1 | // © 2016 and later: Unicode, Inc. and others. |
2 | // License & terms of use: http://www.unicode.org/copyright.html |
3 | /* |
4 | ****************************************************************************** |
5 | * Copyright (C) 1997-2015, International Business Machines |
6 | * Corporation and others. All Rights Reserved. |
7 | ****************************************************************************** |
8 | * file name: nfsubs.cpp |
9 | * encoding: UTF-8 |
10 | * tab size: 8 (not used) |
11 | * indentation:4 |
12 | * |
13 | * Modification history |
14 | * Date Name Comments |
15 | * 10/11/2001 Doug Ported from ICU4J |
16 | */ |
17 | |
18 | #include <stdio.h> |
19 | #include "utypeinfo.h" // for 'typeid' to work |
20 | |
21 | #include "nfsubs.h" |
22 | #include "fmtableimp.h" |
23 | #include "putilimp.h" |
24 | #include "number_decimalquantity.h" |
25 | |
26 | #if U_HAVE_RBNF |
27 | |
28 | static const UChar gLessThan = 0x003c; |
29 | static const UChar gEquals = 0x003d; |
30 | static const UChar gGreaterThan = 0x003e; |
31 | static const UChar gPercent = 0x0025; |
32 | static const UChar gPound = 0x0023; |
33 | static const UChar gZero = 0x0030; |
34 | static const UChar gSpace = 0x0020; |
35 | |
36 | static const UChar gEqualsEquals[] = |
37 | { |
38 | 0x3D, 0x3D, 0 |
39 | }; /* "==" */ |
40 | static const UChar gGreaterGreaterGreaterThan[] = |
41 | { |
42 | 0x3E, 0x3E, 0x3E, 0 |
43 | }; /* ">>>" */ |
44 | static const UChar gGreaterGreaterThan[] = |
45 | { |
46 | 0x3E, 0x3E, 0 |
47 | }; /* ">>" */ |
48 | |
49 | U_NAMESPACE_BEGIN |
50 | |
51 | using number::impl::DecimalQuantity; |
52 | |
53 | class SameValueSubstitution : public NFSubstitution { |
54 | public: |
55 | SameValueSubstitution(int32_t pos, |
56 | const NFRuleSet* ruleset, |
57 | const UnicodeString& description, |
58 | UErrorCode& status); |
59 | virtual ~SameValueSubstitution(); |
60 | |
61 | virtual int64_t transformNumber(int64_t number) const { return number; } |
62 | virtual double transformNumber(double number) const { return number; } |
63 | virtual double composeRuleValue(double newRuleValue, double /*oldRuleValue*/) const { return newRuleValue; } |
64 | virtual double calcUpperBound(double oldUpperBound) const { return oldUpperBound; } |
65 | virtual UChar tokenChar() const { return (UChar)0x003d; } // '=' |
66 | |
67 | public: |
68 | static UClassID getStaticClassID(void); |
69 | virtual UClassID getDynamicClassID(void) const; |
70 | }; |
71 | |
72 | SameValueSubstitution::~SameValueSubstitution() {} |
73 | |
74 | class MultiplierSubstitution : public NFSubstitution { |
75 | int64_t divisor; |
76 | |
77 | public: |
78 | MultiplierSubstitution(int32_t _pos, |
79 | const NFRule *rule, |
80 | const NFRuleSet* _ruleSet, |
81 | const UnicodeString& description, |
82 | UErrorCode& status) |
83 | : NFSubstitution(_pos, _ruleSet, description, status), divisor(rule->getDivisor()) |
84 | { |
85 | if (divisor == 0) { |
86 | status = U_PARSE_ERROR; |
87 | } |
88 | } |
89 | virtual ~MultiplierSubstitution(); |
90 | |
91 | virtual void setDivisor(int32_t radix, int16_t exponent, UErrorCode& status) { |
92 | divisor = util64_pow(radix, exponent); |
93 | |
94 | if(divisor == 0) { |
95 | status = U_PARSE_ERROR; |
96 | } |
97 | } |
98 | |
99 | virtual UBool operator==(const NFSubstitution& rhs) const; |
100 | |
101 | virtual int64_t transformNumber(int64_t number) const { |
102 | return number / divisor; |
103 | } |
104 | |
105 | virtual double transformNumber(double number) const { |
106 | if (getRuleSet()) { |
107 | return uprv_floor(number / divisor); |
108 | } else { |
109 | return number / divisor; |
110 | } |
111 | } |
112 | |
113 | virtual double composeRuleValue(double newRuleValue, double /*oldRuleValue*/) const { |
114 | return newRuleValue * divisor; |
115 | } |
116 | |
117 | virtual double calcUpperBound(double /*oldUpperBound*/) const { return static_cast<double>(divisor); } |
118 | |
119 | virtual UChar tokenChar() const { return (UChar)0x003c; } // '<' |
120 | |
121 | public: |
122 | static UClassID getStaticClassID(void); |
123 | virtual UClassID getDynamicClassID(void) const; |
124 | }; |
125 | |
126 | MultiplierSubstitution::~MultiplierSubstitution() {} |
127 | |
128 | class ModulusSubstitution : public NFSubstitution { |
129 | int64_t divisor; |
130 | const NFRule* ruleToUse; |
131 | public: |
132 | ModulusSubstitution(int32_t pos, |
133 | const NFRule* rule, |
134 | const NFRule* rulePredecessor, |
135 | const NFRuleSet* ruleSet, |
136 | const UnicodeString& description, |
137 | UErrorCode& status); |
138 | virtual ~ModulusSubstitution(); |
139 | |
140 | virtual void setDivisor(int32_t radix, int16_t exponent, UErrorCode& status) { |
141 | divisor = util64_pow(radix, exponent); |
142 | |
143 | if (divisor == 0) { |
144 | status = U_PARSE_ERROR; |
145 | } |
146 | } |
147 | |
148 | virtual UBool operator==(const NFSubstitution& rhs) const; |
149 | |
150 | virtual void doSubstitution(int64_t number, UnicodeString& toInsertInto, int32_t pos, int32_t recursionCount, UErrorCode& status) const; |
151 | virtual void doSubstitution(double number, UnicodeString& toInsertInto, int32_t pos, int32_t recursionCount, UErrorCode& status) const; |
152 | |
153 | virtual int64_t transformNumber(int64_t number) const { return number % divisor; } |
154 | virtual double transformNumber(double number) const { return uprv_fmod(number, static_cast<double>(divisor)); } |
155 | |
156 | virtual UBool doParse(const UnicodeString& text, |
157 | ParsePosition& parsePosition, |
158 | double baseValue, |
159 | double upperBound, |
160 | UBool lenientParse, |
161 | uint32_t nonNumericalExecutedRuleMask, |
162 | Formattable& result) const; |
163 | |
164 | virtual double composeRuleValue(double newRuleValue, double oldRuleValue) const { |
165 | return oldRuleValue - uprv_fmod(oldRuleValue, static_cast<double>(divisor)) + newRuleValue; |
166 | } |
167 | |
168 | virtual double calcUpperBound(double /*oldUpperBound*/) const { return static_cast<double>(divisor); } |
169 | |
170 | virtual UBool isModulusSubstitution() const { return TRUE; } |
171 | |
172 | virtual UChar tokenChar() const { return (UChar)0x003e; } // '>' |
173 | |
174 | virtual void toString(UnicodeString& result) const; |
175 | |
176 | public: |
177 | static UClassID getStaticClassID(void); |
178 | virtual UClassID getDynamicClassID(void) const; |
179 | }; |
180 | |
181 | ModulusSubstitution::~ModulusSubstitution() {} |
182 | |
183 | class IntegralPartSubstitution : public NFSubstitution { |
184 | public: |
185 | IntegralPartSubstitution(int32_t _pos, |
186 | const NFRuleSet* _ruleSet, |
187 | const UnicodeString& description, |
188 | UErrorCode& status) |
189 | : NFSubstitution(_pos, _ruleSet, description, status) {} |
190 | virtual ~IntegralPartSubstitution(); |
191 | |
192 | virtual int64_t transformNumber(int64_t number) const { return number; } |
193 | virtual double transformNumber(double number) const { return uprv_floor(number); } |
