// © 2016 and later: Unicode, Inc. and others. // License & terms of use: http://www.unicode.org/copyright.html /*********************************************************************** * COPYRIGHT: * Copyright (c) 1997-2015, International Business Machines Corporation * and others. All Rights Reserved. ***********************************************************************/ #include "unicode/utypes.h" #if !UCONFIG_NO_FORMATTING #include "unicode/datefmt.h" #include "unicode/smpdtfmt.h" #include "unicode/gregocal.h" #include "dtfmtrtts.h" #include "caltest.h" #include "cstring.h" #include #include // ***************************************************************************** // class DateFormatRoundTripTest // ***************************************************************************** // Useful for turning up subtle bugs: Change the following to TRUE, recompile, // and run while at lunch. // Warning -- makes test run infinite loop!!! #ifndef INFINITE #define INFINITE 0 #endif // Define this to test just a single locale //#define TEST_ONE_LOC "cs_CZ" // If SPARSENESS is > 0, we don't run each exhaustive possibility. // There are 24 total possible tests per each locale. A SPARSENESS // of 12 means we run half of them. A SPARSENESS of 23 means we run // 1 of them. SPARSENESS _must_ be in the range 0..23. int32_t DateFormatRoundTripTest::SPARSENESS = 0; int32_t DateFormatRoundTripTest::TRIALS = 4; int32_t DateFormatRoundTripTest::DEPTH = 5; DateFormatRoundTripTest::DateFormatRoundTripTest() : dateFormat(0) { } DateFormatRoundTripTest::~DateFormatRoundTripTest() { delete dateFormat; } #define CASE(id,test) case id: name = #test; if (exec) { logln(#test "---"); logln((UnicodeString)""); test(); } break; void DateFormatRoundTripTest::runIndexedTest( int32_t index, UBool exec, const char* &name, char* par ) { optionv = (par && *par=='v'); switch (index) { CASE(0,TestDateFormatRoundTrip) CASE(1, TestCentury) default: name = ""; break; } } UBool DateFormatRoundTripTest::failure(UErrorCode status, const char* msg) { if(U_FAILURE(status)) { errln(UnicodeString("FAIL: ") + msg + " failed, error " + u_errorName(status)); return TRUE; } return FALSE; } UBool DateFormatRoundTripTest::failure(UErrorCode status, const char* msg, const UnicodeString& str) { if(U_FAILURE(status)) { UnicodeString escaped; escape(str,escaped); errln(UnicodeString("FAIL: ") + msg + " failed, error " + u_errorName(status) + ", str=" + escaped); return TRUE; } return FALSE; } void DateFormatRoundTripTest::TestCentury() { UErrorCode status = U_ZERO_ERROR; Locale locale("es_PA"); UnicodeString pattern = "MM/dd/yy hh:mm:ss a z"; SimpleDateFormat fmt(pattern, locale, status); if (U_FAILURE(status)) { dataerrln("Fail: construct SimpleDateFormat: %s", u_errorName(status)); return; } UDate date[] = {-55018555891590.05, 0, 0}; UnicodeString result[2]; fmt.format(date[0], result[0]); date[1] = fmt.parse(result[0], status); fmt.format(date[1], result[1]); date[2] = fmt.parse(result[1], status); /* This test case worked OK by accident before. date[1] != date[0], * because we use -80/+20 year window for 2-digit year parsing. * (date[0] is in year 1926, date[1] is in year 2026.) result[1] set * by the first format call returns "07/13/26 07:48:28 p.m. PST", * which is correct, because DST was not used in year 1926 in zone * America/Los_Angeles. When this is parsed, date[1] becomes a time * in 2026, which is "07/13/26 08:48:28 p.m. PDT". There was a zone * offset calculation bug that observed DST in 1926, which was resolved. * Before the bug was resolved, result[0] == result[1] was true, * but after the bug fix, the expected result