1/*
2*******************************************************************************
3* Copyright (C) 2007-2014, International Business Machines Corporation and
4* others. All Rights Reserved.
5*******************************************************************************
6*
7* File DTPTNGEN.CPP
8*
9*******************************************************************************
10*/
11
12#include "unicode/utypes.h"
13#if !UCONFIG_NO_FORMATTING
14
15#include "unicode/datefmt.h"
16#include "unicode/decimfmt.h"
17#include "unicode/dtfmtsym.h"
18#include "unicode/dtptngen.h"
19#include "unicode/msgfmt.h"
20#include "unicode/smpdtfmt.h"
21#include "unicode/udat.h"
22#include "unicode/udatpg.h"
23#include "unicode/uniset.h"
24#include "unicode/uloc.h"
25#include "unicode/ures.h"
26#include "unicode/ustring.h"
27#include "unicode/rep.h"
28#include "cpputils.h"
29#include "mutex.h"
30#include "cmemory.h"
31#include "cstring.h"
32#include "locbased.h"
33#include "gregoimp.h"
34#include "hash.h"
35#include "uresimp.h"
36#include "dtptngen_impl.h"
37#include "shareddatetimepatterngenerator.h"
38#include "unifiedcache.h"
39
40#if U_CHARSET_FAMILY==U_EBCDIC_FAMILY
41/**
42 * If we are on EBCDIC, use an iterator which will
43 * traverse the bundles in ASCII order.
44 */
45#define U_USE_ASCII_BUNDLE_ITERATOR
46#define U_SORT_ASCII_BUNDLE_ITERATOR
47#endif
48
49#if defined(U_USE_ASCII_BUNDLE_ITERATOR)
50
51#include "unicode/ustring.h"
52#include "uarrsort.h"
53
54struct UResAEntry {
55    UChar *key;
56    UResourceBundle *item;
57};
58
59struct UResourceBundleAIterator {
60    UResourceBundle  *bund;
61    UResAEntry *entries;
62    int32_t num;
63    int32_t cursor;
64};
65
66/* Must be C linkage to pass function pointer to the sort function */
67
68U_CDECL_BEGIN
69
70static int32_t U_CALLCONV
71ures_a_codepointSort(const void *context, const void *left, const void *right) {
72    //CompareContext *cmp=(CompareContext *)context;
73    return u_strcmp(((const UResAEntry *)left)->key,
74                    ((const UResAEntry *)right)->key);
75}
76
77U_CDECL_END
78
79static void ures_a_open(UResourceBundleAIterator *aiter, UResourceBundle *bund, UErrorCode *status) {
80    if(U_FAILURE(*status)) {
81        return;
82    }
83    aiter->bund = bund;
84    aiter->num = ures_getSize(aiter->bund);
85    aiter->cursor = 0;
86#if !defined(U_SORT_ASCII_BUNDLE_ITERATOR)
87    aiter->entries = NULL;
88#else
89    aiter->entries = (UResAEntry*)uprv_malloc(sizeof(UResAEntry)*aiter->num);
90    for(int i=0;i<aiter->num;i++) {
91        aiter->entries[i].item = ures_getByIndex(aiter->bund, i, NULL, status);
92        const char *akey = ures_getKey(aiter->entries[i].item);
93        int32_t len = uprv_strlen(akey)+1;
94        aiter->entries[i].key = (UChar*)uprv_malloc(len*sizeof(UChar));
95        u_charsToUChars(akey, aiter->entries[i].key, len);
96    }
97    uprv_sortArray(aiter->entries, aiter->num, sizeof(UResAEntry), ures_a_codepointSort, NULL, TRUE, status);
98#endif
99}
100
101static void ures_a_close(UResourceBundleAIterator *aiter) {
102#if defined(U_SORT_ASCII_BUNDLE_ITERATOR)
103    for(int i=0;i<aiter->num;i++) {
104        uprv_free(aiter->entries[i].key);
105        ures_close(aiter->entries[i].item);
106    }
107#endif
108}
109
110static const UChar *ures_a_getNextString(UResourceBundleAIterator *aiter, int32_t *len, const char **key, UErrorCode *err) {
111#if !defined(U_SORT_ASCII_BUNDLE_ITERATOR)
112    return ures_getNextString(aiter->bund, len, key, err);
113#else
114    if(U_FAILURE(*err)) return NULL;
115    UResourceBundle *item = aiter->entries[aiter->cursor].item;
116    const UChar* ret = ures_getString(item, len, err);
117    *key = ures_getKey(item);
118    aiter->cursor++;
119    return ret;
120#endif
121}
122
123
124#endif
125
126
127U_NAMESPACE_BEGIN
128
129SharedDateTimePatternGenerator::~SharedDateTimePatternGenerator() {
130    delete ptr;
131}
132
133template<> U_I18N_API
134const SharedDateTimePatternGenerator *LocaleCacheKey<SharedDateTimePatternGenerator>::createObject(
135        const void * /*creationContext*/, UErrorCode &status) const {
136    DateTimePatternGenerator *fmt = DateTimePatternGenerator::internalMakeInstance(fLoc, status);
137    if (U_FAILURE(status)) {
138        return NULL;
139    }
140    SharedDateTimePatternGenerator *result = new SharedDateTimePatternGenerator(fmt);
141    if (result == NULL) {
142        delete fmt;
143        status = U_MEMORY_ALLOCATION_ERROR;
144        return NULL;
145    }
146    result->addRef();
147    return result;
148}
149
150
151// *****************************************************************************
152// class DateTimePatternGenerator
153// *****************************************************************************
154static const UChar Canonical_Items[] = {
155    // GyQMwWEdDFHmsSv
156    CAP_G, LOW_Y, CAP_Q, CAP_M, LOW_W, CAP_W, CAP_E, LOW_D, CAP_D, CAP_F,
157    CAP_H, LOW_M, LOW_S, CAP_S, LOW_V, 0
158};
159
160static const dtTypeElem dtTypes[] = {
161    // patternChar, field, type, minLen, weight
162    {CAP_G, UDATPG_ERA_FIELD, DT_SHORT, 1, 3,},
163    {CAP_G, UDATPG_ERA_FIELD, DT_LONG, 4, 0},
164    {LOW_Y, UDATPG_YEAR_FIELD, DT_NUMERIC, 1, 20},
165    {CAP_Y, UDATPG_YEAR_FIELD, DT_NUMERIC + DT_DELTA, 1, 20},
166    {LOW_U, UDATPG_YEAR_FIELD, DT_NUMERIC + 2*DT_DELTA, 1, 20},
167    {LOW_R, UDATPG_YEAR_FIELD, DT_NUMERIC + 3*DT_DELTA, 1, 20},
168    {CAP_U, UDATPG_YEAR_FIELD, DT_SHORT, 1, 3},
169    {CAP_U, UDATPG_YEAR_FIELD, DT_LONG, 4, 0},
170    {CAP_U, UDATPG_YEAR_FIELD, DT_NARROW, 5, 0},
171    {CAP_Q, UDATPG_QUARTER_FIELD, DT_NUMERIC, 1, 2},
172    {CAP_Q, UDATPG_QUARTER_FIELD, DT_SHORT, 3, 0},
173    {CAP_Q, UDATPG_QUARTER_FIELD, DT_LONG, 4, 0},
174    {LOW_Q, UDATPG_QUARTER_FIELD, DT_NUMERIC + DT_DELTA, 1, 2},
175    {LOW_Q, UDATPG_QUARTER_FIELD, DT_SHORT + DT_DELTA, 3, 0},
176    {LOW_Q, UDATPG_QUARTER_FIELD, DT_LONG + DT_DELTA, 4, 0},
177    {CAP_M, UDATPG_MONTH_FIELD, DT_NUMERIC, 1, 2},
178    {CAP_M, UDATPG_MONTH_FIELD, DT_SHORT, 3, 0},
179    {CAP_M, UDATPG_MONTH_FIELD, DT_LONG, 4, 0},
180    {CAP_M, UDATPG_MONTH_FIELD, DT_NARROW, 5, 0},
181    {CAP_L, UDATPG_MONTH_FIELD, DT_NUMERIC + DT_DELTA, 1, 2},
182    {CAP_L, UDATPG_MONTH_FIELD, DT_SHORT - DT_DELTA, 3, 0},
183    {CAP_L, UDATPG_MONTH_FIELD, DT_LONG - DT_DELTA, 4, 0},
184    {CAP_L, UDATPG_MONTH_FIELD, DT_NARROW - DT_DELTA, 5, 0},
185    {LOW_L, UDATPG_MONTH_FIELD, DT_NUMERIC + DT_DELTA, 1, 1},
186    {LOW_W, UDATPG_WEEK_OF_YEAR_FIELD, DT_NUMERIC, 1, 2},
187    {CAP_W, UDATPG_WEEK_OF_MONTH_FIELD, DT_NUMERIC + DT_DELTA, 1, 0},
188    {CAP_E, UDATPG_WEEKDAY_FIELD, DT_SHORT, 1, 3},
189    {CAP_E, UDATPG_WEEKDAY_FIELD, DT_LONG, 4, 0},
190    {CAP_E, UDATPG_WEEKDAY_FIELD, DT_NARROW, 5, 0},
191    {LOW_C, UDATPG_WEEKDAY_FIELD, DT_NUMERIC + 2*DT_DELTA, 1, 2},
192    {LOW_C, UDATPG_WEEKDAY_FIELD, DT_SHORT - 2*DT_DELTA, 3, 0},
193    {LOW_C, UDATPG_WEEKDAY_FIELD, DT_LONG - 2*DT_DELTA, 4, 0},
194    {LOW_C, UDATPG_WEEKDAY_FIELD, DT_NARROW - 2*DT_DELTA, 5, 0},
195    {LOW_E, UDATPG_WEEKDAY_FIELD, DT_NUMERIC + DT_DELTA, 1, 2}, // LOW_E is currently not used in CLDR data, should not be canonical
196    {LOW_E, UDATPG_WEEKDAY_FIELD, DT_SHORT - DT_DELTA, 3, 0},
197    {LOW_E, UDATPG_WEEKDAY_FIELD, DT_LONG - DT_DELTA, 4, 0},
198    {LOW_E, UDATPG_WEEKDAY_FIELD, DT_NARROW - DT_DELTA, 5, 0},
199    {LOW_D, UDATPG_DAY_FIELD, DT_NUMERIC, 1, 2},
200    {CAP_D, UDATPG_DAY_OF_YEAR_FIELD, DT_NUMERIC + DT_DELTA, 1, 3},
201    {CAP_F, UDATPG_DAY_OF_WEEK_IN_MONTH_FIELD, DT_NUMERIC + 2*DT_DELTA, 1, 0},
202    {LOW_G, UDATPG_DAY_FIELD, DT_NUMERIC + 3*DT_DELTA, 1, 20}, // really internal use, so we don't care
203    {LOW_A, UDATPG_DAYPERIOD_FIELD, DT_SHORT, 1, 0},
204    {CAP_H, UDATPG_HOUR_FIELD, DT_NUMERIC + 10*DT_DELTA, 1, 2}, // 24 hour
205    {LOW_K, UDATPG_HOUR_FIELD, DT_NUMERIC + 11*DT_DELTA, 1, 2}, // 24 hour
206    {LOW_H, UDATPG_HOUR_FIELD, DT_NUMERIC, 1, 2}, // 12 hour
207    {CAP_K, UDATPG_HOUR_FIELD, DT_NUMERIC + DT_DELTA, 1, 2}, // 12 hour
208    {LOW_M, UDATPG_MINUTE_FIELD, DT_NUMERIC, 1, 2},
209    {LOW_S, UDATPG_SECOND_FIELD, DT_NUMERIC, 1, 2},
210    {CAP_S, UDATPG_FRACTIONAL_SECOND_FIELD, DT_NUMERIC + DT_DELTA, 1, 1000},
211    {CAP_A, UDATPG_SECOND_FIELD, DT_NUMERIC + 2*DT_DELTA, 1, 1000},
212    {LOW_V, UDATPG_ZONE_FIELD, DT_SHORT - 2*DT_DELTA, 1, 0},
213    {LOW_V, UDATPG_ZONE_FIELD, DT_LONG - 2*DT_DELTA, 4, 0},
214    {LOW_Z, UDATPG_ZONE_FIELD, DT_SHORT, 1, 3},
215    {LOW_Z, UDATPG_ZONE_FIELD, DT_LONG, 4, 0},
216    {CAP_Z, UDATPG_ZONE_FIELD, DT_NARROW - DT_DELTA, 1, 3},
217    {CAP_Z, UDATPG_ZONE_FIELD, DT_LONG - DT_DELTA, 4, 0},
218    {CAP_Z, UDATPG_ZONE_FIELD, DT_SHORT - DT_DELTA, 5, 0},
219    {CAP_O, UDATPG_ZONE_FIELD, DT_SHORT - DT_DELTA, 1, 0},
220    {CAP_O, UDATPG_ZONE_FIELD, DT_LONG - DT_DELTA, 4, 0},
221    {CAP_V, UDATPG_ZONE_FIELD, DT_SHORT - DT_DELTA, 1, 0},
222    {CAP_V, UDATPG_ZONE_FIELD, DT_LONG - DT_DELTA, 2, 0},
223    {CAP_X, UDATPG_ZONE_FIELD, DT_NARROW - DT_DELTA, 1, 0},
224    {CAP_X, UDATPG_ZONE_FIELD, DT_SHORT - DT_DELTA, 2, 0},
