icuSources/i18n/tridpars.cpp

   1 // © 2016 and later: Unicode, Inc. and others.
   2 // License & terms of use: http://www.unicode.org/copyright.html
   3 /*
   4 **********************************************************************
   5 *   Copyright (c) 2002-2014, International Business Machines Corporation
   6 *   and others.  All Rights Reserved.
   7 **********************************************************************
   8 *   Date        Name        Description
   9 *   01/14/2002  aliu        Creation.
  10 **********************************************************************
  11 */
  12
  13 #include "unicode/utypes.h"
  14
  15 #if !UCONFIG_NO_TRANSLITERATION
  16
  17 #include "tridpars.h"
  18 #include "hash.h"
  19 #include "mutex.h"
  20 #include "transreg.h"
  21 #include "uassert.h"
  22 #include "ucln_in.h"
  23 #include "unicode/parsepos.h"
  24 #include "unicode/translit.h"
  25 #include "unicode/uchar.h"
  26 #include "unicode/uniset.h"
  27 #include "unicode/unistr.h"
  28 #include "unicode/utrans.h"
  29 #include "util.h"
  30 #include "uvector.h"
  31
  32 U_NAMESPACE_BEGIN
  33
  34 static const UChar ID_DELIM    = 0x003B; // ;
  35 static const UChar TARGET_SEP  = 0x002D; // -
  36 static const UChar VARIANT_SEP = 0x002F; // /
  37 static const UChar OPEN_REV    = 0x0028; // (
  38 static const UChar CLOSE_REV   = 0x0029; // )
  39
  40 //static const UChar EMPTY[]     = {0}; // ""
  41 static const UChar ANY[]       = {65,110,121,0}; // "Any"
  42 static const UChar ANY_NULL[]  = {65,110,121,45,78,117,108,108,0}; // "Any-Null"
  43
  44 static const int32_t FORWARD = UTRANS_FORWARD;
  45 static const int32_t REVERSE = UTRANS_REVERSE;
  46
  47 static Hashtable* SPECIAL_INVERSES = NULL;
  48 static UInitOnce gSpecialInversesInitOnce = U_INITONCE_INITIALIZER;
  49
  50 /**
  51  * The mutex controlling access to SPECIAL_INVERSES
  52  */
  53 static UMutex LOCK = U_MUTEX_INITIALIZER;
  54
  55 TransliteratorIDParser::Specs::Specs(const UnicodeString& s, const UnicodeString& t,
  56                                      const UnicodeString& v, UBool sawS,
  57                                      const UnicodeString& f) {
  58     source = s;
  59     target = t;
  60     variant = v;
  61     sawSource = sawS;
  62     filter = f;
  63 }
  64
  65 TransliteratorIDParser::SingleID::SingleID(const UnicodeString& c, const UnicodeString& b,
  66                                            const UnicodeString& f) {
  67     canonID = c;
  68     basicID = b;
  69     filter = f;
  70 }
  71
  72 TransliteratorIDParser::SingleID::SingleID(const UnicodeString& c, const UnicodeString& b) {
  73     canonID = c;
  74     basicID = b;
  75 }
  76
  77 Transliterator* TransliteratorIDParser::SingleID::createInstance() {
  78     Transliterator* t;
  79     if (basicID.length() == 0) {
  80         t = createBasicInstance(UnicodeString(TRUE, ANY_NULL, 8), &canonID);
  81     } else {
  82         t = createBasicInstance(basicID, &canonID);
  83     }
  84     if (t != NULL) {
  85         if (filter.length() != 0) {
  86             UErrorCode ec = U_ZERO_ERROR;
  87             UnicodeSet *set = new UnicodeSet(filter, ec);
  88             if (U_FAILURE(ec)) {
  89                 delete set;
  90             } else {
  91                 t->adoptFilter(set);
  92             }
  93         }
  94     }
  95     return t;
  96 }
  97
  98
  99 /**
 100  * Parse a single ID, that is, an ID of the general form
 101  * "[f1] s1-t1/v1 ([f2] s2-t3/v2)", with the parenthesized element
 102  * optional, the filters optional, and the variants optional.
 103  * @param id the id to be parsed
 104  * @param pos INPUT-OUTPUT parameter.  On input, the position of
 105  * the first character to parse.  On output, the position after
 106  * the last character parsed.
 107  * @param dir the direction.  If the direction is REVERSE then the
 108  * SingleID is constructed for the reverse direction.
