• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2 ******************************************************************************
3 *   Copyright (C) 1997-2008, International Business Machines
4 *   Corporation and others.  All Rights Reserved.
5 ******************************************************************************
6 *   file name:  nfrs.cpp
7 *   encoding:   US-ASCII
8 *   tab size:   8 (not used)
9 *   indentation:4
10 *
11 * Modification history
12 * Date        Name      Comments
13 * 10/11/2001  Doug      Ported from ICU4J
14 */
15 
16 #include "nfrs.h"
17 
18 #if U_HAVE_RBNF
19 
20 #include "unicode/uchar.h"
21 #include "nfrule.h"
22 #include "nfrlist.h"
23 
24 #ifdef RBNF_DEBUG
25 #include "cmemory.h"
26 #endif
27 
28 #include "util.h"
29 
30 U_NAMESPACE_BEGIN
31 
32 #if 0
33 // euclid's algorithm works with doubles
34 // note, doubles only get us up to one quadrillion or so, which
35 // isn't as much range as we get with longs.  We probably still
36 // want either 64-bit math, or BigInteger.
37 
38 static int64_t
39 util_lcm(int64_t x, int64_t y)
40 {
41     x.abs();
42     y.abs();
43 
44     if (x == 0 || y == 0) {
45         return 0;
46     } else {
47         do {
48             if (x < y) {
49                 int64_t t = x; x = y; y = t;
50             }
51             x -= y * (x/y);
52         } while (x != 0);
53 
54         return y;
55     }
56 }
57 
58 #else
59 /**
60  * Calculates the least common multiple of x and y.
61  */
62 static int64_t
util_lcm(int64_t x,int64_t y)63 util_lcm(int64_t x, int64_t y)
64 {
65     // binary gcd algorithm from Knuth, "The Art of Computer Programming,"
66     // vol. 2, 1st ed., pp. 298-299
67     int64_t x1 = x;
68     int64_t y1 = y;
69 
70     int p2 = 0;
71     while ((x1 & 1) == 0 && (y1 & 1) == 0) {
72         ++p2;
73         x1 >>= 1;
74         y1 >>= 1;
75     }
76 
77     int64_t t;
78     if ((x1 & 1) == 1) {
79         t = -y1;
80     } else {
81         t = x1;
82     }
83 
84     while (t != 0) {
85         while ((t & 1) == 0) {
86             t = t >> 1;
87         }
88         if (t > 0) {
89             x1 = t;
90         } else {
91             y1 = -t;
92         }
93         t = x1 - y1;
94     }
95 
96     int64_t gcd = x1 << p2;
97 
98     // x * y == gcd(x, y) * lcm(x, y)
99     return x / gcd * y;
100 }
101 #endif
102 
103 static const UChar gPercent = 0x0025;
104 static const UChar gColon = 0x003a;
105 static const UChar gSemicolon = 0x003b;
106 static const UChar gLineFeed = 0x000a;
107 
108 static const UChar gFourSpaces[] =
109 {
110     0x20, 0x20, 0x20, 0x20, 0
111 }; /* "    " */
112 static const UChar gPercentPercent[] =
113 {
114     0x25, 0x25, 0
115 }; /* "%%" */
116 
NFRuleSet(UnicodeString * descriptions,int32_t index,UErrorCode & status)117 NFRuleSet::NFRuleSet(UnicodeString* descriptions, int32_t index, UErrorCode& status)
118   : name()
119   , rules(0)
120   , negativeNumberRule(NULL)
121   , fIsFractionRuleSet(FALSE)
122   , fIsPublic(FALSE)
123   , fRecursionCount(0)
124 {
125     for (int i = 0; i < 3; ++i) {
126         fractionRules[i] = NULL;
127     }
128 
129     if (U_FAILURE(status)) {
130         return;
131     }
132 
133     UnicodeString& description = descriptions[index]; // !!! make sure index is valid
134 
135     if (description.length() == 0) {
136         // throw new IllegalArgumentException("Empty rule set description");
137         status = U_PARSE_ERROR;
138         return;
139     }
140 
141     // if the description begins with a rule set name (the rule set
142     // name can be omitted in formatter descriptions that consist
143     // of only one rule set), copy it out into our "name" member
144     // and delete it from the description
145     if (description.charAt(0) == gPercent) {
146         int32_t pos = description.indexOf(gColon);
147         if (pos == -1) {
148             // throw new IllegalArgumentException("Rule set name doesn't end in colon");
149             status = U_PARSE_ERROR;
150         } else {
