icuSources/i18n/smpdtfmt.cpp

   1 /*
   2 *******************************************************************************
   3 * Copyright (C) 1997-2006, International Business Machines Corporation and    *
   4 * others. All Rights Reserved.                                                *
   5 *******************************************************************************
   6 *
   7 * File SMPDTFMT.CPP
   8 *
   9 * Modification History:
  10 *
  11 *   Date        Name        Description
  12 *   02/19/97    aliu        Converted from java.
  13 *   03/31/97    aliu        Modified extensively to work with 50 locales.
  14 *   04/01/97    aliu        Added support for centuries.
  15 *   07/09/97    helena      Made ParsePosition into a class.
  16 *   07/21/98    stephen     Added initializeDefaultCentury.
  17 *                             Removed getZoneIndex (added in DateFormatSymbols)
  18 *                             Removed subParseLong
  19 *                             Removed chk
  20 *  02/22/99     stephen     Removed character literals for EBCDIC safety
  21 *   10/14/99    aliu        Updated 2-digit year parsing so that only "00" thru
  22 *                           "99" are recognized. {j28 4182066}
  23 *   11/15/99    weiv        Added support for week of year/day of week format
  24 ********************************************************************************
  25 */
  26
  27 #include "unicode/utypes.h"
  28
  29 #if !UCONFIG_NO_FORMATTING
  30
  31 #include "unicode/smpdtfmt.h"
  32 #include "unicode/dtfmtsym.h"
  33 #include "unicode/ures.h"
  34 #include "unicode/msgfmt.h"
  35 #include "unicode/calendar.h"
  36 #include "unicode/gregocal.h"
  37 #include "unicode/timezone.h"
  38 #include "unicode/decimfmt.h"
  39 #include "unicode/dcfmtsym.h"
  40 #include "unicode/uchar.h"
  41 #include "unicode/ustring.h"
  42 #include "util.h"
  43 #include "gregoimp.h"
  44 #include "cstring.h"
  45 #include "uassert.h"
  46 #include <float.h>
  47
  48 #if defined( U_DEBUG_CALSVC ) || defined (U_DEBUG_CAL)
  49 #include <stdio.h>
  50 #endif
  51
  52 // *****************************************************************************
  53 // class SimpleDateFormat
  54 // *****************************************************************************
  55
  56 U_NAMESPACE_BEGIN
  57
  58 /**
  59  * Last-resort string to use for "GMT" when constructing time zone strings.
  60  */
  61 // For time zones that have no names, use strings GMT+minutes and
  62 // GMT-minutes. For instance, in France the time zone is GMT+60.
  63 // Also accepted are GMT+H:MM or GMT-H:MM.
  64 static const UChar gGmt[]      = {0x0047, 0x004D, 0x0054, 0x0000};         // "GMT"
  65 static const UChar gGmtPlus[]  = {0x0047, 0x004D, 0x0054, 0x002B, 0x0000}; // "GMT+"
  66 static const UChar gGmtMinus[] = {0x0047, 0x004D, 0x0054, 0x002D, 0x0000}; // "GMT-"
  67
  68 // This is a pattern-of-last-resort used when we can't load a usable pattern out
  69 // of a resource.
  70 static const UChar gDefaultPattern[] =
  71 {
  72     0x79, 0x79, 0x79, 0x79, 0x4D, 0x4D, 0x64, 0x64, 0x20, 0x68, 0x68, 0x3A, 0x6D, 0x6D, 0x20, 0x61, 0
  73 };  /* "yyyyMMdd hh:mm a" */
  74
  75 // This prefix is designed to NEVER MATCH real text, in order to
  76 // suppress the parsing of negative numbers.  Adjust as needed (if
  77 // this becomes valid Unicode).
  78 static const UChar SUPPRESS_NEGATIVE_PREFIX[] = {0xAB00, 0};
  79
  80 /**
  81  * These are the tags we expect to see in normal resource bundle files associated
  82  * with a locale.
  83  */
  84 static const char gDateTimePatternsTag[]="DateTimePatterns";
  85
  86 UOBJECT_DEFINE_RTTI_IMPLEMENTATION(SimpleDateFormat)
  87
  88 static const UChar QUOTE = 0x27; // Single quote
  89
  90 //----------------------------------------------------------------------
  91
  92 SimpleDateFormat::~SimpleDateFormat()
  93 {
  94     delete fSymbols;
  95     delete parsedTimeZone; // sanity check
  96 }
  97
  98 //----------------------------------------------------------------------
  99
 100 SimpleDateFormat::SimpleDateFormat(UErrorCode& status)
 101   :   fLocale(Locale::getDefault()),
 102       fSymbols(NULL),
 103       parsedTimeZone(NULL)
 104 {
 105     construct(kShort, (EStyle) (kShort + kDateOffset), fLocale, status);
 106     initializeDefaultCentury();
 107 }
 108
 109 //----------------------------------------------------------------------
 110
 111 SimpleDateFormat::SimpleDateFormat(const UnicodeString& pattern,
 112                                    UErrorCode &status)
 113 :   fPattern(pattern),
 114     fLocale(Locale::getDefault()),
 115     fSymbols(NULL),
 116     parsedTimeZone(NULL)
 117 {
 118     initializeSymbols(fLocale, initializeCalendar(NULL,fLocale,status), status);
 119     initialize(fLocale, status);
 120     initializeDefaultCentury();
 121 }
 122
 123 //----------------------------------------------------------------------
 124
 125 SimpleDateFormat::SimpleDateFormat(const UnicodeString& pattern,
 126                                    const Locale& locale,
 127                                    UErrorCode& status)
 128 :   fPattern(pattern),
 129     fLocale(locale),
 130     parsedTimeZone(NULL)
 131 {
 132     initializeSymbols(fLocale, initializeCalendar(NULL,fLocale,status), status);
 133     initialize(fLocale, status);
 134     initializeDefaultCentury();
 135 }
 136
 137 //----------------------------------------------------------------------
 138
 139 SimpleDateFormat::SimpleDateFormat(const UnicodeString& pattern,
 140                                    DateFormatSymbols* symbolsToAdopt,
 141                                    UErrorCode& status)
 142 :   fPattern(pattern),
 143     fLocale(Locale::getDefault()),
 144     fSymbols(symbolsToAdopt),
 145     parsedTimeZone(NULL)
 146 {
 147     initializeCalendar(NULL,fLocale,status);
 148     initialize(fLocale, status);
 149     initializeDefaultCentury();
 150 }
 151
 152 //----------------------------------------------------------------------
 153
 154 SimpleDateFormat::SimpleDateFormat(const UnicodeString& pattern,
 155                                    const DateFormatSymbols& symbols,
 156                                    UErrorCode& status)
 157 :   fPattern(pattern),
 158     fLocale(Locale::getDefault()),
 159     fSymbols(new DateFormatSymbols(symbols)),
 160     parsedTimeZone(NULL)
 161 {
 162     initializeCalendar(NULL, fLocale, status);
 163     initialize(fLocale, status);
 164     initializeDefaultCentury();
 165 }
 166
 167 //----------------------------------------------------------------------
 168
 169 // Not for public consumption; used by DateFormat
 170 SimpleDateFormat::SimpleDateFormat(EStyle timeStyle,
 171                                    EStyle dateStyle,
 172                                    const Locale& locale,
 173                                    UErrorCode& status)
 174 :   fLocale(locale),
 175     fSymbols(NULL),
 176     parsedTimeZone(NULL)
 177 {
 178     construct(timeStyle, dateStyle, fLocale, status);
 179     if(U_SUCCESS(status)) {
 180       initializeDefaultCentury();
 181     }
 182 }
 183
 184 //----------------------------------------------------------------------
 185
 186 /**
 187  * Not for public consumption; used by DateFormat.  This constructor
 188  * never fails.  If the resource data is not available, it uses the
 189  * the last resort symbols.
 190  */
 191 SimpleDateFormat::SimpleDateFormat(const Locale& locale,
 192                                    UErrorCode& status)
 193 :   fPattern(gDefaultPattern),
 194     fLocale(locale),
 195     fSymbols(NULL),
 196     parsedTimeZone(NULL)
 197 {
 198     if (U_FAILURE(status)) return;
 199     initializeSymbols(fLocale, initializeCalendar(NULL, fLocale, status),status);
 200     if (U_FAILURE(status))
 201     {
 202         status = U_ZERO_ERROR;
 203         delete fSymbols;
 204         // This constructor doesn't fail; it uses last resort data
 205         fSymbols = new DateFormatSymbols(status);
 206         /* test for NULL */
 207         if (fSymbols == 0) {
 208             status = U_MEMORY_ALLOCATION_ERROR;
 209             return;
 210         }
 211     }
 212
 213     initialize(fLocale, status);
 214     if(U_SUCCESS(status)) {
 215       initializeDefaultCentury();
 216     }
 217 }
 218
 219 //----------------------------------------------------------------------
 220
 221 SimpleDateFormat::SimpleDateFormat(const SimpleDateFormat& other)
 222 :   DateFormat(other),
 223     fSymbols(NULL),
 224     parsedTimeZone(NULL)
 225 {
 226     *this = other;
 227 }
 228
 229 //----------------------------------------------------------------------
 230
 231 SimpleDateFormat& SimpleDateFormat::operator=(const SimpleDateFormat& other)
 232 {
 233     DateFormat::operator=(other);
 234
 235     delete fSymbols;
 236     fSymbols = NULL;
 237
 238     delete parsedTimeZone; parsedTimeZone = NULL;
 239
 240     if (other.fSymbols)
 241         fSymbols = new DateFormatSymbols(*other.fSymbols);
 242
 243     fDefaultCenturyStart         = other.fDefaultCenturyStart;
 244     fDefaultCenturyStartYear     = other.fDefaultCenturyStartYear;
 245     fHaveDefaultCentury          = other.fHaveDefaultCentury;
 246
 247     fPattern = other.fPattern;
 248
 249     return *this;
 250 }
 251
 252 //----------------------------------------------------------------------
 253
 254 Format*
 255 SimpleDateFormat::clone() const
 256 {
 257     return new SimpleDateFormat(*this);
 258 }
 259
 260 //----------------------------------------------------------------------
 261
 262 UBool
 263 SimpleDateFormat::operator==(const Format& other) const
 264 {
 265     if (DateFormat::operator==(other)) {
 266         // DateFormat::operator== guarantees following cast is safe
 267         SimpleDateFormat* that = (SimpleDateFormat*)&other;
 268         return (fPattern             == that->fPattern &&
 269                 fSymbols             != NULL && // Check for pathological object
 270                 that->fSymbols       != NULL && // Check for pathological object
 271                 *fSymbols            == *that->fSymbols &&
 272                 fHaveDefaultCentury  == that->fHaveDefaultCentury &&
 273                 fDefaultCenturyStart == that->fDefaultCenturyStart);
 274     }
 275     return FALSE;
 276 }
 277
 278 //----------------------------------------------------------------------
 279
 280 void SimpleDateFormat::construct(EStyle timeStyle,
 281                                  EStyle dateStyle,
 282                                  const Locale& locale,
 283                                  UErrorCode& status)
 284 {
 285     // called by several constructors to load pattern data from the resources
 286     if (U_FAILURE(status)) return;
 287
 288     // We will need the calendar to know what type of symbols to load.
