icuSources/i18n/tridpars.cpp

   1 /*
   2 **********************************************************************
   3 *   Copyright (c) 2002-2006, International Business Machines Corporation
   4 *   and others.  All Rights Reserved.
   5 **********************************************************************
   6 *   Date        Name        Description
   7 *   01/14/2002  aliu        Creation.
   8 **********************************************************************
   9 */
  10
  11 #include "unicode/utypes.h"
  12
  13 #if !UCONFIG_NO_TRANSLITERATION
  14
  15 #include "tridpars.h"
  16 #include "hash.h"
  17 #include "mutex.h"
  18 #include "ucln_in.h"
  19 #include "unicode/parsepos.h"
  20 #include "unicode/translit.h"
  21 #include "unicode/uchar.h"
  22 #include "unicode/uniset.h"
  23 #include "unicode/unistr.h"
  24 #include "unicode/utrans.h"
  25 #include "util.h"
  26 #include "uvector.h"
  27
  28 U_NAMESPACE_BEGIN
  29
  30 static const UChar ID_DELIM    = 0x003B; // ;
  31 static const UChar TARGET_SEP  = 0x002D; // -
  32 static const UChar VARIANT_SEP = 0x002F; // /
  33 static const UChar OPEN_REV    = 0x0028; // (
  34 static const UChar CLOSE_REV   = 0x0029; // )
  35
  36 //static const UChar EMPTY[]     = {0}; // ""
  37 static const UChar ANY[]       = {65,110,121,0}; // "Any"
  38 static const UChar ANY_NULL[]  = {65,110,121,45,78,117,108,108,0}; // "Any-Null"
  39
  40 static const int32_t FORWARD = UTRANS_FORWARD;
  41 static const int32_t REVERSE = UTRANS_REVERSE;
  42
  43 static Hashtable* SPECIAL_INVERSES = NULL;
  44
  45 /**
  46  * The mutex controlling access to SPECIAL_INVERSES
  47  */
  48 static UMTX LOCK = 0;
  49
  50 TransliteratorIDParser::Specs::Specs(const UnicodeString& s, const UnicodeString& t,
  51                                      const UnicodeString& v, UBool sawS,
  52                                      const UnicodeString& f) {
  53     source = s;
  54     target = t;
  55     variant = v;
  56     sawSource = sawS;
  57     filter = f;
  58 }
  59
  60 TransliteratorIDParser::SingleID::SingleID(const UnicodeString& c, const UnicodeString& b,
  61                                            const UnicodeString& f) {
  62     canonID = c;
  63     basicID = b;
  64     filter = f;
  65 }
  66
  67 TransliteratorIDParser::SingleID::SingleID(const UnicodeString& c, const UnicodeString& b) {
  68     canonID = c;
  69     basicID = b;
  70 }
  71
  72 Transliterator* TransliteratorIDParser::SingleID::createInstance() {
  73     Transliterator* t;
  74     if (basicID.length() == 0) {
  75         t = createBasicInstance(ANY_NULL, &canonID);
  76     } else {
  77         t = createBasicInstance(basicID, &canonID);
  78     }
  79     if (t != NULL) {
  80         if (filter.length() != 0) {
  81             UErrorCode ec = U_ZERO_ERROR;
  82             UnicodeSet *set = new UnicodeSet(filter, ec);
  83             if (U_FAILURE(ec)) {
  84                 delete set;
  85             } else {
  86                 t->adoptFilter(set);
  87             }
  88         }
  89     }
  90     return t;
  91 }
  92
  93
  94 /**
  95  * Parse a single ID, that is, an ID of the general form
  96  * "[f1] s1-t1/v1 ([f2] s2-t3/v2)", with the parenthesized element
  97  * optional, the filters optional, and the variants optional.
  98  * @param id the id to be parsed
  99  * @param pos INPUT-OUTPUT parameter.  On input, the position of
 100  * the first character to parse.  On output, the position after
 101  * the last character parsed.
 102  * @param dir the direction.  If the direction is REVERSE then the
 103  * SingleID is constructed for the reverse direction.
 104  * @return a SingleID object or NULL
 105  */
 106 TransliteratorIDParser::SingleID*
 107 TransliteratorIDParser::parseSingleID(const UnicodeString& id, int32_t& pos,
 108                                       int32_t dir, UErrorCode& status) {
 109
 110     int32_t start = pos;
 111
 112     // The ID will be of the form A, A(), A(B), or (B), where
 113     // A and B are filter IDs.
