1 // Copyright (C) 2016 and later: Unicode, Inc. and others. 2 // License & terms of use: http://www.unicode.org/copyright.html 3 /* 4 ************************************************************************ 5 * Copyright (c) 1997-2016, International Business Machines 6 * Corporation and others. All Rights Reserved. 7 ************************************************************************ 8 */ 9 10 #include "unicode/utypes.h" 11 12 #if !UCONFIG_NO_NORMALIZATION 13 14 #include "unicode/uchar.h" 15 #include "unicode/normlzr.h" 16 #include "unicode/uniset.h" 17 #include "unicode/putil.h" 18 #include "cmemory.h" 19 #include "cstring.h" 20 #include "filestrm.h" 21 #include "normconf.h" 22 #include <stdio.h> 23 24 #define CASE(id,test,exec) case id: \ 25 name = #test; \ 26 if (exec) { \ 27 logln(#test "---"); \ 28 logln((UnicodeString)""); \ 29 test(); \ 30 } \ 31 break 32 33 void NormalizerConformanceTest::runIndexedTest(int32_t index, UBool exec, const char* &name, char* /*par*/) { 34 switch (index) { 35 CASE(0, TestConformance, exec); 36 #if !UCONFIG_NO_FILE_IO && !UCONFIG_NO_LEGACY_CONVERSION 37 CASE(1, TestConformance32, exec); 38 #endif 39 // CASE(2, TestCase6); 40 default: name = ""; break; 41 } 42 } 43 44 #define FIELD_COUNT 5 45 46 NormalizerConformanceTest::NormalizerConformanceTest() : 47 normalizer(UnicodeString(), UNORM_NFC) {} 48 49 NormalizerConformanceTest::~NormalizerConformanceTest() {} 50 51 // more interesting conformance test cases, not in the unicode.org NormalizationTest.txt 52 static const char *moreCases[]={ 53 // Markus 2001aug30 54 "0061 0332 0308;00E4 0332;0061 0332 0308;00E4 0332;0061 0332 0308; # Markus 0", 55 56 // Markus 2001oct26 - test edge case for iteration: U+0f73.cc==0 but decomposition.lead.cc==129 57 "0061 0301 0F73;00E1 0F71 0F72;0061 0F71 0F72 0301;00E1 0F71 0F72;0061 0F71 0F72 0301; # Markus 1" 58 }; 59 60 void NormalizerConformanceTest::compare(const UnicodeString& s1, const UnicodeString& s2){ 61 UErrorCode status=U_ZERO_ERROR; 62 // TODO: Re-enable this tests after UTC fixes UAX 21 63 if(s1.indexOf((UChar32)0x0345)>=0)return; 64 if(Normalizer::compare(s1,s2,U_FOLD_CASE_DEFAULT,status)!=0){ 65 errln("Normalizer::compare() failed for s1: " + prettify(s1) + " s2: " +prettify(s2)); 66 } 67 } 68 69 FileStream * 70 NormalizerConformanceTest::openNormalizationTestFile(const char *filename) { 71 char unidataPath[2000]; 72 const char *folder; 73 FileStream *input; 74 UErrorCode errorCode; 75 76 // look inside ICU_DATA first 77 folder=pathToDataDirectory(); 78 if(folder!=NULL) { 79 strcpy(unidataPath, folder); 80 strcat(unidataPath, "unidata" U_FILE_SEP_STRING); 81 strcat(unidataPath, filename); 82 input=T_FileStream_open(unidataPath, "rb"); 83 if(input!=NULL) { 84 return input; 85 } 86 } 87 88 // find icu/source/data/unidata relative to the test data 89 errorCode=U_ZERO_ERROR; 90 folder=loadTestData(errorCode); 91 if(U_SUCCESS(errorCode)) { 92 strcpy(unidataPath, folder); 93 strcat(unidataPath, U_FILE_SEP_STRING ".." U_FILE_SEP_STRING ".." 94 U_FILE_SEP_STRING ".." U_FILE_SEP_STRING ".." 95 U_FILE_SEP_STRING "data" U_FILE_SEP_STRING "unidata" U_FILE_SEP_STRING); 96 strcat(unidataPath, filename); 97 input=T_FileStream_open(unidataPath, "rb"); 98 if(input!