194 | virtual double composeRuleValue(double newRuleValue, double oldRuleValue) const { return newRuleValue + oldRuleValue; } |
195 | virtual double calcUpperBound(double /*oldUpperBound*/) const { return DBL_MAX; } |
196 | virtual UChar tokenChar() const { return (UChar)0x003c; } // '<' |
197 | |
198 | public: |
199 | static UClassID getStaticClassID(void); |
200 | virtual UClassID getDynamicClassID(void) const; |
201 | }; |
202 | |
203 | IntegralPartSubstitution::~IntegralPartSubstitution() {} |
204 | |
205 | class FractionalPartSubstitution : public NFSubstitution { |
206 | UBool byDigits; |
207 | UBool useSpaces; |
208 | enum { kMaxDecimalDigits = 8 }; |
209 | public: |
210 | FractionalPartSubstitution(int32_t pos, |
211 | const NFRuleSet* ruleSet, |
212 | const UnicodeString& description, |
213 | UErrorCode& status); |
214 | virtual ~FractionalPartSubstitution(); |
215 | |
216 | virtual UBool operator==(const NFSubstitution& rhs) const; |
217 | |
218 | virtual void doSubstitution(double number, UnicodeString& toInsertInto, int32_t pos, int32_t recursionCount, UErrorCode& status) const; |
219 | virtual void doSubstitution(int64_t /*number*/, UnicodeString& /*toInsertInto*/, int32_t /*_pos*/, int32_t /*recursionCount*/, UErrorCode& /*status*/) const {} |
220 | virtual int64_t transformNumber(int64_t /*number*/) const { return 0; } |
221 | virtual double transformNumber(double number) const { return number - uprv_floor(number); } |
222 | |
223 | virtual UBool doParse(const UnicodeString& text, |
224 | ParsePosition& parsePosition, |
225 | double baseValue, |
226 | double upperBound, |
227 | UBool lenientParse, |
228 | uint32_t nonNumericalExecutedRuleMask, |
229 | Formattable& result) const; |
230 | |
231 | virtual double composeRuleValue(double newRuleValue, double oldRuleValue) const { return newRuleValue + oldRuleValue; } |
232 | virtual double calcUpperBound(double /*oldUpperBound*/) const { return 0.0; } |
233 | virtual UChar tokenChar() const { return (UChar)0x003e; } // '>' |
234 | |
235 | public: |
236 | static UClassID getStaticClassID(void); |
237 | virtual UClassID getDynamicClassID(void) const; |
238 | }; |
239 | |
240 | FractionalPartSubstitution::~FractionalPartSubstitution() {} |
241 | |
242 | class AbsoluteValueSubstitution : public NFSubstitution { |
243 | public: |
244 | AbsoluteValueSubstitution(int32_t _pos, |
245 | const NFRuleSet* _ruleSet, |
246 | const UnicodeString& description, |
247 | UErrorCode& status) |
248 | : NFSubstitution(_pos, _ruleSet, description, status) {} |
249 | virtual ~AbsoluteValueSubstitution(); |
250 | |
251 | virtual int64_t transformNumber(int64_t number) const { return number >= 0 ? number : -number; } |
252 | virtual double transformNumber(double number) const { return uprv_fabs(number); } |
253 | virtual double composeRuleValue(double newRuleValue, double /*oldRuleValue*/) const { return -newRuleValue; } |
254 | virtual double calcUpperBound(double /*oldUpperBound*/) const { return DBL_MAX; } |
255 | virtual UChar tokenChar() const { return (UChar)0x003e; } // '>' |
256 | |
257 | public: |
258 | static UClassID getStaticClassID(void); |
259 | virtual UClassID getDynamicClassID(void) const; |
260 | }; |
261 | |
262 | AbsoluteValueSubstitution::~AbsoluteValueSubstitution() {} |
263 | |
264 | class NumeratorSubstitution : public NFSubstitution { |
265 | double denominator; |
266 | int64_t ldenominator; |
267 | UBool withZeros; |
268 | public: |
269 | static inline UnicodeString fixdesc(const UnicodeString& desc) { |
270 | if (desc.endsWith(LTLT, 2)) { |
271 | UnicodeString result(desc, 0, desc.length()-1); |
272 | return result; |
273 | } |
274 | return desc; |
275 | } |
276 | NumeratorSubstitution(int32_t _pos, |
277 | double _denominator, |
278 | NFRuleSet* _ruleSet, |
279 | const UnicodeString& description, |
280 | UErrorCode& status) |
281 | : NFSubstitution(_pos, _ruleSet, fixdesc(description), status), denominator(_denominator) |
282 | { |
283 | ldenominator = util64_fromDouble(denominator); |
284 | withZeros = description.endsWith(LTLT, 2); |
285 | } |
286 | virtual ~NumeratorSubstitution(); |
287 | |
288 | virtual UBool operator==(const NFSubstitution& rhs) const; |
289 | |
290 | virtual int64_t transformNumber(int64_t number) const { return number * ldenominator; } |
291 | virtual double transformNumber(double number) const { return uprv_round(number * denominator); } |
292 | |
293 | virtual void doSubstitution(int64_t /*number*/, UnicodeString& /*toInsertInto*/, int32_t /*_pos*/, int32_t /*recursionCount*/, UErrorCode& /*status*/) const {} |
294 | virtual void doSubstitution(double number, UnicodeString& toInsertInto, int32_t pos, int32_t recursionCount, UErrorCode& status) const; |
295 | virtual UBool doParse(const UnicodeString& text, |
296 | ParsePosition& parsePosition, |
297 | double baseValue, |
298 | double upperBound, |
299 | UBool /*lenientParse*/, |
300 | uint32_t nonNumericalExecutedRuleMask, |
301 | Formattable& result) const; |
302 | |
303 | virtual double composeRuleValue(double newRuleValue, double oldRuleValue) const { return newRuleValue / oldRuleValue; } |
304 | virtual double calcUpperBound(double /*oldUpperBound*/) const { return denominator; } |
305 | virtual UChar tokenChar() const { return (UChar)0x003c; } // '<' |
306 | private: |
307 | static const UChar LTLT[2]; |
308 | |
309 | public: |
310 | static UClassID getStaticClassID(void); |
311 | virtual UClassID getDynamicClassID(void) const; |
312 | }; |
313 | |
314 | NumeratorSubstitution::~NumeratorSubstitution() {} |
315 | |
316 | NFSubstitution* |
317 | NFSubstitution::makeSubstitution(int32_t pos, |
318 | const NFRule* rule, |
319 | const NFRule* predecessor, |
320 | const NFRuleSet* ruleSet, |
321 | const RuleBasedNumberFormat* formatter, |
322 | const UnicodeString& description, |
323 | UErrorCode& status) |
324 | { |
325 | // if the description is empty, return a NullSubstitution |
326 | if (description.length() == 0) { |
327 | return NULL; |
328 | } |
329 | |
330 | switch (description.charAt(0)) { |
331 | // if the description begins with '<'... |
332 | case gLessThan: |
333 | // throw an exception if the rule is a negative number |
334 | // rule |
335 | if (rule->getBaseValue() == NFRule::kNegativeNumberRule) { |
336 | // throw new IllegalArgumentException("<< not allowed in negative-number rule"); |
337 | status = U_PARSE_ERROR; |
338 | return NULL; |
339 | } |
340 | |
341 | // if the rule is a fraction rule, return an |
342 | // IntegralPartSubstitution |
343 | else if (rule->getBaseValue() == NFRule::kImproperFractionRule |
344 | || rule->getBaseValue() == NFRule::kProperFractionRule |
345 | || rule->getBaseValue() == NFRule::kMasterRule) { |
346 | return new IntegralPartSubstitution(pos, ruleSet, description, status); |
347 | } |
348 | |
349 | // if the rule set containing the rule is a fraction |
350 | // rule set, return a NumeratorSubstitution |
351 | else if (ruleSet->isFractionRuleSet()) { |
352 | return new NumeratorSubstitution(pos, (double)rule->getBaseValue(), |
353 | formatter->getDefaultRuleSet(), description, status); |