is actually * result[0] != result[1]. -Yoshito */ /* TODO: We need to review this code and clarify what we really * want to test here. */ //if (date[1] != date[2] || result[0] != result[1]) { if (date[1] != date[2]) { errln("Round trip failure: \"%S\" (%f), \"%S\" (%f)", result[0].getBuffer(), date[1], result[1].getBuffer(), date[2]); } } // == void DateFormatRoundTripTest::TestDateFormatRoundTrip() { UErrorCode status = U_ZERO_ERROR; getFieldCal = Calendar::createInstance(status); if (U_FAILURE(status)) { dataerrln("Fail: Calendar::createInstance: %s", u_errorName(status)); return; } int32_t locCount = 0; const Locale *avail = DateFormat::getAvailableLocales(locCount); logln("DateFormat available locales: %d", locCount); if(quick) { SPARSENESS = 18; logln("Quick mode: only testing SPARSENESS = 18"); } TimeZone *tz = TimeZone::createDefault(); UnicodeString temp; logln("Default TimeZone: " + tz->getID(temp)); delete tz; #ifdef TEST_ONE_LOC // define this to just test ONE locale. Locale loc(TEST_ONE_LOC); test(loc); #if INFINITE for(;;) { test(loc); } #endif #else # if INFINITE // Special infinite loop test mode for finding hard to reproduce errors Locale loc = Locale::getDefault(); logln("ENTERING INFINITE TEST LOOP FOR Locale: " + loc.getDisplayName(temp)); for(;;) test(loc); # else test(Locale::getDefault()); #if 1 // installed locales for (int i=0; i < locCount; ++i) { test(avail[i]); } #endif #if 1 // special locales int32_t jCount = CalendarTest::testLocaleCount(); for (int32_t j=0; j < jCount; ++j) { test(Locale(CalendarTest::testLocaleID(j))); } #endif # endif #endif delete getFieldCal; } static const char *styleName(DateFormat::EStyle s) { switch(s) { case DateFormat::SHORT: return "SHORT"; case DateFormat::MEDIUM: return "MEDIUM"; case DateFormat::LONG: return "LONG"; case DateFormat::FULL: return "FULL"; // case DateFormat::DEFAULT: return "DEFAULT"; case DateFormat::DATE_OFFSET: return "DATE_OFFSET"; case DateFormat::NONE: return "NONE"; case DateFormat::DATE_TIME: return "DATE_TIME"; default: return "Unknown"; } } void DateFormatRoundTripTest::test(const Locale& loc) { UnicodeString temp; #if !INFINITE logln("Locale: " + loc.getDisplayName(temp)); #endif // Total possibilities = 24 // 4 date // 4 time // 16 date-time UBool TEST_TABLE [24];//= new boolean[24]; int32_t i = 0; for(i = 0; i < 24; ++i) TEST_TABLE[i] = TRUE; // If we have some sparseness, implement it here. Sparseness decreases // test time by eliminating some tests, up to 23. for(i = 0; i < SPARSENESS; ) { int random = (int)(randFraction() * 24); if (random >= 0 && random < 24 && TEST_TABLE[i]) { TEST_TABLE[i] = FALSE; ++i; } } int32_t itable = 0; int32_t style = 0; for(style = DateFormat::FULL; style <= DateFormat::SHORT; ++style) { if(TEST_TABLE[itable++]) { logln("Testing style " + UnicodeString(styleName((DateFormat::EStyle)style))); DateFormat *df = DateFormat::createDateInstance((DateFormat::EStyle)style, loc); if(df == NULL) { errln(UnicodeString("Could not DF::createDateInstance ") + UnicodeString(styleName((DateFormat::EStyle)style)) + " Locale: " + loc.getDisplayName(temp)); } else { test(df, loc); delete df; } } } for(style = DateFormat::FULL; style <= DateFormat::SHORT; ++style) { if (TEST_TABLE[itable++]) { logln("Testing style " + UnicodeString(styleName((DateFormat::EStyle)style))); DateFormat *df = DateFormat::createTimeInstance((DateFormat::EStyle)style, loc); if(df == NULL) { errln(UnicodeString("Could not DF::createTimeInstance ") + UnicodeString(styleName((DateFormat::EStyle)style)) + " Locale: " + loc.getDisplayName(temp)); } else { test(df, loc, TRUE); delete df; } } } for(int32_t