 289     initializeCalendar(NULL, locale, status);
 290     if (U_FAILURE(status)) return;
 291
 292     CalendarData calData(locale, fCalendar?fCalendar->getType():NULL, status);
 293     UResourceBundle *dateTimePatterns = calData.getByKey(gDateTimePatternsTag, status);
 294     if (U_FAILURE(status)) return;
 295
 296     if (ures_getSize(dateTimePatterns) <= kDateTime)
 297     {
 298         status = U_INVALID_FORMAT_ERROR;
 299         return;
 300     }
 301
 302     setLocaleIDs(ures_getLocaleByType(dateTimePatterns, ULOC_VALID_LOCALE, &status),
 303                  ures_getLocaleByType(dateTimePatterns, ULOC_ACTUAL_LOCALE, &status));
 304
 305     // create a symbols object from the locale
 306     initializeSymbols(locale,fCalendar, status);
 307     if (U_FAILURE(status)) return;
 308     /* test for NULL */
 309     if (fSymbols == 0) {
 310         status = U_MEMORY_ALLOCATION_ERROR;
 311         return;
 312     }
 313
 314     const UChar *resStr;
 315     int32_t resStrLen = 0;
 316
 317     // if the pattern should include both date and time information, use the date/time
 318     // pattern string as a guide to tell use how to glue together the appropriate date
 319     // and time pattern strings.  The actual gluing-together is handled by a convenience
 320     // method on MessageFormat.
 321     if ((timeStyle != kNone) && (dateStyle != kNone))
 322     {
 323         Formattable timeDateArray[2];
 324
 325         // use Formattable::adoptString() so that we can use fastCopyFrom()
 326         // instead of Formattable::setString()'s unaware, safe, deep string clone
 327         // see Jitterbug 2296
 328         resStr = ures_getStringByIndex(dateTimePatterns, (int32_t)timeStyle, &resStrLen, &status);
 329         timeDateArray[0].adoptString(new UnicodeString(TRUE, resStr, resStrLen));
 330         resStr = ures_getStringByIndex(dateTimePatterns, (int32_t)dateStyle, &resStrLen, &status);
 331         timeDateArray[1].adoptString(new UnicodeString(TRUE, resStr, resStrLen));
 332
 333         resStr = ures_getStringByIndex(dateTimePatterns, (int32_t)kDateTime, &resStrLen, &status);
 334         MessageFormat::format(UnicodeString(TRUE, resStr, resStrLen), timeDateArray, 2, fPattern, status);
 335     }
 336     // if the pattern includes just time data or just date date, load the appropriate
 337     // pattern string from the resources
 338     // setTo() - see DateFormatSymbols::assignArray comments
 339     else if (timeStyle != kNone) {
 340         resStr = ures_getStringByIndex(dateTimePatterns, (int32_t)timeStyle, &resStrLen, &status);
 341         fPattern.setTo(TRUE, resStr, resStrLen);
 342     }
 343     else if (dateStyle != kNone) {
 344         resStr = ures_getStringByIndex(dateTimePatterns, (int32_t)dateStyle, &resStrLen, &status);
 345         fPattern.setTo(TRUE, resStr, resStrLen);
 346     }
 347
 348     // and if it includes _neither_, that's an error
 349     else
 350         status = U_INVALID_FORMAT_ERROR;
 351
 352     // finally, finish initializing by creating a Calendar and a NumberFormat
 353     initialize(locale, status);
 354 }
 355
 356 //----------------------------------------------------------------------
 357
 358 Calendar*
 359 SimpleDateFormat::initializeCalendar(TimeZone* adoptZone, const Locale& locale, UErrorCode& status)
 360 {
 361     if(!U_FAILURE(status)) {
 362         fCalendar = Calendar::createInstance(adoptZone?adoptZone:TimeZone::createDefault(), locale, status);
 363     }
 364     if (U_SUCCESS(status) && fCalendar == NULL) {
 365         status = U_MEMORY_ALLOCATION_ERROR;
 366     }
 367     return fCalendar;
 368 }
 369
 370 void
 371 SimpleDateFormat::initializeSymbols(const Locale& locale, Calendar* calendar, UErrorCode& status)
 372 {
 373   if(U_FAILURE(status)) {
 374     fSymbols = NULL;
 375   } else {
 376     // pass in calendar type - use NULL (default) if no calendar set (or err).
 377     fSymbols = new DateFormatSymbols(locale, calendar?calendar->getType() :NULL , status);
 378   }
 379 }
 380
 381 void
 382 SimpleDateFormat::initialize(const Locale& locale,
 383                              UErrorCode& status)
 384 {
 385     if (U_FAILURE(status)) return;
 386
 387     // We don't need to check that the row count is >= 1, since all 2d arrays have at
 388     // least one row
 389     fNumberFormat = NumberFormat::createInstance(locale, status);
 390     if (fNumberFormat != NULL && U_SUCCESS(status))
 391     {
 392         // no matter what the locale's default number format looked like, we want
 393         // to modify it so that it doesn't use thousands separators, doesn't always
 394         // show the decimal point, and recognizes integers only when parsing
 395
 396         fNumberFormat->setGroupingUsed(FALSE);
 397         if (fNumberFormat->getDynamicClassID() == DecimalFormat::getStaticClassID())
 398             ((DecimalFormat*)fNumberFormat)->setDecimalSeparatorAlwaysShown(FALSE);
 399         fNumberFormat->setParseIntegerOnly(TRUE);
 400         fNumberFormat->setMinimumFractionDigits(0); // To prevent "Jan 1.00, 1997.00"
 401     }
 402     else if (U_SUCCESS(status))
 403     {
 404         status = U_MISSING_RESOURCE_ERROR;
 405     }
 406 }
 407
 408 /* Initialize the fields we use to disambiguate ambiguous years. Separate
 409  * so we can call it from readObject().
 410  */
 411 void SimpleDateFormat::initializeDefaultCentury()
 412 {
 413   if(fCalendar) {
 414     fHaveDefaultCentury = fCalendar->haveDefaultCentury();
 415     if(fHaveDefaultCentury) {
 416       fDefaultCenturyStart = fCalendar->defaultCenturyStart();
 417       fDefaultCenturyStartYear = fCalendar->defaultCenturyStartYear();
 418     } else {
 419       fDefaultCenturyStart = DBL_MIN;
 420       fDefaultCenturyStartYear = -1;
 421     }
 422   }
 423 }
 424
 425 /* Define one-century window into which to disambiguate dates using
 426  * two-digit years. Make public in JDK 1.2.
 427  */
 428 void SimpleDateFormat::parseAmbiguousDatesAsAfter(UDate startDate, UErrorCode& status)
 429 {
 430     if(U_FAILURE(status)) {
 431         return;
 432     }
 433     if(!fCalendar) {
 434       status = U_ILLEGAL_ARGUMENT_ERROR;
 435       return;
 436     }
 437
 438     fCalendar->setTime(startDate, status);
 439     if(U_SUCCESS(status)) {
 440         fHaveDefaultCentury = TRUE;
 441         fDefaultCenturyStart = startDate;
 442         fDefaultCenturyStartYear = fCalendar->get(UCAL_YEAR, status);
 443     }
 444 }
 445
 446 //----------------------------------------------------------------------
 447
 448 UnicodeString&
 449 SimpleDateFormat::format(Calendar& cal, UnicodeString& appendTo, FieldPosition& pos) const
 450 {
 451     UErrorCode status = U_ZERO_ERROR;
 452     pos.setBeginIndex(0);
 453     pos.setEndIndex(0);
 454
 455     UBool inQuote = FALSE;
 456     UChar prevCh = 0;
 457     int32_t count = 0;
 458
 459     // loop through the pattern string character by character
 460     for (int32_t i = 0; i < fPattern.length() && U_SUCCESS(status); ++i) {
 461         UChar ch = fPattern[i];
 462
 463         // Use subFormat() to format a repeated pattern character
 464         // when a different pattern or non-pattern character is seen
 465         if (ch != prevCh && count > 0) {
 466             subFormat(appendTo, prevCh, count, pos, cal, status);
 467             count = 0;
 468         }
 469         if (ch == QUOTE) {
 470             // Consecutive single quotes are a single quote literal,
 471             // either outside of quotes or between quotes
 472             if ((i+1) < fPattern.length() && fPattern[i+1] == QUOTE) {
 473                 appendTo += (UChar)QUOTE;
 474                 ++i;
 475             } else {
 476                 inQuote = ! inQuote;
 477             }
 478         }
 479         else if ( ! inQuote && ((ch >= 0x0061 /*'a'*/ && ch <= 0x007A /*'z'*/)
 480                     || (ch >= 0x0041 /*'A'*/ && ch <= 0x005A /*'Z'*/))) {
 481             // ch is a date-time pattern character to be interpreted
 482             // by subFormat(); count the number of times it is repeated
 483             prevCh = ch;
 484             ++count;
 485         }
 486         else {
 487             // Append quoted characters and unquoted non-pattern characters
 488             appendTo += ch;
 489         }
 490     }
 491
 492     // Format the last item in the pattern, if any
 493     if (count > 0) {
 494         subFormat(appendTo, prevCh, count, pos, cal, status);
 495     }
 496
 497     // and if something failed (e.g., an invalid format character), reset our FieldPosition
 498     // to (0, 0) to show that
 499     // {sfb} look at this later- are these being set correctly?
 500     if (U_FAILURE(status)) {
 501         pos.setBeginIndex(0);
 502         pos.setEndIndex(0);
 503     }
 504
 505     return appendTo;
 506 }
 507
 508 UnicodeString&
 509 SimpleDateFormat::format(const Formattable& obj,
 510                          UnicodeString& appendTo,
 511                          FieldPosition& pos,
 512                          UErrorCode& status) const
 513 {
 514     // this is just here to get around the hiding problem
 515     // (the previous format() override would hide the version of
 516     // format() on DateFormat that this function correspond to, so we
 517     // have to redefine it here)
 518     return DateFormat::format(obj, appendTo, pos, status);
 519 }
 520
 521 //----------------------------------------------------------------------
 522
 523 // Map index into pattern character string to Calendar field number.