 109  * @return a SingleID object or NULL
 110  */
 111 TransliteratorIDParser::SingleID*
 112 TransliteratorIDParser::parseSingleID(const UnicodeString& id, int32_t& pos,
 113                                       int32_t dir, UErrorCode& status) {
 114
 115     int32_t start = pos;
 116
 117     // The ID will be of the form A, A(), A(B), or (B), where
 118     // A and B are filter IDs.
 119     Specs* specsA = NULL;
 120     Specs* specsB = NULL;
 121     UBool sawParen = FALSE;
 122
 123     // On the first pass, look for (B) or ().  If this fails, then
 124     // on the second pass, look for A, A(B), or A().
 125     for (int32_t pass=1; pass<=2; ++pass) {
 126         if (pass == 2) {
 127             specsA = parseFilterID(id, pos, TRUE);
 128             if (specsA == NULL) {
 129                 pos = start;
 130                 return NULL;
 131             }
 132         }
 133         if (ICU_Utility::parseChar(id, pos, OPEN_REV)) {
 134             sawParen = TRUE;
 135             if (!ICU_Utility::parseChar(id, pos, CLOSE_REV)) {
 136                 specsB = parseFilterID(id, pos, TRUE);
 137                 // Must close with a ')'
 138                 if (specsB == NULL || !ICU_Utility::parseChar(id, pos, CLOSE_REV)) {
 139                     delete specsA;
 140                     pos = start;
 141                     return NULL;
 142                 }
 143             }
 144             break;
 145         }
 146     }
 147
 148     // Assemble return results
 149     SingleID* single;
 150     if (sawParen) {
 151         if (dir == FORWARD) {
 152             SingleID* b = specsToID(specsB, FORWARD);
 153             single = specsToID(specsA, FORWARD);
 154             // Null pointers check
 155             if (b == NULL || single == NULL) {
 156                 delete b;
 157                 delete single;
 158                 status = U_MEMORY_ALLOCATION_ERROR;
 159                 return NULL;
 160             }
 161             single->canonID.append(OPEN_REV)
 162                 .append(b->canonID).append(CLOSE_REV);
 163             if (specsA != NULL) {
 164                 single->filter = specsA->filter;
 165             }
 166             delete b;
 167         } else {
 168             SingleID* a = specsToID(specsA, FORWARD);
 169             single = specsToID(specsB, FORWARD);
 170             // Check for null pointer.
 171             if (a == NULL || single == NULL) {
 172                 delete a;
 173                 delete single;
 174                 status = U_MEMORY_ALLOCATION_ERROR;
 175                 return NULL;
 176             }
 177             single->canonID.append(OPEN_REV)
 178                 .append(a->canonID).append(CLOSE_REV);
 179             if (specsB != NULL) {
 180                 single->filter = specsB->filter;
 181             }
 182             delete a;
 183         }
 184     } else {
 185         // assert(specsA != NULL);
 186         if (dir == FORWARD) {
 187             single = specsToID(specsA, FORWARD);
 188         } else {
 189             single = specsToSpecialInverse(*specsA, status);
 190             if (single == NULL) {
 191                 single = specsToID(specsA, REVERSE);
 192             }
 193         }
 194         // Check for NULL pointer
 195         if (single == NULL) {
 196                 status = U_MEMORY_ALLOCATION_ERROR;
 197                 return NULL;
 198         }
 199         single->filter = specsA->filter;
 200     }
 201
 202     delete specsA;
 203     delete specsB;
 204
 205     return single;
 206 }
 207
 208 /**
 209  * Parse a filter ID, that is, an ID of the general form
 210  * "[f1] s1-t1/v1", with the filters optional, and the variants optional.
 211  * @param id the id to be parsed
 212  * @param pos INPUT-OUTPUT parameter.  On input, the position of
 213  * the first character to parse.  On output, the position after
 214  * the last character parsed.
 215  * @return a SingleID object or null if the parse fails
 216  */
 217 TransliteratorIDParser::SingleID*
 218 TransliteratorIDParser::parseFilterID(const UnicodeString& id, int32_t& pos) {
 219
 220     int32_t start = pos;
 221
 222     Specs* specs = parseFilterID(id, pos, TRUE);
 223     if (specs == NULL) {
 224         pos = start;
 225         return NULL;
 226     }
 227
 228     // Assemble return results
 229     SingleID* single = specsToID(specs, FORWARD);
 230     if (single != NULL) {
 231         single->filter = specs->filter;
 232     }
 233     delete specs;
 234     return single;
 235 }
 236
 237 /**
 238  * Parse a global filter of the form "[f]" or "([f])", depending
 239  * on 'withParens'.