dstyle = DateFormat::FULL; dstyle <= DateFormat::SHORT; ++dstyle) { for(int32_t tstyle = DateFormat::FULL; tstyle <= DateFormat::SHORT; ++tstyle) { if(TEST_TABLE[itable++]) { logln("Testing dstyle" + UnicodeString(styleName((DateFormat::EStyle)dstyle)) + ", tstyle" + UnicodeString(styleName((DateFormat::EStyle)tstyle)) ); DateFormat *df = DateFormat::createDateTimeInstance((DateFormat::EStyle)dstyle, (DateFormat::EStyle)tstyle, loc); if(df == NULL) { dataerrln(UnicodeString("Could not DF::createDateTimeInstance ") + UnicodeString(styleName((DateFormat::EStyle)dstyle)) + ", tstyle" + UnicodeString(styleName((DateFormat::EStyle)tstyle)) + "Locale: " + loc.getDisplayName(temp)); } else { test(df, loc); delete df; } } } } } void DateFormatRoundTripTest::test(DateFormat *fmt, const Locale &origLocale, UBool timeOnly) { UnicodeString pat; if(fmt->getDynamicClassID() != SimpleDateFormat::getStaticClassID()) { errln("DateFormat wasn't a SimpleDateFormat"); return; } UBool isGregorian = FALSE; UErrorCode minStatus = U_ZERO_ERROR; if(fmt->getCalendar() == NULL) { errln((UnicodeString)"DateFormatRoundTripTest::test, DateFormat getCalendar() returns null for " + origLocale.getName()); return; } UDate minDate = CalendarTest::minDateOfCalendar(*fmt->getCalendar(), isGregorian, minStatus); if(U_FAILURE(minStatus)) { errln((UnicodeString)"Failure getting min date for " + origLocale.getName()); return; } //logln(UnicodeString("Min date is ") + fullFormat(minDate) + " for " + origLocale.getName()); pat = ((SimpleDateFormat*)fmt)->toPattern(pat); // NOTE TO MAINTAINER // This indexOf check into the pattern needs to be refined to ignore // quoted characters. Currently, this isn't a problem with the locale // patterns we have, but it may be a problem later. UBool hasEra = (pat.indexOf(UnicodeString("G")) != -1); UBool hasZoneDisplayName = (pat.indexOf(UnicodeString("z")) != -1) || (pat.indexOf(UnicodeString("v")) != -1) || (pat.indexOf(UnicodeString("V")) != -1); // Because patterns contain incomplete data representing the Date, // we must be careful of how we do the roundtrip. We start with // a randomly generated Date because they're easier to generate. // From this we get a string. The string is our real starting point, // because this string should parse the same way all the time. Note // that it will not necessarily parse back to the original date because // of incompleteness in patterns. For example, a time-only pattern won't // parse back to the same date. //try { for(int i = 0; i < TRIALS; ++i) { UDate *d = new UDate [DEPTH]; UnicodeString *s = new UnicodeString[DEPTH]; if(isGregorian == TRUE) { d[0] = generateDate(); } else { d[0] = generateDate(minDate); } UErrorCode status = U_ZERO_ERROR; // We go through this loop until we achieve a match or until // the maximum loop count is reached. We record the points at // which the date and the string starts to match. Once matching // starts, it should continue. int loop; int dmatch = 0; // d[dmatch].getTime() == d[dmatch-1].getTime() int smatch = 0; // s[smatch].equals(s[smatch-1]) for(loop = 0; loop < DEPTH; ++loop) { if (loop > 0) { d[loop] = fmt->parse(s[loop-1], status); if(U_FAILURE(status)) { ParsePosition ppos; (void)fmt->parse(s[loop-1], ppos); failure(status, "fmt->parse", s[loop-1]+" in locale: " + origLocale.getName() + " with pattern: " + pat + " has errIndex: " + ppos.getErrorIndex()); status = U_ZERO_ERROR; /* any error would have been reported */ } } s[loop] = fmt->format(d[loop], s[loop]); // For displaying which date is being tested //logln(s[loop] + " = " + fullFormat(d[loop])); if(s[loop].length() == 0) { errln("FAIL: fmt->format