354 | } |
355 | |
356 | // otherwise, return a MultiplierSubstitution |
357 | else { |
358 | return new MultiplierSubstitution(pos, rule, ruleSet, |
359 | description, status); |
360 | } |
361 | |
362 | // if the description begins with '>'... |
363 | case gGreaterThan: |
364 | // if the rule is a negative-number rule, return |
365 | // an AbsoluteValueSubstitution |
366 | if (rule->getBaseValue() == NFRule::kNegativeNumberRule) { |
367 | return new AbsoluteValueSubstitution(pos, ruleSet, description, status); |
368 | } |
369 | |
370 | // if the rule is a fraction rule, return a |
371 | // FractionalPartSubstitution |
372 | else if (rule->getBaseValue() == NFRule::kImproperFractionRule |
373 | || rule->getBaseValue() == NFRule::kProperFractionRule |
374 | || rule->getBaseValue() == NFRule::kMasterRule) { |
375 | return new FractionalPartSubstitution(pos, ruleSet, description, status); |
376 | } |
377 | |
378 | // if the rule set owning the rule is a fraction rule set, |
379 | // throw an exception |
380 | else if (ruleSet->isFractionRuleSet()) { |
381 | // throw new IllegalArgumentException(">> not allowed in fraction rule set"); |
382 | status = U_PARSE_ERROR; |
383 | return NULL; |
384 | } |
385 | |
386 | // otherwise, return a ModulusSubstitution |
387 | else { |
388 | return new ModulusSubstitution(pos, rule, predecessor, |
389 | ruleSet, description, status); |
390 | } |
391 | |
392 | // if the description begins with '=', always return a |
393 | // SameValueSubstitution |
394 | case gEquals: |
395 | return new SameValueSubstitution(pos, ruleSet, description, status); |
396 | |
397 | // and if it's anything else, throw an exception |
398 | default: |
399 | // throw new IllegalArgumentException("Illegal substitution character"); |
400 | status = U_PARSE_ERROR; |
401 | } |
402 | return NULL; |
403 | } |
404 | |
405 | NFSubstitution::NFSubstitution(int32_t _pos, |
406 | const NFRuleSet* _ruleSet, |
407 | const UnicodeString& description, |
408 | UErrorCode& status) |
409 | : pos(_pos), ruleSet(NULL), numberFormat(NULL) |
410 | { |
411 | // the description should begin and end with the same character. |
412 | // If it doesn't that's a syntax error. Otherwise, |
413 | // makeSubstitution() was the only thing that needed to know |
414 | // about these characters, so strip them off |
415 | UnicodeString workingDescription(description); |
416 | if (description.length() >= 2 |
417 | && description.charAt(0) == description.charAt(description.length() - 1)) |
418 | { |
419 | workingDescription.remove(description.length() - 1, 1); |
420 | workingDescription.remove(0, 1); |
421 | } |
422 | else if (description.length() != 0) { |
423 | // throw new IllegalArgumentException("Illegal substitution syntax"); |
424 | status = U_PARSE_ERROR; |
425 | return; |
426 | } |
427 | |
428 | if (workingDescription.length() == 0) { |
429 | // if the description was just two paired token characters |
430 | // (i.e., "<<" or ">>"), it uses the rule set it belongs to to |
431 | // format its result |
432 | this->ruleSet = _ruleSet; |
433 | } |
434 | else if (workingDescription.charAt(0) == gPercent) { |
435 | // if the description contains a rule set name, that's the rule |
436 | // set we use to format the result: get a reference to the |
437 | // names rule set |
438 | this->ruleSet = _ruleSet->getOwner()->findRuleSet(workingDescription, status); |
439 | } |
440 | else if (workingDescription.charAt(0) == gPound || workingDescription.charAt(0) ==gZero) { |
441 | // if the description begins with 0 or #, treat it as a |
442 | // DecimalFormat pattern, and initialize a DecimalFormat with |
443 | // that pattern (then set it to use the DecimalFormatSymbols |
444 | // belonging to our formatter) |
445 | const DecimalFormatSymbols* sym = _ruleSet->getOwner()->getDecimalFormatSymbols(); |
446 | if (!sym) { |
447 | status = U_MISSING_RESOURCE_ERROR; |
448 | return; |
449 | } |
450 | DecimalFormat *tempNumberFormat = new DecimalFormat(workingDescription, *sym, status); |
451 | /* test for NULL */ |
452 | if (!tempNumberFormat) { |
453 | status = U_MEMORY_ALLOCATION_ERROR; |
454 | return; |
455 | } |
456 | if (U_FAILURE(status)) { |
457 | delete tempNumberFormat; |
458 | return; |
459 | } |
460 | this->numberFormat = tempNumberFormat; |
461 | } |
462 | else if (workingDescription.charAt(0) == gGreaterThan) { |
463 | // if the description is ">>>", this substitution bypasses the |
464 | // usual rule-search process and always uses the rule that precedes |
465 | // it in its own rule set's rule list (this is used for place-value |
466 | // notations: formats where you want to see a particular part of |
467 | // a number even when it's 0) |
468 | |
469 | // this causes problems when >>> is used in a frationalPartSubstitution |
470 | // this->ruleSet = NULL; |
471 | this->ruleSet = _ruleSet; |
472 | this->numberFormat = NULL; |
473 | } |
474 | else { |
475 | // and of the description is none of these things, it's a syntax error |
476 | |
477 | // throw new IllegalArgumentException("Illegal substitution syntax"); |
478 | status = U_PARSE_ERROR; |
479 | } |
480 | } |
481 | |
482 | NFSubstitution::~NFSubstitution() |
483 | { |
484 | delete numberFormat; |
485 | numberFormat = NULL; |
486 | } |
487 | |
488 | /** |
489 | * Set's the substitution's divisor. Used by NFRule.setBaseValue(). |
490 | * A no-op for all substitutions except multiplier and modulus |
491 | * substitutions. |
492 | * @param radix The radix of the divisor |
493 | * @param exponent The exponent of the divisor |
494 | */ |
495 | void |
496 | NFSubstitution::setDivisor(int32_t /*radix*/, int16_t /*exponent*/, UErrorCode& /*status*/) { |
497 | // a no-op for all substitutions except multiplier and modulus substitutions |
498 | } |
499 | |
500 | void |
501 | NFSubstitution::setDecimalFormatSymbols(const DecimalFormatSymbols &newSymbols, UErrorCode& /*status*/) { |
502 | if (numberFormat != NULL) { |
503 | numberFormat->setDecimalFormatSymbols(newSymbols); |
504 | } |
505 | } |
506 | |
507 | //----------------------------------------------------------------------- |
508 | // boilerplate |
509 | //----------------------------------------------------------------------- |
510 | |
511 | UOBJECT_DEFINE_RTTI_IMPLEMENTATION(NFSubstitution) |
512 | |
513 | /** |
514 | * Compares two substitutions for equality |
515 | * @param The substitution to compare this one to |
516 | * @return true if the two substitutions are functionally equivalent |
517 | */ |
518 | UBool |
519 | NFSubstitution::operator==(const NFSubstitution& rhs) const |
520 | { |
521 | // compare class and all of the fields all substitutions have |
522 | // in common |
523 | // this should be called by subclasses before their own equality tests |
524 | return typeid(*this) == typeid(rhs) |
525 | && pos == rhs.pos |
526 | && (ruleSet == NULL) == (rhs.ruleSet == NULL) |
527 | // && ruleSet == rhs.ruleSet causes circularity, other checks to make instead? |
528 | && (numberFormat == NULL |
529 | ? (rhs.numberFormat == NULL) |