 114     Specs* specsA = NULL;
 115     Specs* specsB = NULL;
 116     UBool sawParen = FALSE;
 117
 118     // On the first pass, look for (B) or ().  If this fails, then
 119     // on the second pass, look for A, A(B), or A().
 120     for (int32_t pass=1; pass<=2; ++pass) {
 121         if (pass == 2) {
 122             specsA = parseFilterID(id, pos, TRUE);
 123             if (specsA == NULL) {
 124                 pos = start;
 125                 return NULL;
 126             }
 127         }
 128         if (ICU_Utility::parseChar(id, pos, OPEN_REV)) {
 129             sawParen = TRUE;
 130             if (!ICU_Utility::parseChar(id, pos, CLOSE_REV)) {
 131                 specsB = parseFilterID(id, pos, TRUE);
 132                 // Must close with a ')'
 133                 if (specsB == NULL || !ICU_Utility::parseChar(id, pos, CLOSE_REV)) {
 134                     delete specsA;
 135                     pos = start;
 136                     return NULL;
 137                 }
 138             }
 139             break;
 140         }
 141     }
 142
 143     // Assemble return results
 144     SingleID* single;
 145     if (sawParen) {
 146         if (dir == FORWARD) {
 147             SingleID* b = specsToID(specsB, FORWARD);
 148             single = specsToID(specsA, FORWARD);
 149             single->canonID.append(OPEN_REV)
 150                 .append(b->canonID).append(CLOSE_REV);
 151             if (specsA != NULL) {
 152                 single->filter = specsA->filter;
 153             }
 154             delete b;
 155         } else {
 156             SingleID* a = specsToID(specsA, FORWARD);
 157             single = specsToID(specsB, FORWARD);
 158             single->canonID.append(OPEN_REV)
 159                 .append(a->canonID).append(CLOSE_REV);
 160             if (specsB != NULL) {
 161                 single->filter = specsB->filter;
 162             }
 163             delete a;
 164         }
 165     } else {
 166         // assert(specsA != NULL);
 167         if (dir == FORWARD) {
 168             single = specsToID(specsA, FORWARD);
 169         } else {
 170             single = specsToSpecialInverse(*specsA, status);
 171             if (single == NULL) {
 172                 single = specsToID(specsA, REVERSE);
 173             }
 174         }
 175         single->filter = specsA->filter;
 176     }
 177
 178     delete specsA;
 179     delete specsB;
 180
 181     return single;
 182 }
 183
 184 /**
 185  * Parse a filter ID, that is, an ID of the general form
 186  * "[f1] s1-t1/v1", with the filters optional, and the variants optional.
 187  * @param id the id to be parsed
 188  * @param pos INPUT-OUTPUT parameter.  On input, the position of
 189  * the first character to parse.  On output, the position after
 190  * the last character parsed.
 191  * @return a SingleID object or null if the parse fails
 192  */
 193 TransliteratorIDParser::SingleID*
 194 TransliteratorIDParser::parseFilterID(const UnicodeString& id, int32_t& pos) {
 195
 196     int32_t start = pos;
 197
 198     Specs* specs = parseFilterID(id, pos, TRUE);
 199     if (specs == NULL) {
 200         pos = start;
 201         return NULL;
 202     }
 203
 204     // Assemble return results
 205     SingleID* single = specsToID(specs, FORWARD);
 206     single->filter = specs->filter;
 207     delete specs;
 208     return single;
 209 }
 210
 211 /**
 212  * Parse a global filter of the form "[f]" or "([f])", depending
 213  * on 'withParens'.
 214  * @param id the pattern the parse
 215  * @param pos INPUT-OUTPUT parameter.  On input, the position of
 216  * the first character to parse.  On output, the position after
 217  * the last character parsed.
 218  * @param dir the direction.
 219  * @param withParens INPUT-OUTPUT parameter.  On entry, if
 220  * withParens is 0, then parens are disallowed.  If it is 1,
 221  * then parens are requires.  If it is -1, then parens are
 222  * optional, and the return result will be set to 0 or 1.