gave 0-length string in " + pat + " with number " + d[loop] + " in locale " + origLocale.getName()); } if(loop > 0) { if(smatch == 0) { UBool match = s[loop] == s[loop-1]; if(smatch == 0) { if(match) smatch = loop; } else if( ! match) errln("FAIL: String mismatch after match"); } if(dmatch == 0) { // {sfb} watch out here, this might not work UBool match = d[loop]/*.getTime()*/ == d[loop-1]/*.getTime()*/; if(dmatch == 0) { if(match) dmatch = loop; } else if( ! match) errln("FAIL: Date mismatch after match"); } if(smatch != 0 && dmatch != 0) break; } } // At this point loop == DEPTH if we've failed, otherwise loop is the // max(smatch, dmatch), that is, the index at which we have string and // date matching. // Date usually matches in 2. Exceptions handled below. int maxDmatch = 2; int maxSmatch = 1; if (dmatch > maxDmatch) { // Time-only pattern with zone information and a starting date in PST. if(timeOnly && hasZoneDisplayName) { int32_t startRaw, startDst; fmt->getTimeZone().getOffset(d[0], FALSE, startRaw, startDst, status); failure(status, "TimeZone::getOffset"); // if the start offset is greater than the offset on Jan 1, 1970 // in PST, then need one more round trip. There are two cases // fall into this category. The start date is 1) DST or // 2) LMT (GMT-07:52:58). if (startRaw + startDst > -28800000) { maxDmatch = 3; maxSmatch = 2; } } } // String usually matches in 1. Exceptions are checked for here. if(smatch > maxSmatch) { // Don't compute unless necessary UBool in0; // Starts in BC, with no era in pattern if( ! hasEra && getField(d[0], UCAL_ERA) == GregorianCalendar::BC) maxSmatch = 2; // Starts in DST, no year in pattern else if((in0=fmt->getTimeZone().inDaylightTime(d[0], status)) && ! failure(status, "gettingDaylightTime") && pat.indexOf(UnicodeString("yyyy")) == -1) maxSmatch = 2; // If we start not in DST, but transition into DST else if (!in0 && fmt->getTimeZone().inDaylightTime(d[1], status) && !failure(status, "gettingDaylightTime")) maxSmatch = 2; // Two digit year with no time zone change, // unless timezone isn't used or we aren't close to the DST changover else if (pat.indexOf(UnicodeString("y")) != -1 && pat.indexOf(UnicodeString("yyyy")) == -1 && getField(d[0], UCAL_YEAR) != getField(d[dmatch], UCAL_YEAR) && !failure(status, "error status [smatch>maxSmatch]") && ((hasZoneDisplayName && (fmt->getTimeZone().inDaylightTime(d[0], status) == fmt->getTimeZone().inDaylightTime(d[dmatch], status) || getField(d[0], UCAL_MONTH) == UCAL_APRIL || getField(d[0], UCAL_MONTH) == UCAL_OCTOBER)) || !hasZoneDisplayName) ) { maxSmatch = 2; } // If zone display name is used, fallback format might be used before 1970 else if (hasZoneDisplayName && d[0] < 0) { maxSmatch = 2; } else if (timeOnly && !isGregorian && hasZoneDisplayName && maxSmatch == 1) { int32_t startRaw, startDst; fmt->getTimeZone().getOffset(d[1], FALSE, startRaw, startDst, status); failure(status, "TimeZone::getOffset"); // If the calendar type is not Gregorian and the pattern is time only, // the calendar implementation may use a date before 1970 as day 0. // In this case, time zone offset of the default year might be // different from the one at 1970-01-01 in PST and string match requires // one more iteration. if (startRaw + startDst != -28800000) { maxSmatch = 2; } } } /* * Special case for Japanese and Buddhist (could have large negative years) * Also, Hebrew calendar need help handling leap month. */ if(dmatch > maxDmatch || smatch > maxSmatch) { const char *type = fmt->getCalendar()->getType(); if(!strcmp(type,"japanese") || (!strcmp(type,"buddhist"))) { maxSmatch = 4; maxDmatch = 4; } else