225    {CAP_X, UDATPG_ZONE_FIELD, DT_LONG - DT_DELTA, 4, 0},
226    {LOW_X, UDATPG_ZONE_FIELD, DT_NARROW - DT_DELTA, 1, 0},
227    {LOW_X, UDATPG_ZONE_FIELD, DT_SHORT - DT_DELTA, 2, 0},
228    {LOW_X, UDATPG_ZONE_FIELD, DT_LONG - DT_DELTA, 4, 0},
229    {0, UDATPG_FIELD_COUNT, 0, 0, 0} , // last row of dtTypes[]
230 };
231
232static const char* const CLDR_FIELD_APPEND[] = {
233    "Era", "Year", "Quarter", "Month", "Week", "*", "Day-Of-Week", "Day", "*", "*", "*",
234    "Hour", "Minute", "Second", "*", "Timezone"
235};
236
237static const char* const CLDR_FIELD_NAME[] = {
238    "era", "year", "quarter", "month", "week", "*", "weekday", "*", "*", "day", "dayperiod",
239    "hour", "minute", "second", "*", "zone"
240};
241
242static const char* const Resource_Fields[] = {
243    "day", "dayperiod", "era", "hour", "minute", "month", "second", "week",
244    "weekday", "year", "zone", "quarter" };
245
246// For appendItems
247static const UChar UDATPG_ItemFormat[]= {0x7B, 0x30, 0x7D, 0x20, 0x251C, 0x7B, 0x32, 0x7D, 0x3A,
248    0x20, 0x7B, 0x31, 0x7D, 0x2524, 0};  // {0} \u251C{2}: {1}\u2524
249
250//static const UChar repeatedPatterns[6]={CAP_G, CAP_E, LOW_Z, LOW_V, CAP_Q, 0}; // "GEzvQ"
251
252static const char DT_DateTimePatternsTag[]="DateTimePatterns";
253static const char DT_DateTimeCalendarTag[]="calendar";
254static const char DT_DateTimeGregorianTag[]="gregorian";
255static const char DT_DateTimeAppendItemsTag[]="appendItems";
256static const char DT_DateTimeFieldsTag[]="fields";
257static const char DT_DateTimeAvailableFormatsTag[]="availableFormats";
258//static const UnicodeString repeatedPattern=UnicodeString(repeatedPatterns);
259
260UOBJECT_DEFINE_RTTI_IMPLEMENTATION(DateTimePatternGenerator)
261UOBJECT_DEFINE_RTTI_IMPLEMENTATION(DTSkeletonEnumeration)
262UOBJECT_DEFINE_RTTI_IMPLEMENTATION(DTRedundantEnumeration)
263
264DateTimePatternGenerator*  U_EXPORT2
265DateTimePatternGenerator::createInstance(UErrorCode& status) {
266    return createInstance(Locale::getDefault(), status);
267}
268
269DateTimePatternGenerator* U_EXPORT2
270DateTimePatternGenerator::createInstance(const Locale& locale, UErrorCode& status) {
271    const SharedDateTimePatternGenerator *shared = NULL;
272    UnifiedCache::getByLocale(locale, shared, status);
273    if (U_FAILURE(status)) {
274        return NULL;
275    }
276    DateTimePatternGenerator *result = new DateTimePatternGenerator(**shared);
277    shared->removeRef();
278    if (result == NULL) {
279        status = U_MEMORY_ALLOCATION_ERROR;
280    }
281    return result;
282}
283
284
285DateTimePatternGenerator* U_EXPORT2
286DateTimePatternGenerator::internalMakeInstance(const Locale& locale, UErrorCode& status) {
287    DateTimePatternGenerator *result = new DateTimePatternGenerator(locale, status);
288    if (result == NULL) {
289        status = U_MEMORY_ALLOCATION_ERROR;
290        return NULL;
291    }
292    if (U_FAILURE(status)) {
293        delete result;
294        return NULL;
295    }
296    return result;
297}
298
299DateTimePatternGenerator*  U_EXPORT2
300DateTimePatternGenerator::createEmptyInstance(UErrorCode& status) {
301    DateTimePatternGenerator *result = new DateTimePatternGenerator(status);
302    if (result == NULL) {
303        status = U_MEMORY_ALLOCATION_ERROR;
304    }
305    if (U_FAILURE(status)) {
306        delete result;
307        result = NULL;
308    }
309    return result;
310}
311
312DateTimePatternGenerator::DateTimePatternGenerator(UErrorCode &status) :
313    skipMatcher(NULL),
314    fAvailableFormatKeyHash(NULL)
315{
316    fp = new FormatParser();
317    dtMatcher = new DateTimeMatcher();
318    distanceInfo = new DistanceInfo();
319    patternMap = new PatternMap();
320    if (fp == NULL || dtMatcher == NULL || distanceInfo == NULL || patternMap == NULL) {
321        status = U_MEMORY_ALLOCATION_ERROR;
322    }
323}
324
325DateTimePatternGenerator::DateTimePatternGenerator(const Locale& locale, UErrorCode &status) :
326    skipMatcher(NULL),
327    fAvailableFormatKeyHash(NULL)
328{
329    fp = new FormatParser();
330    dtMatcher = new DateTimeMatcher();
331    distanceInfo = new DistanceInfo();
332    patternMap = new PatternMap();
333    if (fp == NULL || dtMatcher == NULL || distanceInfo == NULL || patternMap == NULL) {
334        status = U_MEMORY_ALLOCATION_ERROR;
335    }
336    else {
337        initData(locale, status);
338    }
339}
340
341DateTimePatternGenerator::DateTimePatternGenerator(const DateTimePatternGenerator& other) :
342    UObject(),
343    skipMatcher(NULL),
344    fAvailableFormatKeyHash(NULL)
345{
346    fp = new FormatParser();
347    dtMatcher = new DateTimeMatcher();
348    distanceInfo = new DistanceInfo();
349    patternMap = new PatternMap();
350    *this=other;
351}
352
353DateTimePatternGenerator&
354DateTimePatternGenerator::operator=(const DateTimePatternGenerator& other) {
355    // reflexive case
356    if (&other == this) {
357        return *this;
358    }
359    pLocale = other.pLocale;
360    fDefaultHourFormatChar = other.fDefaultHourFormatChar;
361    *fp = *(other.fp);
362    dtMatcher->copyFrom(other.dtMatcher->skeleton);
363    *distanceInfo = *(other.distanceInfo);
364    dateTimeFormat = other.dateTimeFormat;
365    decimal = other.decimal;
366    // NUL-terminate for the C API.
367    dateTimeFormat.getTerminatedBuffer();
368    decimal.getTerminatedBuffer();
369    delete skipMatcher;
370    if ( other.skipMatcher == NULL ) {
371        skipMatcher = NULL;
372    }
373    else {
374        skipMatcher = new DateTimeMatcher(*other.skipMatcher);
375    }
376    for (int32_t i=0; i< UDATPG_FIELD_COUNT; ++i ) {
377        appendItemFormats[i] = other.appendItemFormats[i];
378        appendItemNames[i] = other.appendItemNames[i];
379        // NUL-terminate for the C API.
380        appendItemFormats[i].getTerminatedBuffer();
381        appendItemNames[i].getTerminatedBuffer();
382    }
383    UErrorCode status = U_ZERO_ERROR;
384    patternMap->copyFrom(*other.patternMap, status);
385    copyHashtable(other.fAvailableFormatKeyHash, status);
386    return *this;
387}
388
389
390UBool
391DateTimePatternGenerator::operator==(const DateTimePatternGenerator& other) const {
392    if (this == &other) {
393        return TRUE;
394    }
395    if ((pLocale==other.pLocale) && (patternMap->equals(*other.patternMap)) &&
396        (dateTimeFormat==other.dateTimeFormat) && (decimal==other.decimal)) {
397        for ( int32_t i=0 ; i<UDATPG_FIELD_COUNT; ++i ) {
398           if ((appendItemFormats[i] != other.appendItemFormats[i]) ||
399               (appendItemNames[i] != other.appendItemNames[i]) ) {
400               return FALSE;
401           }
402        }
403        return TRUE;
404    }
405    else {
406        return FALSE;
407    }
408}
409
410UBool
411DateTimePatternGenerator::operator!=(const DateTimePatternGenerator& other) const {
412    return  !operator==(other);
413}
414
415DateTimePatternGenerator::~DateTimePatternGenerator() {
416    if (fAvailableFormatKeyHash!=NULL) {
417        delete fAvailableFormatKeyHash;
418    }
419
420    if (fp != NULL) delete fp;
421    if (dtMatcher != NULL) delete dtMatcher;
422    if (distanceInfo != NULL) delete distanceInfo;
423    if (patternMap != NULL) delete patternMap;
424    if (skipMatcher != NULL) delete skipMatcher;
425}
426
427void
428DateTimePatternGenerator::initData(const Locale& locale, UErrorCode &status) {
429    //const char *baseLangName = locale.getBaseName(); // unused
430
431    skipMatcher = NULL;
432    fAvailableFormatKeyHash=NULL;
433    addCanonicalItems();
434    addICUPatterns(locale, status);
435    if (U_FAILURE(status)) {
436        return;
437    }
438    addCLDRData(locale, status);
439    setDateTimeFromCalendar(locale, status);
440    setDecimalSymbols(locale, status);
441} // DateTimePatternGenerator::initData
442
443UnicodeString
444DateTimePatternGenerator::getSkeleton(const UnicodeString& pattern, UErrorCode&
445/*status*/) {
446    dtMatcher->set(pattern, fp);
447    return dtMatcher->getSkeletonPtr()->getSkeleton();
448}
449
450UnicodeString
451DateTimePatternGenerator::getBaseSkeleton(const UnicodeString& pattern, UErrorCode& /*status*/) {
452    dtMatcher->set(pattern, fp);
453    return dtMatcher->getSkeletonPtr()->getBaseSkeleton();
454}
455
456void
457DateTimePatternGenerator::addICUPatterns(const Locale& locale, UErrorCode& status) {
458    UnicodeString dfPattern;
459    UnicodeString conflictingString;
460    DateFormat* df;
461
462    if (U_FAILURE(status)) {
463        return;
464    }
465
466    // Load with ICU patterns
467    for (int32_t i=DateFormat::kFull; i<=DateFormat::kShort; i++) {
468        DateFormat::EStyle style = (DateFormat::EStyle)i;
469        df = DateFormat::createDateInstance(style, locale);
470        SimpleDateFormat* sdf;
471        if (df != NULL && (sdf = dynamic_cast<SimpleDateFormat*>(df)) != NULL) {
472            addPattern(sdf->toPattern(dfPattern), FALSE, conflictingString, status);
473        }
474        // TODO Maybe we should return an error when the date format isn't simple.