 223  * @param canonID OUTPUT parameter.  The pattern for the filter
 224  * added to the canonID, either at the end, if dir is FORWARD, or
 225  * at the start, if dir is REVERSE.  The pattern will be enclosed
 226  * in parentheses if appropriate, and will be suffixed with an
 227  * ID_DELIM character.  May be NULL.
 228  * @return a UnicodeSet object or NULL.  A non-NULL results
 229  * indicates a successful parse, regardless of whether the filter
 230  * applies to the given direction.  The caller should discard it
 231  * if withParens != (dir == REVERSE).
 232  */
 233 UnicodeSet* TransliteratorIDParser::parseGlobalFilter(const UnicodeString& id, int32_t& pos,
 234                                                       int32_t dir,
 235                                                       int32_t& withParens,
 236                                                       UnicodeString* canonID) {
 237     UnicodeSet* filter = NULL;
 238     int32_t start = pos;
 239
 240     if (withParens == -1) {
 241         withParens = ICU_Utility::parseChar(id, pos, OPEN_REV) ? 1 : 0;
 242     } else if (withParens == 1) {
 243         if (!ICU_Utility::parseChar(id, pos, OPEN_REV)) {
 244             pos = start;
 245             return NULL;
 246         }
 247     }
 248
 249     ICU_Utility::skipWhitespace(id, pos, TRUE);
 250
 251     if (UnicodeSet::resemblesPattern(id, pos)) {
 252         ParsePosition ppos(pos);
 253         UErrorCode ec = U_ZERO_ERROR;
 254         filter = new UnicodeSet(id, ppos, USET_IGNORE_SPACE, NULL, ec);
 255         /* test for NULL */
 256         if (filter == 0) {
 257             pos = start;
 258             return 0;
 259         }
 260         if (U_FAILURE(ec)) {
 261             delete filter;
 262             pos = start;
 263             return NULL;
 264         }
 265
 266         UnicodeString pattern;
 267         id.extractBetween(pos, ppos.getIndex(), pattern);
 268         pos = ppos.getIndex();
 269
 270         if (withParens == 1 && !ICU_Utility::parseChar(id, pos, CLOSE_REV)) {
 271             pos = start;
 272             return NULL;
 273         }
 274
 275         // In the forward direction, append the pattern to the
 276         // canonID.  In the reverse, insert it at zero, and invert
 277         // the presence of parens ("A" <-> "(A)").
 278         if (canonID != NULL) {
 279             if (dir == FORWARD) {
 280                 if (withParens == 1) {
 281                     pattern.insert(0, OPEN_REV);
 282                     pattern.append(CLOSE_REV);
 283                 }
 284                 canonID->append(pattern).append(ID_DELIM);
 285             } else {
 286                 if (withParens == 0) {
 287                     pattern.insert(0, OPEN_REV);
 288                     pattern.append(CLOSE_REV);
 289                 }
 290                 canonID->insert(0, pattern);
 291                 canonID->insert(pattern.length(), ID_DELIM);
 292             }
 293         }
 294     }
 295
 296     return filter;
 297 }
 298
 299 U_CDECL_BEGIN
 300 static void U_CALLCONV _deleteSingleID(void* obj) {
 301     delete (TransliteratorIDParser::SingleID*) obj;
 302 }
 303
 304 static void U_CALLCONV _deleteTransliteratorTrIDPars(void* obj) {
 305     delete (Transliterator*) obj;
 306 }
 307 U_CDECL_END
 308
 309 /**
 310  * Parse a compound ID, consisting of an optional forward global
 311  * filter, a separator, one or more single IDs delimited by
 312  * separators, an an optional reverse global filter.  The
 313  * separator is a semicolon.  The global filters are UnicodeSet
 314  * patterns.  The reverse global filter must be enclosed in
 315  * parentheses.
 316  * @param id the pattern the parse
 317  * @param dir the direction.
 318  * @param canonID OUTPUT parameter that receives the canonical ID,
 319  * consisting of canonical IDs for all elements, as returned by
 320  * parseSingleID(), separated by semicolons.  Previous contents
 321  * are discarded.
 322  * @param list OUTPUT parameter that receives a list of SingleID
 323  * objects representing the parsed IDs.  Previous contents are
 324  * discarded.
 325  * @param globalFilter OUTPUT parameter that receives a pointer to
 326  * a newly created global filter for this ID in this direction, or
 327  * NULL if there is none.
