i, j, opt[k].name, _signString(result), _signString(refResult), 1087 U_SUCCESS(errorCode) ? "" : u_errorName(errorCode)); 1088 } 1089 1090 // test UnicodeString::caseCompare - same internal implementation function 1091 if(opt[k].options&U_COMPARE_IGNORE_CASE) { 1092 errorCode=U_ZERO_ERROR; 1093 result=s[i].caseCompare(s[j], opt[k].options); 1094 refResult=ref_case_compare(s[i], s[j], opt[k].options); 1095 if(_sign(result)!=_sign(refResult)) { 1096 errln("UniStr::caseCompare(%d, %d, %s)%s should be %s %s", 1097 i, j, opt[k].name, _signString(result), _signString(refResult), 1098 U_SUCCESS(errorCode) ? "" : u_errorName(errorCode)); 1099 } 1100 } 1101 } 1102 } 1103 } 1104 1105 // test cases with i and I to make sure Turkic works 1106 static const UChar iI[]={ 0x49, 0x69, 0x130, 0x131 }; 1107 UnicodeSet iSet, set; 1108 1109 UnicodeString s1, s2; 1110 1111 const Normalizer2Impl *nfcImpl=Normalizer2Factory::getNFCImpl(errorCode); 1112 if(U_FAILURE(errorCode) || !nfcImpl->ensureCanonIterData(errorCode)) { 1113 dataerrln("Normalizer2Factory::getNFCImpl().ensureCanonIterData() failed: %s", 1114 u_errorName(errorCode)); 1115 return; 1116 } 1117 1118 // collect all sets into one for contiguous output 1119 for(i=0; i<LENGTHOF(iI); ++i) { 1120 if(nfcImpl->getCanonStartSet(iI[i], iSet)) { 1121 set.addAll(iSet); 1122 } 1123 } 1124 1125 // test all of these precomposed characters 1126 const Normalizer2 *nfcNorm2=Normalizer2Factory::getNFCInstance(errorCode); 1127 UnicodeSetIterator it(set); 1128 while(it.next() && !it.isString()) { 1129 UChar32 c=it.getCodepoint(); 1130 if(!nfcNorm2->getDecomposition(c, s2)) { 1131 dataerrln("NFC.getDecomposition(i-composite U+%04lx) failed", (long)c); 1132 return; 1133 } 1134 1135 s1.setTo(c); 1136 for(k=0; k<LENGTHOF(opt); ++k) { 1137 // test Normalizer::compare 1138 errorCode=U_ZERO_ERROR; 1139 result=_norm_compare(s1, s2, opt[k].options, errorCode); 1140 refResult=ref_norm_compare(s1, s2, opt[k].options, errorCode); 1141 if(_sign(result)!=_sign(refResult)) { 1142 errln("Normalizer::compare(U+%04x with its NFD, %s)%s should be %s %s", 1143 c, opt[k].name, _signString(result), _signString(refResult), 1144 U_SUCCESS(errorCode) ? "" : u_errorName(errorCode)); 1145 } 1146 1147 // test UnicodeString::caseCompare - same internal implementation function 1148 if(opt[k].options&U_COMPARE_IGNORE_CASE) { 1149 errorCode=U_ZERO_ERROR; 1150 result=s1.caseCompare(s2, opt[k].options); 1151 refResult=ref_case_compare(s1, s2, opt[k].options); 1152 if(_sign(result)!=_sign(refResult)) { 1153 errln("UniStr::caseCompare(U+%04x with its NFD, %s)%s should be %s %s", 1154 c, opt[k].name, _signString(result), _signString(refResult), 1155 U_SUCCESS(errorCode) ? "" : u_errorName(errorCode)); 1156 } 1157 } 1158 } 1159 } 1160 1161 // test getDecomposition() for some characters that do not decompose 1162 if( nfcNorm2->getDecomposition(0x20, s2) || 1163 nfcNorm2->getDecomposition(0x4e00, s2) || 1164 nfcNorm2->getDecomposition(0x20002, s2) 1165 ) { 1166 errln("NFC.getDecomposition() returns TRUE for characters which do not have decompositions"); 1167 } 1168 1169 // test FilteredNormalizer2::getDecomposition() 1170 UnicodeSet filter(UNICODE_STRING_SIMPLE("[^\\u00a0-\\u00ff]"), errorCode); 1171 FilteredNormalizer2 fn2(*nfcNorm2, filter); 1172 if( fn2.getDecomposition(0xe4, s1) || !fn2.getDecomposition(0x100, s2) || 1173 s2.length()!