=NULL) { 99 return input; 100 } 101 } 102 103 // look in icu/source/test/testdata/out/build 104 errorCode=U_ZERO_ERROR; 105 folder=loadTestData(errorCode); 106 if(U_SUCCESS(errorCode)) { 107 strcpy(unidataPath, folder); 108 strcat(unidataPath, U_FILE_SEP_STRING); 109 strcat(unidataPath, filename); 110 input=T_FileStream_open(unidataPath, "rb"); 111 if(input!=NULL) { 112 return input; 113 } 114 } 115 116 // look in icu/source/test/testdata 117 errorCode=U_ZERO_ERROR; 118 folder=loadTestData(errorCode); 119 if(U_SUCCESS(errorCode)) { 120 strcpy(unidataPath, folder); 121 strcat(unidataPath, U_FILE_SEP_STRING ".." U_FILE_SEP_STRING ".." U_FILE_SEP_STRING); 122 strcat(unidataPath, filename); 123 input=T_FileStream_open(unidataPath, "rb"); 124 if(input!=NULL) { 125 return input; 126 } 127 } 128 129 // find icu/source/data/unidata relative to U_TOPSRCDIR 130 #if defined(U_TOPSRCDIR) 131 strcpy(unidataPath, U_TOPSRCDIR U_FILE_SEP_STRING "data" U_FILE_SEP_STRING "unidata" U_FILE_SEP_STRING); 132 strcat(unidataPath, filename); 133 input=T_FileStream_open(unidataPath, "rb"); 134 if(input!=NULL) { 135 return input; 136 } 137 138 strcpy(unidataPath, U_TOPSRCDIR U_FILE_SEP_STRING "test" U_FILE_SEP_STRING "testdata" U_FILE_SEP_STRING); 139 strcat(unidataPath, filename); 140 input=T_FileStream_open(unidataPath, "rb"); 141 if(input!=NULL) { 142 return input; 143 } 144 #endif 145 146 dataerrln("Failed to open %s", filename); 147 return NULL; 148 } 149 150 /** 151 * Test the conformance of Normalizer to 152 * http://www.unicode.org/Public/UNIDATA/NormalizationTest.txt 153 */ 154 void NormalizerConformanceTest::TestConformance() { 155 TestConformance(openNormalizationTestFile("NormalizationTest.txt"), 0); 156 } 157 158 void NormalizerConformanceTest::TestConformance32() { 159 TestConformance(openNormalizationTestFile("NormalizationTest-3.2.0.txt"), UNORM_UNICODE_3_2); 160 } 161 162 void NormalizerConformanceTest::TestConformance(FileStream *input, int32_t options) { 163 enum { BUF_SIZE = 1024 }; 164 char lineBuf[BUF_SIZE]; 165 UnicodeString fields[FIELD_COUNT]; 166 UErrorCode status = U_ZERO_ERROR; 167 int32_t passCount = 0; 168 int32_t failCount = 0; 169 UChar32 c; 170 171 if(input==NULL) { 172 return; 173 } 174 175 // UnicodeSet for all code points that are not mentioned in NormalizationTest.txt 176 UnicodeSet other(0, 0x10ffff); 177 178 int32_t count, countMoreCases = UPRV_LENGTHOF(moreCases); 179 for (count = 1;;++count) { 180 if (!T_FileStream_eof(input)) { 181 T_FileStream_readLine(input, lineBuf, (int32_t)sizeof(lineBuf)); 182 } else { 183 // once NormalizationTest.txt is finished, use moreCases[] 184 if(count > countMoreCases) { 185 count = 0; 186 } else if(count == countMoreCases) { 187 // all done 188 break; 189 } 190 uprv_strcpy(lineBuf, moreCases[count]); 191 } 192 if (lineBuf[0] == 0 || lineBuf[0] == '\n' || lineBuf[0] == '\r') continue; 193 194 // Expect 5 columns of this format: 195 // 1E0C;1E0C;0044 0323;1E0C;0044 0323; # <comments> 196 197 // Parse out the comment. 