 328  * @return TRUE if the parse succeeds, that is, if the entire
 329  * id is consumed without syntax error.
 330  */
 331 UBool TransliteratorIDParser::parseCompoundID(const UnicodeString& id, int32_t dir,
 332                                               UnicodeString& canonID,
 333                                               UVector& list,
 334                                               UnicodeSet*& globalFilter) {
 335     UErrorCode ec = U_ZERO_ERROR;
 336     int32_t i;
 337     int32_t pos = 0;
 338     int32_t withParens = 1;
 339     list.removeAllElements();
 340     UnicodeSet* filter;
 341     globalFilter = NULL;
 342     canonID.truncate(0);
 343
 344     // Parse leading global filter, if any
 345     withParens = 0; // parens disallowed
 346     filter = parseGlobalFilter(id, pos, dir, withParens, &canonID);
 347     if (filter != NULL) {
 348         if (!ICU_Utility::parseChar(id, pos, ID_DELIM)) {
 349             // Not a global filter; backup and resume
 350             canonID.truncate(0);
 351             pos = 0;
 352         }
 353         if (dir == FORWARD) {
 354             globalFilter = filter;
 355         } else {
 356             delete filter;
 357         }
 358         filter = NULL;
 359     }
 360
 361     UBool sawDelimiter = TRUE;
 362     for (;;) {
 363         SingleID* single = parseSingleID(id, pos, dir, ec);
 364         if (single == NULL) {
 365             break;
 366         }
 367         if (dir == FORWARD) {
 368             list.addElement(single, ec);
 369         } else {
 370             list.insertElementAt(single, 0, ec);
 371         }
 372         if (U_FAILURE(ec)) {
 373             goto FAIL;
 374         }
 375         if (!ICU_Utility::parseChar(id, pos, ID_DELIM)) {
 376             sawDelimiter = FALSE;
 377             break;
 378         }
 379     }
 380
 381     if (list.size() == 0) {
 382         goto FAIL;
 383     }
 384
 385     // Construct canonical ID
 386     for (i=0; i<list.size(); ++i) {
 387         SingleID* single = (SingleID*) list.elementAt(i);
 388         canonID.append(single->canonID);
 389         if (i != (list.size()-1)) {
 390             canonID.append(ID_DELIM);
 391         }
 392     }
 393
 394     // Parse trailing global filter, if any, and only if we saw
 395     // a trailing delimiter after the IDs.
 396     if (sawDelimiter) {
 397         withParens = 1; // parens required
 398         filter = parseGlobalFilter(id, pos, dir, withParens, &canonID);
 399         if (filter != NULL) {
 400             // Don't require trailing ';', but parse it if present
 401             ICU_Utility::parseChar(id, pos, ID_DELIM);
 402
 403             if (dir == REVERSE) {
 404                 globalFilter = filter;
 405             } else {
 406                 delete filter;
 407             }
 408             filter = NULL;
 409         }
 410     }
 411
 412     // Trailing unparsed text is a syntax error
 413     ICU_Utility::skipWhitespace(id, pos, TRUE);
 414     if (pos != id.length()) {
 415         goto FAIL;
 416     }
 417
 418     return TRUE;
 419
 420  FAIL:
 421     UObjectDeleter *save = list.setDeleter(_deleteSingleID);
 422     list.removeAllElements();
 423     list.setDeleter(save);
 424     delete globalFilter;
 425     globalFilter = NULL;
 426     return FALSE;
 427 }
 428
 429 /**
 430  * Convert the elements of the 'list' vector, which are SingleID
 431  * objects, into actual Transliterator objects.  In the course of
 432  * this, some (or all) entries may be removed.  If all entries
 433  * are removed, the NULL transliterator will be added.
 434  *
 435  * Delete entries with empty basicIDs; these are generated by
 436  * elements like "(A)" in the forward direction, or "A()" in
 437  * the reverse.  THIS MAY RESULT IN AN EMPTY VECTOR.  Convert
 438  * SingleID entries to actual transliterators.
 439  *
 440  * @param list vector of SingleID objects.  On exit, vector
 441  * of one or more Transliterators.
 442  * @return new value of insertIndex.  The index will shift if
 443  * there are empty items, like "(Lower)", with indices less than
 444  * insertIndex.