=2 || s2[0]!=0x41 || s2[1]!=0x304 1174 ) { 1175 errln("FilteredNormalizer2(NFC, ^A0-FF).getDecomposition() failed"); 1176 } 1177 } 1178 1179 // verify that case-folding does not un-FCD strings 1180 int32_t 1181 BasicNormalizerTest::countFoldFCDExceptions(uint32_t foldingOptions) { 1182 UnicodeString s, fold, d; 1183 UChar32 c; 1184 int32_t count; 1185 uint8_t cc, trailCC, foldCC, foldTrailCC; 1186 UNormalizationCheckResult qcResult; 1187 int8_t category; 1188 UBool isNFD; 1189 UErrorCode errorCode; 1190 1191 logln("Test if case folding may un-FCD a string (folding options %04lx)", foldingOptions); 1192 1193 count=0; 1194 for(c=0; c<=0x10ffff; ++c) { 1195 errorCode = U_ZERO_ERROR; 1196 category=u_charType(c); 1197 if(category==U_UNASSIGNED) { 1198 continue; // skip unassigned code points 1199 } 1200 if(c==0xac00) { 1201 c=0xd7a3; // skip Hangul - no case folding there 1202 continue; 1203 } 1204 // skip Han blocks - no case folding there either 1205 if(c==0x3400) { 1206 c=0x4db5; 1207 continue; 1208 } 1209 if(c==0x4e00) { 1210 c=0x9fa5; 1211 continue; 1212 } 1213 if(c==0x20000) { 1214 c=0x2a6d6; 1215 continue; 1216 } 1217 1218 s.setTo(c); 1219 1220 // get leading and trailing cc for c 1221 Normalizer::decompose(s, FALSE, 0, d, errorCode); 1222 isNFD= s==d; 1223 cc=u_getCombiningClass(d.char32At(0)); 1224 trailCC=u_getCombiningClass(d.char32At(d.length()-1)); 1225 1226 // get leading and trailing cc for the case-folding of c 1227 s.foldCase(foldingOptions); 1228 Normalizer::decompose(s, FALSE, 0, d, errorCode); 1229 foldCC=u_getCombiningClass(d.char32At(0)); 1230 foldTrailCC=u_getCombiningClass(d.char32At(d.length()-1)); 1231 1232 qcResult=Normalizer::quickCheck(s, UNORM_FCD, errorCode); 1233 1234 if (U_FAILURE(errorCode)) { 1235 ++count; 1236 dataerrln("U+%04lx: Failed with error %s", u_errorName(errorCode)); 1237 } 1238 1239 // bad: 1240 // - character maps to empty string: adjacent characters may then need reordering 1241 // - folding has different leading/trailing cc's, and they don't become just 0 1242 // - folding itself is not FCD 1243 if( qcResult!=UNORM_YES || 1244 s.isEmpty() || 1245 (cc!=foldCC && foldCC!=0) || (trailCC!=foldTrailCC && foldTrailCC!=0) 1246 ) { 1247 ++count; 1248 dataerrln("U+%04lx: case-folding may un-FCD a string (folding options %04lx)", c, foldingOptions); 1249 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); 1250 continue; 1251 } 1252 1253 // also bad: 1254 // if a code point is in NFD but its case folding is not, then 1255 // unorm_compare will also fail 1256 if(isNFD && UNORM_YES!=Normalizer::quickCheck(s, UNORM_NFD, errorCode)) { 1257 ++count; 1258 errln("U+%04lx: case-folding un-NFDs this character (folding options %04lx)", c, foldingOptions); 1259 } 1260 } 1261 1262 logln("There are %ld code points for which case-folding may un-FCD a string (folding options %04lx)", count, foldingOptions); 1263 return count; 1264 } 1265 1266 void 1267 BasicNormalizerTest::FindFoldFCDExceptions() { 1268 int32_t count; 1269 1270 count=countFoldFCDExceptions(0); 1271 count+=countFoldFCDExceptions(U_FOLD_CASE_EXCLUDE_SPECIAL_I); 1272 if(count>0) { 1273 /* 1274 * If case-folding un-FCDs any strings, then unorm_compare() must be 1275 * re-implemented. 