151             name.setTo(description, 0, pos);
152             while (pos < description.length() && uprv_isRuleWhiteSpace(description.charAt(++pos))) {
153             }
154             description.remove(0, pos);
155         }
156     } else {
157         name.setTo(UNICODE_STRING_SIMPLE("%default"));
158     }
159 
160     if (description.length() == 0) {
161         // throw new IllegalArgumentException("Empty rule set description");
162         status = U_PARSE_ERROR;
163     }
164 
165     fIsPublic = name.indexOf(gPercentPercent) != 0;
166 
167     // all of the other members of NFRuleSet are initialized
168     // by parseRules()
169 }
170 
171 void
parseRules(UnicodeString & description,const RuleBasedNumberFormat * owner,UErrorCode & status)172 NFRuleSet::parseRules(UnicodeString& description, const RuleBasedNumberFormat* owner, UErrorCode& status)
173 {
174     // start by creating a Vector whose elements are Strings containing
175     // the descriptions of the rules (one rule per element).  The rules
176     // are separated by semicolons (there's no escape facility: ALL
177     // semicolons are rule delimiters)
178 
179     if (U_FAILURE(status)) {
180         return;
181     }
182 
183     // dlf - the original code kept a separate description array for no reason,
184     // so I got rid of it.  The loop was too complex so I simplified it.
185 
186     UnicodeString currentDescription;
187     int32_t oldP = 0;
188     while (oldP < description.length()) {
189         int32_t p = description.indexOf(gSemicolon, oldP);
190         if (p == -1) {
191             p = description.length();
192         }
193         currentDescription.setTo(description, oldP, p - oldP);
194         NFRule::makeRules(currentDescription, this, rules.last(), owner, rules, status);
195         oldP = p + 1;
196     }
197 
198     // for rules that didn't specify a base value, their base values
199     // were initialized to 0.  Make another pass through the list and
200     // set all those rules' base values.  We also remove any special
201     // rules from the list and put them into their own member variables
202     int64_t defaultBaseValue = 0;
203 
204     // (this isn't a for loop because we might be deleting items from
205     // the vector-- we want to make sure we only increment i when
206     // we _didn't_ delete aything from the vector)
207     uint32_t i = 0;
208     while (i < rules.size()) {
209         NFRule* rule = rules[i];
210 
211         switch (rule->getType()) {
212             // if the rule's base value is 0, fill in a default
213             // base value (this will be 1 plus the preceding
214             // rule's base value for regular rule sets, and the
215             // same as the preceding rule's base value in fraction
216             // rule sets)
217         case NFRule::kNoBase:
218             rule->setBaseValue(defaultBaseValue, status);
219             if (!isFractionRuleSet()) {
220                 ++defaultBaseValue;
221             }
222             ++i;
223             break;
224 
225             // if it's the negative-number rule, copy it into its own
226             // data member and delete it from the list
227         case NFRule::kNegativeNumberRule:
228             negativeNumberRule = rules.remove(i);
229             break;
230 
231             // if it's the improper fraction rule, copy it into the
232             // correct element of fractionRules
233         case NFRule::kImproperFractionRule:
234             fractionRules[0] = rules.remove(i);
235             break;
236 
237             // if it's the proper fraction rule, copy it into the
238             // correct element of fractionRules
239         case NFRule::kProperFractionRule:
240             fractionRules[1] = rules.remove(i);
241             break;
242 
243             // if it's the master rule, copy it into the
244             // correct element of fractionRules
245         case NFRule::kMasterRule:
246             fractionRules[2] = rules.remove(i);