530 | : (*numberFormat == *rhs.numberFormat)); |
531 | } |
532 | |
533 | /** |
534 | * Returns a textual description of the substitution |
535 | * @return A textual description of the substitution. This might |
536 | * not be identical to the description it was created from, but |
537 | * it'll produce the same result. |
538 | */ |
539 | void |
540 | NFSubstitution::toString(UnicodeString& text) const |
541 | { |
542 | // use tokenChar() to get the character at the beginning and |
543 | // end of the substitutin token. In between them will go |
544 | // either the name of the rule set it uses, or the pattern of |
545 | // the DecimalFormat it uses |
546 | text.remove(); |
547 | text.append(tokenChar()); |
548 | |
549 | UnicodeString temp; |
550 | if (ruleSet != NULL) { |
551 | ruleSet->getName(temp); |
552 | } else if (numberFormat != NULL) { |
553 | numberFormat->toPattern(temp); |
554 | } |
555 | text.append(temp); |
556 | text.append(tokenChar()); |
557 | } |
558 | |
559 | //----------------------------------------------------------------------- |
560 | // formatting |
561 | //----------------------------------------------------------------------- |
562 | |
563 | /** |
564 | * Performs a mathematical operation on the number, formats it using |
565 | * either ruleSet or decimalFormat, and inserts the result into |
566 | * toInsertInto. |
567 | * @param number The number being formatted. |
568 | * @param toInsertInto The string we insert the result into |
569 | * @param pos The position in toInsertInto where the owning rule's |
570 | * rule text begins (this value is added to this substitution's |
571 | * position to determine exactly where to insert the new text) |
572 | */ |
573 | void |
574 | NFSubstitution::doSubstitution(int64_t number, UnicodeString& toInsertInto, int32_t _pos, int32_t recursionCount, UErrorCode& status) const |
575 | { |
576 | if (ruleSet != NULL) { |
577 | // Perform a transformation on the number that is dependent |
578 | // on the type of substitution this is, then just call its |
579 | // rule set's format() method to format the result |
580 | ruleSet->format(transformNumber(number), toInsertInto, _pos + this->pos, recursionCount, status); |
581 | } else if (numberFormat != NULL) { |
582 | if (number <= MAX_INT64_IN_DOUBLE) { |
583 | // or perform the transformation on the number (preserving |
584 | // the result's fractional part if the formatter it set |
585 | // to show it), then use that formatter's format() method |
586 | // to format the result |
587 | double numberToFormat = transformNumber((double)number); |
588 | if (numberFormat->getMaximumFractionDigits() == 0) { |
589 | numberToFormat = uprv_floor(numberToFormat); |
590 | } |
591 | |
592 | UnicodeString temp; |
593 | numberFormat->format(numberToFormat, temp, status); |
594 | toInsertInto.insert(_pos + this->pos, temp); |
595 | } |
596 | else { |
597 | // We have gone beyond double precision. Something has to give. |
598 | // We're favoring accuracy of the large number over potential rules |
599 | // that round like a CompactDecimalFormat, which is not a common use case. |
600 | // |
601 | // Perform a transformation on the number that is dependent |
602 | // on the type of substitution this is, then just call its |
603 | // rule set's format() method to format the result |
604 | int64_t numberToFormat = transformNumber(number); |
605 | UnicodeString temp; |
606 | numberFormat->format(numberToFormat, temp, status); |
607 | toInsertInto.insert(_pos + this->pos, temp); |
608 | } |
609 | } |
610 | } |
611 | |
612 | /** |
613 | * Performs a mathematical operation on the number, formats it using |
614 | * either ruleSet or decimalFormat, and inserts the result into |
615 | * toInsertInto. |
616 | * @param number The number being formatted. |
617 | * @param toInsertInto The string we insert the result into |
618 | * @param pos The position in toInsertInto where the owning rule's |
619 | * rule text begins (this value is added to this substitution's |
620 | * position to determine exactly where to insert the new text) |
621 | */ |
622 | void |
623 | NFSubstitution::doSubstitution(double number, UnicodeString& toInsertInto, int32_t _pos, int32_t recursionCount, UErrorCode& status) const { |
624 | // perform a transformation on the number being formatted that |
625 | // is dependent on the type of substitution this is |
626 | double numberToFormat = transformNumber(number); |
627 | |
628 | if (uprv_isInfinite(numberToFormat)) { |
629 | // This is probably a minus rule. Combine it with an infinite rule. |
630 | const NFRule *infiniteRule = ruleSet->findDoubleRule(uprv_getInfinity()); |
631 | infiniteRule->doFormat(numberToFormat, toInsertInto, _pos + this->pos, recursionCount, status); |
632 | return; |
633 | } |
634 | |
635 | // if the result is an integer, from here on out we work in integer |
636 | // space (saving time and memory and preserving accuracy) |
637 | if (numberToFormat == uprv_floor(numberToFormat) && ruleSet != NULL) { |
638 | ruleSet->format(util64_fromDouble(numberToFormat), toInsertInto, _pos + this->pos, recursionCount, status); |
639 | |
640 | // if the result isn't an integer, then call either our rule set's |
641 | // format() method or our DecimalFormat's format() method to |
642 | // format the result |
643 | } else { |
644 | if (ruleSet != NULL) { |
645 | ruleSet->format(numberToFormat, toInsertInto, _pos + this->pos, recursionCount, status); |
646 | } else if (numberFormat != NULL) { |
647 | UnicodeString temp; |
648 | numberFormat->format(numberToFormat, temp); |
649 | toInsertInto.insert(_pos + this->pos, temp); |
650 | } |
651 | } |
652 | } |
653 | |
654 | |
655 | //----------------------------------------------------------------------- |
656 | // parsing |
657 | //----------------------------------------------------------------------- |
658 | |
659 | #ifdef RBNF_DEBUG |
660 | #include <stdio.h> |
661 | #endif |
662 | |
663 | /** |
664 | * Parses a string using the rule set or DecimalFormat belonging |
665 | * to this substitution. If there's a match, a mathematical |
666 | * operation (the inverse of the one used in formatting) is |
667 | * performed on the result of the parse and the value passed in |
668 | * and returned as the result. The parse position is updated to |
669 | * point to the first unmatched character in the string. |
670 | * @param text The string to parse |
671 | * @param parsePosition On entry, ignored, but assumed to be 0. |
672 | * On exit, this is updated to point to the first unmatched |
673 | * character (or 0 if the substitution didn't match) |
674 | * @param baseValue A partial parse result that should be |
675 | * combined with the result of this parse |
676 | * @param upperBound When searching the rule set for a rule |
677 | * matching the string passed in, only rules with base values |
678 | * lower than this are considered |
679 | * @param lenientParse If true and matching against rules fails, |
680 | * the substitution will also try matching the text against |