475        delete df;
476        if (U_FAILURE(status)) {
477            return;
478        }
479
480        df = DateFormat::createTimeInstance(style, locale);
481        if (df != NULL && (sdf = dynamic_cast<SimpleDateFormat*>(df)) != NULL) {
482            addPattern(sdf->toPattern(dfPattern), FALSE, conflictingString, status);
483            // HACK for hh:ss
484            if ( i==DateFormat::kMedium ) {
485                hackPattern = dfPattern;
486            }
487        }
488        // TODO Maybe we should return an error when the date format isn't simple.
489        delete df;
490        if (U_FAILURE(status)) {
491            return;
492        }
493    }
494}
495
496void
497DateTimePatternGenerator::hackTimes(const UnicodeString& hackPattern, UErrorCode& status)  {
498    UnicodeString conflictingString;
499
500    fp->set(hackPattern);
501    UnicodeString mmss;
502    UBool gotMm=FALSE;
503    for (int32_t i=0; i<fp->itemNumber; ++i) {
504        UnicodeString field = fp->items[i];
505        if ( fp->isQuoteLiteral(field) ) {
506            if ( gotMm ) {
507               UnicodeString quoteLiteral;
508               fp->getQuoteLiteral(quoteLiteral, &i);
509               mmss += quoteLiteral;
510            }
511        }
512        else {
513            if (fp->isPatternSeparator(field) && gotMm) {
514                mmss+=field;
515            }
516            else {
517                UChar ch=field.charAt(0);
518                if (ch==LOW_M) {
519                    gotMm=TRUE;
520                    mmss+=field;
521                }
522                else {
523                    if (ch==LOW_S) {
524                        if (!gotMm) {
525                            break;
526                        }
527                        mmss+= field;
528                        addPattern(mmss, FALSE, conflictingString, status);
529                        break;
530                    }
531                    else {
532                        if (gotMm || ch==LOW_Z || ch==CAP_Z || ch==LOW_V || ch==CAP_V) {
533                            break;
534                        }
535                    }
536                }
537            }
538        }
539    }
540}
541
542#define ULOC_LOCALE_IDENTIFIER_CAPACITY (ULOC_FULLNAME_CAPACITY + 1 + ULOC_KEYWORD_AND_VALUES_CAPACITY)
543
544static const UChar hourFormatChars[] = { CAP_H, LOW_H, CAP_K, LOW_K, 0 }; // HhKk, the hour format characters
545
546void
547DateTimePatternGenerator::addCLDRData(const Locale& locale, UErrorCode& err) {
548    UResourceBundle *rb, *calTypeBundle, *calBundle;
549    UResourceBundle *patBundle, *fieldBundle, *fBundle;
550    UnicodeString rbPattern, value, field;
551    UnicodeString conflictingPattern;
552    const char *key=NULL;
553    int32_t i;
554
555    UnicodeString defaultItemFormat(TRUE, UDATPG_ItemFormat, UPRV_LENGTHOF(UDATPG_ItemFormat)-1);  // Read-only alias.
556
557    err = U_ZERO_ERROR;
558
559    fDefaultHourFormatChar = 0;
560    for (i=0; i<UDATPG_FIELD_COUNT; ++i ) {
561        appendItemNames[i]=CAP_F;
562        if (i<10) {
563            appendItemNames[i]+=(UChar)(i+0x30);
564        }
565        else {
566            appendItemNames[i]+=(UChar)0x31;
567            appendItemNames[i]+=(UChar)(i-10 + 0x30);
568        }
569        // NUL-terminate for the C API.
570        appendItemNames[i].getTerminatedBuffer();
571    }
572
573    rb = ures_open(NULL, locale.getName(), &err);
574    if (rb == NULL || U_FAILURE(err)) {
575        return;
576    }
577    const char *curLocaleName=ures_getLocaleByType(rb, ULOC_ACTUAL_LOCALE, &err);
578    const char * calendarTypeToUse = DT_DateTimeGregorianTag; // initial default
579    char         calendarType[ULOC_KEYWORDS_CAPACITY]; // to be filled in with the type to use, if all goes well
580    if ( U_SUCCESS(err) ) {
581        char    localeWithCalendarKey[ULOC_LOCALE_IDENTIFIER_CAPACITY];
582        // obtain a locale that always has the calendar key value that should be used
583        (void)ures_getFunctionalEquivalent(localeWithCalendarKey, ULOC_LOCALE_IDENTIFIER_CAPACITY, NULL,
584                                            "calendar", "calendar", locale.getName(), NULL, FALSE, &err);
585        localeWithCalendarKey[ULOC_LOCALE_IDENTIFIER_CAPACITY-1] = 0; // ensure null termination
586        // now get the calendar key value from that locale
587        int32_t calendarTypeLen = uloc_getKeywordValue(localeWithCalendarKey, "calendar", calendarType, ULOC_KEYWORDS_CAPACITY, &err);
588        if (U_SUCCESS(err) && calendarTypeLen < ULOC_KEYWORDS_CAPACITY) {
589            calendarTypeToUse = calendarType;
590        }
591        err = U_ZERO_ERROR;
592    }
593    calBundle = ures_getByKeyWithFallback(rb, DT_DateTimeCalendarTag, NULL, &err);
594    calTypeBundle = ures_getByKeyWithFallback(calBundle, calendarTypeToUse, NULL, &err);
595
596    key=NULL;
597    int32_t dtCount=0;
598    patBundle = ures_getByKeyWithFallback(calTypeBundle, DT_DateTimePatternsTag, NULL, &err);
599    while (U_SUCCESS(err)) {
600        rbPattern = ures_getNextUnicodeString(patBundle, &key, &err);
601        dtCount++;
602        if (rbPattern.length()==0 ) {
603            break;  // no more pattern
604        }
605        else {
606            if (dtCount==9) {
607                setDateTimeFormat(rbPattern);
608            } else if (dtCount==4) { // short time format
609                // set fDefaultHourFormatChar to the hour format character from this pattern
610                int32_t tfIdx, tfLen = rbPattern.length();
611                UBool ignoreChars = FALSE;
612                for (tfIdx = 0; tfIdx < tfLen; tfIdx++) {
613                    UChar tfChar = rbPattern.charAt(tfIdx);
614                    if ( tfChar == SINGLE_QUOTE ) {
615                        ignoreChars = !ignoreChars; // toggle (handle quoted literals & '' for single quote)
616                    } else if ( !ignoreChars && u_strchr(hourFormatChars, tfChar) != NULL ) {
617                        fDefaultHourFormatChar = tfChar;
618                        break;
619                    }
620                }
621            }
622        }
623    }
624    ures_close(patBundle);
625
626    err = U_ZERO_ERROR;
627    patBundle = ures_getByKeyWithFallback(calTypeBundle, DT_DateTimeAppendItemsTag, NULL, &err);
628    key=NULL;
629    UnicodeString itemKey;
630    while (U_SUCCESS(err)) {
631        rbPattern = ures_getNextUnicodeString(patBundle, &key, &err);
632        if (rbPattern.length()==0 ) {
633            break;  // no more pattern
634        }
635        else {
636            setAppendItemFormat(getAppendFormatNumber(key), rbPattern);
637        }
638    }
639    ures_close(patBundle);
640
641    key=NULL;
642    err = U_ZERO_ERROR;
643    fBundle = ures_getByKeyWithFallback(rb, DT_DateTimeFieldsTag, NULL, &err);
644    for (i=0; i<MAX_RESOURCE_FIELD; ++i) {
645        err = U_ZERO_ERROR;
646        patBundle = ures_getByKeyWithFallback(fBundle, Resource_Fields[i], NULL, &err);
647        fieldBundle = ures_getByKeyWithFallback(patBundle, "dn", NULL, &err);
648        rbPattern = ures_getNextUnicodeString(fieldBundle, &key, &err);
649        ures_close(fieldBundle);
650        ures_close(patBundle);
651        if (rbPattern.length()==0 ) {
652            continue;
653        }
654        else {
655            setAppendItemName(getAppendNameNumber(Resource_Fields[i]), rbPattern);
656        }
657    }
658    ures_close(fBundle);
659
660    // add available formats
661    UBool firstTimeThrough = TRUE;
662    err = U_ZERO_ERROR;
663    initHashtable(err);
664    UBool override = TRUE;
665    while (TRUE) {
666        // At the start of the loop:
667        // - rb is the open resource bundle for the current locale being processed,
668        //   whose actual name is in curLocaleName.
669        // - if U_SUCCESS(err), then calBundle and calTypeBundle are open;
670        //   process contents of calTypeBundle, then close calBundle and calTypeBundle.
671        if (U_SUCCESS(err)) {
672            // process contents of calTypeBundle
673            patBundle = ures_getByKeyWithFallback(calTypeBundle, DT_DateTimeAvailableFormatsTag, NULL, &err);
674            if (U_SUCCESS(err)) {
675                int32_t numberKeys = ures_getSize(patBundle);
676                int32_t len;
677                const UChar *retPattern;
678                key=NULL;
679#if defined(U_USE_ASCII_BUNDLE_ITERATOR)
680                UResourceBundleAIterator aiter;
681                ures_a_open(&aiter, patBundle, &err);
682#endif
683                for(i=0; i<numberKeys; ++i) {
684#if defined(U_USE_ASCII_BUNDLE_ITERATOR)
685                    retPattern=ures_a_getNextString(&aiter, &len, &key, &err);
686#else
687                    retPattern=ures_getNextString(patBundle, &len, &key, &err);
688#endif
689                    UnicodeString format=UnicodeString(retPattern);
690                    UnicodeString retKey=UnicodeString(key, -1, US_INV);
691                    if ( firstTimeThrough || !isAvailableFormatSet(retKey) ) {
692                        setAvailableFormat(retKey, err);
693                        // Add pattern with its associated skeleton. Override any duplicate derived from std patterns,
694                        // but not a previous availableFormats entry:
695                        addPatternWithSkeleton(format, &retKey, override, conflictingPattern, err);
696                    }
697                }
698#if defined(U_USE_ASCII_BUNDLE_ITERATOR)
699                ures_a_close(&aiter);
700#endif
701                ures_close(patBundle);
702            }
703            firstTimeThrough = FALSE;
704            // close calBundle and calTypeBundle
705            ures_close(calTypeBundle);
706            ures_close(calBundle);
707        }
708        if (uprv_strcmp(curLocaleName,"root")==0 || uprv_strlen(curLocaleName)==0) {
709            // we just finished handling root, nothing more to check
710            ures_close(rb);
711            break;
712        }
713        // Find the name of the appropriate parent locale (from %%Parent if present, else
714        // uloc_getParent on the actual locale name)
715        // (It would be nice to have a ures function that did this...)