247             break;
248 
249             // if it's a regular rule that already knows its base value,
250             // check to make sure the rules are in order, and update
251             // the default base value for the next rule
252         default:
253             if (rule->getBaseValue() < defaultBaseValue) {
254                 // throw new IllegalArgumentException("Rules are not in order");
255                 status = U_PARSE_ERROR;
256                 return;
257             }
258             defaultBaseValue = rule->getBaseValue();
259             if (!isFractionRuleSet()) {
260                 ++defaultBaseValue;
261             }
262             ++i;
263             break;
264         }
265     }
266 }
267 
~NFRuleSet()268 NFRuleSet::~NFRuleSet()
269 {
270     delete negativeNumberRule;
271     delete fractionRules[0];
272     delete fractionRules[1];
273     delete fractionRules[2];
274 }
275 
276 static UBool
util_equalRules(const NFRule * rule1,const NFRule * rule2)277 util_equalRules(const NFRule* rule1, const NFRule* rule2)
278 {
279     if (rule1) {
280         if (rule2) {
281             return *rule1 == *rule2;
282         }
283     } else if (!rule2) {
284         return TRUE;
285     }
286     return FALSE;
287 }
288 
289 UBool
operator ==(const NFRuleSet & rhs) const290 NFRuleSet::operator==(const NFRuleSet& rhs) const
291 {
292     if (rules.size() == rhs.rules.size() &&
293         fIsFractionRuleSet == rhs.fIsFractionRuleSet &&
294         name == rhs.name &&
295         util_equalRules(negativeNumberRule, rhs.negativeNumberRule) &&
296         util_equalRules(fractionRules[0], rhs.fractionRules[0]) &&
297         util_equalRules(fractionRules[1], rhs.fractionRules[1]) &&
298         util_equalRules(fractionRules[2], rhs.fractionRules[2])) {
299 
300         for (uint32_t i = 0; i < rules.size(); ++i) {
301             if (*rules[i] != *rhs.rules[i]) {
302                 return FALSE;
303             }
304         }
305         return TRUE;
306     }
307     return FALSE;
308 }
309 
310 #define RECURSION_LIMIT 50
311 
312 void
format(int64_t number,UnicodeString & toAppendTo,int32_t pos) const313 NFRuleSet::format(int64_t number, UnicodeString& toAppendTo, int32_t pos) const
314 {
315     NFRule *rule = findNormalRule(number);
316     if (rule) { // else error, but can't report it
317         NFRuleSet* ncThis = (NFRuleSet*)this;
318         if (ncThis->fRecursionCount++ >= RECURSION_LIMIT) {
319             // stop recursion
320             ncThis->fRecursionCount = 0;
321         } else {
322             rule->doFormat(number, toAppendTo, pos);
323             ncThis->fRecursionCount--;
324         }
325     }
326 }
327 
328 void
format(double number,UnicodeString & toAppendTo,int32_t pos) const329 NFRuleSet::format(double number, UnicodeString& toAppendTo, int32_t pos) const
330 {
331     NFRule *rule = findDoubleRule(number);
332     if (rule) { // else error, but can't report it
333         NFRuleSet* ncThis = (NFRuleSet*)this;
334         if (ncThis->fRecursionCount++ >= RECURSION_LIMIT) {
335             // stop recursion
336             ncThis->fRecursionCount = 0;
337         } else {
338             rule->doFormat(number, toAppendTo, pos);
339             ncThis->fRecursionCount--;
340         }
341     }
342 }
343 
344 NFRule*
findDoubleRule(double number) const345 NFRuleSet::findDoubleRule(double number) const
346 {
347     // if this is a fraction rule set, use findFractionRuleSetRule()
348     if (isFractionRuleSet()) {
349         return findFractionRuleSetRule(number);
350     }
351 
352     // if the number is negative, return the negative number rule
353     // (if there isn't a negative-number rule, we pretend it's a
354     // positive number)
355     if (number < 0) {
356         if (negativeNumberRule) {
357             return  negativeNumberRule;
358         } else {
359             number = -number;
360         }
361     }
362 
363     // if the number isn't an integer, we use one of the fraction rules...