 240  * @param id the pattern the parse
 241  * @param pos INPUT-OUTPUT parameter.  On input, the position of
 242  * the first character to parse.  On output, the position after
 243  * the last character parsed.
 244  * @param dir the direction.
 245  * @param withParens INPUT-OUTPUT parameter.  On entry, if
 246  * withParens is 0, then parens are disallowed.  If it is 1,
 247  * then parens are requires.  If it is -1, then parens are
 248  * optional, and the return result will be set to 0 or 1.
 249  * @param canonID OUTPUT parameter.  The pattern for the filter
 250  * added to the canonID, either at the end, if dir is FORWARD, or
 251  * at the start, if dir is REVERSE.  The pattern will be enclosed
 252  * in parentheses if appropriate, and will be suffixed with an
 253  * ID_DELIM character.  May be NULL.
 254  * @return a UnicodeSet object or NULL.  A non-NULL results
 255  * indicates a successful parse, regardless of whether the filter
 256  * applies to the given direction.  The caller should discard it
 257  * if withParens != (dir == REVERSE).
 258  */
 259 UnicodeSet* TransliteratorIDParser::parseGlobalFilter(const UnicodeString& id, int32_t& pos,
 260                                                       int32_t dir,
 261                                                       int32_t& withParens,
 262                                                       UnicodeString* canonID) {
 263     UnicodeSet* filter = NULL;
 264     int32_t start = pos;
 265
 266     if (withParens == -1) {
 267         withParens = ICU_Utility::parseChar(id, pos, OPEN_REV) ? 1 : 0;
 268     } else if (withParens == 1) {
 269         if (!ICU_Utility::parseChar(id, pos, OPEN_REV)) {
 270             pos = start;
 271             return NULL;
 272         }
 273     }
 274
 275     ICU_Utility::skipWhitespace(id, pos, TRUE);
 276
 277     if (UnicodeSet::resemblesPattern(id, pos)) {
 278         ParsePosition ppos(pos);
 279         UErrorCode ec = U_ZERO_ERROR;
 280         filter = new UnicodeSet(id, ppos, USET_IGNORE_SPACE, NULL, ec);
 281         /* test for NULL */
 282         if (filter == 0) {
 283             pos = start;
 284             return 0;
 285         }
 286         if (U_FAILURE(ec)) {
 287             delete filter;
 288             pos = start;
 289             return NULL;
 290         }
 291
 292         UnicodeString pattern;
 293         id.extractBetween(pos, ppos.getIndex(), pattern);
 294         pos = ppos.getIndex();
 295
 296         if (withParens == 1 && !ICU_Utility::parseChar(id, pos, CLOSE_REV)) {
 297             pos = start;
 298             return NULL;
 299         }
 300
 301         // In the forward direction, append the pattern to the
 302         // canonID.  In the reverse, insert it at zero, and invert
 303         // the presence of parens ("A" <-> "(A)").
 304         if (canonID != NULL) {
 305             if (dir == FORWARD) {
 306                 if (withParens == 1) {
 307                     pattern.insert(0, OPEN_REV);
 308                     pattern.append(CLOSE_REV);
 309                 }
 310                 canonID->append(pattern).append(ID_DELIM);
 311             } else {
 312                 if (withParens == 0) {
 313                     pattern.insert(0, OPEN_REV);
 314                     pattern.append(CLOSE_REV);
 315                 }
 316                 canonID->insert(0, pattern);
 317                 canonID->insert(pattern.length(), ID_DELIM);
 318             }
 319         }
 320     }
 321
 322     return filter;
 323 }
 324
 325 U_CDECL_BEGIN
 326 static void U_CALLCONV _deleteSingleID(void* obj) {
 327     delete (TransliteratorIDParser::SingleID*) obj;
 328 }
 329
 330 static void U_CALLCONV _deleteTransliteratorTrIDPars(void* obj) {
 331     delete (Transliterator*) obj;
 332 }
 333 U_CDECL_END
 334
 335 /**
 336  * Parse a compound ID, consisting of an optional forward global
 337  * filter, a separator, one or more single IDs delimited by
 338  * separators, an an optional reverse global filter.  The
 339  * separator is a semicolon.  The global filters are UnicodeSet
 340  * patterns.  The reverse global filter must be enclosed in
 341  * parentheses.
 342  * @param id the pattern the parse
 343  * @param dir the direction.