 524 const UCalendarDateFields
 525 SimpleDateFormat::fgPatternIndexToCalendarField[] =
 526 {
 527     /*GyM*/ UCAL_ERA, UCAL_YEAR, UCAL_MONTH,
 528     /*dkH*/ UCAL_DATE, UCAL_HOUR_OF_DAY, UCAL_HOUR_OF_DAY,
 529     /*msS*/ UCAL_MINUTE, UCAL_SECOND, UCAL_MILLISECOND,
 530     /*EDF*/ UCAL_DAY_OF_WEEK, UCAL_DAY_OF_YEAR, UCAL_DAY_OF_WEEK_IN_MONTH,
 531     /*wWa*/ UCAL_WEEK_OF_YEAR, UCAL_WEEK_OF_MONTH, UCAL_AM_PM,
 532     /*hKz*/ UCAL_HOUR, UCAL_HOUR, UCAL_ZONE_OFFSET,
 533     /*Yeu*/ UCAL_YEAR_WOY, UCAL_DOW_LOCAL, UCAL_EXTENDED_YEAR,
 534     /*gAZ*/ UCAL_JULIAN_DAY, UCAL_MILLISECONDS_IN_DAY, UCAL_ZONE_OFFSET,
 535     /*v*/   UCAL_ZONE_OFFSET,
 536     /*c*/   UCAL_DAY_OF_WEEK,
 537     /*L*/   UCAL_MONTH,
 538     /*Q*/   UCAL_MONTH,
 539     /*q*/   UCAL_MONTH,
 540 };
 541
 542 // Map index into pattern character string to DateFormat field number
 543 const UDateFormatField
 544 SimpleDateFormat::fgPatternIndexToDateFormatField[] = {
 545     /*GyM*/ UDAT_ERA_FIELD, UDAT_YEAR_FIELD, UDAT_MONTH_FIELD,
 546     /*dkH*/ UDAT_DATE_FIELD, UDAT_HOUR_OF_DAY1_FIELD, UDAT_HOUR_OF_DAY0_FIELD,
 547     /*msS*/ UDAT_MINUTE_FIELD, UDAT_SECOND_FIELD, UDAT_FRACTIONAL_SECOND_FIELD,
 548     /*EDF*/ UDAT_DAY_OF_WEEK_FIELD, UDAT_DAY_OF_YEAR_FIELD, UDAT_DAY_OF_WEEK_IN_MONTH_FIELD,
 549     /*wWa*/ UDAT_WEEK_OF_YEAR_FIELD, UDAT_WEEK_OF_MONTH_FIELD, UDAT_AM_PM_FIELD,
 550     /*hKz*/ UDAT_HOUR1_FIELD, UDAT_HOUR0_FIELD, UDAT_TIMEZONE_FIELD,
 551     /*Yeu*/ UDAT_YEAR_WOY_FIELD, UDAT_DOW_LOCAL_FIELD, UDAT_EXTENDED_YEAR_FIELD,
 552     /*gAZ*/ UDAT_JULIAN_DAY_FIELD, UDAT_MILLISECONDS_IN_DAY_FIELD, UDAT_TIMEZONE_RFC_FIELD,
 553     /*v*/   UDAT_TIMEZONE_GENERIC_FIELD,
 554     /*c*/   UDAT_STANDALONE_DAY_FIELD,
 555     /*L*/   UDAT_STANDALONE_MONTH_FIELD,
 556     /*Q*/   UDAT_QUARTER_FIELD,
 557     /*q*/   UDAT_STANDALONE_QUARTER_FIELD,
 558 };
 559
 560 //----------------------------------------------------------------------
 561
 562 /**
 563  * Append symbols[value] to dst.  Make sure the array index is not out
 564  * of bounds.
 565  */
 566 static inline void
 567 _appendSymbol(UnicodeString& dst,
 568               int32_t value,
 569               const UnicodeString* symbols,
 570               int32_t symbolsCount) {
 571     U_ASSERT(0 <= value && value < symbolsCount);
 572     if (0 <= value && value < symbolsCount) {
 573         dst += symbols[value];
 574     }
 575 }
 576
 577 //---------------------------------------------------------------------
 578 inline void SimpleDateFormat::appendGMT(UnicodeString &appendTo, Calendar& cal, UErrorCode& status) const{
 579     int32_t value = cal.get(UCAL_ZONE_OFFSET, status) +
 580     cal.get(UCAL_DST_OFFSET, status);
 581
 582     if (value < 0) {
 583         appendTo += gGmtMinus;
 584         value = -value; // suppress the '-' sign for text display.
 585     }else{
 586         appendTo += gGmtPlus;
 587     }
 588
 589     zeroPaddingNumber(appendTo, (int32_t)(value/U_MILLIS_PER_HOUR), 2, 2);
 590     appendTo += (UChar)0x003A /*':'*/;
 591     zeroPaddingNumber(appendTo, (int32_t)((value%U_MILLIS_PER_HOUR)/U_MILLIS_PER_MINUTE), 2, 2);
 592 }
 593
 594 //---------------------------------------------------------------------
 595 void
 596 SimpleDateFormat::subFormat(UnicodeString &appendTo,
 597                             UChar ch,
 598                             int32_t count,
 599                             FieldPosition& pos,
 600                             Calendar& cal,
 601                             UErrorCode& status) const
 602 {
 603     if (U_FAILURE(status)) {
 604         return;
 605     }
 606
 607     // this function gets called by format() to produce the appropriate substitution
 608     // text for an individual pattern symbol (e.g., "HH" or "yyyy")
 609
 610     UChar *patternCharPtr = u_strchr(DateFormatSymbols::getPatternUChars(), ch);
 611     UDateFormatField patternCharIndex;
 612     const int32_t maxIntCount = 10;
 613     int32_t beginOffset = appendTo.length();
 614
 615     // if the pattern character is unrecognized, signal an error and dump out
 616     if (patternCharPtr == NULL)
 617     {
 618         status = U_INVALID_FORMAT_ERROR;
 619         return;
 620     }
 621
 622     patternCharIndex = (UDateFormatField)(patternCharPtr - DateFormatSymbols::getPatternUChars());
 623     UCalendarDateFields field = fgPatternIndexToCalendarField[patternCharIndex];
 624     int32_t value = cal.get(field, status);
 625     if (U_FAILURE(status)) {
 626         return;
 627     }
 628
 629     switch (patternCharIndex) {
 630
 631     // for any "G" symbol, write out the appropriate era string
 632     // "GGGG" is wide era name, anything else is abbreviated name
 633     case UDAT_ERA_FIELD:
 634         if (count >= 4)
 635            _appendSymbol(appendTo, value, fSymbols->fEraNames, fSymbols->fEraNamesCount);
 636         else
 637            _appendSymbol(appendTo, value, fSymbols->fEras, fSymbols->fErasCount);
 638         break;
 639
 640     // for "yyyy", write out the whole year; for "yy", write out the last 2 digits
 641     case UDAT_YEAR_FIELD:
 642     case UDAT_YEAR_WOY_FIELD:
 643         if (count >= 4)
 644             zeroPaddingNumber(appendTo, value, 4, maxIntCount);
 645         else if(count == 1)
 646             zeroPaddingNumber(appendTo, value, count, maxIntCount);
 647         else
 648             zeroPaddingNumber(appendTo, value, 2, 2);
 649         break;  // TODO: this needs to be synced with Java, with GCL/Shanghai's work
 650
 651     // for "MMMM", write out the whole month name, for "MMM", write out the month
 652     // abbreviation, for "M" or "MM", write out the month as a number with the
 653     // appropriate number of digits
 654     // for "MMMMM", use the narrow form
 655     case UDAT_MONTH_FIELD:
 656         if (count == 5)
 657             _appendSymbol(appendTo, value, fSymbols->fNarrowMonths,
 658                           fSymbols->fNarrowMonthsCount);
 659         else if (count == 4)
 660             _appendSymbol(appendTo, value, fSymbols->fMonths,
 661                           fSymbols->fMonthsCount);
 662         else if (count == 3)
 663             _appendSymbol(appendTo, value, fSymbols->fShortMonths,
 664                           fSymbols->fShortMonthsCount);
 665         else
 666             zeroPaddingNumber(appendTo, value + 1, count, maxIntCount);
 667         break;
 668
 669     // for "LLLL", write out the whole month name, for "LLL", write out the month
 670     // abbreviation, for "L" or "LL", write out the month as a number with the
 671     // appropriate number of digits
 672     // for "LLLLL", use the narrow form
 673     case UDAT_STANDALONE_MONTH_FIELD:
 674         if (count == 5)
 675             _appendSymbol(appendTo, value, fSymbols->fStandaloneNarrowMonths,
 676                           fSymbols->fStandaloneNarrowMonthsCount);
 677         else if (count == 4)
 678             _appendSymbol(appendTo, value, fSymbols->fStandaloneMonths,
 679                           fSymbols->fStandaloneMonthsCount);
 680         else if (count == 3)
 681             _appendSymbol(appendTo, value, fSymbols->fStandaloneShortMonths,
 682                           fSymbols->fStandaloneShortMonthsCount);
 683         else
 684             zeroPaddingNumber(appendTo, value + 1, count, maxIntCount);
 685         break;
 686
 687     // for "k" and "kk", write out the hour, adjusting midnight to appear as "24"
 688     case UDAT_HOUR_OF_DAY1_FIELD:
 689         if (value == 0)
 690             zeroPaddingNumber(appendTo, cal.getMaximum(UCAL_HOUR_OF_DAY) + 1, count, maxIntCount);
 691         else
 692             zeroPaddingNumber(appendTo, value, count, maxIntCount);
 693         break;
 694
 695     case UDAT_FRACTIONAL_SECOND_FIELD:
 696         // Fractional seconds left-justify
 697         {
 698             fNumberFormat->setMinimumIntegerDigits((count > 3) ? 3 : count);
 699             fNumberFormat->setMaximumIntegerDigits(maxIntCount);
 700             if (count == 1) {
 701                 value = (value + 50) / 100;
 702             } else if (count == 2) {
 703                 value = (value + 5) / 10;
 704             }
 705             FieldPosition p(0);
 706             fNumberFormat->format(value, appendTo, p);
 707             if (count > 3) {
 708                 fNumberFormat->setMinimumIntegerDigits(count - 3);
 709                 fNumberFormat->format((int32_t)0, appendTo, p);
 710             }
 711         }
 712         break;
 713
 714     // for "EEE", write out the abbreviated day-of-the-week name
 715     // for "EEEE", write out the wide day-of-the-week name
 716     // for "EEEEE", use the narrow day-of-the-week name
 717     case UDAT_DAY_OF_WEEK_FIELD:
 718         if (count == 5)
 719             _appendSymbol(appendTo, value, fSymbols->fNarrowWeekdays,
 720                           fSymbols->fNarrowWeekdaysCount);
 721         else if (count == 4)
 722             _appendSymbol(appendTo, value, fSymbols->fWeekdays,
 723                           fSymbols->fWeekdaysCount);
 724         else
 725             _appendSymbol(appendTo, value, fSymbols->fShortWeekdays,
 726                           fSymbols->fShortWeekdaysCount);
 727         break;
 728
 729     // for "ccc", write out the abbreviated day-of-the-week name
 730     // for "cccc", write out the wide day-of-the-week name
 731     // for "ccccc", use the narrow day-of-the-week name
 732     case UDAT_STANDALONE_DAY_FIELD:
 733         if (count == 5)
 734             _appendSymbol(appendTo, value, fSymbols->fStandaloneNarrowWeekdays,
 735                           fSymbols->fStandaloneNarrowWeekdaysCount);
 736         else if (count == 4)
 737             _appendSymbol(appendTo, value, fSymbols->fStandaloneWeekdays,
 738                           fSymbols->fStandaloneWeekdaysCount);
 739         else if (count == 3)
 740             _appendSymbol(appendTo, value, fSymbols->fStandaloneShortWeekdays,
 741                           fSymbols->fStandaloneShortWeekdaysCount);
 742         else
 743             zeroPaddingNumber(appendTo, value, 1, maxIntCount);
 744         break;
 745
 746     // for and "a" symbol, write out the whole AM/PM string
 747     case UDAT_AM_PM_FIELD:
 748         _appendSymbol(appendTo, value, fSymbols->fAmPms,
 749                       fSymbols->fAmPmsCount);
 750         break;
 751
 752     // for "h" and "hh", write out the hour, adjusting noon and midnight to show up
 753     // as "12"
 754     case UDAT_HOUR1_FIELD:
 755         if (value == 0)
 756             zeroPaddingNumber(appendTo, cal.getLeastMaximum(UCAL_HOUR) + 1, count, maxIntCount);
 757         else
 758             zeroPaddingNumber(appendTo, value, count, maxIntCount);
 759         break;
 760
 761     // for the "z" symbols, we have to check our time zone data first.  If we have a
 762     // localized name for the time zone, then "zzzz" / "zzz" indicate whether
 763     // daylight time is in effect (long/short) and "zz" / "z" do not (long/short).