364     if (number != uprv_floor(number)) {
365         // if the number is between 0 and 1, return the proper
366         // fraction rule
367         if (number < 1 && fractionRules[1]) {
368             return fractionRules[1];
369         }
370         // otherwise, return the improper fraction rule
371         else if (fractionRules[0]) {
372             return fractionRules[0];
373         }
374     }
375 
376     // if there's a master rule, use it to format the number
377     if (fractionRules[2]) {
378         return fractionRules[2];
379     }
380 
381     // and if we haven't yet returned a rule, use findNormalRule()
382     // to find the applicable rule
383     int64_t r = util64_fromDouble(number + 0.5);
384     return findNormalRule(r);
385 }
386 
387 NFRule *
findNormalRule(int64_t number) const388 NFRuleSet::findNormalRule(int64_t number) const
389 {
390     // if this is a fraction rule set, use findFractionRuleSetRule()
391     // to find the rule (we should only go into this clause if the
392     // value is 0)
393     if (fIsFractionRuleSet) {
394         return findFractionRuleSetRule((double)number);
395     }
396 
397     // if the number is negative, return the negative-number rule
398     // (if there isn't one, pretend the number is positive)
399     if (number < 0) {
400         if (negativeNumberRule) {
401             return negativeNumberRule;
402         } else {
403             number = -number;
404         }
405     }
406 
407     // we have to repeat the preceding two checks, even though we
408     // do them in findRule(), because the version of format() that
409     // takes a long bypasses findRule() and goes straight to this
410     // function.  This function does skip the fraction rules since
411     // we know the value is an integer (it also skips the master
412     // rule, since it's considered a fraction rule.  Skipping the
413     // master rule in this function is also how we avoid infinite
414     // recursion)
415 
416     // {dlf} unfortunately this fails if there are no rules except
417     // special rules.  If there are no rules, use the master rule.
418 
419     // binary-search the rule list for the applicable rule
420     // (a rule is used for all values from its base value to
421     // the next rule's base value)
422     int32_t hi = rules.size();
423     if (hi > 0) {
424         int32_t lo = 0;
425 
426         while (lo < hi) {
427             int32_t mid = (lo + hi) / 2;
428             if (rules[mid]->getBaseValue() == number) {
429                 return rules[mid];
430             }
431             else if (rules[mid]->getBaseValue() > number) {
432                 hi = mid;
433             }
434             else {
435                 lo = mid + 1;
436             }
437         }
438         if (hi == 0) { // bad rule set, minimum base > 0
439             return NULL; // want to throw exception here
440         }
441 
442         NFRule *result = rules[hi - 1];
443 
444         // use shouldRollBack() to see whether we need to invoke the
445         // rollback rule (see shouldRollBack()'s documentation for
446         // an explanation of the rollback rule).  If we do, roll back
447         // one rule and return that one instead of the one we'd normally
448         // return
449         if (result->shouldRollBack((double)number)) {
450             if (hi == 1) { // bad rule set, no prior rule to rollback to from this base
451                 return NULL;
452             }
453             result = rules[hi - 2];
454         }
455         return result;
456     }
457     // else use the master rule
458     return fractionRules[2];
459 }
460 
461 /**
462  * If this rule is a fraction rule set, this function is used by
463  * findRule() to select the most appropriate rule for formatting
464  * the number.  Basically, the base value of each rule in the rule
465  * set is treated as the denominator of a fraction.  Whichever
466  * denominator can produce the fraction closest in value to the
467  * number passed in is the result.  If there's a tie, the earlier
468  * one in the list wins.  (If there are two rules in a row with the
469  * same base value, the first one is used when the numerator of the
470  * fraction would be 1, and the second rule is used the rest of the
471  * time.
472  * @param number The number being formatted (which will always be
473  * a number between 0 and 1)
474  * @return The rule to use to format this number
475  */
476 NFRule*
findFractionRuleSetRule(double number) const477 NFRuleSet::findFractionRuleSetRule(double number) const
478 {
479     // the obvious way to do this (multiply the value being formatted
480     // by each rule's base value until you get an integral result)
481     // doesn't work because of rounding error.  This method is more
482     // accurate
483 
484     // find the least common multiple of the rules' base values
485     // and multiply this by the number being formatted.  This is
486     // all the precision we need, and we can do all of the rest
487     // of the math using integer arithmetic
488     int64_t leastCommonMultiple = rules[0]->getBaseValue();
489     int64_t numerator;
490     {
491         for (uint32_t i = 1; i < rules.size(); ++i) {
492             leastCommonMultiple = util_lcm(leastCommonMultiple, rules[i]->getBaseValue());
493         }
494         numerator = util64_fromDouble(number * (double)leastCommonMultiple + 0.5);
495     }
496     // for each rule, do the following...