716        err = U_ZERO_ERROR;
717        char parentLocale[ULOC_FULLNAME_CAPACITY];
718        int32_t locNameLen;
719        const UChar * parentUName = ures_getStringByKey(rb, "%%Parent", &locNameLen, &err);
720        if (U_SUCCESS(err) && err != U_USING_FALLBACK_WARNING && locNameLen < ULOC_FULLNAME_CAPACITY) {
721            u_UCharsToChars(parentUName, parentLocale, locNameLen + 1);
722        } else {
723            err = U_ZERO_ERROR;
724            uloc_getParent(curLocaleName, parentLocale, ULOC_FULLNAME_CAPACITY, &err);
725            if (U_FAILURE(err) || err == U_STRING_NOT_TERMINATED_WARNING) {
726                // just fallback to root, since we are not already there
727                parentLocale[0] = 0;
728                err = U_ZERO_ERROR;
729            }
730        }
731        // Close current locale bundle
732        ures_close(rb);
733        // And open its parent, which becomes the new current locale being processed
734        rb = ures_open(NULL, parentLocale, &err);
735        if ( U_FAILURE(err) ) {
736            err = U_ZERO_ERROR;
737            break;
738        }
739        // Get the name of the parent / new current locale
740        curLocaleName=ures_getLocaleByType(rb, ULOC_ACTUAL_LOCALE, &err);
741        if ( U_FAILURE(err) ) {
742            curLocaleName = parentLocale;
743            err = U_ZERO_ERROR;
744        }
745        if (uprv_strcmp(curLocaleName,"root")==0 || uprv_strlen(curLocaleName)==0) {
746            override = FALSE;
747        }
748        // Open calBundle and calTypeBundle
749        calBundle = ures_getByKeyWithFallback(rb, DT_DateTimeCalendarTag, NULL, &err);
750        if (U_SUCCESS(err)) {
751            calTypeBundle = ures_getByKeyWithFallback(calBundle, calendarTypeToUse, NULL, &err);
752            if ( U_FAILURE(err) ) {
753                ures_close(calBundle);
754            }
755        }
756        // Go to the top of the loop to process contents of calTypeBundle
757    }
758
759    if (hackPattern.length()>0) {
760        hackTimes(hackPattern, err);
761    }
762}
763
764void
765DateTimePatternGenerator::initHashtable(UErrorCode& err) {
766    if (fAvailableFormatKeyHash!=NULL) {
767        return;
768    }
769    if ((fAvailableFormatKeyHash = new Hashtable(FALSE, err))==NULL) {
770        err=U_MEMORY_ALLOCATION_ERROR;
771        return;
772    }
773}
774
775
776void
777DateTimePatternGenerator::setAppendItemFormat(UDateTimePatternField field, const UnicodeString& value) {
778    appendItemFormats[field] = value;
779    // NUL-terminate for the C API.
780    appendItemFormats[field].getTerminatedBuffer();
781}
782
783const UnicodeString&
784DateTimePatternGenerator::getAppendItemFormat(UDateTimePatternField field) const {
785    return appendItemFormats[field];
786}
787
788void
789DateTimePatternGenerator::setAppendItemName(UDateTimePatternField field, const UnicodeString& value) {
790    appendItemNames[field] = value;
791    // NUL-terminate for the C API.
792    appendItemNames[field].getTerminatedBuffer();
793}
794
795const UnicodeString&
796DateTimePatternGenerator:: getAppendItemName(UDateTimePatternField field) const {
797    return appendItemNames[field];
798}
799
800void
801DateTimePatternGenerator::getAppendName(UDateTimePatternField field, UnicodeString& value) {
802    value = SINGLE_QUOTE;
803    value += appendItemNames[field];
804    value += SINGLE_QUOTE;
805}
806
807UnicodeString
808DateTimePatternGenerator::getBestPattern(const UnicodeString& patternForm, UErrorCode& status) {
809    return getBestPattern(patternForm, UDATPG_MATCH_NO_OPTIONS, status);
810}
811
812UnicodeString
813DateTimePatternGenerator::getBestPattern(const UnicodeString& patternForm, UDateTimePatternMatchOptions options, UErrorCode& status) {
814    const UnicodeString *bestPattern=NULL;
815    UnicodeString dtFormat;
816    UnicodeString resultPattern;
817    int32_t flags = kDTPGNoFlags;
818
819    int32_t dateMask=(1<<UDATPG_DAYPERIOD_FIELD) - 1;
820    int32_t timeMask=(1<<UDATPG_FIELD_COUNT) - 1 - dateMask;
821
822    // Replace hour metacharacters 'j' and 'J', set flags as necessary
823    UnicodeString patternFormCopy = UnicodeString(patternForm);
824    int32_t patPos, patLen = patternFormCopy.length();
825    UBool inQuoted = FALSE;
826    for (patPos = 0; patPos < patLen; patPos++) {
827        UChar patChr = patternFormCopy.charAt(patPos);
828        if (patChr == SINGLE_QUOTE) {
829            inQuoted = !inQuoted;
830        } else if (!inQuoted) {
831            if (patChr == LOW_J) {
832                patternFormCopy.setCharAt(patPos, fDefaultHourFormatChar);
833            } else if (patChr == CAP_J) {
834                // Get pattern for skeleton with H, then replace H or k
835                // with fDefaultHourFormatChar (if different)
836                patternFormCopy.setCharAt(patPos, CAP_H);
837                flags |= kDTPGSkeletonUsesCapJ;
838            }
839        }
840    }
841
842    resultPattern.remove();
843    dtMatcher->set(patternFormCopy, fp);
844    const PtnSkeleton* specifiedSkeleton=NULL;
845    bestPattern=getBestRaw(*dtMatcher, -1, distanceInfo, &specifiedSkeleton);
846    if ( distanceInfo->missingFieldMask==0 && distanceInfo->extraFieldMask==0 ) {
847        resultPattern = adjustFieldTypes(*bestPattern, specifiedSkeleton, flags, options);
848
849        return resultPattern;
850    }
851    int32_t neededFields = dtMatcher->getFieldMask();
852    UnicodeString datePattern=getBestAppending(neededFields & dateMask, flags, options);
853    UnicodeString timePattern=getBestAppending(neededFields & timeMask, flags, options);
854    if (datePattern.length()==0) {
855        if (timePattern.length()==0) {
856            resultPattern.remove();
857        }
858        else {
859            return timePattern;
860        }
861    }
862    if (timePattern.length()==0) {
863        return datePattern;
864    }
865    resultPattern.remove();
866    status = U_ZERO_ERROR;
867    dtFormat=getDateTimeFormat();
868    Formattable dateTimeObject[] = { timePattern, datePattern };
869    resultPattern = MessageFormat::format(dtFormat, dateTimeObject, 2, resultPattern, status );
870    return resultPattern;
871}
872
873UnicodeString
874DateTimePatternGenerator::replaceFieldTypes(const UnicodeString& pattern,
875                                            const UnicodeString& skeleton,
876                                            UErrorCode& status) {
877    return replaceFieldTypes(pattern, skeleton, UDATPG_MATCH_NO_OPTIONS, status);
878}
879
880UnicodeString
881DateTimePatternGenerator::replaceFieldTypes(const UnicodeString& pattern,
882                                            const UnicodeString& skeleton,
883                                            UDateTimePatternMatchOptions options,
884                                            UErrorCode& /*status*/) {
885    dtMatcher->set(skeleton, fp);
886    UnicodeString result = adjustFieldTypes(pattern, NULL, kDTPGNoFlags, options);
887    return result;
888}
889
890void
891DateTimePatternGenerator::setDecimal(const UnicodeString& newDecimal) {
892    this->decimal = newDecimal;
893    // NUL-terminate for the C API.
894    this->decimal.getTerminatedBuffer();
895}
896
897const UnicodeString&
898DateTimePatternGenerator::getDecimal() const {
899    return decimal;
900}
901
902void
903DateTimePatternGenerator::addCanonicalItems() {
904    UnicodeString  conflictingPattern;
905    UErrorCode status = U_ZERO_ERROR;
906
907    for (int32_t i=0; i<UDATPG_FIELD_COUNT; i++) {
908        addPattern(UnicodeString(Canonical_Items[i]), FALSE, conflictingPattern, status);
909    }
910}
911
912void
913DateTimePatternGenerator::setDateTimeFormat(const UnicodeString& dtFormat) {
914    dateTimeFormat = dtFormat;
915    // NUL-terminate for the C API.
916    dateTimeFormat.getTerminatedBuffer();
917}
918
919const UnicodeString&
920DateTimePatternGenerator::getDateTimeFormat() const {
921    return dateTimeFormat;
922}
923
924void
925DateTimePatternGenerator::setDateTimeFromCalendar(const Locale& locale, UErrorCode& status) {
926    const UChar *resStr;
927    int32_t resStrLen = 0;
928
929    Calendar* fCalendar = Calendar::createInstance(locale, status);
930    CalendarData calData(locale, fCalendar?fCalendar->getType():NULL, status);
931    UResourceBundle *dateTimePatterns = calData.getByKey(DT_DateTimePatternsTag, status);
932    if (U_FAILURE(status)) return;
933
934    if (ures_getSize(dateTimePatterns) <= DateFormat::kDateTime)
935    {
936        status = U_INVALID_FORMAT_ERROR;
937        return;
938    }
939    resStr = ures_getStringByIndex(dateTimePatterns, (int32_t)DateFormat::kDateTime, &resStrLen, &status);
940    setDateTimeFormat(UnicodeString(TRUE, resStr, resStrLen));
941
942    delete fCalendar;
943}
944
945void
946DateTimePatternGenerator::setDecimalSymbols(const Locale& locale, UErrorCode& status) {
947    DecimalFormatSymbols dfs = DecimalFormatSymbols(locale, status);
948    if(U_SUCCESS(status)) {
949        decimal = dfs.getSymbol(DecimalFormatSymbols::kDecimalSeparatorSymbol);
950        // NUL-terminate for the C API.
951        decimal.getTerminatedBuffer();
952    }
953}
954
955UDateTimePatternConflict
956DateTimePatternGenerator::addPattern(
957    const UnicodeString& pattern,
958    UBool override,
959    UnicodeString &conflictingPattern,
960    UErrorCode& status)
961{
962    return addPatternWithSkeleton(pattern, NULL, override, conflictingPattern, status);
963}
964
965// For DateTimePatternGenerator::addPatternWithSkeleton -
966// If skeletonToUse is specified, then an availableFormats entry is being added. In this case:
967// 1. We pass that skeleton to matcher.set instead of having it derive a skeleton from the pattern.
968// 2. If the new entry's skeleton or basePattern does match an existing entry but that entry also had a skeleton specified
969// (i.e. it was also from availableFormats), then the new entry does not override it regardless of the value of the override
970// parameter. This prevents later availableFormats entries from a parent locale overriding earlier ones from the actual
971// specified locale. However, availableFormats entries *should* override entries with matching skeleton whose skeleton was
972// derived (i.e. entries derived from the standard date/time patters for the specified locale).
973// 3. When adding the pattern (patternMap->add), we set a new boolean to indicate that the added entry had a
974// specified skeleton (which sets a new field in the PtnElem in the PatternMap).
975UDateTimePatternConflict
976DateTimePatternGenerator::addPatternWithSkeleton(
977    const UnicodeString& pattern,
978    const UnicodeString* skeletonToUse,
979    UBool override,
980    UnicodeString& conflictingPattern,
981    UErrorCode& status)
982{
983
984    UnicodeString basePattern;
985    PtnSkeleton   skeleton;
986    UDateTimePatternConflict conflictingStatus = UDATPG_NO_CONFLICT;
987
988    DateTimeMatcher matcher;
989    if ( skeletonToUse == NULL ) {
990        matcher.set(pattern, fp, skeleton);
991        matcher.getBasePattern(basePattern);
992    } else {
993        matcher.set(*skeletonToUse, fp, skeleton); // no longer trims skeleton fields to max len 3, per #7930
994        matcher.getBasePattern(basePattern); // or perhaps instead: basePattern = *skeletonToUse;
995    }
996    // We only care about base conflicts - and replacing the pattern associated with a base - if:
997    // 1. the conflicting previous base pattern did *not* have an explicit skeleton; in that case the previous
998    // base + pattern combination was derived from either (a) a canonical item, (b) a standard format, or
999    // (c) a pattern specified programmatically with a previous call to addPattern (which would only happen
1000    // if we are getting here from a subsequent call to addPattern).
1001    // 2. a skeleton is specified for the current pattern, but override=false; in that case we are checking
1002    // availableFormats items from root, which should not override any previous entry with the same base.