 764     // If we don't have a localized time zone name,
 765     // then the time zone shows up as "GMT+hh:mm" or "GMT-hh:mm" (where "hh:mm" is the
 766     // offset from GMT) regardless of how many z's were in the pattern symbol
 767     case UDAT_TIMEZONE_FIELD:
 768     case UDAT_TIMEZONE_GENERIC_FIELD: {
 769         UnicodeString str;
 770         UnicodeString zid;
 771         UnicodeString displayString;
 772         zid = fSymbols->getZoneID(cal.getTimeZone().getID(str), zid, status);
 773         if(U_FAILURE(status)){
 774             break;
 775         }
 776         if (zid.length() == 0) {
 777             appendGMT(appendTo, cal, status);
 778         }
 779         else {
 780
 781             if (patternCharIndex == UDAT_TIMEZONE_GENERIC_FIELD) {
 782                 if(count < 4){
 783                     fSymbols->getZoneString(zid, DateFormatSymbols::TIMEZONE_SHORT_GENERIC, displayString, status);
 784                 }else{
 785                     fSymbols->getZoneString(zid, DateFormatSymbols::TIMEZONE_LONG_GENERIC, displayString, status);
 786                 }
 787             } else {
 788                 if (cal.get(UCAL_DST_OFFSET, status) != 0) {
 789                     if(count < 4){
 790                         fSymbols->getZoneString(zid, DateFormatSymbols::TIMEZONE_SHORT_DAYLIGHT, displayString, status);
 791                     }else{
 792                         fSymbols->getZoneString(zid, DateFormatSymbols::TIMEZONE_LONG_DAYLIGHT, displayString, status);
 793                     }
 794                 }else{
 795                     if(count < 4){
 796                         fSymbols->getZoneString(zid, DateFormatSymbols::TIMEZONE_SHORT_STANDARD, displayString, status);
 797                     }else{
 798                         fSymbols->getZoneString(zid, DateFormatSymbols::TIMEZONE_LONG_STANDARD, displayString, status);
 799                     }
 800                 }
 801             }
 802             if(displayString.length()==0){
 803                 appendGMT(appendTo, cal, status);
 804             }else{
 805                 appendTo += displayString;
 806             }
 807         }
 808       }
 809     break;
 810
 811     case 23: // 'Z' - TIMEZONE_RFC
 812         {
 813             UChar sign = 43/*'+'*/;
 814             value = (cal.get(UCAL_ZONE_OFFSET, status) +
 815                      cal.get(UCAL_DST_OFFSET, status)) / U_MILLIS_PER_MINUTE;
 816             if (value < 0) {
 817                 value = -value;
 818                 sign = 45/*'-'*/;
 819             }
 820             value = (value / 60) * 100 + (value % 60); // minutes => KKmm
 821             appendTo += sign;
 822             zeroPaddingNumber(appendTo, value, 4, 4);
 823         }
 824         break;
 825
 826     case UDAT_QUARTER_FIELD:
 827         if (count >= 4)
 828             _appendSymbol(appendTo, value/3, fSymbols->fQuarters,
 829                           fSymbols->fQuartersCount);
 830         else if (count == 3)
 831             _appendSymbol(appendTo, value/3, fSymbols->fShortQuarters,
 832                           fSymbols->fShortQuartersCount);
 833         else
 834             zeroPaddingNumber(appendTo, (value/3) + 1, count, maxIntCount);
 835         break;
 836
 837     case UDAT_STANDALONE_QUARTER_FIELD:
 838         if (count >= 4)
 839             _appendSymbol(appendTo, value/3, fSymbols->fStandaloneQuarters,
 840                           fSymbols->fStandaloneQuartersCount);
 841         else if (count == 3)
 842             _appendSymbol(appendTo, value/3, fSymbols->fStandaloneShortQuarters,
 843                           fSymbols->fStandaloneShortQuartersCount);
 844         else
 845             zeroPaddingNumber(appendTo, (value/3) + 1, count, maxIntCount);
 846         break;
 847
 848
 849     // all of the other pattern symbols can be formatted as simple numbers with
 850     // appropriate zero padding
 851     default:
 852         zeroPaddingNumber(appendTo, value, count, maxIntCount);
 853         break;
 854     }
 855
 856     // if the field we're formatting is the one the FieldPosition says it's interested
 857     // in, fill in the FieldPosition with this field's positions
 858     if (pos.getBeginIndex() == pos.getEndIndex() &&
 859         pos.getField() == fgPatternIndexToDateFormatField[patternCharIndex]) {
 860         pos.setBeginIndex(beginOffset);
 861         pos.setEndIndex(appendTo.length());
 862     }
 863 }
 864
 865 //----------------------------------------------------------------------
 866
 867 void
 868 SimpleDateFormat::zeroPaddingNumber(UnicodeString &appendTo, int32_t value, int32_t minDigits, int32_t maxDigits) const
 869 {
 870     FieldPosition pos(0);
 871
 872     fNumberFormat->setMinimumIntegerDigits(minDigits);
 873     fNumberFormat->setMaximumIntegerDigits(maxDigits);
 874     fNumberFormat->format(value, appendTo, pos);  // 3rd arg is there to speed up processing
 875 }
 876
 877 //----------------------------------------------------------------------
 878
 879 /**
 880  * Format characters that indicate numeric fields.  The character
 881  * at index 0 is treated specially.
 882  */
 883 static const UChar NUMERIC_FORMAT_CHARS[] = {0x4D, 0x79, 0x75, 0x64, 0x68, 0x48, 0x6D, 0x73, 0x53, 0x44, 0x46, 0x77, 0x57, 0x6B, 0x4B, 0x00}; /* "MyudhHmsSDFwWkK" */
 884
 885 /**
 886  * Return true if the given format character, occuring count
 887  * times, represents a numeric field.
 888  */
 889 UBool SimpleDateFormat::isNumeric(UChar formatChar, int32_t count) {
 890     UnicodeString s(NUMERIC_FORMAT_CHARS);
 891     int32_t i = s.indexOf(formatChar);
 892     return (i > 0 || (i == 0 && count < 3));
 893 }
 894
 895 void
 896 SimpleDateFormat::parse(const UnicodeString& text, Calendar& cal, ParsePosition& parsePos) const
 897 {
 898     int32_t pos = parsePos.getIndex();
 899     int32_t start = pos;
 900     UBool ambiguousYear[] = { FALSE };
 901     int32_t count = 0;
 902
 903     // hack, clear parsedTimeZone, cast away const
 904     delete parsedTimeZone;
 905     ((SimpleDateFormat*)this)->parsedTimeZone = NULL;
 906
 907     // For parsing abutting numeric fields. 'abutPat' is the
 908     // offset into 'pattern' of the first of 2 or more abutting
 909     // numeric fields.  'abutStart' is the offset into 'text'
 910     // where parsing the fields begins. 'abutPass' starts off as 0
 911     // and increments each time we try to parse the fields.
 912     int32_t abutPat = -1; // If >=0, we are in a run of abutting numeric fields
 913     int32_t abutStart = 0;
 914     int32_t abutPass = 0;
 915     UBool inQuote = FALSE;
 916
 917     const UnicodeString numericFormatChars(NUMERIC_FORMAT_CHARS);
 918
 919     for (int32_t i=0; i<fPattern.length(); ++i) {
 920         UChar ch = fPattern.charAt(i);
 921
 922         // Handle alphabetic field characters.