497     int64_t tempDifference;
498     int64_t difference = util64_fromDouble(uprv_maxMantissa());
499     int32_t winner = 0;
500     for (uint32_t i = 0; i < rules.size(); ++i) {
501         // "numerator" is the numerator of the fraction if the
502         // denominator is the LCD.  The numerator if the rule's
503         // base value is the denominator is "numerator" times the
504         // base value divided bythe LCD.  Here we check to see if
505         // that's an integer, and if not, how close it is to being
506         // an integer.
507         tempDifference = numerator * rules[i]->getBaseValue() % leastCommonMultiple;
508 
509 
510         // normalize the result of the above calculation: we want
511         // the numerator's distance from the CLOSEST multiple
512         // of the LCD
513         if (leastCommonMultiple - tempDifference < tempDifference) {
514             tempDifference = leastCommonMultiple - tempDifference;
515         }
516 
517         // if this is as close as we've come, keep track of how close
518         // that is, and the line number of the rule that did it.  If
519         // we've scored a direct hit, we don't have to look at any more
520         // rules
521         if (tempDifference < difference) {
522             difference = tempDifference;
523             winner = i;
524             if (difference == 0) {
525                 break;
526             }
527         }
528     }
529 
530     // if we have two successive rules that both have the winning base
531     // value, then the first one (the one we found above) is used if
532     // the numerator of the fraction is 1 and the second one is used if
533     // the numerator of the fraction is anything else (this lets us
534     // do things like "one third"/"two thirds" without haveing to define
535     // a whole bunch of extra rule sets)
536     if ((unsigned)(winner + 1) < rules.size() &&
537         rules[winner + 1]->getBaseValue() == rules[winner]->getBaseValue()) {
538         double n = ((double)rules[winner]->getBaseValue()) * number;
539         if (n < 0.5 || n >= 2) {
540             ++winner;
541         }
542     }
543 
544     // finally, return the winning rule
545     return rules[winner];
546 }
547 
548 /**
549  * Parses a string.  Matches the string to be parsed against each
550  * of its rules (with a base value less than upperBound) and returns
551  * the value produced by the rule that matched the most charcters
552  * in the source string.
553  * @param text The string to parse
554  * @param parsePosition The initial position is ignored and assumed
555  * to be 0.  On exit, this object has been updated to point to the
556  * first character position this rule set didn't consume.
557  * @param upperBound Limits the rules that can be allowed to match.
558  * Only rules whose base values are strictly less than upperBound
559  * are considered.
560  * @return The numerical result of parsing this string.  This will
561  * be the matching rule's base value, composed appropriately with
562  * the results of matching any of its substitutions.  The object
563  * will be an instance of Long if it's an integral value; otherwise,
564  * it will be an instance of Double.  This function always returns
565  * a valid object: If nothing matched the input string at all,
566  * this function returns new Long(0), and the parse position is
567  * left unchanged.
568  */
569 #ifdef RBNF_DEBUG
570 #include <stdio.h>
571 
dumpUS(FILE * f,const UnicodeString & us)572 static void dumpUS(FILE* f, const UnicodeString& us) {
573   int len = us.length();
574   char* buf = (char *)uprv_malloc((len+1)*sizeof(char)); //new char[len+1];
575   if (buf != NULL) {
576 	  us.extract(0, len, buf);
577 	  buf[len] = 0;
578 	  fprintf(f, "%s", buf);
579 	  uprv_free(buf); //delete[] buf;
580   }
581 }
582 #endif
583 
584 UBool
parse(const UnicodeString & text,ParsePosition & pos,double upperBound,Formattable & result) const585 NFRuleSet::parse(const UnicodeString& text, ParsePosition& pos, double upperBound, Formattable& result) const
586 {
587     // try matching each rule in the rule set against the text being
588     // parsed.  Whichever one matches the most characters is the one
589     // that determines the value we return.