 344  * @param canonID OUTPUT parameter that receives the canonical ID,
 345  * consisting of canonical IDs for all elements, as returned by
 346  * parseSingleID(), separated by semicolons.  Previous contents
 347  * are discarded.
 348  * @param list OUTPUT parameter that receives a list of SingleID
 349  * objects representing the parsed IDs.  Previous contents are
 350  * discarded.
 351  * @param globalFilter OUTPUT parameter that receives a pointer to
 352  * a newly created global filter for this ID in this direction, or
 353  * NULL if there is none.
 354  * @return TRUE if the parse succeeds, that is, if the entire
 355  * id is consumed without syntax error.
 356  */
 357 UBool TransliteratorIDParser::parseCompoundID(const UnicodeString& id, int32_t dir,
 358                                               UnicodeString& canonID,
 359                                               UVector& list,
 360                                               UnicodeSet*& globalFilter) {
 361     UErrorCode ec = U_ZERO_ERROR;
 362     int32_t i;
 363     int32_t pos = 0;
 364     int32_t withParens = 1;
 365     list.removeAllElements();
 366     UnicodeSet* filter;
 367     globalFilter = NULL;
 368     canonID.truncate(0);
 369
 370     // Parse leading global filter, if any
 371     withParens = 0; // parens disallowed
 372     filter = parseGlobalFilter(id, pos, dir, withParens, &canonID);
 373     if (filter != NULL) {
 374         if (!ICU_Utility::parseChar(id, pos, ID_DELIM)) {
 375             // Not a global filter; backup and resume
 376             canonID.truncate(0);
 377             pos = 0;
 378         }
 379         if (dir == FORWARD) {
 380             globalFilter = filter;
 381         } else {
 382             delete filter;
 383         }
 384         filter = NULL;
 385     }
 386
 387     UBool sawDelimiter = TRUE;
 388     for (;;) {
 389         SingleID* single = parseSingleID(id, pos, dir, ec);
 390         if (single == NULL) {
 391             break;
 392         }
 393         if (dir == FORWARD) {
 394             list.addElement(single, ec);
 395         } else {
 396             list.insertElementAt(single, 0, ec);
 397         }
 398         if (U_FAILURE(ec)) {
 399             goto FAIL;
 400         }
 401         if (!ICU_Utility::parseChar(id, pos, ID_DELIM)) {
 402             sawDelimiter = FALSE;
 403             break;
 404         }
 405     }
 406
 407     if (list.size() == 0) {
 408         goto FAIL;
 409     }
 410
 411     // Construct canonical ID
 412     for (i=0; i<list.size(); ++i) {
 413         SingleID* single = (SingleID*) list.elementAt(i);
 414         canonID.append(single->canonID);
 415         if (i != (list.size()-1)) {
 416             canonID.append(ID_DELIM);
 417         }
 418     }
 419
 420     // Parse trailing global filter, if any, and only if we saw
 421     // a trailing delimiter after the IDs.
 422     if (sawDelimiter) {
 423         withParens = 1; // parens required
 424         filter = parseGlobalFilter(id, pos, dir, withParens, &canonID);
 425         if (filter != NULL) {
 426             // Don't require trailing ';', but parse it if present
 427             ICU_Utility::parseChar(id, pos, ID_DELIM);
 428
 429             if (dir == REVERSE) {
 430                 globalFilter = filter;
 431             } else {
 432                 delete filter;
 433             }
 434             filter = NULL;
 435         }
 436     }
 437
 438     // Trailing unparsed text is a syntax error
 439     ICU_Utility::skipWhitespace(id, pos, TRUE);
 440     if (pos != id.length()) {
 441         goto FAIL;
 442     }
 443
 444     return TRUE;
 445
 446  FAIL:
 447     UObjectDeleter *save = list.setDeleter(_deleteSingleID);
 448     list.removeAllElements();
 449     list.setDeleter(save);
 450     delete globalFilter;
 451     globalFilter = NULL;
 452     return FALSE;
 453 }
 454
 455 /**
 456  * Convert the elements of the 'list' vector, which are SingleID
 457  * objects, into actual Transliterator objects.  In the course of
 458  * this, some (or all) entries may be removed.  If all entries
 459  * are removed, the NULL transliterator will be added.
 460  *
 461  * Delete entries with empty basicIDs; these are generated by
 462  * elements like "(A)" in the forward direction, or "A()" in
 463  * the reverse.  THIS MAY RESULT IN AN EMPTY VECTOR.  Convert
 464  * SingleID entries to actual transliterators.
 465  *
 466  * @param list vector of SingleID objects.  On exit, vector
 467  * of one or more Transliterators.