681 | * numerals using a default-costructed NumberFormat. If false, |
682 | * no extra work is done. (This value is false whenever the |
683 | * formatter isn't in lenient-parse mode, but is also false |
684 | * under some conditions even when the formatter _is_ in |
685 | * lenient-parse mode.) |
686 | * @return If there's a match, this is the result of composing |
687 | * baseValue with whatever was returned from matching the |
688 | * characters. This will be either a Long or a Double. If there's |
689 | * no match this is new Long(0) (not null), and parsePosition |
690 | * is left unchanged. |
691 | */ |
692 | UBool |
693 | NFSubstitution::doParse(const UnicodeString& text, |
694 | ParsePosition& parsePosition, |
695 | double baseValue, |
696 | double upperBound, |
697 | UBool lenientParse, |
698 | uint32_t nonNumericalExecutedRuleMask, |
699 | Formattable& result) const |
700 | { |
701 | #ifdef RBNF_DEBUG |
702 | fprintf(stderr, "<nfsubs> %x bv: %g ub: %g\n" , this, baseValue, upperBound); |
703 | #endif |
704 | // figure out the highest base value a rule can have and match |
705 | // the text being parsed (this varies according to the type of |
706 | // substitutions: multiplier, modulus, and numerator substitutions |
707 | // restrict the search to rules with base values lower than their |
708 | // own; same-value substitutions leave the upper bound wherever |
709 | // it was, and the others allow any rule to match |
710 | upperBound = calcUpperBound(upperBound); |
711 | |
712 | // use our rule set to parse the text. If that fails and |
713 | // lenient parsing is enabled (this is always false if the |
714 | // formatter's lenient-parsing mode is off, but it may also |
715 | // be false even when the formatter's lenient-parse mode is |
716 | // on), then also try parsing the text using a default- |
717 | // constructed NumberFormat |
718 | if (ruleSet != NULL) { |
719 | ruleSet->parse(text, parsePosition, upperBound, nonNumericalExecutedRuleMask, result); |
720 | if (lenientParse && !ruleSet->isFractionRuleSet() && parsePosition.getIndex() == 0) { |
721 | UErrorCode status = U_ZERO_ERROR; |
722 | NumberFormat* fmt = NumberFormat::createInstance(status); |
723 | if (U_SUCCESS(status)) { |
724 | fmt->parse(text, result, parsePosition); |
725 | } |
726 | delete fmt; |
727 | } |
728 | |
729 | // ...or use our DecimalFormat to parse the text |
730 | } else if (numberFormat != NULL) { |
731 | numberFormat->parse(text, result, parsePosition); |
732 | } |
733 | |
734 | // if the parse was successful, we've already advanced the caller's |
735 | // parse position (this is the one function that doesn't have one |
736 | // of its own). Derive a parse result and return it as a Long, |
737 | // if possible, or a Double |
738 | if (parsePosition.getIndex() != 0) { |
739 | UErrorCode status = U_ZERO_ERROR; |
740 | double tempResult = result.getDouble(status); |
741 | |
742 | // composeRuleValue() produces a full parse result from |
743 | // the partial parse result passed to this function from |
744 | // the caller (this is either the owning rule's base value |
745 | // or the partial result obtained from composing the |
746 | // owning rule's base value with its other substitution's |
747 | // parse result) and the partial parse result obtained by |
748 | // matching the substitution (which will be the same value |
749 | // the caller would get by parsing just this part of the |
750 | // text with RuleBasedNumberFormat.parse() ). How the two |
751 | // values are used to derive the full parse result depends |
752 | // on the types of substitutions: For a regular rule, the |
753 | // ultimate result is its multiplier substitution's result |
754 | // times the rule's divisor (or the rule's base value) plus |
755 | // the modulus substitution's result (which will actually |
756 | // supersede part of the rule's base value). For a negative- |
757 | // number rule, the result is the negative of its substitution's |
758 | // result. For a fraction rule, it's the sum of its two |
759 | // substitution results. For a rule in a fraction rule set, |
760 | // it's the numerator substitution's result divided by |
761 | // the rule's base value. Results from same-value substitutions |
762 | // propagate back upard, and null substitutions don't affect |
763 | // the result. |
764 | tempResult = composeRuleValue(tempResult, baseValue); |
765 | result.setDouble(tempResult); |
766 | return TRUE; |
767 | // if the parse was UNsuccessful, return 0 |
768 | } else { |
769 | result.setLong(0); |
770 | return FALSE; |
771 | } |
772 | } |
773 | |
774 | /** |
775 | * Returns true if this is a modulus substitution. (We didn't do this |
776 | * with instanceof partially because it causes source files to |
777 | * proliferate and partially because we have to port this to C++.) |
778 | * @return true if this object is an instance of ModulusSubstitution |
779 | */ |
780 | UBool |
781 | NFSubstitution::isModulusSubstitution() const { |
782 | return FALSE; |
783 | } |
784 | |
785 | //=================================================================== |
786 | // SameValueSubstitution |
787 | //=================================================================== |
788 | |
789 | /** |
790 | * A substitution that passes the value passed to it through unchanged. |
791 | * Represented by == in rule descriptions. |
792 | */ |
793 | SameValueSubstitution::SameValueSubstitution(int32_t _pos, |
794 | const NFRuleSet* _ruleSet, |
795 | const UnicodeString& description, |
796 | UErrorCode& status) |
797 | : NFSubstitution(_pos, _ruleSet, description, status) |
798 | { |
799 | if (0 == description.compare(gEqualsEquals, 2)) { |
800 | // throw new IllegalArgumentException("== is not a legal token"); |
801 | status = U_PARSE_ERROR; |
802 | } |
803 | } |
804 | |
805 | UOBJECT_DEFINE_RTTI_IMPLEMENTATION(SameValueSubstitution) |
806 | |
807 | //=================================================================== |
808 | // MultiplierSubstitution |
809 | //=================================================================== |
810 | |
811 | UOBJECT_DEFINE_RTTI_IMPLEMENTATION(MultiplierSubstitution) |
812 | |
813 | UBool MultiplierSubstitution::operator==(const NFSubstitution& rhs) const |
814 | { |
815 | return NFSubstitution::operator==(rhs) && |
816 | divisor == ((const MultiplierSubstitution*)&rhs)->divisor; |
817 | } |
818 | |
819 | |
820 | //=================================================================== |
821 | // ModulusSubstitution |
822 | //=================================================================== |
823 | |
824 | /** |
825 | * A substitution that divides the number being formatted by the its rule's |
826 | * divisor and formats the remainder. Represented by ">>" in a |
827 | * regular rule. |
828 | */ |
829 | ModulusSubstitution::ModulusSubstitution(int32_t _pos, |
830 | const NFRule* rule, |
831 | const NFRule* predecessor, |
832 | const NFRuleSet* _ruleSet, |
833 | const UnicodeString& description, |
834 | UErrorCode& status) |