198 if (lineBuf[0] == '#') continue; 199 200 // Read separator lines starting with '@' 201 if (lineBuf[0] == '@') { 202 logln(lineBuf); 203 continue; 204 } 205 206 // Parse out the fields 207 if (!hexsplit(lineBuf, ';', fields, FIELD_COUNT)) { 208 errln((UnicodeString)"Unable to parse line " + count); 209 break; // Syntax error 210 } 211 212 // Remove a single code point from the "other" UnicodeSet 213 if(fields[0].length()==fields[0].moveIndex32(0, 1)) { 214 c=fields[0].char32At(0); 215 if(0xac20<=c && c<=0xd73f && quick) { 216 // not an exhaustive test run: skip most Hangul syllables 217 if(c==0xac20) { 218 other.remove(0xac20, 0xd73f); 219 } 220 continue; 221 } 222 other.remove(c); 223 } 224 225 if (checkConformance(fields, lineBuf, options, status)) { 226 ++passCount; 227 } else { 228 ++failCount; 229 if(status == U_FILE_ACCESS_ERROR) { 230 dataerrln("Something is wrong with the normalizer, skipping the rest of the test."); 231 break; 232 } 233 } 234 if ((count % 1000) == 0) { 235 logln("Line %d", count); 236 } 237 } 238 239 T_FileStream_close(input); 240 241 /* 242 * Test that all characters that are not mentioned 243 * as single code points in column 1 244 * do not change under any normalization. 245 */ 246 247 // remove U+ffff because that is the end-of-iteration sentinel value 248 other.remove(0xffff); 249 250 for(c=0; c<=0x10ffff; quick ? c+=113 : ++c) { 251 if(0x30000<=c && c<0xe0000) { 252 c=0xe0000; 253 } 254 if(!other.contains(c)) { 255 continue; 256 } 257 258 fields[0]=fields[1]=fields[2]=fields[3]=fields[4].setTo(c); 259 sprintf(lineBuf, "not mentioned code point U+%04lx", (long)c); 260 261 if (checkConformance(fields, lineBuf, options, status)) { 262 ++passCount; 263 } else { 264 ++failCount; 265 if(status == U_FILE_ACCESS_ERROR) { 266 dataerrln("Something is wrong with the normalizer, skipping the rest of the test.: %s", u_errorName(status)); 267 break; 268 } 269 } 270 if ((c % 0x1000) == 0) { 271 logln("Code point U+%04lx", c); 272 } 273 } 274 275 if (failCount != 0) { 276 dataerrln((UnicodeString)"Total: " + failCount + " lines/code points failed, " + 277 passCount + " lines/code points passed"); 278 } else { 279 logln((UnicodeString)"Total: " + passCount + " lines/code points passed"); 280 } 281 } 282 283 /** 284 * Verify the conformance of the given line of the Unicode 285 * normalization (UTR 15) test suite file. For each line, 286 * there are five columns, corresponding to field[0]..field[4]. 287 * 288 * The following invariants must be true for all conformant implementations 289 * c2 == NFC(c1) == NFC(c2) == NFC(c3) 290 * c3 == NFD(c1) == NFD(c2) == NFD(c3) 291 * c4 == NFKC(c1) == NFKC(c2) == NFKC(c3) == NFKC(c4) == NFKC(c5) 292 * c5 == NFKD(c1) == NFKD(c2) == NFKD(c3) == NFKD(c4) == NFKD(c5) 293 * 294 * @param field the 5 columns 295 * @param line the source line from the test suite file 296 * @return true if the test passes 297 */ 298 UBool NormalizerConformanceTest::checkConformance(const UnicodeString* field, 299 const char *line, 300 int32_t options, 301 UErrorCode &status) { 302 UBool pass = TRUE, result; 303 //UErrorCode status = U_ZERO_ERROR; 304 UnicodeString out, fcd; 305 int32_t fieldNum; 306 307 for (int32_t i=0; i<FIELD_COUNT; ++i) { 308 fieldNum = i+1; 309 if (i<3) { 310 Normalizer::normalize(field[i], UNORM_NFC, options, out, status); 311 if (U_FAILURE(status)) { 312 dataerrln("Error running normalize UNORM_NFC: %s", u_errorName(status)); 313 } else { 314 pass &= assertEqual("C", field[i], out, field[1], "c2!=C(c", fieldNum); 315 iterativeNorm(field[i], UNORM_NFC, options, out, +1); 316 pass &= assertEqual("C(+1)", field[i], out, field[1], "c2!