1003    UBool entryHadSpecifiedSkeleton;
1004    const UnicodeString *duplicatePattern = patternMap->getPatternFromBasePattern(basePattern, entryHadSpecifiedSkeleton);
1005    if (duplicatePattern != NULL && (!entryHadSpecifiedSkeleton || (skeletonToUse != NULL && !override))) {
1006        conflictingStatus = UDATPG_BASE_CONFLICT;
1007        conflictingPattern = *duplicatePattern;
1008        if (!override) {
1009            return conflictingStatus;
1010        }
1011    }
1012    // The only time we get here with override=true and skeletonToUse!=null is when adding availableFormats
1013    // items from CLDR data. In that case, we don't want an item from a parent locale to replace an item with
1014    // same skeleton from the specified locale, so skip the current item if skeletonWasSpecified is true for
1015    // the previously-specified conflicting item.
1016    const PtnSkeleton* entrySpecifiedSkeleton = NULL;
1017    duplicatePattern = patternMap->getPatternFromSkeleton(skeleton, &entrySpecifiedSkeleton);
1018    if (duplicatePattern != NULL ) {
1019        conflictingStatus = UDATPG_CONFLICT;
1020        conflictingPattern = *duplicatePattern;
1021        if (!override || (skeletonToUse != NULL && entrySpecifiedSkeleton != NULL)) {
1022            return conflictingStatus;
1023        }
1024    }
1025    patternMap->add(basePattern, skeleton, pattern, skeletonToUse != NULL, status);
1026    if(U_FAILURE(status)) {
1027        return conflictingStatus;
1028    }
1029
1030    return UDATPG_NO_CONFLICT;
1031}
1032
1033
1034UDateTimePatternField
1035DateTimePatternGenerator::getAppendFormatNumber(const char* field) const {
1036    for (int32_t i=0; i<UDATPG_FIELD_COUNT; ++i ) {
1037        if (uprv_strcmp(CLDR_FIELD_APPEND[i], field)==0) {
1038            return (UDateTimePatternField)i;
1039        }
1040    }
1041    return UDATPG_FIELD_COUNT;
1042}
1043
1044UDateTimePatternField
1045DateTimePatternGenerator::getAppendNameNumber(const char* field) const {
1046    for (int32_t i=0; i<UDATPG_FIELD_COUNT; ++i ) {
1047        if (uprv_strcmp(CLDR_FIELD_NAME[i],field)==0) {
1048            return (UDateTimePatternField)i;
1049        }
1050    }
1051    return UDATPG_FIELD_COUNT;
1052}
1053
1054const UnicodeString*
1055DateTimePatternGenerator::getBestRaw(DateTimeMatcher& source,
1056                                     int32_t includeMask,
1057                                     DistanceInfo* missingFields,
1058                                     const PtnSkeleton** specifiedSkeletonPtr) {
1059    int32_t bestDistance = 0x7fffffff;
1060    DistanceInfo tempInfo;
1061    const UnicodeString *bestPattern=NULL;
1062    const PtnSkeleton* specifiedSkeleton=NULL;
1063
1064    PatternMapIterator it;
1065    for (it.set(*patternMap); it.hasNext(); ) {
1066        DateTimeMatcher trial = it.next();
1067        if (trial.equals(skipMatcher)) {
1068            continue;
1069        }
1070        int32_t distance=source.getDistance(trial, includeMask, tempInfo);
1071        if (distance<bestDistance) {
1072            bestDistance=distance;
1073            bestPattern=patternMap->getPatternFromSkeleton(*trial.getSkeletonPtr(), &specifiedSkeleton);
1074            missingFields->setTo(tempInfo);
1075            if (distance==0) {
1076                break;
1077            }
1078        }
1079    }
1080
1081    // If the best raw match had a specified skeleton and that skeleton was requested by the caller,
1082    // then return it too. This generally happens when the caller needs to pass that skeleton
1083    // through to adjustFieldTypes so the latter can do a better job.
1084    if (bestPattern && specifiedSkeletonPtr) {
1085        *specifiedSkeletonPtr = specifiedSkeleton;
1086    }
1087    return bestPattern;
1088}
1089
1090UnicodeString
1091DateTimePatternGenerator::adjustFieldTypes(const UnicodeString& pattern,
1092                                           const PtnSkeleton* specifiedSkeleton,
1093                                           int32_t flags,
1094                                           UDateTimePatternMatchOptions options) {
1095    UnicodeString newPattern;
1096    fp->set(pattern);
1097    for (int32_t i=0; i < fp->itemNumber; i++) {
1098        UnicodeString field = fp->items[i];
1099        if ( fp->isQuoteLiteral(field) ) {
1100
1101            UnicodeString quoteLiteral;
1102            fp->getQuoteLiteral(quoteLiteral, &i);
1103            newPattern += quoteLiteral;
1104        }
1105        else {
1106            if (fp->isPatternSeparator(field)) {
1107                newPattern+=field;
1108                continue;
1109            }
1110            int32_t canonicalIndex = fp->getCanonicalIndex(field);
1111            if (canonicalIndex < 0) {
1112                newPattern+=field;
1113                continue;  // don't adjust
1114            }
1115            const dtTypeElem *row = &dtTypes[canonicalIndex];
1116            int32_t typeValue = row->field;
1117            if ((flags & kDTPGFixFractionalSeconds) != 0 && typeValue == UDATPG_SECOND_FIELD) {
1118                UnicodeString newField=dtMatcher->skeleton.original[UDATPG_FRACTIONAL_SECOND_FIELD];
1119                field = field + decimal + newField;
1120            } else if (dtMatcher->skeleton.type[typeValue]!=0) {
1121                    // Here:
1122                    // - "reqField" is the field from the originally requested skeleton, with length
1123                    // "reqFieldLen".
1124                    // - "field" is the field from the found pattern.
1125                    //
1126                    // The adjusted field should consist of characters from the originally requested
1127                    // skeleton, except in the case of UDATPG_HOUR_FIELD or UDATPG_MONTH_FIELD or
1128                    // UDATPG_WEEKDAY_FIELD or UDATPG_YEAR_FIELD, in which case it should consist
1129                    // of characters from the  found pattern.
1130                    //
1131                    // The length of the adjusted field (adjFieldLen) should match that in the originally
1132                    // requested skeleton, except that in the following cases the length of the adjusted field
1133                    // should match that in the found pattern (i.e. the length of this pattern field should
1134                    // not be adjusted):
1135                    // 1. typeValue is UDATPG_HOUR_FIELD/MINUTE/SECOND and the corresponding bit in options is
1136                    //    not set (ticket #7180). Note, we may want to implement a similar change for other
1137                    //    numeric fields (MM, dd, etc.) so the default behavior is to get locale preference for
1138                    //    field length, but options bits can be used to override this.
1139                    // 2. There is a specified skeleton for the found pattern and one of the following is true:
1140                    //    a) The length of the field in the skeleton (skelFieldLen) is equal to reqFieldLen.
1141                    //    b) The pattern field is numeric and the skeleton field is not, or vice versa.
1142
1143                    UnicodeString reqField = dtMatcher->skeleton.original[typeValue];
1144                    int32_t reqFieldLen = reqField.length();
1145                    if (reqField.charAt(0) == CAP_E && reqFieldLen < 3)
1146                        reqFieldLen = 3; // 1-3 for E are equivalent to 3 for c,e
1147                    int32_t adjFieldLen = reqFieldLen;
1148                    if ( (typeValue==UDATPG_HOUR_FIELD && (options & UDATPG_MATCH_HOUR_FIELD_LENGTH)==0) ||
1149                         (typeValue==UDATPG_MINUTE_FIELD && (options & UDATPG_MATCH_MINUTE_FIELD_LENGTH)==0) ||
1150                         (typeValue==UDATPG_SECOND_FIELD && (options & UDATPG_MATCH_SECOND_FIELD_LENGTH)==0) ) {
1151                         adjFieldLen = field.length();
1152                    } else if (specifiedSkeleton) {
1153                        UnicodeString skelField = specifiedSkeleton->original[typeValue];
1154                        int32_t skelFieldLen = skelField.length();
1155                        UBool patFieldIsNumeric = (row->type > 0);
1156                        UBool skelFieldIsNumeric = (specifiedSkeleton->type[typeValue] > 0);
1157                        if (skelFieldLen == reqFieldLen || (patFieldIsNumeric && !skelFieldIsNumeric) || (skelFieldIsNumeric && !patFieldIsNumeric)) {
1158                            // don't adjust the field length in the found pattern
1159                            adjFieldLen = field.length();
1160                        }
1161                    }
1162                    UChar c = (typeValue!= UDATPG_HOUR_FIELD && typeValue!= UDATPG_MONTH_FIELD &&
1163                               typeValue!= UDATPG_WEEKDAY_FIELD && (typeValue!= UDATPG_YEAR_FIELD || reqField.charAt(0)==CAP_Y))?
1164                        reqField.charAt(0): field.charAt(0);
1165                    if (typeValue == UDATPG_HOUR_FIELD && (flags & kDTPGSkeletonUsesCapJ) != 0) {
1166                        c = fDefaultHourFormatChar;
1167                    }
1168                    field.remove();
1169                    for (int32_t i=adjFieldLen; i>0; --i) {
1170                        field+=c;
1171                    }
1172            }
1173            newPattern+=field;
1174        }
1175    }
1176    return newPattern;
1177}
1178
1179UnicodeString
1180DateTimePatternGenerator::getBestAppending(int32_t missingFields, int32_t flags, UDateTimePatternMatchOptions options) {
1181    UnicodeString  resultPattern, tempPattern;
1182    UErrorCode err=U_ZERO_ERROR;
1183    int32_t lastMissingFieldMask=0;
1184    if (missingFields!=0) {
1185        resultPattern=UnicodeString();
1186        const PtnSkeleton* specifiedSkeleton=NULL;
1187        tempPattern = *getBestRaw(*dtMatcher, missingFields, distanceInfo, &specifiedSkeleton);
1188        resultPattern = adjustFieldTypes(tempPattern, specifiedSkeleton, flags, options);
1189        if ( distanceInfo->missingFieldMask==0 ) {
1190            return resultPattern;
1191        }
1192        while (distanceInfo->missingFieldMask!=0) { // precondition: EVERY single field must work!