835 | : NFSubstitution(_pos, _ruleSet, description, status) |
836 | , divisor(rule->getDivisor()) |
837 | , ruleToUse(NULL) |
838 | { |
839 | // the owning rule's divisor controls the behavior of this |
840 | // substitution: rather than keeping a backpointer to the rule, |
841 | // we keep a copy of the divisor |
842 | |
843 | if (divisor == 0) { |
844 | status = U_PARSE_ERROR; |
845 | } |
846 | |
847 | if (0 == description.compare(gGreaterGreaterGreaterThan, 3)) { |
848 | // the >>> token doesn't alter how this substituion calculates the |
849 | // values it uses for formatting and parsing, but it changes |
850 | // what's done with that value after it's obtained: >>> short- |
851 | // circuits the rule-search process and goes straight to the |
852 | // specified rule to format the substitution value |
853 | ruleToUse = predecessor; |
854 | } |
855 | } |
856 | |
857 | UOBJECT_DEFINE_RTTI_IMPLEMENTATION(ModulusSubstitution) |
858 | |
859 | UBool ModulusSubstitution::operator==(const NFSubstitution& rhs) const |
860 | { |
861 | return NFSubstitution::operator==(rhs) && |
862 | divisor == ((const ModulusSubstitution*)&rhs)->divisor && |
863 | ruleToUse == ((const ModulusSubstitution*)&rhs)->ruleToUse; |
864 | } |
865 | |
866 | //----------------------------------------------------------------------- |
867 | // formatting |
868 | //----------------------------------------------------------------------- |
869 | |
870 | |
871 | /** |
872 | * If this is a >>> substitution, use ruleToUse to fill in |
873 | * the substitution. Otherwise, just use the superclass function. |
874 | * @param number The number being formatted |
875 | * @toInsertInto The string to insert the result of this substitution |
876 | * into |
877 | * @param pos The position of the rule text in toInsertInto |
878 | */ |
879 | void |
880 | ModulusSubstitution::doSubstitution(int64_t number, UnicodeString& toInsertInto, int32_t _pos, int32_t recursionCount, UErrorCode& status) const |
881 | { |
882 | // if this isn't a >>> substitution, just use the inherited version |
883 | // of this function (which uses either a rule set or a DecimalFormat |
884 | // to format its substitution value) |
885 | if (ruleToUse == NULL) { |
886 | NFSubstitution::doSubstitution(number, toInsertInto, _pos, recursionCount, status); |
887 | |
888 | // a >>> substitution goes straight to a particular rule to |
889 | // format the substitution value |
890 | } else { |
891 | int64_t numberToFormat = transformNumber(number); |
892 | ruleToUse->doFormat(numberToFormat, toInsertInto, _pos + getPos(), recursionCount, status); |
893 | } |
894 | } |
895 | |
896 | /** |
897 | * If this is a >>> substitution, use ruleToUse to fill in |
898 | * the substitution. Otherwise, just use the superclass function. |
899 | * @param number The number being formatted |
900 | * @toInsertInto The string to insert the result of this substitution |
901 | * into |
902 | * @param pos The position of the rule text in toInsertInto |
903 | */ |
904 | void |
905 | ModulusSubstitution::doSubstitution(double number, UnicodeString& toInsertInto, int32_t _pos, int32_t recursionCount, UErrorCode& status) const |
906 | { |
907 | // if this isn't a >>> substitution, just use the inherited version |
908 | // of this function (which uses either a rule set or a DecimalFormat |
909 | // to format its substitution value) |
910 | if (ruleToUse == NULL) { |
911 | NFSubstitution::doSubstitution(number, toInsertInto, _pos, recursionCount, status); |
912 | |
913 | // a >>> substitution goes straight to a particular rule to |
914 | // format the substitution value |
915 | } else { |
916 | double numberToFormat = transformNumber(number); |
917 | |
918 | ruleToUse->doFormat(numberToFormat, toInsertInto, _pos + getPos(), recursionCount, status); |
919 | } |
920 | } |
921 | |
922 | //----------------------------------------------------------------------- |
923 | // parsing |
924 | //----------------------------------------------------------------------- |
925 | |
926 | /** |
927 | * If this is a >>> substitution, match only against ruleToUse. |
928 | * Otherwise, use the superclass function. |
929 | * @param text The string to parse |
930 | * @param parsePosition Ignored on entry, updated on exit to point to |
931 | * the first unmatched character. |
932 | * @param baseValue The partial parse result prior to calling this |
933 | * routine. |
934 | */ |
935 | UBool |
936 | ModulusSubstitution::doParse(const UnicodeString& text, |
937 | ParsePosition& parsePosition, |
938 | double baseValue, |
939 | double upperBound, |
940 | UBool lenientParse, |
941 | uint32_t nonNumericalExecutedRuleMask, |
942 | Formattable& result) const |
943 | { |
944 | // if this isn't a >>> substitution, we can just use the |
945 | // inherited parse() routine to do the parsing |
946 | if (ruleToUse == NULL) { |
947 | return NFSubstitution::doParse(text, parsePosition, baseValue, upperBound, lenientParse, nonNumericalExecutedRuleMask, result); |
948 | |
949 | // but if it IS a >>> substitution, we have to do it here: we |
950 | // use the specific rule's doParse() method, and then we have to |
951 | // do some of the other work of NFRuleSet.parse() |
952 | } else { |
953 | ruleToUse->doParse(text, parsePosition, FALSE, upperBound, nonNumericalExecutedRuleMask, result); |
954 | |
955 | if (parsePosition.getIndex() != 0) { |
956 | UErrorCode status = U_ZERO_ERROR; |
957 | double tempResult = result.getDouble(status); |
958 | tempResult = composeRuleValue(tempResult, baseValue); |
959 | result.setDouble(tempResult); |
960 | } |
961 | |
962 | return TRUE; |
963 | } |
964 | } |
965 | /** |
966 | * Returns a textual description of the substitution |
967 | * @return A textual description of the substitution. This might |
968 | * not be identical to the description it was created from, but |
969 | * it'll produce the same result. |
970 | */ |
971 | void |
972 | ModulusSubstitution::toString(UnicodeString& text) const |
973 | { |
974 | // use tokenChar() to get the character at the beginning and |
975 | // end of the substitutin token. In between them will go |
976 | // either the name of the rule set it uses, or the pattern of |
977 | // the DecimalFormat it uses |
978 | |
979 | if ( ruleToUse != NULL ) { // Must have been a >>> substitution. |
980 | text.remove(); |
981 | text.append(tokenChar()); |
982 | text.append(tokenChar()); |
983 | text.append(tokenChar()); |
984 | } else { // Otherwise just use the super-class function. |
985 | NFSubstitution::toString(text); |
986 | } |
987 | } |
988 | //=================================================================== |
989 | // IntegralPartSubstitution |
990 | //=================================================================== |
991 | |
992 | UOBJECT_DEFINE_RTTI_IMPLEMENTATION(IntegralPartSubstitution) |
993 | |
994 | |
995 | //=================================================================== |
996 | // FractionalPartSubstitution |
997 | //=================================================================== |