 923         if (!inQuote && ((ch >= 0x41 && ch <= 0x5A) || (ch >= 0x61 && ch <= 0x7A))) { // [A-Za-z]
 924             int32_t fieldPat = i;
 925
 926             // Count the length of this field specifier
 927             count = 1;
 928             while ((i+1)<fPattern.length() &&
 929                    fPattern.charAt(i+1) == ch) {
 930                 ++count;
 931                 ++i;
 932             }
 933
 934             if (isNumeric(ch, count)) {
 935                 if (abutPat < 0) {
 936                     // Determine if there is an abutting numeric field.  For
 937                     // most fields we can just look at the next characters,
 938                     // but the 'm' field is either numeric or text,
 939                     // depending on the count, so we have to look ahead for
 940                     // that field.
 941                     if ((i+1)<fPattern.length()) {
 942                         UBool abutting;
 943                         UChar nextCh = fPattern.charAt(i+1);
 944                         int32_t k = numericFormatChars.indexOf(nextCh);
 945                         if (k == 0) {
 946                             int32_t j = i+2;
 947                             while (j<fPattern.length() &&
 948                                    fPattern.charAt(j) == nextCh) {
 949                                 ++j;
 950                             }
 951                             abutting = (j-i) < 4; // nextCount < 3
 952                         } else {
 953                             abutting = k > 0;
 954                         }
 955
 956                         // Record the start of a set of abutting numeric
 957                         // fields.
 958                         if (abutting) {
 959                             abutPat = fieldPat;
 960                             abutStart = pos;
 961                             abutPass = 0;
 962                         }
 963                     }
 964                 }
 965             } else {
 966                 abutPat = -1; // End of any abutting fields
 967             }
 968
 969             // Handle fields within a run of abutting numeric fields.  Take
 970             // the pattern "HHmmss" as an example. We will try to parse
 971             // 2/2/2 characters of the input text, then if that fails,
 972             // 1/2/2.  We only adjust the width of the leftmost field; the
 973             // others remain fixed.  This allows "123456" => 12:34:56, but
 974             // "12345" => 1:23:45.  Likewise, for the pattern "yyyyMMdd" we
 975             // try 4/2/2, 3/2/2, 2/2/2, and finally 1/2/2.
 976             if (abutPat >= 0) {
 977                 // If we are at the start of a run of abutting fields, then
 978                 // shorten this field in each pass.  If we can't shorten
 979                 // this field any more, then the parse of this set of
 980                 // abutting numeric fields has failed.
 981                 if (fieldPat == abutPat) {
 982                     count -= abutPass++;
 983                     if (count == 0) {
 984                         parsePos.setIndex(start);
 985                         parsePos.setErrorIndex(pos);
 986                         return;
 987                     }
 988                 }
 989
 990                 pos = subParse(text, pos, ch, count,
 991                                TRUE, FALSE, ambiguousYear, cal);
 992
 993                 // If the parse fails anywhere in the run, back up to the
 994                 // start of the run and retry.
 995                 if (pos < 0) {
 996                     i = abutPat - 1;
 997                     pos = abutStart;
 998                     continue;
 999                 }
1000             }
1001
1002             // Handle non-numeric fields and non-abutting numeric
1003             // fields.
1004             else {
1005                 int32_t s = pos;
1006                 pos = subParse(text, pos, ch, count,
1007                                FALSE, TRUE, ambiguousYear, cal);
1008
1009                 if (pos < 0) {
1010                     parsePos.setErrorIndex(s);
1011                     parsePos.setIndex(start);
1012                     return;
1013                 }
1014             }
1015         }
1016
1017         // Handle literal pattern characters.  These are any
1018         // quoted characters and non-alphabetic unquoted
1019         // characters.
1020         else {
1021
1022             abutPat = -1; // End of any abutting fields
1023
1024             // Handle quotes.  Two consecutive quotes is a quote
1025             // literal, inside or outside of quotes.  Otherwise a
1026             // quote indicates entry or exit from a quoted region.
1027             if (ch == QUOTE) {
1028                 // Match a quote literal '' within OR outside of quotes
1029                 if ((i+1)<fPattern.length() && fPattern.charAt(i+1)==ch) {
1030                     ++i; // Skip over doubled quote
1031                     // Fall through and treat quote as a literal
1032                 } else {
1033                     // Enter or exit quoted region
1034                     inQuote = !inQuote;
1035                     continue;
1036                 }
1037             }
1038
1039             // A run of white space in the pattern matches a run
1040             // of white space in the input text.
1041             if (uprv_isRuleWhiteSpace(ch)) {
1042                 // Advance over run in pattern
1043                 while ((i+1)<fPattern.length() &&
1044                        uprv_isRuleWhiteSpace(fPattern.charAt(i+1))) {
1045                     ++i;
1046                 }
1047
1048                 // Advance over run in input text
1049                 int32_t s = pos;
1050                 while (pos<text.length() &&
1051                        u_isUWhiteSpace(text.charAt(pos))) {
1052                     ++pos;
1053                 }
1054
1055                 // Must see at least one white space char in input
1056                 if (pos > s) {
1057                     continue;
1058                 }
1059             } else if (pos<text.length() && text.charAt(pos)==ch) {
1060                 // Match a literal
1061                 ++pos;
1062                 continue;
1063             }
1064
1065             // We fall through to this point if the match fails
1066             parsePos.setIndex(start);
1067             parsePos.setErrorIndex(pos);
1068             return;
1069         }
1070     }
1071
1072     // At this point the fields of Calendar have been set.  Calendar
1073     // will fill in default values for missing fields when the time
1074     // is computed.
1075
1076     parsePos.setIndex(pos);
1077
1078     // This part is a problem:  When we call parsedDate.after, we compute the time.
1079     // Take the date April 3 2004 at 2:30 am.  When this is first set up, the year
1080     // will be wrong if we're parsing a 2-digit year pattern.  It will be 1904.
1081     // April 3 1904 is a Sunday (unlike 2004) so it is the DST onset day.  2:30 am
1082     // is therefore an "impossible" time, since the time goes from 1:59 to 3:00 am
1083     // on that day.  It is therefore parsed out to fields as 3:30 am.  Then we
1084     // add 100 years, and get April 3 2004 at 3:30 am.  Note that April 3 2004 is
1085     // a Saturday, so it can have a 2:30 am -- and it should. [LIU]
1086     /*
1087         UDate parsedDate = calendar.getTime();
1088         if( ambiguousYear[0] && !parsedDate.after(fDefaultCenturyStart) ) {
1089             calendar.add(Calendar.YEAR, 100);
1090             parsedDate = calendar.getTime();
1091         }
1092     */
1093     // Because of the above condition, save off the fields in case we need to readjust.
1094     // The procedure we use here is not particularly efficient, but there is no other
1095     // way to do this given the API restrictions present in Calendar.  We minimize
1096     // inefficiency by only performing this computation when it might apply, that is,
1097     // when the two-digit year is equal to the start year, and thus might fall at the
1098     // front or the back of the default century.  This only works because we adjust
1099     // the year correctly to start with in other cases -- see subParse().
1100     UErrorCode status = U_ZERO_ERROR;
1101     if (ambiguousYear[0] || parsedTimeZone != NULL) // If this is true then the two-digit year == the default start year
1102     {
1103         // We need a copy of the fields, and we need to avoid triggering a call to
1104         // complete(), which will recalculate the fields.  Since we can't access
1105         // the fields[] array in Calendar, we clone the entire object.  This will
1106         // stop working if Calendar.clone() is ever rewritten to call complete().
1107         Calendar *copy = cal.clone();
1108         if (ambiguousYear[0]) {
1109             UDate parsedDate = copy->getTime(status);
1110             // {sfb} check internalGetDefaultCenturyStart
1111             if (fHaveDefaultCentury && (parsedDate < fDefaultCenturyStart)) {
1112                 // We can't use add here because that does a complete() first.
1113                 cal.set(UCAL_YEAR, fDefaultCenturyStartYear + 100);
1114             }
1115         }
1116
1117         if (parsedTimeZone != NULL) {
1118             TimeZone *tz = parsedTimeZone;
1119
1120             // the calendar represents the parse as gmt time
1121             // we need to turn this into local time, so we add the raw offset
1122             // then we ask the timezone to handle this local time
1123             int32_t rawOffset = 0;
1124             int32_t dstOffset = 0;
1125             tz->getOffset(copy->getTime(status)+tz->getRawOffset(), TRUE,
1126                 rawOffset, dstOffset, status);
1127             if (U_SUCCESS(status)) {
1128                 cal.set(UCAL_ZONE_OFFSET, rawOffset);
1129                 cal.set(UCAL_DST_OFFSET, dstOffset);
1130             }
1131         }
1132
1133         delete copy;
1134     }
1135
1136     // If any Calendar calls failed, we pretend that we
1137     // couldn't parse the string, when in reality this isn't quite accurate--
1138     // we did parse it; the Calendar calls just failed.
1139     if (U_FAILURE(status)) {
1140         parsePos.setErrorIndex(pos);
1141         parsePos.setIndex(start);
1142     }
1143 }
1144
1145 UDate
1146 SimpleDateFormat::parse( const UnicodeString& text,
1147                          ParsePosition& pos) const {
1148     // redefined here because the other parse() function hides this function's
1149     // cunterpart on DateFormat
1150     return DateFormat::parse(text, pos);
1151 }
1152
1153 UDate
1154 SimpleDateFormat::parse(const UnicodeString& text, UErrorCode& status) const
1155 {
1156     // redefined here because the other parse() function hides this function's
1157     // counterpart on DateFormat
1158     return DateFormat::parse(text, status);
1159 }
1160 //----------------------------------------------------------------------
1161
1162 int32_t SimpleDateFormat::matchQuarterString(const UnicodeString& text,
1163                               int32_t start,
1164                               UCalendarDateFields field,
1165                               const UnicodeString* data,
1166                               int32_t dataCount,
1167                               Calendar& cal) const
1168 {
1169     int32_t i = 0;
1170     int32_t count = dataCount;
1171
1172     // There may be multiple strings in the data[] array which begin with
1173     // the same prefix (e.g., Cerven and Cervenec (June and July) in Czech).
1174     // We keep track of the longest match, and return that.  Note that this
1175     // unfortunately requires us to test all array elements.
1176     int32_t bestMatchLength = 0, bestMatch = -1;
1177
1178     // {sfb} kludge to support case-insensitive comparison
1179     // {markus 2002oct11} do not just use caseCompareBetween because we do not know
1180     // the length of the match after case folding
1181     // {alan 20040607} don't case change the whole string, since the length
1182     // can change
1183     // TODO we need a case-insensitive startsWith function
1184     UnicodeString lcase, lcaseText;
1185     text.extract(start, INT32_MAX, lcaseText);
1186     lcaseText.foldCase();
1187
1188     for (; i < count; ++i)
1189     {
1190         // Always compare if we have no match yet; otherwise only compare
1191         // against potentially better matches (longer strings).