=C(c", fieldNum); 317 iterativeNorm(field[i], UNORM_NFC, options, out, -1); 318 pass &= assertEqual("C(-1)", field[i], out, field[1], "c2!=C(c", fieldNum); 319 } 320 321 Normalizer::normalize(field[i], UNORM_NFD, options, out, status); 322 if (U_FAILURE(status)) { 323 dataerrln("Error running normalize UNORM_NFD: %s", u_errorName(status)); 324 } else { 325 pass &= assertEqual("D", field[i], out, field[2], "c3!=D(c", fieldNum); 326 iterativeNorm(field[i], UNORM_NFD, options, out, +1); 327 pass &= assertEqual("D(+1)", field[i], out, field[2], "c3!=D(c", fieldNum); 328 iterativeNorm(field[i], UNORM_NFD, options, out, -1); 329 pass &= assertEqual("D(-1)", field[i], out, field[2], "c3!=D(c", fieldNum); 330 } 331 } 332 Normalizer::normalize(field[i], UNORM_NFKC, options, out, status); 333 if (U_FAILURE(status)) { 334 dataerrln("Error running normalize UNORM_NFKC: %s", u_errorName(status)); 335 } else { 336 pass &= assertEqual("KC", field[i], out, field[3], "c4!=KC(c", fieldNum); 337 iterativeNorm(field[i], UNORM_NFKC, options, out, +1); 338 pass &= assertEqual("KC(+1)", field[i], out, field[3], "c4!=KC(c", fieldNum); 339 iterativeNorm(field[i], UNORM_NFKC, options, out, -1); 340 pass &= assertEqual("KC(-1)", field[i], out, field[3], "c4!=KC(c", fieldNum); 341 } 342 343 Normalizer::normalize(field[i], UNORM_NFKD, options, out, status); 344 if (U_FAILURE(status)) { 345 dataerrln("Error running normalize UNORM_NFKD: %s", u_errorName(status)); 346 } else { 347 pass &= assertEqual("KD", field[i], out, field[4], "c5!=KD(c", fieldNum); 348 iterativeNorm(field[i], UNORM_NFKD, options, out, +1); 349 pass &= assertEqual("KD(+1)", field[i], out, field[4], "c5!=KD(c", fieldNum); 350 iterativeNorm(field[i], UNORM_NFKD, options, out, -1); 351 pass &= assertEqual("KD(-1)", field[i], out, field[4], "c5!=KD(c", fieldNum); 352 } 353 } 354 compare(field[1],field[2]); 355 compare(field[0],field[1]); 356 // test quick checks 357 if(UNORM_NO == Normalizer::quickCheck(field[1], UNORM_NFC, options, status)) { 358 errln("Normalizer error: quickCheck(NFC(s), UNORM_NFC) is UNORM_NO"); 359 pass = FALSE; 360 } 361 if(UNORM_NO == Normalizer::quickCheck(field[2], UNORM_NFD, options, status)) { 362 errln("Normalizer error: quickCheck(NFD(s), UNORM_NFD) is UNORM_NO"); 363 pass = FALSE; 364 } 365 if(UNORM_NO == Normalizer::quickCheck(field[3], UNORM_NFKC, options, status)) { 366 errln("Normalizer error: quickCheck(NFKC(s), UNORM_NFKC) is UNORM_NO"); 367 pass = FALSE; 368 } 369 if(UNORM_NO == Normalizer::quickCheck(field[4], UNORM_NFKD, options, status)) { 370 errln("Normalizer error: quickCheck(NFKD(s), UNORM_NFKD) is UNORM_NO"); 371 pass = FALSE; 372 } 373 374 // branch on options==0 for better code coverage 375 if(options==0) { 376 result = Normalizer::isNormalized(field[1], UNORM_NFC, status); 377 } else { 378 result = Normalizer::isNormalized(field[1], UNORM_NFC, options, status); 379 } 380 if(!result) { 381 dataerrln("Normalizer error: isNormalized(NFC(s), UNORM_NFC) is FALSE"); 382 pass = FALSE; 383 } 384 if(field[0]!=field[1] && Normalizer::isNormalized(field[0], UNORM_NFC, options, status)) { 385 errln("Normalizer error: isNormalized(s, UNORM_NFC) is TRUE"); 386 pass = FALSE; 387 } 388 if(!Normalizer::isNormalized(field[3], UNORM_NFKC, options, status)) { 389 dataerrln("Normalizer error: isNormalized(NFKC(s), UNORM_NFKC) is FALSE"); 390 pass = FALSE; 391 } 392 if(field[0]!