590 
591     result.setLong(0);
592 
593     // dump out if there's no text to parse
594     if (text.length() == 0) {
595         return 0;
596     }
597 
598     ParsePosition highWaterMark;
599     ParsePosition workingPos = pos;
600 
601 #ifdef RBNF_DEBUG
602     fprintf(stderr, "<nfrs> %x '", this);
603     dumpUS(stderr, name);
604     fprintf(stderr, "' text '");
605     dumpUS(stderr, text);
606     fprintf(stderr, "'\n");
607     fprintf(stderr, "  parse negative: %d\n", this, negativeNumberRule != 0);
608 #endif
609 
610     // start by trying the negative number rule (if there is one)
611     if (negativeNumberRule) {
612         Formattable tempResult;
613 #ifdef RBNF_DEBUG
614         fprintf(stderr, "  <nfrs before negative> %x ub: %g\n", negativeNumberRule, upperBound);
615 #endif
616         UBool success = negativeNumberRule->doParse(text, workingPos, 0, upperBound, tempResult);
617 #ifdef RBNF_DEBUG
618         fprintf(stderr, "  <nfrs after negative> success: %d wpi: %d\n", success, workingPos.getIndex());
619 #endif
620         if (success && workingPos.getIndex() > highWaterMark.getIndex()) {
621             result = tempResult;
622             highWaterMark = workingPos;
623         }
624         workingPos = pos;
625     }
626 #ifdef RBNF_DEBUG
627     fprintf(stderr, "<nfrs> continue fractional with text '");
628     dumpUS(stderr, text);
629     fprintf(stderr, "' hwm: %d\n", highWaterMark.getIndex());
630 #endif
631     // then try each of the fraction rules
632     {
633         for (int i = 0; i < 3; i++) {
634             if (fractionRules[i]) {
635                 Formattable tempResult;
636                 UBool success = fractionRules[i]->doParse(text, workingPos, 0, upperBound, tempResult);
637                 if (success && (workingPos.getIndex() > highWaterMark.getIndex())) {
638                     result = tempResult;
639                     highWaterMark = workingPos;
640                 }
641                 workingPos = pos;
642             }
643         }
644     }
645 #ifdef RBNF_DEBUG
646     fprintf(stderr, "<nfrs> continue other with text '");
647     dumpUS(stderr, text);
648     fprintf(stderr, "' hwm: %d\n", highWaterMark.getIndex());
649 #endif
650 
651     // finally, go through the regular rules one at a time.  We start
652     // at the end of the list because we want to try matching the most
653     // sigificant rule first (this helps ensure that we parse
654     // "five thousand three hundred six" as
655     // "(five thousand) (three hundred) (six)" rather than
656     // "((five thousand three) hundred) (six)").  Skip rules whose
657     // base values are higher than the upper bound (again, this helps
658     // limit ambiguity by making sure the rules that match a rule's
659     // are less significant than the rule containing the substitutions)/
660     {
661         int64_t ub = util64_fromDouble(upperBound);
662 #ifdef RBNF_DEBUG
663         {
664             char ubstr[64];
665             util64_toa(ub, ubstr, 64);
666             char ubstrhex[64];
667             util64_toa(ub, ubstrhex, 64, 16);
668             fprintf(stderr, "ub: %g, i64: %s (%s)\n", upperBound, ubstr, ubstrhex);
669         }
670 #endif
671         for (int32_t i = rules.size(); --i >= 0 && highWaterMark.getIndex() < text.length();) {
672             if ((!fIsFractionRuleSet) && (rules[i]->getBaseValue() >= ub)) {
673                 continue;
674             }
675             Formattable tempResult;
676             UBool success = rules[i]->doParse(text, workingPos, fIsFractionRuleSet, upperBound, tempResult);
677             if (success && workingPos.getIndex() > highWaterMark.getIndex()) {
678                 result = tempResult;
679                 highWaterMark = workingPos;
680             }
681             workingPos = pos;
682         }
683     }
684 #ifdef RBNF_DEBUG
685     fprintf(stderr, "<nfrs> exit\n");
686 #endif
687     // finally, update the parse postion we were passed to point to the
688     // first character we didn't use, and return the result that
689     // corresponds to that string of characters
690     pos = highWaterMark;
691 
692     return 1;
693 }
694 
695 void
appendRules(UnicodeString & result) const696 NFRuleSet::appendRules(UnicodeString& result) const
697 {
698     // the rule set name goes first...
699     result.append(name);
700     result.append(gColon);
701     result.append(gLineFeed);
702 
703     // followed by the regular rules...