1193            if ( lastMissingFieldMask == distanceInfo->missingFieldMask ) {
1194                break;  // cannot find the proper missing field
1195            }
1196            if (((distanceInfo->missingFieldMask & UDATPG_SECOND_AND_FRACTIONAL_MASK)==UDATPG_FRACTIONAL_MASK) &&
1197                ((missingFields & UDATPG_SECOND_AND_FRACTIONAL_MASK) == UDATPG_SECOND_AND_FRACTIONAL_MASK)) {
1198                resultPattern = adjustFieldTypes(resultPattern, specifiedSkeleton, flags | kDTPGFixFractionalSeconds, options);
1199                distanceInfo->missingFieldMask &= ~UDATPG_FRACTIONAL_MASK;
1200                continue;
1201            }
1202            int32_t startingMask = distanceInfo->missingFieldMask;
1203            tempPattern = *getBestRaw(*dtMatcher, distanceInfo->missingFieldMask, distanceInfo, &specifiedSkeleton);
1204            tempPattern = adjustFieldTypes(tempPattern, specifiedSkeleton, flags, options);
1205            int32_t foundMask=startingMask& ~distanceInfo->missingFieldMask;
1206            int32_t topField=getTopBitNumber(foundMask);
1207            UnicodeString appendName;
1208            getAppendName((UDateTimePatternField)topField, appendName);
1209            const Formattable formatPattern[] = {
1210                resultPattern,
1211                tempPattern,
1212                appendName
1213            };
1214            UnicodeString emptyStr;
1215            resultPattern = MessageFormat::format(appendItemFormats[topField], formatPattern, 3, emptyStr, err);
1216            lastMissingFieldMask = distanceInfo->missingFieldMask;
1217        }
1218    }
1219    return resultPattern;
1220}
1221
1222int32_t
1223DateTimePatternGenerator::getTopBitNumber(int32_t foundMask) {
1224    if ( foundMask==0 ) {
1225        return 0;
1226    }
1227    int32_t i=0;
1228    while (foundMask!=0) {
1229        foundMask >>=1;
1230        ++i;
1231    }
1232    if (i-1 >UDATPG_ZONE_FIELD) {
1233        return UDATPG_ZONE_FIELD;
1234    }
1235    else
1236        return i-1;
1237}
1238
1239void
1240DateTimePatternGenerator::setAvailableFormat(const UnicodeString &key, UErrorCode& err)
1241{
1242    fAvailableFormatKeyHash->puti(key, 1, err);
1243}
1244
1245UBool
1246DateTimePatternGenerator::isAvailableFormatSet(const UnicodeString &key) const {
1247    return (UBool)(fAvailableFormatKeyHash->geti(key) == 1);
1248}
1249
1250void
1251DateTimePatternGenerator::copyHashtable(Hashtable *other, UErrorCode &status) {
1252
1253    if (other == NULL) {
1254        return;
1255    }
1256    if (fAvailableFormatKeyHash != NULL) {
1257        delete fAvailableFormatKeyHash;
1258        fAvailableFormatKeyHash = NULL;
1259    }
1260    initHashtable(status);
1261    if(U_FAILURE(status)){
1262        return;
1263    }
1264    int32_t pos = UHASH_FIRST;
1265    const UHashElement* elem = NULL;
1266    // walk through the hash table and create a deep clone
1267    while((elem = other->nextElement(pos))!= NULL){
1268        const UHashTok otherKeyTok = elem->key;
1269        UnicodeString* otherKey = (UnicodeString*)otherKeyTok.pointer;
1270        fAvailableFormatKeyHash->puti(*otherKey, 1, status);
1271        if(U_FAILURE(status)){
1272            return;
1273        }
1274    }
1275}
1276
1277StringEnumeration*
1278DateTimePatternGenerator::getSkeletons(UErrorCode& status) const {
1279    StringEnumeration* skeletonEnumerator = new DTSkeletonEnumeration(*patternMap, DT_SKELETON, status);
1280    return skeletonEnumerator;
1281}
1282
1283const UnicodeString&
1284DateTimePatternGenerator::getPatternForSkeleton(const UnicodeString& skeleton) const {
1285    PtnElem *curElem;
1286
1287    if (skeleton.length() ==0) {
1288        return emptyString;
1289    }
1290    curElem = patternMap->getHeader(skeleton.charAt(0));
1291    while ( curElem != NULL ) {
1292        if ( curElem->skeleton->getSkeleton()==skeleton ) {
1293            return curElem->pattern;
1294        }
1295        curElem=curElem->next;
1296    }
1297    return emptyString;
1298}
1299
1300StringEnumeration*
1301DateTimePatternGenerator::getBaseSkeletons(UErrorCode& status) const {
1302    StringEnumeration* baseSkeletonEnumerator = new DTSkeletonEnumeration(*patternMap, DT_BASESKELETON, status);
1303    return baseSkeletonEnumerator;
1304}
1305
1306StringEnumeration*
1307DateTimePatternGenerator::getRedundants(UErrorCode& status) {
1308    StringEnumeration* output = new DTRedundantEnumeration();
1309    const UnicodeString *pattern;
1310    PatternMapIterator it;
1311    for (it.set(*patternMap); it.hasNext(); ) {
1312        DateTimeMatcher current = it.next();
1313        pattern = patternMap->getPatternFromSkeleton(*(it.getSkeleton()));
1314        if ( isCanonicalItem(*pattern) ) {
1315            continue;
1316        }
1317        if ( skipMatcher == NULL ) {
1318            skipMatcher = new DateTimeMatcher(current);
1319        }
1320        else {
1321            *skipMatcher = current;
1322        }
1323        UnicodeString trial = getBestPattern(current.getPattern(), status);
1324        if (trial == *pattern) {
1325            ((DTRedundantEnumeration *)output)->add(*pattern, status);
1326        }
1327        if (current.equals(skipMatcher)) {
1328            continue;
1329        }
1330    }
1331    return output;
1332}
1333
1334UBool
1335DateTimePatternGenerator::isCanonicalItem(const UnicodeString& item) const {
1336    if ( item.length() != 1 ) {
1337        return FALSE;
1338    }
1339    for (int32_t i=0; i<UDATPG_FIELD_COUNT; ++i) {
1340        if (item.charAt(0)==Canonical_Items[i]) {
1341            return TRUE;
1342        }
1343    }
1344    return FALSE;
1345}
1346
1347
1348DateTimePatternGenerator*
1349DateTimePatternGenerator::clone() const {
1350    return new DateTimePatternGenerator(*this);
1351}
1352
1353PatternMap::PatternMap() {
1354   for (int32_t i=0; i < MAX_PATTERN_ENTRIES; ++i ) {
1355      boot[i]=NULL;
1356   }
1357   isDupAllowed = TRUE;
1358}
1359
1360void
1361PatternMap::copyFrom(const PatternMap& other, UErrorCode& status) {
1362    this->isDupAllowed = other.isDupAllowed;
1363    for (int32_t bootIndex=0; bootIndex<MAX_PATTERN_ENTRIES; ++bootIndex ) {
1364        PtnElem *curElem, *otherElem, *prevElem=NULL;
1365        otherElem = other.boot[bootIndex];
1366        while (otherElem!=NULL) {
1367            if ((curElem = new PtnElem(otherElem->basePattern, otherElem->pattern))==NULL) {
1368                // out of memory
1369                status = U_MEMORY_ALLOCATION_ERROR;
1370                return;
1371            }
1372            if ( this->boot[bootIndex]== NULL ) {
1373                this->boot[bootIndex] = curElem;
1374            }
1375            if ((curElem->skeleton=new PtnSkeleton(*(otherElem->skeleton))) == NULL ) {
1376                // out of memory
1377                status = U_MEMORY_ALLOCATION_ERROR;
1378                return;
1379            }
1380            curElem->skeletonWasSpecified = otherElem->skeletonWasSpecified;
1381            if (prevElem!=NULL) {
1382                prevElem->next=curElem;
1383            }
1384            curElem->next=NULL;
1385            prevElem = curElem;
1386            otherElem = otherElem->next;
1387        }
1388
1389    }
1390}
1391
1392PtnElem*
1393PatternMap::getHeader(UChar baseChar) {
1394    PtnElem* curElem;
1395
1396    if ( (baseChar >= CAP_A) && (baseChar <= CAP_Z) ) {
1397         curElem = boot[baseChar-CAP_A];
1398    }
1399    else {
1400        if ( (baseChar >=LOW_A) && (baseChar <= LOW_Z) ) {
1401            curElem = boot[26+baseChar-LOW_A];
1402        }
1403        else {
1404            return NULL;
1405        }
1406    }
1407    return curElem;
1408}
1409
1410PatternMap::~PatternMap() {
1411   for (int32_t i=0; i < MAX_PATTERN_ENTRIES; ++i ) {
1412       if (boot[i]!=NULL ) {
1413           delete boot[i];
1414           boot[i]=NULL;
1415       }
1416   }
1417}  // PatternMap destructor
1418
1419void
1420PatternMap::add(const UnicodeString& basePattern,
1421                const PtnSkeleton& skeleton,
1422                const UnicodeString& value,// mapped pattern value
1423                UBool skeletonWasSpecified,
1424                UErrorCode &status) {
1425    UChar baseChar = basePattern.charAt(0);
1426    PtnElem *curElem, *baseElem;
1427    status = U_ZERO_ERROR;
1428
1429    // the baseChar must be A-Z or a-z
1430    if ((baseChar >= CAP_A) && (baseChar <= CAP_Z)) {
1431        baseElem = boot[baseChar-CAP_A];
1432    }
1433    else {
1434        if ((baseChar >=LOW_A) && (baseChar <= LOW_Z)) {
1435            baseElem = boot[26+baseChar-LOW_A];
1436         }
1437         else {
1438             status = U_ILLEGAL_CHARACTER;
1439             return;
1440         }
1441    }
1442
1443    if (baseElem == NULL) {
1444        if ((curElem = new PtnElem(basePattern, value)) == NULL ) {
1445            // out of memory
1446            status = U_MEMORY_ALLOCATION_ERROR;
1447            return;
1448        }
1449        if (baseChar >= LOW_A) {
1450            boot[26 + (baseChar-LOW_A)] = curElem;
1451        }
1452        else {
1453            boot[baseChar-CAP_A] = curElem;
1454        }
1455        curElem->skeleton = new PtnSkeleton(skeleton);
1456        curElem->skeletonWasSpecified = skeletonWasSpecified;
1457    }
1458    if ( baseElem != NULL ) {
1459        curElem = getDuplicateElem(basePattern, skeleton, baseElem);
1460
1461        if (curElem == NULL) {
1462            // add new element to the list.
1463            curElem = baseElem;
1464            while( curElem -> next != NULL )
1465            {
1466                curElem = curElem->next;
1467            }
1468            if ((curElem->next = new PtnElem(basePattern, value)) == NULL ) {
1469                // out of memory
1470                status = U_MEMORY_ALLOCATION_ERROR;
1471                return;
1472            }
1473            curElem=curElem->next;
1474            curElem->skeleton = new PtnSkeleton(skeleton);
1475            curElem->skeletonWasSpecified = skeletonWasSpecified;
1476        }
1477        else {
1478            // Pattern exists in the list already.
1479            if ( !isDupAllowed ) {
1480                return;
1481            }
1482            // Overwrite the value.
1483            curElem->pattern = value;
1484            // It was a bug that we were not doing the following previously,
1485            // though that bug hid other problems by making things partly work.
1486            curElem->skeletonWasSpecified = skeletonWasSpecified;
1487        }
1488    }
1489}  // PatternMap::add
1490
1491// Find the pattern from the given basePattern string.
1492const UnicodeString *
1493PatternMap::getPatternFromBasePattern(UnicodeString& basePattern, UBool& skeletonWasSpecified) { // key to search for
1494   PtnElem *curElem;
1495
1496   if ((curElem=getHeader(basePattern.charAt(0)))==NULL) {
1497       return NULL;  // no match
1498   }
1499
1500   do  {
1501       if ( basePattern.compare(curElem->basePattern)==0 ) {
1502          skeletonWasSpecified = curElem->skeletonWasSpecified;
1503          return &(curElem->pattern);
1504       }
1505       curElem=curElem->next;
1506   }while (curElem != NULL);
1507
1508   return NULL;
1509}  // PatternMap::getFromBasePattern
1510
1511
1512// Find the pattern from the given skeleton.
1513// At least when this is called from getBestRaw & addPattern (in which case specifiedSkeletonPtr is non-NULL),
1514// the comparison should be based on skeleton.original (which is unique and tied to the distance measurement in bestRaw)
1515// and not skeleton.baseOriginal (which is not unique); otherwise we may pick a different skeleton than the one with the
1516// optimum distance value in getBestRaw. When this is called from public getRedundants (specifiedSkeletonPtr is NULL),
1517// for now it will continue to compare based on baseOriginal so as not to change the behavior unnecessarily.