998 | |
999 | |
1000 | /** |
1001 | * Constructs a FractionalPartSubstitution. This object keeps a flag |
1002 | * telling whether it should format by digits or not. In addition, |
1003 | * it marks the rule set it calls (if any) as a fraction rule set. |
1004 | */ |
1005 | FractionalPartSubstitution::FractionalPartSubstitution(int32_t _pos, |
1006 | const NFRuleSet* _ruleSet, |
1007 | const UnicodeString& description, |
1008 | UErrorCode& status) |
1009 | : NFSubstitution(_pos, _ruleSet, description, status) |
1010 | , byDigits(FALSE) |
1011 | , useSpaces(TRUE) |
1012 | |
1013 | { |
1014 | // akk, ruleSet can change in superclass constructor |
1015 | if (0 == description.compare(gGreaterGreaterThan, 2) || |
1016 | 0 == description.compare(gGreaterGreaterGreaterThan, 3) || |
1017 | _ruleSet == getRuleSet()) { |
1018 | byDigits = TRUE; |
1019 | if (0 == description.compare(gGreaterGreaterGreaterThan, 3)) { |
1020 | useSpaces = FALSE; |
1021 | } |
1022 | } else { |
1023 | // cast away const |
1024 | ((NFRuleSet*)getRuleSet())->makeIntoFractionRuleSet(); |
1025 | } |
1026 | } |
1027 | |
1028 | //----------------------------------------------------------------------- |
1029 | // formatting |
1030 | //----------------------------------------------------------------------- |
1031 | |
1032 | /** |
1033 | * If in "by digits" mode, fills in the substitution one decimal digit |
1034 | * at a time using the rule set containing this substitution. |
1035 | * Otherwise, uses the superclass function. |
1036 | * @param number The number being formatted |
1037 | * @param toInsertInto The string to insert the result of formatting |
1038 | * the substitution into |
1039 | * @param pos The position of the owning rule's rule text in |
1040 | * toInsertInto |
1041 | */ |
1042 | void |
1043 | FractionalPartSubstitution::doSubstitution(double number, UnicodeString& toInsertInto, |
1044 | int32_t _pos, int32_t recursionCount, UErrorCode& status) const |
1045 | { |
1046 | // if we're not in "byDigits" mode, just use the inherited |
1047 | // doSubstitution() routine |
1048 | if (!byDigits) { |
1049 | NFSubstitution::doSubstitution(number, toInsertInto, _pos, recursionCount, status); |
1050 | |
1051 | // if we're in "byDigits" mode, transform the value into an integer |
1052 | // by moving the decimal point eight places to the right and |
1053 | // pulling digits off the right one at a time, formatting each digit |
1054 | // as an integer using this substitution's owning rule set |
1055 | // (this is slower, but more accurate, than doing it from the |
1056 | // other end) |
1057 | } else { |
1058 | // int32_t numberToFormat = (int32_t)uprv_round(transformNumber(number) * uprv_pow(10, kMaxDecimalDigits)); |
1059 | // // this flag keeps us from formatting trailing zeros. It starts |
1060 | // // out false because we're pulling from the right, and switches |
1061 | // // to true the first time we encounter a non-zero digit |
1062 | // UBool doZeros = FALSE; |
1063 | // for (int32_t i = 0; i < kMaxDecimalDigits; i++) { |
1064 | // int64_t digit = numberToFormat % 10; |
1065 | // if (digit != 0 || doZeros) { |
1066 | // if (doZeros && useSpaces) { |
1067 | // toInsertInto.insert(_pos + getPos(), gSpace); |
1068 | // } |
1069 | // doZeros = TRUE; |
1070 | // getRuleSet()->format(digit, toInsertInto, _pos + getPos()); |
1071 | // } |
1072 | // numberToFormat /= 10; |
1073 | // } |
1074 | |
1075 | DecimalQuantity dl; |
1076 | dl.setToDouble(number); |
1077 | dl.roundToMagnitude(-20, UNUM_ROUND_HALFEVEN, status); // round to 20 fraction digits. |
1078 | |
1079 | UBool pad = FALSE; |
1080 | for (int32_t didx = dl.getLowerDisplayMagnitude(); didx<0; didx++) { |
1081 | // Loop iterates over fraction digits, starting with the LSD. |
1082 | // include both real digits from the number, and zeros |
1083 | // to the left of the MSD but to the right of the decimal point. |
1084 | if (pad && useSpaces) { |
1085 | toInsertInto.insert(_pos + getPos(), gSpace); |
1086 | } else { |
1087 | pad = TRUE; |
1088 | } |
1089 | int64_t digit = dl.getDigit(didx); |
1090 | getRuleSet()->format(digit, toInsertInto, _pos + getPos(), recursionCount, status); |
1091 | } |
1092 | |
1093 | if (!pad) { |
1094 | // hack around lack of precision in digitlist. if we would end up with |
1095 | // "foo point" make sure we add a " zero" to the end. |
1096 | getRuleSet()->format((int64_t)0, toInsertInto, _pos + getPos(), recursionCount, status); |
1097 | } |
1098 | } |
1099 | } |
1100 | |
1101 | //----------------------------------------------------------------------- |
1102 | // parsing |
1103 | //----------------------------------------------------------------------- |
1104 | |
1105 | /** |
1106 | * If in "by digits" mode, parses the string as if it were a string |
1107 | * of individual digits; otherwise, uses the superclass function. |
1108 | * @param text The string to parse |
1109 | * @param parsePosition Ignored on entry, but updated on exit to point |
1110 | * to the first unmatched character |
1111 | * @param baseValue The partial parse result prior to entering this |
1112 | * function |
1113 | * @param upperBound Only consider rules with base values lower than |
1114 | * this when filling in the substitution |
1115 | * @param lenientParse If true, try matching the text as numerals if |
1116 | * matching as words doesn't work |
1117 | * @return If the match was successful, the current partial parse |
1118 | * result; otherwise new Long(0). The result is either a Long or |
1119 | * a Double. |
1120 | */ |
1121 | |
1122 | UBool |
1123 | FractionalPartSubstitution::doParse(const UnicodeString& text, |
1124 | ParsePosition& parsePosition, |
1125 | double baseValue, |
1126 | double /*upperBound*/, |
1127 | UBool lenientParse, |
1128 | uint32_t nonNumericalExecutedRuleMask, |
1129 | Formattable& resVal) const |
1130 | { |
1131 | // if we're not in byDigits mode, we can just use the inherited |
1132 | // doParse() |
1133 | if (!byDigits) { |
1134 | return NFSubstitution::doParse(text, parsePosition, baseValue, 0, lenientParse, nonNumericalExecutedRuleMask, resVal); |
1135 | |
1136 | // if we ARE in byDigits mode, parse the text one digit at a time |
1137 | // using this substitution's owning rule set (we do this by setting |
1138 | // upperBound to 10 when calling doParse() ) until we reach |
1139 | // nonmatching text |
1140 | } else { |
1141 | UnicodeString workText(text); |
1142 | ParsePosition workPos(1); |
1143 | double result = 0; |
1144 | int32_t digit; |
1145 | // double p10 = 0.1; |
1146 | |
1147 | DecimalQuantity dl; |
1148 | int32_t totalDigits = 0; |
1149 | NumberFormat* fmt = NULL; |
1150 | while (workText.length() > 0 && workPos.getIndex() != 0) { |
1151 | workPos.setIndex(0); |
1152 | Formattable temp; |
1153 | getRuleSet()->parse(workText, workPos, 10, nonNumericalExecutedRuleMask, temp); |
1154 | UErrorCode status = U_ZERO_ERROR; |
1155 | digit = temp.getLong(status); |
1156 | // digit = temp.getType() == Formattable::kLong ? |
1157 | // temp.getLong() : |