if(!strcmp(type,"hebrew")) { maxSmatch = 3; maxDmatch = 3; } else if (timeOnly && uprv_strcmp(origLocale.getName(),"ar_JO@calendar=islamic-civil")==0 && logKnownIssue("21049", "ar_JO@calendar=islamic-civil timeOnly roundtrip converges too slowly")) { // For some reason, for time-only tests, ar_JO@calendar=islamic-civil is no // longer converging to a match as fast as expected above. Investigate with // ICU-21022, but meanwhile allow more cycles for convergence. maxSmatch = 2; maxDmatch = 3; } } // Use @v to see verbose results on successful cases UBool fail = (dmatch > maxDmatch || smatch > maxSmatch); if (optionv || fail) { if (fail) { errln(UnicodeString("\nFAIL: Pattern: ") + pat + " in Locale: " + origLocale.getName() + "\nget dmatch: " + dmatch + " (expected max " + maxDmatch + "), smatch: " + smatch + " (expected max " + maxSmatch + ")"); } else { errln(UnicodeString("\nOk: Pattern: ") + pat + " in Locale: " + origLocale.getName()); } logln("Date iters until match=%d (max allowed=%d), string iters until match=%d (max allowed=%d)", dmatch,maxDmatch, smatch, maxSmatch); for(int j = 0; j <= loop && j < DEPTH; ++j) { UnicodeString temp; FieldPosition pos(FieldPosition::DONT_CARE); errln((j>0?" P> ":" ") + fullFormat(d[j]) + " F> " + escape(s[j], temp) + UnicodeString(" d=") + d[j] + (j > 0 && d[j]/*.getTime()*/==d[j-1]/*.getTime()*/?" d==":"") + (j > 0 && s[j] == s[j-1]?" s==":"")); } } delete[] d; delete[] s; } /*} catch (ParseException e) { errln("Exception: " + e.getMessage()); logln(e.toString()); }*/ } const UnicodeString& DateFormatRoundTripTest::fullFormat(UDate d) { UErrorCode ec = U_ZERO_ERROR; if (dateFormat == 0) { dateFormat = new SimpleDateFormat((UnicodeString)"EEE MMM dd HH:mm:ss.SSS zzz yyyy G", ec); if (U_FAILURE(ec) || dateFormat == 0) { fgStr = "[FAIL: SimpleDateFormat constructor]"; delete dateFormat; dateFormat = 0; return fgStr; } } fgStr.truncate(0); dateFormat->format(d, fgStr); return fgStr; } /** * Return a field of the given date */ int32_t DateFormatRoundTripTest::getField(UDate d, int32_t f) { // Should be synchronized, but we're single threaded so it's ok UErrorCode status = U_ZERO_ERROR; getFieldCal->setTime(d, status); failure(status, "getfieldCal->setTime"); int32_t ret = getFieldCal->get((UCalendarDateFields)f, status); failure(status, "getfieldCal->get"); return ret; } UnicodeString& DateFormatRoundTripTest::escape(const UnicodeString& src, UnicodeString& dst ) { dst.remove(); for (int32_t i = 0; i < src.length(); ++i) { UChar32 c = src.char32At(i); if (c >= 0x10000) { ++i; } if (c < 0x0080) { dst += c; } else { dst += UnicodeString("["); char buf [12]; sprintf(buf, "%#04x", c); dst += UnicodeString(buf); dst += UnicodeString("]"); } } return dst; } #define U_MILLIS_PER_YEAR (365.25 * 24 * 60 * 60 * 1000) UDate DateFormatRoundTripTest::generateDate(UDate minDate) { // Bring range in conformance to generateDate() below. if(minDate < (U_MILLIS_PER_YEAR * -(4000-1970))) { minDate = (U_MILLIS_PER_YEAR * -(4000-1970)); } for(int i=0;i<8;i++) { double a = randFraction(); // Range from (min) to (8000-1970) AD double dateRange = (0.0 - minDate) + (U_MILLIS_PER_YEAR + (8000-1970)); a *= dateRange; // Now offset from minDate a += minDate; // Last sanity check if(a>=minDate) { return a; } } return minDate; } UDate DateFormatRoundTripTest::generateDate() { double a = randFraction(); // Now 'a' ranges from 0..1; scale it to range from 0 to 8000 years a *= 8000; // Range from (4000-1970) BC to (8000-1970) AD a -= 4000; // Now scale up to ms a *= 365.25 * 24 * 60 * 60 * 1000; //return new Date((long)a); return a; } #endif /* #if !UCONFIG_NO_FORMATTING */ //eof