1276 * It currently assumes that one can check for FCD then case-fold 1277 * and then still have FCD strings for raw decomposition without reordering. 1278 */ 1279 dataerrln("error: There are %ld code points for which case-folding may un-FCD a string for all folding options.\n" 1280 "See comment in BasicNormalizerTest::FindFoldFCDExceptions()!", count); 1281 } 1282 } 1283 1284 static void 1285 initExpectedSkippables(UnicodeSet skipSets[UNORM_MODE_COUNT], UErrorCode &errorCode) { 1286 skipSets[UNORM_NFD].applyPattern( 1287 UNICODE_STRING_SIMPLE("[[:NFD_QC=Yes:]&[:ccc=0:]]"), errorCode); 1288 skipSets[UNORM_NFC].applyPattern( 1289 UNICODE_STRING_SIMPLE("[[:NFC_QC=Yes:]&[:ccc=0:]-[:HST=LV:]]"), errorCode); 1290 skipSets[UNORM_NFKD].applyPattern( 1291 UNICODE_STRING_SIMPLE("[[:NFKD_QC=Yes:]&[:ccc=0:]]"), errorCode); 1292 skipSets[UNORM_NFKC].applyPattern( 1293 UNICODE_STRING_SIMPLE("[[:NFKC_QC=Yes:]&[:ccc=0:]-[:HST=LV:]]"), errorCode); 1294 1295 // Remove from the NFC and NFKC sets all those characters that change 1296 // when a back-combining character is added. 1297 // First, get all of the back-combining characters and their combining classes. 1298 UnicodeSet combineBack("[:NFC_QC=Maybe:]", errorCode); 1299 int32_t numCombineBack=combineBack.size(); 1300 int32_t *combineBackCharsAndCc=new int32_t[numCombineBack*2]; 1301 UnicodeSetIterator iter(combineBack); 1302 for(int32_t i=0; i<numCombineBack; ++i) { 1303 iter.next(); 1304 UChar32 c=iter.getCodepoint(); 1305 combineBackCharsAndCc[2*i]=c; 1306 combineBackCharsAndCc[2*i+1]=u_getCombiningClass(c); 1307 } 1308 1309 // We need not look at control codes, Han characters nor Hangul LVT syllables because they 1310 // do not combine forward. LV syllables are already removed. 1311 UnicodeSet notInteresting("[[:C:][:Unified_Ideograph:][:HST=LVT:]]", errorCode); 1312 LocalPointer<UnicodeSet> unsure(&((UnicodeSet *)(skipSets[UNORM_NFC].clone()))->removeAll(notInteresting)); 1313 // System.out.format("unsure.size()=%d\n", unsure.size()); 1314 1315 // For each character about which we are unsure, see if it changes when we add 1316 // one of the back-combining characters. 1317 const Normalizer2 *norm2=Normalizer2::getInstance(NULL, "nfc", UNORM2_COMPOSE, errorCode); 1318 UnicodeString s; 1319 iter.reset(*unsure); 1320 while(iter.next()) { 1321 UChar32 c=iter.getCodepoint(); 1322 s.setTo(c); 1323 int32_t cLength=s.length(); 1324 int32_t tccc=u_getIntPropertyValue(c, UCHAR_TRAIL_CANONICAL_COMBINING_CLASS); 1325 for(int32_t i=0; i<numCombineBack; ++i) { 1326 // If c's decomposition ends with a character with non-zero combining class, then 1327 // c can only change if it combines with a character with a non-zero combining class. 1328 int32_t cc2=combineBackCharsAndCc[2*i+1]; 1329 if(tccc==0 || cc2!