1518const UnicodeString *
1519PatternMap::getPatternFromSkeleton(PtnSkeleton& skeleton, const PtnSkeleton** specifiedSkeletonPtr) { // key to search for
1520   PtnElem *curElem;
1521
1522   if (specifiedSkeletonPtr) {
1523       *specifiedSkeletonPtr = NULL;
1524   }
1525
1526   // find boot entry
1527   UChar baseChar='\0';
1528   for (int32_t i=0; i<UDATPG_FIELD_COUNT; ++i) {
1529       if (skeleton.baseOriginal[i].length() !=0 ) {
1530           baseChar = skeleton.baseOriginal[i].charAt(0);
1531           break;
1532       }
1533   }
1534
1535   if ((curElem=getHeader(baseChar))==NULL) {
1536       return NULL;  // no match
1537   }
1538
1539   do  {
1540       int32_t i=0;
1541       if (specifiedSkeletonPtr != NULL) { // called from DateTimePatternGenerator::getBestRaw or addPattern, use original
1542           for (i=0; i<UDATPG_FIELD_COUNT; ++i) {
1543               if (curElem->skeleton->original[i].compare(skeleton.original[i]) != 0 )
1544               {
1545                   break;
1546               }
1547           }
1548       } else { // called from DateTimePatternGenerator::getRedundants, use baseOriginal
1549           for (i=0; i<UDATPG_FIELD_COUNT; ++i) {
1550               if (curElem->skeleton->baseOriginal[i].compare(skeleton.baseOriginal[i]) != 0 )
1551               {
1552                   break;
1553               }
1554           }
1555       }
1556       if (i == UDATPG_FIELD_COUNT) {
1557           if (specifiedSkeletonPtr && curElem->skeletonWasSpecified) {
1558               *specifiedSkeletonPtr = curElem->skeleton;
1559           }
1560           return &(curElem->pattern);
1561       }
1562       curElem=curElem->next;
1563   }while (curElem != NULL);
1564
1565   return NULL;
1566}
1567
1568UBool
1569PatternMap::equals(const PatternMap& other) {
1570    if ( this==&other ) {
1571        return TRUE;
1572    }
1573    for (int32_t bootIndex=0; bootIndex<MAX_PATTERN_ENTRIES; ++bootIndex ) {
1574        if ( boot[bootIndex]==other.boot[bootIndex] ) {
1575            continue;
1576        }
1577        if ( (boot[bootIndex]==NULL)||(other.boot[bootIndex]==NULL) ) {
1578            return FALSE;
1579        }
1580        PtnElem *otherElem = other.boot[bootIndex];
1581        PtnElem *myElem = boot[bootIndex];
1582        while ((otherElem!=NULL) || (myElem!=NULL)) {
1583            if ( myElem == otherElem ) {
1584                break;
1585            }
1586            if ((otherElem==NULL) || (myElem==NULL)) {
1587                return FALSE;
1588            }
1589            if ( (myElem->basePattern != otherElem->basePattern) ||
1590                 (myElem->pattern != otherElem->pattern) ) {
1591                return FALSE;
1592            }
1593            if ((myElem->skeleton!=otherElem->skeleton)&&
1594                !myElem->skeleton->equals(*(otherElem->skeleton))) {
1595                return FALSE;
1596            }
1597            myElem = myElem->next;
1598            otherElem=otherElem->next;
1599        }
1600    }
1601    return TRUE;
1602}
1603
1604// find any key existing in the mapping table already.
1605// return TRUE if there is an existing key, otherwise return FALSE.
1606PtnElem*
1607PatternMap::getDuplicateElem(
1608            const UnicodeString &basePattern,
1609            const PtnSkeleton &skeleton,
1610            PtnElem *baseElem)  {
1611   PtnElem *curElem;
1612
1613   if ( baseElem == (PtnElem *)NULL )  {
1614         return (PtnElem*)NULL;
1615   }
1616   else {
1617         curElem = baseElem;
1618   }
1619   do {
1620     if ( basePattern.compare(curElem->basePattern)==0 ) {
1621        UBool isEqual=TRUE;
1622        for (int32_t i=0; i<UDATPG_FIELD_COUNT; ++i) {
1623            if (curElem->skeleton->type[i] != skeleton.type[i] ) {
1624                isEqual=FALSE;
1625                break;
1626            }
1627        }
1628        if (isEqual) {
1629            return curElem;
1630        }
1631     }
1632     curElem = curElem->next;
1633   } while( curElem != (PtnElem *)NULL );
1634
1635   // end of the list
1636   return (PtnElem*)NULL;
1637
1638}  // PatternMap::getDuplicateElem
1639
1640DateTimeMatcher::DateTimeMatcher(void) {
1641}
1642
1643DateTimeMatcher::~DateTimeMatcher() {}
1644
1645DateTimeMatcher::DateTimeMatcher(const DateTimeMatcher& other) {
1646    copyFrom(other.skeleton);
1647}
1648
1649
1650void
1651DateTimeMatcher::set(const UnicodeString& pattern, FormatParser* fp) {
1652    PtnSkeleton localSkeleton;
1653    return set(pattern, fp, localSkeleton);
1654}
1655
1656void
1657DateTimeMatcher::set(const UnicodeString& pattern, FormatParser* fp, PtnSkeleton& skeletonResult) {
1658    int32_t i;
1659    for (i=0; i<UDATPG_FIELD_COUNT; ++i) {
1660        skeletonResult.type[i]=NONE;
1661    }
1662    fp->set(pattern);
1663    for (i=0; i < fp->itemNumber; i++) {
1664        UnicodeString field = fp->items[i];
1665        if ( field.charAt(0) == LOW_A ) {
1666            continue;  // skip 'a'
1667        }
1668
1669        if ( fp->isQuoteLiteral(field) ) {
1670            UnicodeString quoteLiteral;
1671            fp->getQuoteLiteral(quoteLiteral, &i);
1672            continue;
1673        }
1674        int32_t canonicalIndex = fp->getCanonicalIndex(field);
1675        if (canonicalIndex < 0 ) {
1676            continue;
1677        }
1678        const dtTypeElem *row = &dtTypes[canonicalIndex];
1679        int32_t typeValue = row->field;
1680        skeletonResult.original[typeValue]=field;
1681        UChar repeatChar = row->patternChar;
1682        int32_t repeatCount = row->minLen; // #7930 removes cap at 3
1683        while (repeatCount-- > 0) {
1684            skeletonResult.baseOriginal[typeValue] += repeatChar;
1685        }
1686        int16_t subTypeValue = row->type;
1687        if ( row->type > 0) {
1688            subTypeValue += field.length();
1689        }
1690        skeletonResult.type[typeValue] = subTypeValue;
1691    }
1692    copyFrom(skeletonResult);
1693}
1694
1695void
1696DateTimeMatcher::getBasePattern(UnicodeString &result ) {
1697    result.remove(); // Reset the result first.
1698    for (int32_t i=0; i<UDATPG_FIELD_COUNT; ++i ) {
1699        if (skeleton.baseOriginal[i].length()!=0) {
1700            result += skeleton.baseOriginal[i];
1701        }
1702    }
1703}
1704
1705UnicodeString
1706DateTimeMatcher::getPattern() {
1707    UnicodeString result;
1708
1709    for (int32_t i=0; i<UDATPG_FIELD_COUNT; ++i ) {
1710        if (skeleton.original[i].length()!=0) {
1711            result += skeleton.original[i];
1712        }
1713    }
1714    return result;
1715}
1716
1717int32_t
1718DateTimeMatcher::getDistance(const DateTimeMatcher& other, int32_t includeMask, DistanceInfo& distanceInfo) {
1719    int32_t result=0;
1720    distanceInfo.clear();
1721    for (int32_t i=0; i<UDATPG_FIELD_COUNT; ++i ) {
1722        int32_t myType = (includeMask&(1<<i))==0 ? 0 : skeleton.type[i];
1723        int32_t otherType = other.skeleton.type[i];
1724        if (myType==otherType) {
1725            continue;
1726        }
1727        if (myType==0) {// and other is not
1728            result += EXTRA_FIELD;
1729            distanceInfo.addExtra(i);
1730        }
1731        else {
1732            if (otherType==0) {
1733                result += MISSING_FIELD;
1734                distanceInfo.addMissing(i);
1735            }
1736            else {
1737                result += abs(myType - otherType);
1738            }
1739        }
1740
1741    }
1742    return result;
1743}
1744
1745void
1746DateTimeMatcher::copyFrom(const PtnSkeleton& newSkeleton) {
1747    for (int32_t i=0; i<UDATPG_FIELD_COUNT; ++i) {
1748        this->skeleton.type[i]=newSkeleton.type[i];
1749        this->skeleton.original[i]=newSkeleton.original[i];
1750        this->skeleton.baseOriginal[i]=newSkeleton.baseOriginal[i];
1751    }
1752}
1753
1754void
1755DateTimeMatcher::copyFrom() {
1756    // same as clear
1757    for (int32_t i=0; i<UDATPG_FIELD_COUNT; ++i) {
1758        this->skeleton.type[i]=0;
1759        this->skeleton.original[i].remove();
1760        this->skeleton.baseOriginal[i].remove();
1761    }
1762}
1763
1764UBool
1765DateTimeMatcher::equals(const DateTimeMatcher* other) const {
1766    if (other==NULL) {
1767        return FALSE;
1768    }
1769    for (int32_t i=0; i<UDATPG_FIELD_COUNT; ++i) {
1770        if (this->skeleton.original[i]!=other->skeleton.original[i] ) {
1771            return FALSE;
1772        }
1773    }
1774    return TRUE;
1775}
1776
1777int32_t
1778DateTimeMatcher::getFieldMask() {
1779    int32_t result=0;
1780
1781    for (int32_t i=0; i<UDATPG_FIELD_COUNT; ++i) {
1782        if (skeleton.type[i]!=0) {
1783            result |= (1<<i);
1784        }
1785    }
1786    return result;
1787}
1788
1789PtnSkeleton*
1790DateTimeMatcher::getSkeletonPtr() {
1791    return &skeleton;
1792}
1793
1794FormatParser::FormatParser () {
1795    status = START;
1796    itemNumber=0;
1797}
1798
1799
1800FormatParser::~FormatParser () {
1801}
1802
1803
1804// Find the next token with the starting position and length
1805// Note: the startPos may
1806FormatParser::TokenStatus
1807FormatParser::setTokens(const UnicodeString& pattern, int32_t startPos, int32_t *len) {
1808    int32_t  curLoc = startPos;
1809    if ( curLoc >= pattern.length()) {
1810        return DONE;
1811    }
1812    // check the current char is between A-Z or a-z
1813    do {
1814        UChar c=pattern.charAt(curLoc);
1815        if ( (c>=CAP_A && c<=CAP_Z) || (c>=LOW_A && c<=LOW_Z) ) {
1816           curLoc++;
1817        }
1818        else {
1819               startPos = curLoc;
1820               *len=1;
1821               return ADD_TOKEN;
1822        }
1823
1824        if ( pattern.charAt(curLoc)!= pattern.charAt(startPos) ) {
1825            break;  // not the same token
1826        }
1827    } while(curLoc <= pattern.length());
1828    *len = curLoc-startPos;
1829    return ADD_TOKEN;
1830}
1831
1832void
1833FormatParser::set(const UnicodeString& pattern) {
1834    int32_t startPos=0;
1835    TokenStatus result=START;
1836    int32_t len=0;
1837    itemNumber =0;
1838
1839    do {
1840        result = setTokens( pattern, startPos, &len );
1841        if ( result == ADD_TOKEN )
1842        {
1843            items[itemNumber++] = UnicodeString(pattern, startPos, len );
1844            startPos += len;
1845        }
1846        else {
1847            break;
1848        }
1849    } while (result==ADD_TOKEN && itemNumber < MAX_DT_TOKEN);
1850}
1851
1852int32_t
1853FormatParser::getCanonicalIndex(const UnicodeString& s, UBool strict) {
1854    int32_t len = s.length();
1855    if (len == 0) {
1856        return -1;
1857    }
1858    UChar ch = s.charAt(0);
1859
1860    // Verify that all are the same character.
1861    for (int32_t l = 1; l < len; l++) {
1862        if (ch != s.charAt(l)) {
1863            return -1;
1864        }
1865    }
1866    int32_t i = 0;
1867    int32_t bestRow = -1;
1868    while (dtTypes[i].patternChar != '\0') {
1869        if ( dtTypes[i].patternChar != ch ) {
1870            ++i;
1871            continue;
1872        }
1873        bestRow = i;
1874        if (dtTypes[i].patternChar != dtTypes[i+1].patternChar) {
1875            return i;
1876        }
1877        if (dtTypes[i+1].minLen <= len) {
1878            ++i;
1879            continue;
1880        }
1881        return i;
1882    }
1883    return strict ? -1 : bestRow;
1884}
1885
1886UBool
1887FormatParser::isQuoteLiteral(const UnicodeString& s) const {
1888    return (UBool)(s.charAt(0)==SINGLE_QUOTE);
1889}
1890
1891// This function aussumes the current itemIndex points to the quote literal.