1192
1193         lcase.fastCopyFrom(data[i]).foldCase();
1194         int32_t length = lcase.length();
1195
1196         if (length > bestMatchLength &&
1197             lcaseText.compareBetween(0, length, lcase, 0, length) == 0)
1198         {
1199             bestMatch = i;
1200             bestMatchLength = length;
1201         }
1202     }
1203     if (bestMatch >= 0)
1204     {
1205         cal.set(field, bestMatch * 3);
1206
1207         // Once we have a match, we have to determine the length of the
1208         // original source string.  This will usually be == the length of
1209         // the case folded string, but it may differ (e.g. sharp s).
1210         lcase.fastCopyFrom(data[bestMatch]).foldCase();
1211
1212         // Most of the time, the length will be the same as the length
1213         // of the string from the locale data.  Sometimes it will be
1214         // different, in which case we will have to figure it out by
1215         // adding a character at a time, until we have a match.  We do
1216         // this all in one loop, where we try 'len' first (at index
1217         // i==0).
1218         int32_t len = data[bestMatch].length(); // 99+% of the time
1219         int32_t n = text.length() - start;
1220         for (i=0; i<=n; ++i) {
1221             int32_t j=i;
1222             if (i == 0) {
1223                 j = len;
1224             } else if (i == len) {
1225                 continue; // already tried this when i was 0
1226             }
1227             text.extract(start, j, lcaseText);
1228             lcaseText.foldCase();
1229             if (lcase == lcaseText) {
1230                 return start + j;
1231             }
1232         }
1233     }
1234
1235     return -start;
1236 }
1237
1238 //----------------------------------------------------------------------
1239
1240 int32_t SimpleDateFormat::matchString(const UnicodeString& text,
1241                               int32_t start,
1242                               UCalendarDateFields field,
1243                               const UnicodeString* data,
1244                               int32_t dataCount,
1245                               Calendar& cal) const
1246 {
1247     int32_t i = 0;
1248     int32_t count = dataCount;
1249
1250     if (field == UCAL_DAY_OF_WEEK) i = 1;
1251
1252     // There may be multiple strings in the data[] array which begin with
1253     // the same prefix (e.g., Cerven and Cervenec (June and July) in Czech).
1254     // We keep track of the longest match, and return that.  Note that this
1255     // unfortunately requires us to test all array elements.
1256     int32_t bestMatchLength = 0, bestMatch = -1;
1257
1258     // {sfb} kludge to support case-insensitive comparison
1259     // {markus 2002oct11} do not just use caseCompareBetween because we do not know
1260     // the length of the match after case folding
1261     // {alan 20040607} don't case change the whole string, since the length
1262     // can change
1263     // TODO we need a case-insensitive startsWith function
1264     UnicodeString lcase, lcaseText;
1265     text.extract(start, INT32_MAX, lcaseText);
1266     lcaseText.foldCase();
1267
1268     for (; i < count; ++i)
1269     {
1270         // Always compare if we have no match yet; otherwise only compare
1271         // against potentially better matches (longer strings).
1272
1273         lcase.fastCopyFrom(data[i]).foldCase();
1274         int32_t length = lcase.length();
1275
1276         if (length > bestMatchLength &&
1277             lcaseText.compareBetween(0, length, lcase, 0, length) == 0)
1278         {
1279             bestMatch = i;
1280             bestMatchLength = length;
1281         }
1282     }
1283     if (bestMatch >= 0)
1284     {
1285         cal.set(field, bestMatch);
1286
1287         // Once we have a match, we have to determine the length of the
1288         // original source string.  This will usually be == the length of
1289         // the case folded string, but it may differ (e.g. sharp s).
1290         lcase.fastCopyFrom(data[bestMatch]).foldCase();
1291
1292         // Most of the time, the length will be the same as the length
1293         // of the string from the locale data.  Sometimes it will be
1294         // different, in which case we will have to figure it out by
1295         // adding a character at a time, until we have a match.  We do
1296         // this all in one loop, where we try 'len' first (at index
1297         // i==0).
1298         int32_t len = data[bestMatch].length(); // 99+% of the time
1299         int32_t n = text.length() - start;
1300         for (i=0; i<=n; ++i) {
1301             int32_t j=i;
1302             if (i == 0) {
1303                 j = len;
1304             } else if (i == len) {
1305                 continue; // already tried this when i was 0
1306             }
1307             text.extract(start, j, lcaseText);
1308             lcaseText.foldCase();
1309             if (lcase == lcaseText) {
1310                 return start + j;
1311             }
1312         }
1313     }
1314
1315     return -start;
1316 }
1317
1318 //----------------------------------------------------------------------
1319
1320 void
1321 SimpleDateFormat::set2DigitYearStart(UDate d, UErrorCode& status)
1322 {
1323     parseAmbiguousDatesAsAfter(d, status);
1324 }
1325
1326 /**
1327  * Private member function that converts the parsed date strings into
1328  * timeFields. Returns -start (for ParsePosition) if failed.
1329  * @param text the time text to be parsed.
1330  * @param start where to start parsing.
1331  * @param ch the pattern character for the date field text to be parsed.
1332  * @param count the count of a pattern character.
1333  * @return the new start position if matching succeeded; a negative number
1334  * indicating matching failure, otherwise.
1335  */
1336 int32_t SimpleDateFormat::subParse(const UnicodeString& text, int32_t& start, UChar ch, int32_t count,
1337                            UBool obeyCount, UBool allowNegative, UBool ambiguousYear[], Calendar& cal) const
1338 {
1339     Formattable number;
1340     int32_t value = 0;
1341     int32_t i;
1342     ParsePosition pos(0);
1343     int32_t patternCharIndex;
1344     UnicodeString temp;
1345     UChar *patternCharPtr = u_strchr(DateFormatSymbols::getPatternUChars(), ch);
1346
1347 #if defined (U_DEBUG_CAL)
1348     //fprintf(stderr, "%s:%d - [%c]  st=%d \n", __FILE__, __LINE__, (char) ch, start);
1349 #endif
1350
1351     if (patternCharPtr == NULL) {
1352         return -start;
1353     }
1354
1355     patternCharIndex = (UDateFormatField)(patternCharPtr - DateFormatSymbols::getPatternUChars());
1356
1357     UCalendarDateFields field = fgPatternIndexToCalendarField[patternCharIndex];
1358
1359     // If there are any spaces here, skip over them.  If we hit the end
1360     // of the string, then fail.
1361     for (;;) {
1362         if (start >= text.length()) {
1363             return -start;
1364         }
1365         UChar32 c = text.char32At(start);
1366         if (!u_isUWhiteSpace(c)) {
1367             break;
1368         }
1369         start += UTF_CHAR_LENGTH(c);
1370     }
1371     pos.setIndex(start);
1372
1373     // We handle a few special cases here where we need to parse
1374     // a number value.  We handle further, more generic cases below.  We need
1375     // to handle some of them here because some fields require extra processing on
1376     // the parsed value.
1377     if (patternCharIndex == UDAT_HOUR_OF_DAY1_FIELD ||
1378         patternCharIndex == UDAT_HOUR1_FIELD ||
1379         (patternCharIndex == UDAT_MONTH_FIELD && count <= 2) ||
1380         (patternCharIndex == UDAT_STANDALONE_MONTH_FIELD && count <= 2) ||
1381         (patternCharIndex == UDAT_QUARTER_FIELD && count <= 2) ||
1382         (patternCharIndex == UDAT_STANDALONE_QUARTER_FIELD && count <= 2) ||
1383         patternCharIndex == UDAT_YEAR_FIELD ||
1384         patternCharIndex == UDAT_YEAR_WOY_FIELD ||
1385         patternCharIndex == UDAT_FRACTIONAL_SECOND_FIELD)
1386     {
1387         int32_t parseStart = pos.getIndex();
1388         // It would be good to unify this with the obeyCount logic below,
1389         // but that's going to be difficult.
1390         const UnicodeString* src;
1391
1392         if (obeyCount) {
1393             if ((start+count) > text.length()) {
1394                 return -start;
1395             }
1396
1397             text.extractBetween(0, start + count, temp);
1398             src = &temp;
1399         } else {
1400             src = &text;
1401         }
1402
1403         parseInt(*src, number, pos, allowNegative);
1404
1405         if (pos.getIndex() == parseStart)
1406             return -start;
1407         value = number.getLong();
1408     }
1409
1410     switch (patternCharIndex) {
1411     case UDAT_ERA_FIELD:
1412         if (count == 4) {
1413             return matchString(text, start, UCAL_ERA, fSymbols->fEraNames, fSymbols->fEraNamesCount, cal);
1414         }
1415
1416         return matchString(text, start, UCAL_ERA, fSymbols->fEras, fSymbols->fErasCount, cal);
1417
1418     case UDAT_YEAR_FIELD:
1419         // If there are 3 or more YEAR pattern characters, this indicates
1420         // that the year value is to be treated literally, without any
1421         // two-digit year adjustments (e.g., from "01" to 2001).  Otherwise
1422         // we made adjustments to place the 2-digit year in the proper
1423         // century, for parsed strings from "00" to "99".  Any other string
1424         // is treated literally:  "2250", "-1", "1", "002".
1425         if (count <= 2 && (pos.getIndex() - start) == 2
1426             && u_isdigit(text.charAt(start))
1427             && u_isdigit(text.charAt(start+1)))
1428         {
1429             // Assume for example that the defaultCenturyStart is 6/18/1903.
1430             // This means that two-digit years will be forced into the range
1431             // 6/18/1903 to 6/17/2003.  As a result, years 00, 01, and 02
1432             // correspond to 2000, 2001, and 2002.  Years 04, 05, etc. correspond
1433             // to 1904, 1905, etc.  If the year is 03, then it is 2003 if the
1434             // other fields specify a date before 6/18, or 1903 if they specify a
1435             // date afterwards.  As a result, 03 is an ambiguous year.  All other
1436             // two-digit years are unambiguous.