=0) { 1330 UChar32 c2=combineBackCharsAndCc[2*i]; 1331 s.append(c2); 1332 if(!norm2->isNormalized(s, errorCode)) { 1333 // System.out.format("remove U+%04x (tccc=%d) + U+%04x (cc=%d)\n", c, tccc, c2, cc2); 1334 skipSets[UNORM_NFC].remove(c); 1335 skipSets[UNORM_NFKC].remove(c); 1336 break; 1337 } 1338 s.truncate(cLength); 1339 } 1340 } 1341 } 1342 delete [] combineBackCharsAndCc; 1343 } 1344 1345 void 1346 BasicNormalizerTest::TestSkippable() { 1347 UnicodeSet diff, skipSets[UNORM_MODE_COUNT], expectSets[UNORM_MODE_COUNT]; 1348 UnicodeString s, pattern; 1349 1350 /* build NF*Skippable sets from runtime data */ 1351 IcuTestErrorCode errorCode(*this, "TestSkippable"); 1352 skipSets[UNORM_NFD].applyPattern(UNICODE_STRING_SIMPLE("[:NFD_Inert:]"), errorCode); 1353 skipSets[UNORM_NFKD].applyPattern(UNICODE_STRING_SIMPLE("[:NFKD_Inert:]"), errorCode); 1354 skipSets[UNORM_NFC].applyPattern(UNICODE_STRING_SIMPLE("[:NFC_Inert:]"), errorCode); 1355 skipSets[UNORM_NFKC].applyPattern(UNICODE_STRING_SIMPLE("[:NFKC_Inert:]"), errorCode); 1356 if(errorCode.logDataIfFailureAndReset("UnicodeSet(NF..._Inert) failed")) { 1357 return; 1358 } 1359 1360 /* get expected sets from hardcoded patterns */ 1361 initExpectedSkippables(expectSets, errorCode); 1362 errorCode.assertSuccess(); 1363 1364 for(int32_t i=UNORM_NONE; i<UNORM_MODE_COUNT; ++i) { 1365 if(skipSets[i]!=expectSets[i]) { 1366 errln("error: TestSkippable skipSets[%d]!=expectedSets[%d]\n", i, i); 1367 // Note: This used to depend on hardcoded UnicodeSet patterns generated by 1368 // Mark's unicodetools.com.ibm.text.UCD.NFSkippable, by 1369 // running com.ibm.text.UCD.Main with the option NFSkippable. 1370 // Since ICU 4.6/Unicode 6, we are generating the 1371 // expectSets ourselves in initSkippables(). 1372 1373 s=UNICODE_STRING_SIMPLE("skip-expect="); 1374 (diff=skipSets[i]).removeAll(expectSets[i]).toPattern(pattern, TRUE); 1375 s.append(pattern); 1376 1377 pattern.remove(); 1378 s.append(UNICODE_STRING_SIMPLE("\n\nexpect-skip=")); 1379 (diff=expectSets[i]).removeAll(skipSets[i]).toPattern(pattern, TRUE); 1380 s.append(pattern); 1381 s.append(UNICODE_STRING_SIMPLE("\n\n")); 1382 1383 errln(s); 1384 } 1385 } 1386 } 1387 1388 struct StringPair { const char *input, *expected; }; 1389 1390 void 1391 BasicNormalizerTest::TestCustomComp() { 1392 static const StringPair pairs[]={ 1393 { "\\uD801\\uE000\\uDFFE", "" }, 1394 { "\\uD800\\uD801\\uE000\\uDFFE\\uDFFF", "\\uD7FF\\uFFFF" }, 1395 { "\\uD800\\uD801\\uDFFE\\uDFFF", "\\uD7FF\\U000107FE\\uFFFF" }, 1396 { "\\uE001\\U000110B9\\u0345\\u0308\\u0327", "\\uE002\\U000110B9\\u0327\\u0345" }, 1397 { "\\uE010\\U000F0011\\uE012", "\\uE011\\uE012" }, 1398 { "\\uE010\\U000F0011\\U000F0011\\uE012", "\\uE011\\U000F0010" }, 1399 { "\\uE111\\u1161\\uE112\\u1162", "\\uAE4C\\u1102\\u0062\\u1162" }, 1400 { "\\uFFF3\\uFFF7\\U00010036\\U00010077", "\\U00010037\\U00010037\\uFFF6\\U00010037" } 1401 }; 1402 IcuTestErrorCode errorCode(*this, "BasicNormalizerTest/TestCustomComp"); 1403 const Normalizer2 *customNorm2= 1404 Normalizer2::getInstance(loadTestData(errorCode), "testnorm", 1405 UNORM2_COMPOSE, errorCode); 1406 if(errorCode.logDataIfFailureAndReset("unable to load testdata/testnorm.nrm")) { 1407 return; 1408 } 1409 for(int32_t i=0; i<LENGTHOF(pairs); ++i) { 1410 const StringPair &pair=pairs[i]; 1411 UnicodeString input=UnicodeString(pair.input, -1, US_INV).unescape(); 1412 UnicodeString expected=UnicodeString(pair.expected, -1, US_INV).unescape(); 1413 UnicodeString result=customNorm2->normalize(input, errorCode); 1414 if(result!