1892// Please call isQuoteLiteral prior to this function.
1893void
1894FormatParser::getQuoteLiteral(UnicodeString& quote, int32_t *itemIndex) {
1895    int32_t i=*itemIndex;
1896
1897    quote.remove();
1898    if (items[i].charAt(0)==SINGLE_QUOTE) {
1899        quote += items[i];
1900        ++i;
1901    }
1902    while ( i < itemNumber ) {
1903        if ( items[i].charAt(0)==SINGLE_QUOTE ) {
1904            if ( (i+1<itemNumber) && (items[i+1].charAt(0)==SINGLE_QUOTE)) {
1905                // two single quotes e.g. 'o''clock'
1906                quote += items[i++];
1907                quote += items[i++];
1908                continue;
1909            }
1910            else {
1911                quote += items[i];
1912                break;
1913            }
1914        }
1915        else {
1916            quote += items[i];
1917        }
1918        ++i;
1919    }
1920    *itemIndex=i;
1921}
1922
1923UBool
1924FormatParser::isPatternSeparator(UnicodeString& field) {
1925    for (int32_t i=0; i<field.length(); ++i ) {
1926        UChar c= field.charAt(i);
1927        if ( (c==SINGLE_QUOTE) || (c==BACKSLASH) || (c==SPACE) || (c==COLON) ||
1928             (c==QUOTATION_MARK) || (c==COMMA) || (c==HYPHEN) ||(items[i].charAt(0)==DOT) ) {
1929            continue;
1930        }
1931        else {
1932            return FALSE;
1933        }
1934    }
1935    return TRUE;
1936}
1937
1938DistanceInfo::~DistanceInfo() {}
1939
1940void
1941DistanceInfo::setTo(DistanceInfo &other) {
1942    missingFieldMask = other.missingFieldMask;
1943    extraFieldMask= other.extraFieldMask;
1944}
1945
1946PatternMapIterator::PatternMapIterator() {
1947    bootIndex = 0;
1948    nodePtr = NULL;
1949    patternMap=NULL;
1950    matcher= new DateTimeMatcher();
1951}
1952
1953
1954PatternMapIterator::~PatternMapIterator() {
1955    delete matcher;
1956}
1957
1958void
1959PatternMapIterator::set(PatternMap& newPatternMap) {
1960    this->patternMap=&newPatternMap;
1961}
1962
1963PtnSkeleton*
1964PatternMapIterator::getSkeleton() {
1965    if ( nodePtr == NULL ) {
1966        return NULL;
1967    }
1968    else {
1969        return nodePtr->skeleton;
1970    }
1971}
1972
1973UBool
1974PatternMapIterator::hasNext() {
1975    int32_t headIndex=bootIndex;
1976    PtnElem *curPtr=nodePtr;
1977
1978    if (patternMap==NULL) {
1979        return FALSE;
1980    }
1981    while ( headIndex < MAX_PATTERN_ENTRIES ) {
1982        if ( curPtr != NULL ) {
1983            if ( curPtr->next != NULL ) {
1984                return TRUE;
1985            }
1986            else {
1987                headIndex++;
1988                curPtr=NULL;
1989                continue;
1990            }
1991        }
1992        else {
1993            if ( patternMap->boot[headIndex] != NULL ) {
1994                return TRUE;
1995            }
1996            else {
1997                headIndex++;
1998                continue;
1999            }
2000        }
2001
2002    }
2003    return FALSE;
2004}
2005
2006DateTimeMatcher&
2007PatternMapIterator::next() {
2008    while ( bootIndex < MAX_PATTERN_ENTRIES ) {
2009        if ( nodePtr != NULL ) {
2010            if ( nodePtr->next != NULL ) {
2011                nodePtr = nodePtr->next;
2012                break;
2013            }
2014            else {
2015                bootIndex++;
2016                nodePtr=NULL;
2017                continue;
2018            }
2019        }
2020        else {
2021            if ( patternMap->boot[bootIndex] != NULL ) {
2022                nodePtr = patternMap->boot[bootIndex];
2023                break;
2024            }
2025            else {
2026                bootIndex++;
2027                continue;
2028            }
2029        }
2030    }
2031    if (nodePtr!=NULL) {
2032        matcher->copyFrom(*nodePtr->skeleton);
2033    }
2034    else {
2035        matcher->copyFrom();
2036    }
2037    return *matcher;
2038}
2039
2040PtnSkeleton::PtnSkeleton() {
2041}
2042
2043
2044PtnSkeleton::PtnSkeleton(const PtnSkeleton& other) {
2045    for (int32_t i=0; i<UDATPG_FIELD_COUNT; ++i) {
2046        this->type[i]=other.type[i];
2047        this->original[i]=other.original[i];
2048        this->baseOriginal[i]=other.baseOriginal[i];
2049    }
2050}
2051
2052UBool
2053PtnSkeleton::equals(const PtnSkeleton& other)  {
2054    for (int32_t i=0; i<UDATPG_FIELD_COUNT; ++i) {
2055        if ( (type[i]!= other.type[i]) ||
2056             (original[i]!=other.original[i]) ||
2057             (baseOriginal[i]!=other.baseOriginal[i]) ) {
2058            return FALSE;
2059        }
2060    }
2061    return TRUE;
2062}
2063
2064UnicodeString
2065PtnSkeleton::getSkeleton() {
2066    UnicodeString result;
2067
2068    for(int32_t i=0; i< UDATPG_FIELD_COUNT; ++i) {
2069        if (original[i].length()!=0) {
2070            result += original[i];
2071        }
2072    }
2073    return result;
2074}
2075
2076UnicodeString
2077PtnSkeleton::getBaseSkeleton() {
2078    UnicodeString result;
2079
2080    for(int32_t i=0; i< UDATPG_FIELD_COUNT; ++i) {
2081        if (baseOriginal[i].length()!=0) {
2082            result += baseOriginal[i];
2083        }
2084    }
2085    return result;
2086}
2087
2088PtnSkeleton::~PtnSkeleton() {
2089}
2090
2091PtnElem::PtnElem(const UnicodeString &basePat, const UnicodeString &pat) :
2092basePattern(basePat),
2093skeleton(NULL),
2094pattern(pat),
2095next(NULL)
2096{
2097}
2098
2099PtnElem::~PtnElem() {
2100
2101    if (next!=NULL) {
2102        delete next;
2103    }
2104    delete skeleton;
2105}
2106
2107DTSkeletonEnumeration::DTSkeletonEnumeration(PatternMap &patternMap, dtStrEnum type, UErrorCode& status) {
2108    PtnElem  *curElem;
2109    PtnSkeleton *curSkeleton;
2110    UnicodeString s;
2111    int32_t bootIndex;
2112
2113    pos=0;
2114    fSkeletons = new UVector(status);
2115    if (U_FAILURE(status)) {
2116        delete fSkeletons;
2117        return;
2118    }
2119    for (bootIndex=0; bootIndex<MAX_PATTERN_ENTRIES; ++bootIndex ) {
2120        curElem = patternMap.boot[bootIndex];
2121        while (curElem!=NULL) {
2122            switch(type) {
2123                case DT_BASESKELETON:
2124                    s=curElem->basePattern;
2125                    break;
2126                case DT_PATTERN:
2127                    s=curElem->pattern;
2128                    break;
2129                case DT_SKELETON:
2130                    curSkeleton=curElem->skeleton;
2131                    s=curSkeleton->getSkeleton();
2132                    break;
2133            }
2134            if ( !isCanonicalItem(s) ) {
2135                fSkeletons->addElement(new UnicodeString(s), status);
2136                if (U_FAILURE(status)) {
2137                    delete fSkeletons;
2138                    fSkeletons = NULL;
2139                    return;
2140                }
2141            }
2142            curElem = curElem->next;
2143        }
2144    }
2145    if ((bootIndex==MAX_PATTERN_ENTRIES) && (curElem!=NULL) ) {
2146        status = U_BUFFER_OVERFLOW_ERROR;
2147    }
2148}
2149
2150const UnicodeString*
2151DTSkeletonEnumeration::snext(UErrorCode& status) {
2152    if (U_SUCCESS(status) && pos < fSkeletons->size()) {
2153        return (const UnicodeString*)fSkeletons->elementAt(pos++);
2154    }
2155    return NULL;
2156}
2157
2158void
2159DTSkeletonEnumeration::reset(UErrorCode& /*status*/) {
2160    pos=0;
2161}
2162
2163int32_t
2164DTSkeletonEnumeration::count(UErrorCode& /*status*/) const {
2165   return (fSkeletons==NULL) ? 0 : fSkeletons->size();
2166}
2167
2168UBool
2169DTSkeletonEnumeration::isCanonicalItem(const UnicodeString& item) {
2170    if ( item.length() != 1 ) {
2171        return FALSE;
2172    }
2173    for (int32_t i=0; i<UDATPG_FIELD_COUNT; ++i) {
2174        if (item.charAt(0)==Canonical_Items[i]) {
2175            return TRUE;
2176        }
2177    }
2178    return FALSE;
2179}
2180
2181DTSkeletonEnumeration::~DTSkeletonEnumeration() {
2182    UnicodeString *s;
2183    for (int32_t i=0; i<fSkeletons->size(); ++i) {
2184        if ((s=(UnicodeString *)fSkeletons->elementAt(i))!=NULL) {
2185            delete s;
2186        }
2187    }
2188    delete fSkeletons;
2189}
2190
2191DTRedundantEnumeration::DTRedundantEnumeration() {
2192    pos=0;
2193    fPatterns = NULL;
2194}
2195
2196void
2197DTRedundantEnumeration::add(const UnicodeString& pattern, UErrorCode& status) {
2198    if (U_FAILURE(status)) return;
2199    if (fPatterns == NULL)  {
2200        fPatterns = new UVector(status);
2201        if (U_FAILURE(status)) {
2202            delete fPatterns;
2203            fPatterns = NULL;
2204            return;
2205       }
2206    }
2207    fPatterns->addElement(new UnicodeString(pattern), status);
2208    if (U_FAILURE(status)) {
2209        delete fPatterns;
2210        fPatterns = NULL;
2211        return;
2212    }
2213}
2214
2215const UnicodeString*
2216DTRedundantEnumeration::snext(UErrorCode& status) {
2217    if (U_SUCCESS(status) && pos < fPatterns->size()) {
2218        return (const UnicodeString*)fPatterns->elementAt(pos++);
2219    }
2220    return NULL;
2221}
2222
2223void
2224DTRedundantEnumeration::reset(UErrorCode& /*status*/) {
2225    pos=0;
2226}
2227
2228int32_t
2229DTRedundantEnumeration::count(UErrorCode& /*status*/) const {
2230       return (fPatterns==NULL) ? 0 : fPatterns->size();
2231}
2232
2233UBool
2234DTRedundantEnumeration::isCanonicalItem(const UnicodeString& item) {
2235    if ( item.length() != 1 ) {
2236        return FALSE;
2237    }
2238    for (int32_t i=0; i<UDATPG_FIELD_COUNT; ++i) {
2239        if (item.charAt(0)==Canonical_Items[i]) {
2240            return TRUE;
2241        }
2242    }
2243    return FALSE;
2244}
2245
2246DTRedundantEnumeration::~DTRedundantEnumeration() {
2247    UnicodeString *s;
2248    for (int32_t i=0; i<fPatterns->size(); ++i) {
2249        if ((s=(UnicodeString *)fPatterns->elementAt(i))!=NULL) {
2250            delete s;
2251        }
2252    }
2253    delete fPatterns;
2254}
2255
2256U_NAMESPACE_END
2257
2258
2259#endif /* #if !UCONFIG_NO_FORMATTING */
2260
2261//eof
2262