1437           if(fHaveDefaultCentury) { // check if this formatter even has a pivot year
1438               int32_t ambiguousTwoDigitYear = fDefaultCenturyStartYear % 100;
1439               ambiguousYear[0] = (value == ambiguousTwoDigitYear);
1440               value += (fDefaultCenturyStartYear/100)*100 +
1441                 (value < ambiguousTwoDigitYear ? 100 : 0);
1442             }
1443         }
1444         cal.set(UCAL_YEAR, value);
1445         return pos.getIndex();
1446
1447     case UDAT_YEAR_WOY_FIELD:
1448         // Comment is the same as for UDAT_Year_FIELDs - look above
1449         if (count <= 2 && (pos.getIndex() - start) == 2
1450             && u_isdigit(text.charAt(start))
1451             && u_isdigit(text.charAt(start+1))
1452             && fHaveDefaultCentury )
1453         {
1454             int32_t ambiguousTwoDigitYear = fDefaultCenturyStartYear % 100;
1455             ambiguousYear[0] = (value == ambiguousTwoDigitYear);
1456             value += (fDefaultCenturyStartYear/100)*100 +
1457                 (value < ambiguousTwoDigitYear ? 100 : 0);
1458         }
1459         cal.set(UCAL_YEAR_WOY, value);
1460         return pos.getIndex();
1461
1462     case UDAT_MONTH_FIELD:
1463         if (count <= 2) // i.e., M or MM.
1464         {
1465             // Don't want to parse the month if it is a string
1466             // while pattern uses numeric style: M or MM.
1467             // [We computed 'value' above.]
1468             cal.set(UCAL_MONTH, value - 1);
1469             return pos.getIndex();
1470         } else {
1471             // count >= 3 // i.e., MMM or MMMM
1472             // Want to be able to parse both short and long forms.
1473             // Try count == 4 first:
1474             int32_t newStart = 0;
1475
1476             if ((newStart = matchString(text, start, UCAL_MONTH,
1477                                       fSymbols->fMonths, fSymbols->fMonthsCount, cal)) > 0)
1478                 return newStart;
1479             else // count == 4 failed, now try count == 3
1480                 return matchString(text, start, UCAL_MONTH,
1481                                    fSymbols->fShortMonths, fSymbols->fShortMonthsCount, cal);
1482         }
1483
1484     case UDAT_STANDALONE_MONTH_FIELD:
1485         if (count <= 2) // i.e., L or LL.
1486         {
1487             // Don't want to parse the month if it is a string
1488             // while pattern uses numeric style: M or MM.
1489             // [We computed 'value' above.]
1490             cal.set(UCAL_MONTH, value - 1);
1491             return pos.getIndex();
1492         } else {
1493             // count >= 3 // i.e., LLL or LLLL
1494             // Want to be able to parse both short and long forms.
1495             // Try count == 4 first:
1496             int32_t newStart = 0;
1497
1498             if ((newStart = matchString(text, start, UCAL_MONTH,
1499                                       fSymbols->fStandaloneMonths, fSymbols->fStandaloneMonthsCount, cal)) > 0)
1500                 return newStart;
1501             else // count == 4 failed, now try count == 3
1502                 return matchString(text, start, UCAL_MONTH,
1503                                    fSymbols->fStandaloneShortMonths, fSymbols->fStandaloneShortMonthsCount, cal);
1504         }
1505
1506     case UDAT_HOUR_OF_DAY1_FIELD:
1507         // [We computed 'value' above.]
1508         if (value == cal.getMaximum(UCAL_HOUR_OF_DAY) + 1)
1509             value = 0;
1510         cal.set(UCAL_HOUR_OF_DAY, value);
1511         return pos.getIndex();
1512
1513     case UDAT_FRACTIONAL_SECOND_FIELD:
1514         // Fractional seconds left-justify
1515         i = pos.getIndex() - start;
1516         if (i < 3) {
1517             while (i < 3) {
1518                 value *= 10;
1519                 i++;
1520             }
1521         } else {
1522             int32_t a = 1;
1523             while (i > 3) {
1524                 a *= 10;
1525                 i--;
1526             }
1527             value = (value + (a>>1)) / a;
1528         }
1529         cal.set(UCAL_MILLISECOND, value);
1530         return pos.getIndex();
1531
1532     case UDAT_DAY_OF_WEEK_FIELD:
1533         {
1534             // Want to be able to parse both short and long forms.
1535             // Try count == 4 (DDDD) first:
1536             int32_t newStart = 0;
1537             if ((newStart = matchString(text, start, UCAL_DAY_OF_WEEK,
1538                                       fSymbols->fWeekdays, fSymbols->fWeekdaysCount, cal)) > 0)
1539                 return newStart;
1540             else // DDDD failed, now try DDD
1541                 return matchString(text, start, UCAL_DAY_OF_WEEK,
1542                                    fSymbols->fShortWeekdays, fSymbols->fShortWeekdaysCount, cal);
1543         }
1544
1545     case UDAT_STANDALONE_DAY_FIELD:
1546         {
1547             // Want to be able to parse both short and long forms.
1548             // Try count == 4 (DDDD) first:
1549             int32_t newStart = 0;
1550             if ((newStart = matchString(text, start, UCAL_DAY_OF_WEEK,
1551                                       fSymbols->fStandaloneWeekdays, fSymbols->fStandaloneWeekdaysCount, cal)) > 0)
1552                 return newStart;
1553             else // DDDD failed, now try DDD
1554                 return matchString(text, start, UCAL_DAY_OF_WEEK,
1555                                    fSymbols->fStandaloneShortWeekdays, fSymbols->fStandaloneShortWeekdaysCount, cal);
1556         }
1557
1558     case UDAT_AM_PM_FIELD:
1559         return matchString(text, start, UCAL_AM_PM, fSymbols->fAmPms, fSymbols->fAmPmsCount, cal);
1560
1561     case UDAT_HOUR1_FIELD:
1562         // [We computed 'value' above.]
1563         if (value == cal.getLeastMaximum(UCAL_HOUR)+1)
1564             value = 0;
1565         cal.set(UCAL_HOUR, value);
1566         return pos.getIndex();
1567
1568     case UDAT_QUARTER_FIELD:
1569         if (count <= 2) // i.e., Q or QQ.
1570         {
1571             // Don't want to parse the month if it is a string
1572             // while pattern uses numeric style: Q or QQ.
1573             // [We computed 'value' above.]
1574             cal.set(UCAL_MONTH, (value - 1) * 3);
1575             return pos.getIndex();
1576         } else {
1577             // count >= 3 // i.e., QQQ or QQQQ
1578             // Want to be able to parse both short and long forms.
1579             // Try count == 4 first:
1580             int32_t newStart = 0;
1581
1582             if ((newStart = matchQuarterString(text, start, UCAL_MONTH,
1583                                       fSymbols->fQuarters, fSymbols->fQuartersCount, cal)) > 0)
1584                 return newStart;
1585             else // count == 4 failed, now try count == 3
1586                 return matchQuarterString(text, start, UCAL_MONTH,
1587                                    fSymbols->fShortQuarters, fSymbols->fShortQuartersCount, cal);
1588         }
1589
1590     case UDAT_STANDALONE_QUARTER_FIELD:
1591         if (count <= 2) // i.e., q or qq.
1592         {
1593             // Don't want to parse the month if it is a string
1594             // while pattern uses numeric style: q or q.
1595             // [We computed 'value' above.]
1596             cal.set(UCAL_MONTH, (value - 1) * 3);
1597             return pos.getIndex();
1598         } else {
1599             // count >= 3 // i.e., qqq or qqqq
1600             // Want to be able to parse both short and long forms.
1601             // Try count == 4 first:
1602             int32_t newStart = 0;
1603
1604             if ((newStart = matchQuarterString(text, start, UCAL_MONTH,
1605                                       fSymbols->fStandaloneQuarters, fSymbols->fStandaloneQuartersCount, cal)) > 0)
1606                 return newStart;
1607             else // count == 4 failed, now try count == 3
1608                 return matchQuarterString(text, start, UCAL_MONTH,
1609                                    fSymbols->fStandaloneShortQuarters, fSymbols->fStandaloneShortQuartersCount, cal);
1610         }
1611
1612     case UDAT_TIMEZONE_FIELD:
1613     case UDAT_TIMEZONE_RFC_FIELD:
1614     case UDAT_TIMEZONE_GENERIC_FIELD:
1615         {
1616         // First try to parse generic forms such as GMT-07:00. Do this first
1617         // in case localized DateFormatZoneData contains the string "GMT"
1618         // for a zone; in that case, we don't want to match the first three
1619         // characters of GMT+/-HH:MM etc.
1620
1621         int32_t sign = 0;
1622         int32_t offset;
1623         int32_t gmtLen = u_strlen(gGmt);
1624
1625         // For time zones that have no known names, look for strings
1626         // of the form:
1627         //    GMT[+-]hours:minutes or
1628         //    GMT[+-]hhmm or
1629         //    GMT.
1630
1631         if ((text.length() - start) >= gmtLen &&
1632             (text.caseCompare(start, gmtLen, gGmt, 0, gmtLen, U_FOLD_CASE_DEFAULT)) == 0)
1633           {
1634             cal.set(UCAL_DST_OFFSET, 0);
1635
1636             pos.setIndex(start + gmtLen);
1637
1638             if( text[pos.getIndex()] == 0x002B /*'+'*/ )
1639               sign = 1;
1640             else if( text[pos.getIndex()] == 0x002D /*'-'*/ )
1641               sign = -1;
1642             else {
1643               cal.set(UCAL_ZONE_OFFSET, 0 );
1644               return pos.getIndex();
1645             }
1646
1647             // Look for hours:minutes or hhmm.
1648             pos.setIndex(pos.getIndex() + 1);
1649             int32_t parseStart = pos.getIndex();
1650             Formattable tzNumber;
1651             fNumberFormat->parse(text, tzNumber, pos);
1652             if( pos.getIndex() == parseStart) {
1653               return -start;
1654             }
1655             if( text[pos.getIndex()] == 0x003A /*':'*/ ) {
1656               // This is the hours:minutes case
1657               offset = tzNumber.getLong() * 60;
1658               pos.setIndex(pos.getIndex() + 1);
1659               parseStart = pos.getIndex();
1660               fNumberFormat->parse(text, tzNumber, pos);
1661               if( pos.getIndex() == parseStart) {
1662                 return -start;
1663               }
1664               offset += tzNumber.getLong();
1665             }
1666             else {
1667               // This is the hhmm case.
1668               offset = tzNumber.getLong();
1669               if( offset < 24 )
1670                 offset *= 60;
1671               else
1672                 offset = offset % 100 + offset / 100 * 60;
1673             }
1674
1675             // Fall through for final processing below of 'offset' and 'sign'.
1676           }
1677         else {
1678             // At this point, check for named time zones by looking through
1679             // the locale data from the DateFormatZoneData strings.
1680             // Want to be able to parse both short and long forms.