 445  */
 446 void TransliteratorIDParser::instantiateList(UVector& list,
 447                                                 UErrorCode& ec) {
 448     UVector tlist(ec);
 449     if (U_FAILURE(ec)) {
 450         goto RETURN;
 451     }
 452     tlist.setDeleter(_deleteTransliteratorTrIDPars);
 453
 454     Transliterator* t;
 455     int32_t i;
 456     for (i=0; i<=list.size(); ++i) { // [sic]: i<=list.size()
 457         // We run the loop too long by one, so we can
 458         // do an insert after the last element
 459         if (i==list.size()) {
 460             break;
 461         }
 462
 463         SingleID* single = (SingleID*) list.elementAt(i);
 464         if (single->basicID.length() != 0) {
 465             t = single->createInstance();
 466             if (t == NULL) {
 467                 ec = U_INVALID_ID;
 468                 goto RETURN;
 469             }
 470             tlist.addElement(t, ec);
 471             if (U_FAILURE(ec)) {
 472                 delete t;
 473                 goto RETURN;
 474             }
 475         }
 476     }
 477
 478     // An empty list is equivalent to a NULL transliterator.
 479     if (tlist.size() == 0) {
 480         t = createBasicInstance(ANY_NULL, NULL);
 481         if (t == NULL) {
 482             // Should never happen
 483             ec = U_INTERNAL_TRANSLITERATOR_ERROR;
 484         }
 485         tlist.addElement(t, ec);
 486         if (U_FAILURE(ec)) {
 487             delete t;
 488         }
 489     }
 490
 491  RETURN:
 492
 493     UObjectDeleter *save = list.setDeleter(_deleteSingleID);
 494     list.removeAllElements();
 495
 496     if (U_SUCCESS(ec)) {
 497         list.setDeleter(_deleteTransliteratorTrIDPars);
 498
 499         while (tlist.size() > 0) {
 500             t = (Transliterator*) tlist.orphanElementAt(0);
 501             list.addElement(t, ec);
 502             if (U_FAILURE(ec)) {
 503                 delete t;
 504                 list.removeAllElements();
 505                 break;
 506             }
 507         }
 508     }
 509
 510     list.setDeleter(save);
 511 }
 512
 513 /**
 514  * Parse an ID into pieces.  Take IDs of the form T, T/V, S-T,
 515  * S-T/V, or S/V-T.  If the source is missing, return a source of
 516  * ANY.
 517  * @param id the id string, in any of several forms
 518  * @return an array of 4 strings: source, target, variant, and
 519  * isSourcePresent.  If the source is not present, ANY will be
 520  * given as the source, and isSourcePresent will be NULL.  Otherwise
 521  * isSourcePresent will be non-NULL.  The target may be empty if the
 522  * id is not well-formed.  The variant may be empty.
 523  */
 524 void TransliteratorIDParser::IDtoSTV(const UnicodeString& id,
 525                                      UnicodeString& source,
 526                                      UnicodeString& target,
 527                                      UnicodeString& variant,
 528                                      UBool& isSourcePresent) {
 529     source = ANY;
 530     target.truncate(0);
 531     variant.truncate(0);
 532
 533     int32_t sep = id.indexOf(TARGET_SEP);
 534     int32_t var = id.indexOf(VARIANT_SEP);
 535     if (var < 0) {
 536         var = id.length();
 537     }
 538     isSourcePresent = FALSE;
 539
 540     if (sep < 0) {
 541         // Form: T/V or T (or /V)
 542         id.extractBetween(0, var, target);
 543         id.extractBetween(var, id.length(), variant);
 544     } else if (sep < var) {
 545         // Form: S-T/V or S-T (or -T/V or -T)
 546         if (sep > 0) {
 547             id.extractBetween(0, sep, source);
 548             isSourcePresent = TRUE;
 549         }
 550         id.extractBetween(++sep, var, target);
 551         id.extractBetween(var, id.length(), variant);
 552     } else {
 553         // Form: (S/V-T or /V-T)
 554         if (var > 0) {
 555             id.extractBetween(0, var, source);
 556             isSourcePresent = TRUE;
 557         }
 558         id.extractBetween(var, sep++, variant);
 559         id.extractBetween(sep, id.length(), target);
 560     }
 561
 562     if (variant.length() > 0) {
 563         variant.remove(0, 1);
 564     }
 565 }
 566
 567 /**
 568  * Given source, target, and variant strings, concatenate them into a
 569  * full ID.  If the source is empty, then "Any" will be used for the
 570  * source, so the ID will always be of the form s-t/v or s-t.