704     for (uint32_t i = 0; i < rules.size(); i++) {
705         result.append(gFourSpaces);
706         rules[i]->_appendRuleText(result);
707         result.append(gLineFeed);
708     }
709 
710     // followed by the special rules (if they exist)
711     if (negativeNumberRule) {
712         result.append(gFourSpaces);
713         negativeNumberRule->_appendRuleText(result);
714         result.append(gLineFeed);
715     }
716 
717     {
718         for (uint32_t i = 0; i < 3; ++i) {
719             if (fractionRules[i]) {
720                 result.append(gFourSpaces);
721                 fractionRules[i]->_appendRuleText(result);
722                 result.append(gLineFeed);
723             }
724         }
725     }
726 }
727 
728 // utility functions
729 
util64_fromDouble(double d)730 int64_t util64_fromDouble(double d) {
731     int64_t result = 0;
732     if (!uprv_isNaN(d)) {
733         double mant = uprv_maxMantissa();
734         if (d < -mant) {
735             d = -mant;
736         } else if (d > mant) {
737             d = mant;
738         }
739         UBool neg = d < 0;
740         if (neg) {
741             d = -d;
742         }
743         result = (int64_t)uprv_floor(d);
744         if (neg) {
745             result = -result;
746         }
747     }
748     return result;
749 }
750 
util64_pow(int32_t r,uint32_t e)751 int64_t util64_pow(int32_t r, uint32_t e)  {
752     if (r == 0) {
753         return 0;
754     } else if (e == 0) {
755         return 1;
756     } else {
757         int64_t n = r;
758         while (--e > 0) {
759             n *= r;
760         }
761         return n;
762     }
763 }
764 
765 static const uint8_t asciiDigits[] = {
766     0x30u, 0x31u, 0x32u, 0x33u, 0x34u, 0x35u, 0x36u, 0x37u,
767     0x38u, 0x39u, 0x61u, 0x62u, 0x63u, 0x64u, 0x65u, 0x66u,
768     0x67u, 0x68u, 0x69u, 0x6au, 0x6bu, 0x6cu, 0x6du, 0x6eu,
769     0x6fu, 0x70u, 0x71u, 0x72u, 0x73u, 0x74u, 0x75u, 0x76u,
770     0x77u, 0x78u, 0x79u, 0x7au,
771 };
772 
773 static const UChar kUMinus = (UChar)0x002d;
774 
775 #ifdef RBNF_DEBUG
776 static const char kMinus = '-';
777 
778 static const uint8_t digitInfo[] = {
779         0,     0,     0,     0,     0,     0,     0,     0,
780         0,     0,     0,     0,     0,     0,     0,     0,
781         0,     0,     0,     0,     0,     0,     0,     0,
782         0,     0,     0,     0,     0,     0,     0,     0,
783         0,     0,     0,     0,     0,     0,     0,     0,
784         0,     0,     0,     0,     0,     0,     0,     0,
785     0x80u, 0x81u, 0x82u, 0x83u, 0x84u, 0x85u, 0x86u, 0x87u,
786     0x88u, 0x89u,     0,     0,     0,     0,     0,     0,
787         0, 0x8au, 0x8bu, 0x8cu, 0x8du, 0x8eu, 0x8fu, 0x90u,
788     0x91u, 0x92u, 0x93u, 0x94u, 0x95u, 0x96u, 0x97u, 0x98u,
789     0x99u, 0x9au, 0x9bu, 0x9cu, 0x9du, 0x9eu, 0x9fu, 0xa0u,
790     0xa1u, 0xa2u, 0xa3u,     0,     0,     0,     0,     0,
791         0, 0x8au, 0x8bu, 0x8cu, 0x8du, 0x8eu, 0x8fu, 0x90u,
792     0x91u, 0x92u, 0x93u, 0x94u, 0x95u, 0x96u, 0x97u, 0x98u,
793     0x99u, 0x9au, 0x9bu, 0x9cu, 0x9du, 0x9eu, 0x9fu, 0xa0u,
794     0xa1u, 0xa2u, 0xa3u,     0,     0,     0,     0,     0,
795 };
796 
util64_atoi(const char * str,uint32_t radix)797 int64_t util64_atoi(const char* str, uint32_t radix)
798 {
799     if (radix > 36) {
800         radix = 36;
801     } else if (radix < 2) {
802         radix = 2;
803     }
804     int64_t lradix = radix;
805 
806     int neg = 0;