1681                         // !!! side effect, might set parsedZoneString
1682             UErrorCode status = U_ZERO_ERROR;
1683             int32_t result = subParseZoneString(text, start, cal, status);
1684             if (result != 0) {
1685                 return result;
1686             }
1687
1688             // As a last resort, look for numeric timezones of the form
1689             // [+-]hhmm as specified by RFC 822.  This code is actually
1690             // a little more permissive than RFC 822.  It will try to do
1691             // its best with numbers that aren't strictly 4 digits long
1692             DecimalFormat fmt(UNICODE_STRING_SIMPLE("+####;-####"), status);
1693             if(U_FAILURE(status))
1694                 return -start;
1695             fmt.setParseIntegerOnly(TRUE);
1696             int32_t parseStart = pos.getIndex();
1697             Formattable tzNumber;
1698             fmt.parse( text, tzNumber, pos );
1699             if( pos.getIndex() == parseStart) {
1700                 return -start;   // Wasn't actually a number.
1701             }
1702             offset = tzNumber.getLong();
1703             sign = 1;
1704             if( offset < 0 ) {
1705                 sign = -1;
1706                 offset = -offset;
1707             }
1708             if( offset < 24 )
1709                 offset = offset * 60;
1710             else
1711                 offset = offset % 100 + offset / 100 * 60;
1712
1713             // Fall through for final processing below of 'offset' and 'sign'.
1714         }
1715
1716         // Do the final processing for both of the above cases.  We only
1717         // arrive here if the form GMT+/-... or an RFC 822 form was seen.
1718         if (sign != 0)
1719         {
1720             offset *= U_MILLIS_PER_MINUTE * sign;
1721
1722             if (cal.getTimeZone().useDaylightTime())
1723             {
1724                 cal.set(UCAL_DST_OFFSET, U_MILLIS_PER_HOUR);
1725                 offset -= U_MILLIS_PER_HOUR;
1726             }
1727             cal.set(UCAL_ZONE_OFFSET, offset);
1728
1729             return pos.getIndex();
1730         }
1731
1732         // All efforts to parse a zone failed.
1733         return -start;
1734         }
1735
1736     default:
1737         // Handle "generic" fields
1738         int32_t parseStart = pos.getIndex();
1739         const UnicodeString* src;
1740         if (obeyCount) {
1741             if ((start+count) > text.length()) {
1742                 return -start;
1743             }
1744             text.extractBetween(0, start + count, temp);
1745             src = &temp;
1746         } else {
1747             src = &text;
1748         }
1749         parseInt(*src, number, pos, allowNegative);
1750         if (pos.getIndex() != parseStart) {
1751             cal.set(field, number.getLong());
1752             return pos.getIndex();
1753         }
1754         return -start;
1755     }
1756 }
1757
1758 int32_t
1759 SimpleDateFormat::subParseZoneString(const UnicodeString& text, int32_t start, Calendar& cal, UErrorCode& status) const
1760 {
1761   // At this point, check for named time zones by looking through
1762   // the locale data from the DateFormatZoneData strings.
1763   // Want to be able to parse both short and long forms.
1764
1765   // optimize for calendar's current time zone
1766   TimeZone *tz = NULL;
1767   UnicodeString id;
1768   UnicodeString zid, value;
1769   DateFormatSymbols::TimeZoneTranslationType type = DateFormatSymbols::TIMEZONE_COUNT;
1770   fSymbols->getZoneID(getTimeZone().getID(id), zid, status);
1771   if(zid.length() > 0){
1772       fSymbols->findZoneIDTypeValue(zid, text, start, type, value, status);
1773       if(type != DateFormatSymbols::TIMEZONE_COUNT) {
1774           tz = TimeZone::createTimeZone(zid);
1775       }
1776   }
1777
1778   if(U_FAILURE(status)){
1779       return 0;
1780   }
1781   if (tz != NULL) { // Matched any ?
1782     // always set zone offset, needed to get correct hour in wall time
1783     // when checking daylight savings
1784     cal.set(UCAL_ZONE_OFFSET, tz->getRawOffset());
1785     if (type==DateFormatSymbols::TIMEZONE_SHORT_STANDARD || type==DateFormatSymbols::TIMEZONE_LONG_STANDARD ) {
1786       // standard time
1787       cal.set(UCAL_DST_OFFSET, 0);
1788       delete tz; tz = NULL;
1789     } else if (type==DateFormatSymbols::TIMEZONE_SHORT_DAYLIGHT || type==DateFormatSymbols::TIMEZONE_LONG_DAYLIGHT ) {
1790       // daylight time
1791       // !!! todo - no getDSTSavings() in ICU's timezone
1792       // use the correct DST SAVINGS for the zone.
1793       // cal.set(UCAL_DST_OFFSET, tz->getDSTSavings());
1794       cal.set(UCAL_DST_OFFSET, U_MILLIS_PER_HOUR);
1795       delete tz; tz = NULL;
1796     } else {
1797       // either standard or daylight
1798       // need to finish getting the date, then compute dst offset as appropriate
1799
1800       // !!! hack for api compatibility, can't modify subParse(...) so can't
1801       // pass this back any other way.  cast away const.
1802       ((SimpleDateFormat*)this)->parsedTimeZone = tz;
1803     }
1804
1805     return start + value.length();
1806   }
1807
1808
1809   // complete failure
1810   return 0;
1811 }
1812
1813 /**
1814  * Parse an integer using fNumberFormat.  This method is semantically
1815  * const, but actually may modify fNumberFormat.
1816  */
1817 void SimpleDateFormat::parseInt(const UnicodeString& text,
1818                                 Formattable& number,
1819                                 ParsePosition& pos,
1820                                 UBool allowNegative) const {
1821     UnicodeString oldPrefix;
1822     DecimalFormat* df = NULL;
1823     if (!allowNegative &&
1824         fNumberFormat->getDynamicClassID() == DecimalFormat::getStaticClassID()) {
1825         df = (DecimalFormat*)fNumberFormat;
1826         df->getNegativePrefix(oldPrefix);
1827         df->setNegativePrefix(SUPPRESS_NEGATIVE_PREFIX);
1828     }
1829     fNumberFormat->parse(text, number, pos);
1830     if (df != NULL) {
1831         df->setNegativePrefix(oldPrefix);
1832     }
1833 }
1834
1835 //----------------------------------------------------------------------
1836
1837 void SimpleDateFormat::translatePattern(const UnicodeString& originalPattern,
1838                                         UnicodeString& translatedPattern,
1839                                         const UnicodeString& from,
1840                                         const UnicodeString& to,
1841                                         UErrorCode& status)
1842 {
1843   // run through the pattern and convert any pattern symbols from the version
1844   // in "from" to the corresponding character ion "to".  This code takes
1845   // quoted strings into account (it doesn't try to translate them), and it signals
1846   // an error if a particular "pattern character" doesn't appear in "from".
1847   // Depending on the values of "from" and "to" this can convert from generic
1848   // to localized patterns or localized to generic.
1849   if (U_FAILURE(status))
1850     return;
1851
1852   translatedPattern.remove();
1853   UBool inQuote = FALSE;
1854   for (int32_t i = 0; i < originalPattern.length(); ++i) {
1855     UChar c = originalPattern[i];
1856     if (inQuote) {
1857       if (c == QUOTE)
1858     inQuote = FALSE;
1859     }
1860     else {
1861       if (c == QUOTE)
1862     inQuote = TRUE;
1863       else if ((c >= 0x0061 /*'a'*/ && c <= 0x007A) /*'z'*/
1864            || (c >= 0x0041 /*'A'*/ && c <= 0x005A /*'Z'*/)) {
1865     int32_t ci = from.indexOf(c);
1866     if (ci == -1) {
1867       status = U_INVALID_FORMAT_ERROR;
1868       return;
1869     }
1870     c = to[ci];
1871       }
1872     }
1873     translatedPattern += c;
1874   }
1875   if (inQuote) {
1876     status = U_INVALID_FORMAT_ERROR;
1877     return;
1878   }
1879 }
1880
1881 //----------------------------------------------------------------------
1882
1883 UnicodeString&
1884 SimpleDateFormat::toPattern(UnicodeString& result) const
1885 {
1886     result = fPattern;
1887     return result;
1888 }
1889
1890 //----------------------------------------------------------------------
1891
1892 UnicodeString&
1893 SimpleDateFormat::toLocalizedPattern(UnicodeString& result,
1894                                      UErrorCode& status) const
1895 {
1896     translatePattern(fPattern, result, DateFormatSymbols::getPatternUChars(), fSymbols->fLocalPatternChars, status);
1897     return result;
1898 }
1899
1900 //----------------------------------------------------------------------
1901
1902 void
1903 SimpleDateFormat::applyPattern(const UnicodeString& pattern)
1904 {
1905     fPattern = pattern;
1906 }
1907
1908 //----------------------------------------------------------------------
1909
1910 void
1911 SimpleDateFormat::applyLocalizedPattern(const UnicodeString& pattern,
1912                                         UErrorCode &status)
1913 {
1914     translatePattern(pattern, fPattern, fSymbols->fLocalPatternChars, DateFormatSymbols::getPatternUChars(), status);
1915 }
1916
1917 //----------------------------------------------------------------------
1918
1919 const DateFormatSymbols*
1920 SimpleDateFormat::getDateFormatSymbols() const
1921 {
1922     return fSymbols;
1923 }
1924
1925 //----------------------------------------------------------------------
1926
1927 void
1928 SimpleDateFormat::adoptDateFormatSymbols(DateFormatSymbols* newFormatSymbols)
1929 {
1930     delete fSymbols;
1931     fSymbols = newFormatSymbols;
1932 }
1933
1934 //----------------------------------------------------------------------
1935 void
1936 SimpleDateFormat::setDateFormatSymbols(const DateFormatSymbols& newFormatSymbols)
1937 {
1938     delete fSymbols;
1939     fSymbols = new DateFormatSymbols(newFormatSymbols);
1940 }
1941
1942
1943 //----------------------------------------------------------------------
1944
1945
1946 void SimpleDateFormat::adoptCalendar(Calendar* calendarToAdopt)
1947 {
1948   UErrorCode status = U_ZERO_ERROR;
1949   DateFormat::adoptCalendar(calendarToAdopt);
1950   delete fSymbols;
1951   fSymbols=NULL;
1952   initializeSymbols(fLocale, fCalendar, status);  // we need new symbols
1953   initializeDefaultCentury();  // we need a new century (possibly)
1954 }
1955
1956 U_NAMESPACE_END
1957
1958 #endif /* #if !UCONFIG_NO_FORMATTING */
1959
1960 //eof