=field[3] && Normalizer::isNormalized(field[0], UNORM_NFKC, options, status)) { 393 errln("Normalizer error: isNormalized(s, UNORM_NFKC) is TRUE"); 394 pass = FALSE; 395 } 396 397 // test FCD quick check and "makeFCD" 398 Normalizer::normalize(field[0], UNORM_FCD, options, fcd, status); 399 if(UNORM_NO == Normalizer::quickCheck(fcd, UNORM_FCD, options, status)) { 400 errln("Normalizer error: quickCheck(FCD(s), UNORM_FCD) is UNORM_NO"); 401 pass = FALSE; 402 } 403 if(UNORM_NO == Normalizer::quickCheck(field[2], UNORM_FCD, options, status)) { 404 errln("Normalizer error: quickCheck(NFD(s), UNORM_FCD) is UNORM_NO"); 405 pass = FALSE; 406 } 407 if(UNORM_NO == Normalizer::quickCheck(field[4], UNORM_FCD, options, status)) { 408 errln("Normalizer error: quickCheck(NFKD(s), UNORM_FCD) is UNORM_NO"); 409 pass = FALSE; 410 } 411 412 Normalizer::normalize(fcd, UNORM_NFD, options, out, status); 413 if(out != field[2]) { 414 dataerrln("Normalizer error: NFD(FCD(s))!=NFD(s)"); 415 pass = FALSE; 416 } 417 418 if (U_FAILURE(status)) { 419 dataerrln("Normalizer::normalize returned error status: %s", u_errorName(status)); 420 pass = FALSE; 421 } 422 423 if(field[0]!=field[2]) { 424 // two strings that are canonically equivalent must test 425 // equal under a canonical caseless match 426 // see UAX #21 Case Mappings and Jitterbug 2021 and 427 // Unicode Technical Committee meeting consensus 92-C31 428 int32_t rc; 429 430 status=U_ZERO_ERROR; 431 rc=Normalizer::compare(field[0], field[2], (options<<UNORM_COMPARE_NORM_OPTIONS_SHIFT)|U_COMPARE_IGNORE_CASE, status); 432 if(U_FAILURE(status)) { 433 dataerrln("Normalizer::compare(case-insensitive) sets %s", u_errorName(status)); 434 pass=FALSE; 435 } else if(rc!=0) { 436 errln("Normalizer::compare(original, NFD, case-insensitive) returned %d instead of 0 for equal", rc); 437 pass=FALSE; 438 } 439 } 440 441 if (!pass) { 442 dataerrln("FAIL: %s", line); 443 } 444 return pass; 445 } 446 447 /** 448 * Do a normalization using the iterative API in the given direction. 449 * @param dir either +1 or -1 450 */ 451 void NormalizerConformanceTest::iterativeNorm(const UnicodeString& str, 452 UNormalizationMode mode, int32_t options, 453 UnicodeString& result, 454 int8_t dir) { 455 UErrorCode status = U_ZERO_ERROR; 456 normalizer.setText(str, status); 457 normalizer.setMode(mode); 458 normalizer.setOption(-1, 0); // reset all options 459 normalizer.setOption(options, 1); // set desired options 460 result.truncate(0); 461 if (U_FAILURE(status)) { 462 return; 463 } 464 UChar32 ch; 465 if (dir > 0) { 466 for (ch = normalizer.first(); ch != Normalizer::DONE; 467 ch = normalizer.next()) { 468 result.append(ch); 469 } 470 } else { 471 for (ch = normalizer.last(); ch != Normalizer::DONE; 472 ch = normalizer.previous()) { 473 result.insert(0, ch); 474 } 475 } 476 } 477 478 /** 479 * @param op name of normalization form, e.g., "KC" 480 * @param s string being normalized 481 * @param got value received 482 * @param exp expected value 483 * @param msg description of this test 484 * @param return true if got == exp 485 */ 486 UBool NormalizerConformanceTest::assertEqual(const char *op, 487 const UnicodeString& s, 