 571  */
 572 void TransliteratorIDParser::STVtoID(const UnicodeString& source,
 573                                      const UnicodeString& target,
 574                                      const UnicodeString& variant,
 575                                      UnicodeString& id) {
 576     id = source;
 577     if (id.length() == 0) {
 578         id = ANY;
 579     }
 580     id.append(TARGET_SEP).append(target);
 581     if (variant.length() != 0) {
 582         id.append(VARIANT_SEP).append(variant);
 583     }
 584     // NUL-terminate the ID string for getTerminatedBuffer.
 585     // This prevents valgrind and Purify warnings.
 586     id.append((UChar)0);
 587     id.truncate(id.length()-1);
 588 }
 589
 590 /**
 591  * Register two targets as being inverses of one another.  For
 592  * example, calling registerSpecialInverse("NFC", "NFD", TRUE) causes
 593  * Transliterator to form the following inverse relationships:
 594  *
 595  * <pre>NFC => NFD
 596  * Any-NFC => Any-NFD
 597  * NFD => NFC
 598  * Any-NFD => Any-NFC</pre>
 599  *
 600  * (Without the special inverse registration, the inverse of NFC
 601  * would be NFC-Any.)  Note that NFD is shorthand for Any-NFD, but
 602  * that the presence or absence of "Any-" is preserved.
 603  *
 604  * <p>The relationship is symmetrical; registering (a, b) is
 605  * equivalent to registering (b, a).
 606  *
 607  * <p>The relevant IDs must still be registered separately as
 608  * factories or classes.
 609  *
 610  * <p>Only the targets are specified.  Special inverses always
 611  * have the form Any-Target1 <=> Any-Target2.  The target should
 612  * have canonical casing (the casing desired to be produced when
 613  * an inverse is formed) and should contain no whitespace or other
 614  * extraneous characters.
 615  *
 616  * @param target the target against which to register the inverse
 617  * @param inverseTarget the inverse of target, that is
 618  * Any-target.getInverse() => Any-inverseTarget
 619  * @param bidirectional if TRUE, register the reverse relation
 620  * as well, that is, Any-inverseTarget.getInverse() => Any-target
 621  */
 622 void TransliteratorIDParser::registerSpecialInverse(const UnicodeString& target,
 623                                                     const UnicodeString& inverseTarget,
 624                                                     UBool bidirectional,
 625                                                     UErrorCode &status) {
 626     init(status);
 627     if (U_FAILURE(status)) {
 628         return;
 629     }
 630
 631     // If target == inverseTarget then force bidirectional => FALSE
 632     if (bidirectional && 0==target.caseCompare(inverseTarget, U_FOLD_CASE_DEFAULT)) {
 633         bidirectional = FALSE;
 634     }
 635
 636     umtx_init(&LOCK);
 637     Mutex lock(&LOCK);
 638
 639     SPECIAL_INVERSES->put(target, new UnicodeString(inverseTarget), status);
 640     if (bidirectional) {
 641         SPECIAL_INVERSES->put(inverseTarget, new UnicodeString(target), status);
 642     }
 643 }
 644
 645 //----------------------------------------------------------------
 646 // Private implementation
 647 //----------------------------------------------------------------
 648
 649 /**
 650  * Parse an ID into component pieces.  Take IDs of the form T,
 651  * T/V, S-T, S-T/V, or S/V-T.  If the source is missing, return a
 652  * source of ANY.
 653  * @param id the id string, in any of several forms
 654  * @param pos INPUT-OUTPUT parameter.  On input, pos is the
 655  * offset of the first character to parse in id.  On output,
 656  * pos is the offset after the last parsed character.  If the
 657  * parse failed, pos will be unchanged.
 658  * @param allowFilter2 if TRUE, a UnicodeSet pattern is allowed
 659  * at any location between specs or delimiters, and is returned
 660  * as the fifth string in the array.
 661  * @return a Specs object, or NULL if the parse failed.  If
 662  * neither source nor target was seen in the parsed id, then the
 663  * parse fails.  If allowFilter is TRUE, then the parsed filter
 664  * pattern is returned in the Specs object, otherwise the returned
 665  * filter reference is NULL.  If the parse fails for any reason
 666  * NULL is returned.