807     if (*str == kMinus) {
808         ++str;
809         neg = 1;
810     }
811     int64_t result = 0;
812     uint8_t b;
813     while ((b = digitInfo[*str++]) && ((b &= 0x7f) < radix)) {
814         result *= lradix;
815         result += (int32_t)b;
816     }
817     if (neg) {
818         result = -result;
819     }
820     return result;
821 }
822 
util64_utoi(const UChar * str,uint32_t radix)823 int64_t util64_utoi(const UChar* str, uint32_t radix)
824 {
825     if (radix > 36) {
826         radix = 36;
827     } else if (radix < 2) {
828         radix = 2;
829     }
830     int64_t lradix = radix;
831 
832     int neg = 0;
833     if (*str == kUMinus) {
834         ++str;
835         neg = 1;
836     }
837     int64_t result = 0;
838     UChar c;
839     uint8_t b;
840     while (((c = *str++) < 0x0080) && (b = digitInfo[c]) && ((b &= 0x7f) < radix)) {
841         result *= lradix;
842         result += (int32_t)b;
843     }
844     if (neg) {
845         result = -result;
846     }
847     return result;
848 }
849 
util64_toa(int64_t w,char * buf,uint32_t len,uint32_t radix,UBool raw)850 uint32_t util64_toa(int64_t w, char* buf, uint32_t len, uint32_t radix, UBool raw)
851 {
852     if (radix > 36) {
853         radix = 36;
854     } else if (radix < 2) {
855         radix = 2;
856     }
857     int64_t base = radix;
858 
859     char* p = buf;
860     if (len && (w < 0) && (radix == 10) && !raw) {
861         w = -w;
862         *p++ = kMinus;
863         --len;
864     } else if (len && (w == 0)) {
865         *p++ = (char)raw ? 0 : asciiDigits[0];
866         --len;
867     }
868 
869     while (len && w != 0) {
870         int64_t n = w / base;
871         int64_t m = n * base;
872         int32_t d = (int32_t)(w-m);
873         *p++ = raw ? (char)d : asciiDigits[d];
874         w = n;
875         --len;
876     }
877     if (len) {
878         *p = 0; // null terminate if room for caller convenience
879     }
880 
881     len = p - buf;
882     if (*buf == kMinus) {
883         ++buf;
884     }
885     while (--p > buf) {
886         char c = *p;
887         *p = *buf;
888         *buf = c;
889         ++buf;
890     }
891 
892     return len;
893 }
894 #endif
895 
util64_tou(int64_t w,UChar * buf,uint32_t len,uint32_t radix,UBool raw)896 uint32_t util64_tou(int64_t w, UChar* buf, uint32_t len, uint32_t radix, UBool raw)
897 {
898     if (radix > 36) {
899         radix = 36;
900     } else if (radix < 2) {
901         radix = 2;
902     }
903     int64_t base = radix;
904 
905     UChar* p = buf;
906     if (len && (w < 0) && (radix == 10) && !raw) {
907         w = -w;
908         *p++ = kUMinus;
909         --len;
910     } else if (len && (w == 0)) {
911         *p++ = (UChar)raw ? 0 : asciiDigits[0];
912         --len;
913     }
914 
915     while (len && (w != 0)) {
916         int64_t n = w / base;
917         int64_t m = n * base;
918         int32_t d = (int32_t)(w-m);
919         *p++ = (UChar)(raw ? d : asciiDigits[d]);
920         w = n;
921         --len;
922     }
923     if (len) {
924         *p = 0; // null terminate if room for caller convenience
925     }
926 
927     len = (uint32_t)(p - buf);
928     if (*buf == kUMinus) {
929         ++buf;
930     }
931     while (--p > buf) {
932         UChar c = *p;
933         *p = *buf;
934         *buf = c;
935         ++buf;
936     }
937 
938     return len;
939 }
940 
941 
942 U_NAMESPACE_END
943 
944 /* U_HAVE_RBNF */
945 #endif
946 
947