=expected) { 1415 errln("custom compose Normalizer2 did not normalize input %d as expected", i); 1416 } 1417 } 1418 } 1419 1420 void 1421 BasicNormalizerTest::TestCustomFCC() { 1422 static const StringPair pairs[]={ 1423 { "\\uD801\\uE000\\uDFFE", "" }, 1424 { "\\uD800\\uD801\\uE000\\uDFFE\\uDFFF", "\\uD7FF\\uFFFF" }, 1425 { "\\uD800\\uD801\\uDFFE\\uDFFF", "\\uD7FF\\U000107FE\\uFFFF" }, 1426 // The following expected result is different from CustomComp 1427 // because of only-contiguous composition. 1428 { "\\uE001\\U000110B9\\u0345\\u0308\\u0327", "\\uE001\\U000110B9\\u0327\\u0308\\u0345" }, 1429 { "\\uE010\\U000F0011\\uE012", "\\uE011\\uE012" }, 1430 { "\\uE010\\U000F0011\\U000F0011\\uE012", "\\uE011\\U000F0010" }, 1431 { "\\uE111\\u1161\\uE112\\u1162", "\\uAE4C\\u1102\\u0062\\u1162" }, 1432 { "\\uFFF3\\uFFF7\\U00010036\\U00010077", "\\U00010037\\U00010037\\uFFF6\\U00010037" } 1433 }; 1434 IcuTestErrorCode errorCode(*this, "BasicNormalizerTest/TestCustomFCC"); 1435 const Normalizer2 *customNorm2= 1436 Normalizer2::getInstance(loadTestData(errorCode), "testnorm", 1437 UNORM2_COMPOSE_CONTIGUOUS, errorCode); 1438 if(errorCode.logDataIfFailureAndReset("unable to load testdata/testnorm.nrm")) { 1439 return; 1440 } 1441 for(int32_t i=0; i<LENGTHOF(pairs); ++i) { 1442 const StringPair &pair=pairs[i]; 1443 UnicodeString input=UnicodeString(pair.input, -1, US_INV).unescape(); 1444 UnicodeString expected=UnicodeString(pair.expected, -1, US_INV).unescape(); 1445 UnicodeString result=customNorm2->normalize(input, errorCode); 1446 if(result!=expected) { 1447 errln("custom FCC Normalizer2 did not normalize input %d as expected", i); 1448 } 1449 } 1450 } 1451 1452 /* Improve code coverage of Normalizer2 */ 1453 void 1454 BasicNormalizerTest::TestFilteredNormalizer2Coverage() { 1455 UErrorCode errorCode = U_ZERO_ERROR; 1456 const Normalizer2 *nfcNorm2=Normalizer2Factory::getNFCInstance(errorCode); 1457 if (U_FAILURE(errorCode)) { 1458 dataerrln("Normalizer2Factory::getNFCInstance() call failed - %s", u_errorName(status)); 1459 return; 1460 } 1461 UnicodeSet filter(UNICODE_STRING_SIMPLE("[^\\u00a0-\\u00ff]"), errorCode); 1462 UnicodeString newString1 = UNICODE_STRING_SIMPLE("[^\\u0100-\\u01ff]"); 1463 UnicodeString newString2 = UNICODE_STRING_SIMPLE("[^\\u0200-\\u02ff]"); 1464 FilteredNormalizer2 fn2(*nfcNorm2, filter); 1465 1466 UChar32 char32 = 0x0054; 1467 1468 if (fn2.isInert(char32)) { 1469 errln("FilteredNormalizer2.isInert() failed."); 1470 } 1471 1472 if (fn2.hasBoundaryAfter(char32)) { 1473 errln("FilteredNormalizer2.hasBoundaryAfter() failed."); 1474 } 1475 1476 fn2.append(newString1, newString2, errorCode); 1477 if (U_FAILURE(errorCode)) { 1478 errln("FilteredNormalizer2.append() failed."); 1479 } 1480 } 1481 1482 #endif /* #if !UCONFIG_NO_NORMALIZATION */ 1483