488 const UnicodeString& got, 489 const UnicodeString& exp, 490 const char *msg, 491 int32_t field) 492 { 493 if (exp == got) 494 return TRUE; 495 496 char *sChars, *gotChars, *expChars; 497 UnicodeString sPretty(prettify(s)); 498 UnicodeString gotPretty(prettify(got)); 499 UnicodeString expPretty(prettify(exp)); 500 501 sChars = new char[sPretty.length() + 1]; 502 gotChars = new char[gotPretty.length() + 1]; 503 expChars = new char[expPretty.length() + 1]; 504 505 sPretty.extract(0, sPretty.length(), sChars, sPretty.length() + 1); 506 sChars[sPretty.length()] = 0; 507 gotPretty.extract(0, gotPretty.length(), gotChars, gotPretty.length() + 1); 508 gotChars[gotPretty.length()] = 0; 509 expPretty.extract(0, expPretty.length(), expChars, expPretty.length() + 1); 510 expChars[expPretty.length()] = 0; 511 512 errln(" %s%d)%s(%s)=%s, exp. %s", msg, field, op, sChars, gotChars, expChars); 513 514 delete []sChars; 515 delete []gotChars; 516 delete []expChars; 517 return FALSE; 518 } 519 520 /** 521 * Split a string into pieces based on the given delimiter 522 * character. Then, parse the resultant fields from hex into 523 * characters. That is, "0040 0400;0C00;0899" -> new String[] { 524 * "\u0040\u0400", "\u0C00", "\u0899" }. The output is assumed to 525 * be of the proper length already, and exactly output.length 526 * fields are parsed. If there are too few an exception is 527 * thrown. If there are too many the extras are ignored. 528 * 529 * @return FALSE upon failure 530 */ 531 UBool NormalizerConformanceTest::hexsplit(const char *s, char delimiter, 532 UnicodeString output[], int32_t outputLength) { 533 const char *t = s; 534 char *end = NULL; 535 UChar32 c; 536 int32_t i; 537 for (i=0; i<outputLength; ++i) { 538 // skip whitespace 539 while(*t == ' ' || *t == '\t') { 540 ++t; 541 } 542 543 // read a sequence of code points 544 output[i].remove(); 545 for(;;) { 546 c = (UChar32)uprv_strtoul(t, &end, 16); 547 548 if( (char *)t == end || 549 (uint32_t)c > 0x10ffff || 550 (*end != ' ' && *end != '\t' && *end != delimiter) 551 ) { 552 errln(UnicodeString("Bad field ", "") + (i + 1) + " in " + UnicodeString(s, "")); 553 return FALSE; 554 } 555 556 output[i].append(c); 557 558 t = (const char *)end; 559 560 // skip whitespace 561 while(*t == ' ' || *t == '\t') { 562 ++t; 563 } 564 565 if(*t == delimiter) { 566 ++t; 567 break; 568 } 569 if(*t == 0) { 570 if((i + 1) == outputLength) { 571 return TRUE; 572 } else { 573 errln(UnicodeString("Missing field(s) in ", "") + s + " only " + (i + 1) + " out of " + outputLength); 574 return FALSE; 575 } 576 } 577 } 578 } 579 return TRUE; 580 } 581 582 // Specific tests for debugging. These are generally failures taken from 583 // the conformance file, but culled out to make debugging easier. 584 585 void NormalizerConformanceTest::TestCase6(void) { 586 _testOneLine("0385;0385;00A8 0301;0020 0308 0301;0020 0308 0301;"); 587 } 588 589 void NormalizerConformanceTest::_testOneLine(const char *line) { 590 UErrorCode status = U_ZERO_ERROR; 591 UnicodeString fields[FIELD_COUNT]; 592 if (!hexsplit(line, ';', fields, FIELD_COUNT)) { 593 errln((UnicodeString)"Unable to parse line " + line); 594 } else { 595 checkConformance(fields, line, 0, status); 596 } 597 } 598 599 #endif /* #if !UCONFIG_NO_NORMALIZATION */ 600