 468  * @return new value of insertIndex.  The index will shift if
 469  * there are empty items, like "(Lower)", with indices less than
 470  * insertIndex.
 471  */
 472 void TransliteratorIDParser::instantiateList(UVector& list,
 473                                                 UErrorCode& ec) {
 474     UVector tlist(ec);
 475     if (U_FAILURE(ec)) {
 476         goto RETURN;
 477     }
 478     tlist.setDeleter(_deleteTransliteratorTrIDPars);
 479
 480     Transliterator* t;
 481     int32_t i;
 482     for (i=0; i<=list.size(); ++i) { // [sic]: i<=list.size()
 483         // We run the loop too long by one, so we can
 484         // do an insert after the last element
 485         if (i==list.size()) {
 486             break;
 487         }
 488
 489         SingleID* single = (SingleID*) list.elementAt(i);
 490         if (single->basicID.length() != 0) {
 491             t = single->createInstance();
 492             if (t == NULL) {
 493                 ec = U_INVALID_ID;
 494                 goto RETURN;
 495             }
 496             tlist.addElement(t, ec);
 497             if (U_FAILURE(ec)) {
 498                 delete t;
 499                 goto RETURN;
 500             }
 501         }
 502     }
 503
 504     // An empty list is equivalent to a NULL transliterator.
 505     if (tlist.size() == 0) {
 506         t = createBasicInstance(UnicodeString(TRUE, ANY_NULL, 8), NULL);
 507         if (t == NULL) {
 508             // Should never happen
 509             ec = U_INTERNAL_TRANSLITERATOR_ERROR;
 510         }
 511         tlist.addElement(t, ec);
 512         if (U_FAILURE(ec)) {
 513             delete t;
 514         }
 515     }
 516
 517  RETURN:
 518
 519     UObjectDeleter *save = list.setDeleter(_deleteSingleID);
 520     list.removeAllElements();
 521
 522     if (U_SUCCESS(ec)) {
 523         list.setDeleter(_deleteTransliteratorTrIDPars);
 524
 525         while (tlist.size() > 0) {
 526             t = (Transliterator*) tlist.orphanElementAt(0);
 527             list.addElement(t, ec);
 528             if (U_FAILURE(ec)) {
 529                 delete t;
 530                 list.removeAllElements();
 531                 break;
 532             }
 533         }
 534     }
 535
 536     list.setDeleter(save);
 537 }
 538
 539 /**
 540  * Parse an ID into pieces.  Take IDs of the form T, T/V, S-T,
 541  * S-T/V, or S/V-T.  If the source is missing, return a source of
 542  * ANY.
 543  * @param id the id string, in any of several forms
 544  * @return an array of 4 strings: source, target, variant, and
 545  * isSourcePresent.  If the source is not present, ANY will be
 546  * given as the source, and isSourcePresent will be NULL.  Otherwise
 547  * isSourcePresent will be non-NULL.  The target may be empty if the
 548  * id is not well-formed.  The variant may be empty.
 549  */
 550 void TransliteratorIDParser::IDtoSTV(const UnicodeString& id,
 551                                      UnicodeString& source,
 552                                      UnicodeString& target,
 553                                      UnicodeString& variant,
 554                                      UBool& isSourcePresent) {
 555     source.setTo(ANY, 3);
 556     target.truncate(0);
 557     variant.truncate(0);
 558
 559     int32_t sep = id.indexOf(TARGET_SEP);
 560     int32_t var = id.indexOf(VARIANT_SEP);
 561     if (var < 0) {
 562         var = id.length();
 563     }
 564     isSourcePresent = FALSE;
 565
 566     if (sep < 0) {
 567         // Form: T/V or T (or /V)
 568         id.extractBetween(0, var, target);
 569         id.extractBetween(var, id.length(), variant);
 570     } else if (sep < var) {
 571         // Form: S-T/V or S-T (or -T/V or -T)
 572         if (sep > 0) {
 573             id.extractBetween(0, sep, source);
 574             isSourcePresent = TRUE;
 575         }
 576         id.extractBetween(++sep, var, target);
 577         id.extractBetween(var, id.length(), variant);
 578     } else {
 579         // Form: (S/V-T or /V-T)
 580         if (var > 0) {
 581             id.extractBetween(0, var, source);
 582             isSourcePresent = TRUE;
 583         }
 584         id.extractBetween(var, sep++, variant);
 585         id.extractBetween(sep, id.length(), target);
 586     }
 587
 588     if (variant.length() > 0) {
 589         variant.remove(0, 1);
 590     }
 591 }
 592
 593 /**
 594  * Given source, target, and variant strings, concatenate them into a
 595  * full ID.  If the source is empty, then "Any" will be used for the
 596  * source, so the ID will always be of the form s-t/v or s-t.