 667  */
 668 TransliteratorIDParser::Specs*
 669 TransliteratorIDParser::parseFilterID(const UnicodeString& id, int32_t& pos,
 670                                       UBool allowFilter) {
 671     UnicodeString first;
 672     UnicodeString source;
 673     UnicodeString target;
 674     UnicodeString variant;
 675     UnicodeString filter;
 676     UChar delimiter = 0;
 677     int32_t specCount = 0;
 678     int32_t start = pos;
 679
 680     // This loop parses one of the following things with each
 681     // pass: a filter, a delimiter character (either '-' or '/'),
 682     // or a spec (source, target, or variant).
 683     for (;;) {
 684         ICU_Utility::skipWhitespace(id, pos, TRUE);
 685         if (pos == id.length()) {
 686             break;
 687         }
 688
 689         // Parse filters
 690         if (allowFilter && filter.length() == 0 &&
 691             UnicodeSet::resemblesPattern(id, pos)) {
 692
 693             ParsePosition ppos(pos);
 694             UErrorCode ec = U_ZERO_ERROR;
 695             UnicodeSet set(id, ppos, USET_IGNORE_SPACE, NULL, ec);
 696             if (U_FAILURE(ec)) {
 697                 pos = start;
 698                 return NULL;
 699             }
 700             id.extractBetween(pos, ppos.getIndex(), filter);
 701             pos = ppos.getIndex();
 702             continue;
 703         }
 704
 705         if (delimiter == 0) {
 706             UChar c = id.charAt(pos);
 707             if ((c == TARGET_SEP && target.length() == 0) ||
 708                 (c == VARIANT_SEP && variant.length() == 0)) {
 709                 delimiter = c;
 710                 ++pos;
 711                 continue;
 712             }
 713         }
 714
 715         // We are about to try to parse a spec with no delimiter
 716         // when we can no longer do so (we can only do so at the
 717         // start); break.
 718         if (delimiter == 0 && specCount > 0) {
 719             break;
 720         }
 721
 722         UnicodeString spec = ICU_Utility::parseUnicodeIdentifier(id, pos);
 723         if (spec.length() == 0) {
 724             // Note that if there was a trailing delimiter, we
 725             // consume it.  So Foo-, Foo/, Foo-Bar/, and Foo/Bar-
 726             // are legal.
 727             break;
 728         }
 729
 730         switch (delimiter) {
 731         case 0:
 732             first = spec;
 733             break;
 734         case TARGET_SEP:
 735             target = spec;
 736             break;
 737         case VARIANT_SEP:
 738             variant = spec;
 739             break;
 740         }
 741         ++specCount;
 742         delimiter = 0;
 743     }
 744
 745     // A spec with no prior character is either source or target,
 746     // depending on whether an explicit "-target" was seen.
 747     if (first.length() != 0) {
 748         if (target.length() == 0) {
 749             target = first;
 750         } else {
 751             source = first;
 752         }
 753     }
 754
 755     // Must have either source or target
 756     if (source.length() == 0 && target.length() == 0) {
 757         pos = start;
 758         return NULL;
 759     }
 760
 761     // Empty source or target defaults to ANY
 762     UBool sawSource = TRUE;
 763     if (source.length() == 0) {
 764         source = ANY;
 765         sawSource = FALSE;
 766     }
 767     if (target.length() == 0) {
 768         target = ANY;
 769     }
 770
 771     return new Specs(source, target, variant, sawSource, filter);
 772 }
 773
 774 /**
 775  * Givens a Spec object, convert it to a SingleID object.  The
 776  * Spec object is a more unprocessed parse result.  The SingleID
 777  * object contains information about canonical and basic IDs.
 778  * @return a SingleID; never returns NULL.  Returned object always
 779  * has 'filter' field of NULL.