1158 | // (int32_t)temp.getDouble(); |
1159 | |
1160 | if (lenientParse && workPos.getIndex() == 0) { |
1161 | if (!fmt) { |
1162 | status = U_ZERO_ERROR; |
1163 | fmt = NumberFormat::createInstance(status); |
1164 | if (U_FAILURE(status)) { |
1165 | delete fmt; |
1166 | fmt = NULL; |
1167 | } |
1168 | } |
1169 | if (fmt) { |
1170 | fmt->parse(workText, temp, workPos); |
1171 | digit = temp.getLong(status); |
1172 | } |
1173 | } |
1174 | |
1175 | if (workPos.getIndex() != 0) { |
1176 | dl.appendDigit(static_cast<int8_t>(digit), 0, true); |
1177 | totalDigits++; |
1178 | // result += digit * p10; |
1179 | // p10 /= 10; |
1180 | parsePosition.setIndex(parsePosition.getIndex() + workPos.getIndex()); |
1181 | workText.removeBetween(0, workPos.getIndex()); |
1182 | while (workText.length() > 0 && workText.charAt(0) == gSpace) { |
1183 | workText.removeBetween(0, 1); |
1184 | parsePosition.setIndex(parsePosition.getIndex() + 1); |
1185 | } |
1186 | } |
1187 | } |
1188 | delete fmt; |
1189 | |
1190 | dl.adjustMagnitude(-totalDigits); |
1191 | result = dl.toDouble(); |
1192 | result = composeRuleValue(result, baseValue); |
1193 | resVal.setDouble(result); |
1194 | return TRUE; |
1195 | } |
1196 | } |
1197 | |
1198 | UBool |
1199 | FractionalPartSubstitution::operator==(const NFSubstitution& rhs) const |
1200 | { |
1201 | return NFSubstitution::operator==(rhs) && |
1202 | ((const FractionalPartSubstitution*)&rhs)->byDigits == byDigits; |
1203 | } |
1204 | |
1205 | UOBJECT_DEFINE_RTTI_IMPLEMENTATION(FractionalPartSubstitution) |
1206 | |
1207 | |
1208 | //=================================================================== |
1209 | // AbsoluteValueSubstitution |
1210 | //=================================================================== |
1211 | |
1212 | UOBJECT_DEFINE_RTTI_IMPLEMENTATION(AbsoluteValueSubstitution) |
1213 | |
1214 | //=================================================================== |
1215 | // NumeratorSubstitution |
1216 | //=================================================================== |
1217 | |
1218 | void |
1219 | NumeratorSubstitution::doSubstitution(double number, UnicodeString& toInsertInto, int32_t apos, int32_t recursionCount, UErrorCode& status) const { |
1220 | // perform a transformation on the number being formatted that |
1221 | // is dependent on the type of substitution this is |
1222 | |
1223 | double numberToFormat = transformNumber(number); |
1224 | int64_t longNF = util64_fromDouble(numberToFormat); |
1225 | |
1226 | const NFRuleSet* aruleSet = getRuleSet(); |
1227 | if (withZeros && aruleSet != NULL) { |
1228 | // if there are leading zeros in the decimal expansion then emit them |
1229 | int64_t nf =longNF; |
1230 | int32_t len = toInsertInto.length(); |
1231 | while ((nf *= 10) < denominator) { |
1232 | toInsertInto.insert(apos + getPos(), gSpace); |
1233 | aruleSet->format((int64_t)0, toInsertInto, apos + getPos(), recursionCount, status); |
1234 | } |
1235 | apos += toInsertInto.length() - len; |
1236 | } |
1237 | |
1238 | // if the result is an integer, from here on out we work in integer |
1239 | // space (saving time and memory and preserving accuracy) |
1240 | if (numberToFormat == longNF && aruleSet != NULL) { |
1241 | aruleSet->format(longNF, toInsertInto, apos + getPos(), recursionCount, status); |
1242 | |
1243 | // if the result isn't an integer, then call either our rule set's |
1244 | // format() method or our DecimalFormat's format() method to |
1245 | // format the result |
1246 | } else { |
1247 | if (aruleSet != NULL) { |
1248 | aruleSet->format(numberToFormat, toInsertInto, apos + getPos(), recursionCount, status); |
1249 | } else { |
1250 | UnicodeString temp; |
1251 | getNumberFormat()->format(numberToFormat, temp, status); |
1252 | toInsertInto.insert(apos + getPos(), temp); |
1253 | } |
1254 | } |
1255 | } |
1256 | |
1257 | UBool |
1258 | NumeratorSubstitution::doParse(const UnicodeString& text, |
1259 | ParsePosition& parsePosition, |
1260 | double baseValue, |
1261 | double upperBound, |
1262 | UBool /*lenientParse*/, |
1263 | uint32_t nonNumericalExecutedRuleMask, |
1264 | Formattable& result) const |
1265 | { |
1266 | // we don't have to do anything special to do the parsing here, |
1267 | // but we have to turn lenient parsing off-- if we leave it on, |
1268 | // it SERIOUSLY messes up the algorithm |
1269 | |
1270 | // if withZeros is true, we need to count the zeros |
1271 | // and use that to adjust the parse result |
1272 | UErrorCode status = U_ZERO_ERROR; |
1273 | int32_t zeroCount = 0; |
1274 | UnicodeString workText(text); |
1275 | |
1276 | if (withZeros) { |
1277 | ParsePosition workPos(1); |
1278 | Formattable temp; |
1279 | |
1280 | while (workText.length() > 0 && workPos.getIndex() != 0) { |
1281 | workPos.setIndex(0); |
1282 | getRuleSet()->parse(workText, workPos, 1, nonNumericalExecutedRuleMask, temp); // parse zero or nothing at all |
1283 | if (workPos.getIndex() == 0) { |
1284 | // we failed, either there were no more zeros, or the number was formatted with digits |
1285 | // either way, we're done |
1286 | break; |
1287 | } |
1288 | |
1289 | ++zeroCount; |
1290 | parsePosition.setIndex(parsePosition.getIndex() + workPos.getIndex()); |
1291 | workText.remove(0, workPos.getIndex()); |
1292 | while (workText.length() > 0 && workText.charAt(0) == gSpace) { |
1293 | workText.remove(0, 1); |
1294 | parsePosition.setIndex(parsePosition.getIndex() + 1); |
1295 | } |
1296 | } |
1297 | |
1298 | workText = text; |
1299 | workText.remove(0, (int32_t)parsePosition.getIndex()); |
1300 | parsePosition.setIndex(0); |
1301 | } |
1302 | |
1303 | // we've parsed off the zeros, now let's parse the rest from our current position |
1304 | NFSubstitution::doParse(workText, parsePosition, withZeros ? 1 : baseValue, upperBound, FALSE, nonNumericalExecutedRuleMask, result); |
1305 | |
1306 | if (withZeros) { |
1307 | // any base value will do in this case. is there a way to |
1308 | // force this to not bother trying all the base values? |
1309 | |
1310 | // compute the 'effective' base and prescale the value down |
1311 | int64_t n = result.getLong(status); // force conversion! |
1312 | int64_t d = 1; |
1313 | int32_t pow = 0; |
1314 | while (d <= n) { |
1315 | d *= 10; |
1316 | ++pow; |
1317 | } |
1318 | // now add the zeros |
1319 | while (zeroCount > 0) { |
1320 | d *= 10; |
1321 | --zeroCount; |
1322 | } |
1323 | // d is now our true denominator |
1324 | result.setDouble((double)n/(double)d); |
1325 | } |
1326 | |
1327 | return TRUE; |
1328 | } |
1329 | |
1330 | UBool |
1331 | NumeratorSubstitution::operator==(const NFSubstitution& rhs) const |
1332 | { |
1333 | return NFSubstitution::operator==(rhs) && |
1334 | denominator == ((const NumeratorSubstitution*)&rhs)->denominator; |
1335 | } |
1336 | |
1337 | UOBJECT_DEFINE_RTTI_IMPLEMENTATION(NumeratorSubstitution) |
1338 | |
1339 | const UChar NumeratorSubstitution::LTLT[] = { 0x003c, 0x003c }; |
1340 | |
1341 | U_NAMESPACE_END |
1342 | |
1343 | /* U_HAVE_RBNF */ |
1344 | #endif |
1345 | |
1346 | |