 597  */
 598 void TransliteratorIDParser::STVtoID(const UnicodeString& source,
 599                                      const UnicodeString& target,
 600                                      const UnicodeString& variant,
 601                                      UnicodeString& id) {
 602     id = source;
 603     if (id.length() == 0) {
 604         id.setTo(ANY, 3);
 605     }
 606     id.append(TARGET_SEP).append(target);
 607     if (variant.length() != 0) {
 608         id.append(VARIANT_SEP).append(variant);
 609     }
 610     // NUL-terminate the ID string for getTerminatedBuffer.
 611     // This prevents valgrind and Purify warnings.
 612     id.append((UChar)0);
 613     id.truncate(id.length()-1);
 614 }
 615
 616 /**
 617  * Register two targets as being inverses of one another.  For
 618  * example, calling registerSpecialInverse("NFC", "NFD", TRUE) causes
 619  * Transliterator to form the following inverse relationships:
 620  *
 621  * <pre>NFC => NFD
 622  * Any-NFC => Any-NFD
 623  * NFD => NFC
 624  * Any-NFD => Any-NFC</pre>
 625  *
 626  * (Without the special inverse registration, the inverse of NFC
 627  * would be NFC-Any.)  Note that NFD is shorthand for Any-NFD, but
 628  * that the presence or absence of "Any-" is preserved.
 629  *
 630  * <p>The relationship is symmetrical; registering (a, b) is
 631  * equivalent to registering (b, a).
 632  *
 633  * <p>The relevant IDs must still be registered separately as
 634  * factories or classes.
 635  *
 636  * <p>Only the targets are specified.  Special inverses always
 637  * have the form Any-Target1 <=> Any-Target2.  The target should
 638  * have canonical casing (the casing desired to be produced when
 639  * an inverse is formed) and should contain no whitespace or other
 640  * extraneous characters.
 641  *
 642  * @param target the target against which to register the inverse
 643  * @param inverseTarget the inverse of target, that is
 644  * Any-target.getInverse() => Any-inverseTarget
 645  * @param bidirectional if TRUE, register the reverse relation
 646  * as well, that is, Any-inverseTarget.getInverse() => Any-target
 647  */
 648 void TransliteratorIDParser::registerSpecialInverse(const UnicodeString& target,
 649                                                     const UnicodeString& inverseTarget,
 650                                                     UBool bidirectional,
 651                                                     UErrorCode &status) {
 652     umtx_initOnce(gSpecialInversesInitOnce, init, status);
 653     if (U_FAILURE(status)) {
 654         return;
 655     }
 656
 657     // If target == inverseTarget then force bidirectional => FALSE
 658     if (bidirectional && 0==target.caseCompare(inverseTarget, U_FOLD_CASE_DEFAULT)) {
 659         bidirectional = FALSE;
 660     }
 661
 662     Mutex lock(&LOCK);
 663
 664     UnicodeString *tempus = new UnicodeString(inverseTarget);  // Used for null pointer check before usage.
 665     if (tempus == NULL) {
 666         status = U_MEMORY_ALLOCATION_ERROR;
 667         return;
 668     }
 669     SPECIAL_INVERSES->put(target, tempus, status);
 670     if (bidirectional) {
 671         tempus = new UnicodeString(target);
 672         if (tempus == NULL) {
 673                 status = U_MEMORY_ALLOCATION_ERROR;
 674                 return;
 675         }
 676         SPECIAL_INVERSES->put(inverseTarget, tempus, status);
 677     }
 678 }
 679
 680 //----------------------------------------------------------------
 681 // Private implementation
 682 //----------------------------------------------------------------
 683
 684 /**
 685  * Parse an ID into component pieces.  Take IDs of the form T,
 686  * T/V, S-T, S-T/V, or S/V-T.  If the source is missing, return a
 687  * source of ANY.
 688  * @param id the id string, in any of several forms
 689  * @param pos INPUT-OUTPUT parameter.  On input, pos is the
 690  * offset of the first character to parse in id.  On output,
 691  * pos is the offset after the last parsed character.  If the
 692  * parse failed, pos will be unchanged.
 693  * @param allowFilter2 if TRUE, a UnicodeSet pattern is allowed
 694  * at any location between specs or delimiters, and is returned
 695  * as the fifth string in the array.