 780  */
 781 TransliteratorIDParser::SingleID*
 782 TransliteratorIDParser::specsToID(const Specs* specs, int32_t dir) {
 783     UnicodeString canonID;
 784     UnicodeString basicID;
 785     UnicodeString basicPrefix;
 786     if (specs != NULL) {
 787         UnicodeString buf;
 788         if (dir == FORWARD) {
 789             if (specs->sawSource) {
 790                 buf.append(specs->source).append(TARGET_SEP);
 791             } else {
 792                 basicPrefix = specs->source;
 793                 basicPrefix.append(TARGET_SEP);
 794             }
 795             buf.append(specs->target);
 796         } else {
 797             buf.append(specs->target).append(TARGET_SEP).append(specs->source);
 798         }
 799         if (specs->variant.length() != 0) {
 800             buf.append(VARIANT_SEP).append(specs->variant);
 801         }
 802         basicID = basicPrefix;
 803         basicID.append(buf);
 804         if (specs->filter.length() != 0) {
 805             buf.insert(0, specs->filter);
 806         }
 807         canonID = buf;
 808     }
 809     return new SingleID(canonID, basicID);
 810 }
 811
 812 /**
 813  * Given a Specs object, return a SingleID representing the
 814  * special inverse of that ID.  If there is no special inverse
 815  * then return NULL.
 816  * @return a SingleID or NULL.  Returned object always has
 817  * 'filter' field of NULL.
 818  */
 819 TransliteratorIDParser::SingleID*
 820 TransliteratorIDParser::specsToSpecialInverse(const Specs& specs, UErrorCode &status) {
 821     if (0!=specs.source.caseCompare(ANY, U_FOLD_CASE_DEFAULT)) {
 822         return NULL;
 823     }
 824     init(status);
 825
 826     UnicodeString* inverseTarget;
 827
 828     umtx_init(&LOCK);
 829     umtx_lock(&LOCK);
 830     inverseTarget = (UnicodeString*) SPECIAL_INVERSES->get(specs.target);
 831     umtx_unlock(&LOCK);
 832
 833     if (inverseTarget != NULL) {
 834         // If the original ID contained "Any-" then make the
 835         // special inverse "Any-Foo"; otherwise make it "Foo".
 836         // So "Any-NFC" => "Any-NFD" but "NFC" => "NFD".
 837         UnicodeString buf;
 838         if (specs.filter.length() != 0) {
 839             buf.append(specs.filter);
 840         }
 841         if (specs.sawSource) {
 842             buf.append(ANY).append(TARGET_SEP);
 843         }
 844         buf.append(*inverseTarget);
 845
 846         UnicodeString basicID(ANY);
 847         basicID.append(TARGET_SEP).append(*inverseTarget);
 848
 849         if (specs.variant.length() != 0) {
 850             buf.append(VARIANT_SEP).append(specs.variant);
 851             basicID.append(VARIANT_SEP).append(specs.variant);
 852         }
 853         return new SingleID(buf, basicID);
 854     }
 855     return NULL;
 856 }
 857
 858 /**
 859  * Glue method to get around access problems in C++.  This would
 860  * ideally be inline but we want to avoid a circular header
 861  * dependency.
 862  */
 863 Transliterator* TransliteratorIDParser::createBasicInstance(const UnicodeString& id, const UnicodeString* canonID) {
 864     return Transliterator::createBasicInstance(id, canonID);
 865 }
 866
 867 /**
 868  * Initialize static memory.
 869  */
 870 void TransliteratorIDParser::init(UErrorCode &status) {
 871     if (SPECIAL_INVERSES != NULL) {
 872         return;
 873     }
 874
 875     Hashtable* special_inverses = new Hashtable(TRUE, status);
 876     special_inverses->setValueDeleter(uhash_deleteUnicodeString);
 877
 878     umtx_init(&LOCK);
 879     umtx_lock(&LOCK);
 880     if (SPECIAL_INVERSES == NULL) {
 881         SPECIAL_INVERSES = special_inverses;
 882         special_inverses = NULL;
 883     }
 884     umtx_unlock(&LOCK);
 885     delete special_inverses; /*null instance*/
 886
 887     ucln_i18n_registerCleanup(UCLN_I18N_TRANSLITERATOR, transliterator_cleanup);
 888 }
 889
 890 /**
 891  * Free static memory.
 892  */
 893 void TransliteratorIDParser::cleanup() {
 894     if (SPECIAL_INVERSES) {
 895         delete SPECIAL_INVERSES;
 896         SPECIAL_INVERSES = NULL;
 897     }
 898     umtx_destroy(&LOCK);
 899 }
 900
 901 U_NAMESPACE_END
 902
 903 #endif /* #if !UCONFIG_NO_TRANSLITERATION */
 904
 905 //eof