 696  * @return a Specs object, or NULL if the parse failed.  If
 697  * neither source nor target was seen in the parsed id, then the
 698  * parse fails.  If allowFilter is TRUE, then the parsed filter
 699  * pattern is returned in the Specs object, otherwise the returned
 700  * filter reference is NULL.  If the parse fails for any reason
 701  * NULL is returned.
 702  */
 703 TransliteratorIDParser::Specs*
 704 TransliteratorIDParser::parseFilterID(const UnicodeString& id, int32_t& pos,
 705                                       UBool allowFilter) {
 706     UnicodeString first;
 707     UnicodeString source;
 708     UnicodeString target;
 709     UnicodeString variant;
 710     UnicodeString filter;
 711     UChar delimiter = 0;
 712     int32_t specCount = 0;
 713     int32_t start = pos;
 714
 715     // This loop parses one of the following things with each
 716     // pass: a filter, a delimiter character (either '-' or '/'),
 717     // or a spec (source, target, or variant).
 718     for (;;) {
 719         ICU_Utility::skipWhitespace(id, pos, TRUE);
 720         if (pos == id.length()) {
 721             break;
 722         }
 723
 724         // Parse filters
 725         if (allowFilter && filter.length() == 0 &&
 726             UnicodeSet::resemblesPattern(id, pos)) {
 727
 728             ParsePosition ppos(pos);
 729             UErrorCode ec = U_ZERO_ERROR;
 730             UnicodeSet set(id, ppos, USET_IGNORE_SPACE, NULL, ec);
 731             if (U_FAILURE(ec)) {
 732                 pos = start;
 733                 return NULL;
 734             }
 735             id.extractBetween(pos, ppos.getIndex(), filter);
 736             pos = ppos.getIndex();
 737             continue;
 738         }
 739
 740         if (delimiter == 0) {
 741             UChar c = id.charAt(pos);
 742             if ((c == TARGET_SEP && target.length() == 0) ||
 743                 (c == VARIANT_SEP && variant.length() == 0)) {
 744                 delimiter = c;
 745                 ++pos;
 746                 continue;
 747             }
 748         }
 749
 750         // We are about to try to parse a spec with no delimiter
 751         // when we can no longer do so (we can only do so at the
 752         // start); break.
 753         if (delimiter == 0 && specCount > 0) {
 754             break;
 755         }
 756
 757         UnicodeString spec = ICU_Utility::parseUnicodeIdentifier(id, pos);
 758         if (spec.length() == 0) {
 759             // Note that if there was a trailing delimiter, we
 760             // consume it.  So Foo-, Foo/, Foo-Bar/, and Foo/Bar-
 761             // are legal.
 762             break;
 763         }
 764
 765         switch (delimiter) {
 766         case 0:
 767             first = spec;
 768             break;
 769         case TARGET_SEP:
 770             target = spec;
 771             break;
 772         case VARIANT_SEP:
 773             variant = spec;
 774             break;
 775         }
 776         ++specCount;
 777         delimiter = 0;
 778     }
 779
 780     // A spec with no prior character is either source or target,
 781     // depending on whether an explicit "-target" was seen.
 782     if (first.length() != 0) {
 783         if (target.length() == 0) {
 784             target = first;
 785         } else {
 786             source = first;
 787         }
 788     }
 789
 790     // Must have either source or target
 791     if (source.length() == 0 && target.length() == 0) {
 792         pos = start;
 793         return NULL;
 794     }
 795
 796     // Empty source or target defaults to ANY
 797     UBool sawSource = TRUE;
 798     if (source.length() == 0) {
 799         source.setTo(ANY, 3);
 800         sawSource = FALSE;
 801     }
 802     if (target.length() == 0) {
 803         target.setTo(ANY, 3);
 804     }
 805
 806     return new Specs(source, target, variant, sawSource, filter);
 807 }
 808
 809 /**
 810  * Givens a Spec object, convert it to a SingleID object.  The
 811  * Spec object is a more unprocessed parse result.  The SingleID
 812  * object contains information about canonical and basic IDs.
 813  * @return a SingleID; never returns NULL.  Returned object always
 814  * has 'filter' field of NULL.
 815  */
 816 TransliteratorIDParser::SingleID*
 817 TransliteratorIDParser::specsToID(const Specs* specs, int32_t dir) {
 818     UnicodeString canonID;
 819     UnicodeString basicID;
 820     UnicodeString basicPrefix;
 821     if (specs != NULL) {
 822         UnicodeString buf;
 823         if (dir == FORWARD) {
 824             if (specs->sawSource) {
 825                 buf.append(specs->source).append(TARGET_SEP);
 826             } else {
 827                 basicPrefix = specs->source;
 828                 basicPrefix.append(TARGET_SEP);
 829             }
 830             buf.append(specs->target);
 831         } else {
 832             buf.append(specs->target).append(TARGET_SEP).append(specs->source);
 833         }
 834         if (specs->variant.length() != 0) {
 835             buf.append(VARIANT_SEP).append(specs->variant);
 836         }
 837         basicID = basicPrefix;
 838         basicID.append(buf);
 839         if (specs->filter.length() != 0) {
 840             buf.insert(0, specs->filter);
 841         }
 842         canonID = buf;
 843     }
 844     return new SingleID(canonID, basicID);
 845 }
 846
 847 /**
 848  * Given a Specs object, return a SingleID representing the
 849  * special inverse of that ID.  If there is no special inverse
 850  * then return NULL.
 851  * @return a SingleID or NULL.  Returned object always has
 852  * 'filter' field of NULL.
 853  */
 854 TransliteratorIDParser::SingleID*
 855 TransliteratorIDParser::specsToSpecialInverse(const Specs& specs, UErrorCode &status) {
 856     if (0!=specs.source.caseCompare(ANY, 3, U_FOLD_CASE_DEFAULT)) {
 857         return NULL;
 858     }
 859     umtx_initOnce(gSpecialInversesInitOnce, init, status);
 860     if (U_FAILURE(status)) {
 861         return NULL;
 862     }
 863
 864     UnicodeString* inverseTarget;
 865
 866     umtx_lock(&LOCK);
 867     inverseTarget = (UnicodeString*) SPECIAL_INVERSES->get(specs.target);
 868     umtx_unlock(&LOCK);
 869
 870     if (inverseTarget != NULL) {
 871         // If the original ID contained "Any-" then make the
 872         // special inverse "Any-Foo"; otherwise make it "Foo".
 873         // So "Any-NFC" => "Any-NFD" but "NFC" => "NFD".
 874         UnicodeString buf;
 875         if (specs.filter.length() != 0) {
 876             buf.append(specs.filter);
 877         }
 878         if (specs.sawSource) {
 879             buf.append(ANY, 3).append(TARGET_SEP);
 880         }
 881         buf.append(*inverseTarget);
 882
 883         UnicodeString basicID(TRUE, ANY, 3);
 884         basicID.append(TARGET_SEP).append(*inverseTarget);
 885
 886         if (specs.variant.length() != 0) {
 887             buf.append(VARIANT_SEP).append(specs.variant);
 888             basicID.append(VARIANT_SEP).append(specs.variant);
 889         }
 890         return new SingleID(buf, basicID);
 891     }
 892     return NULL;
 893 }
 894
 895 /**
 896  * Glue method to get around access problems in C++.  This would
 897  * ideally be inline but we want to avoid a circular header
 898  * dependency.
 899  */
 900 Transliterator* TransliteratorIDParser::createBasicInstance(const UnicodeString& id, const UnicodeString* canonID) {
 901     return Transliterator::createBasicInstance(id, canonID);
 902 }
 903
 904 /**
 905  * Initialize static memory. Called through umtx_initOnce only.
 906  */
 907 void U_CALLCONV TransliteratorIDParser::init(UErrorCode &status) {
 908     U_ASSERT(SPECIAL_INVERSES == NULL);
 909     ucln_i18n_registerCleanup(UCLN_I18N_TRANSLITERATOR, utrans_transliterator_cleanup);
 910
 911     SPECIAL_INVERSES = new Hashtable(TRUE, status);
 912     if (SPECIAL_INVERSES == NULL) {
 913         status = U_MEMORY_ALLOCATION_ERROR;
 914         return;
 915     }
 916     SPECIAL_INVERSES->setValueDeleter(uprv_deleteUObject);
 917 }
 918
 919 /**
 920  * Free static memory.
 921  */
 922 void TransliteratorIDParser::cleanup() {
 923     if (SPECIAL_INVERSES) {
 924         delete SPECIAL_INVERSES;
 925         SPECIAL_INVERSES = NULL;
 926     }
 927     gSpecialInversesInitOnce.reset();
 928 }
 929
 930 U_NAMESPACE_END
 931
 932 #endif /* #if !UCONFIG_NO_TRANSLITERATION */
 933
 934 //eof