intl/icu/source/i18n/nfsubs.cpp

changeset 0
6474c204b198
     1.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     1.2 +++ b/intl/icu/source/i18n/nfsubs.cpp	Wed Dec 31 06:09:35 2014 +0100
     1.3 @@ -0,0 +1,1364 @@
     1.4 +/*
     1.5 +******************************************************************************
     1.6 +*   Copyright (C) 1997-2012, International Business Machines
     1.7 +*   Corporation and others.  All Rights Reserved.
     1.8 +******************************************************************************
     1.9 +*   file name:  nfsubs.cpp
    1.10 +*   encoding:   US-ASCII
    1.11 +*   tab size:   8 (not used)
    1.12 +*   indentation:4
    1.13 +*
    1.14 +* Modification history
    1.15 +* Date        Name      Comments
    1.16 +* 10/11/2001  Doug      Ported from ICU4J
    1.17 +*/
    1.18 +
    1.19 +#include <stdio.h>
    1.20 +#include "utypeinfo.h"  // for 'typeid' to work
    1.21 +
    1.22 +#include "nfsubs.h"
    1.23 +#include "digitlst.h"
    1.24 +
    1.25 +#if U_HAVE_RBNF
    1.26 +
    1.27 +static const UChar gLessThan = 0x003c;
    1.28 +static const UChar gEquals = 0x003d;
    1.29 +static const UChar gGreaterThan = 0x003e;
    1.30 +static const UChar gPercent = 0x0025;
    1.31 +static const UChar gPound = 0x0023;
    1.32 +static const UChar gZero = 0x0030;
    1.33 +static const UChar gSpace = 0x0020;
    1.34 +
    1.35 +static const UChar gEqualsEquals[] =
    1.36 +{
    1.37 +    0x3D, 0x3D, 0
    1.38 +}; /* "==" */
    1.39 +static const UChar gGreaterGreaterGreaterThan[] =
    1.40 +{
    1.41 +    0x3E, 0x3E, 0x3E, 0
    1.42 +}; /* ">>>" */
    1.43 +static const UChar gGreaterGreaterThan[] =
    1.44 +{
    1.45 +    0x3E, 0x3E, 0
    1.46 +}; /* ">>" */
    1.47 +
    1.48 +U_NAMESPACE_BEGIN
    1.49 +
    1.50 +class SameValueSubstitution : public NFSubstitution {
    1.51 +public:
    1.52 +    SameValueSubstitution(int32_t pos,
    1.53 +        const NFRuleSet* ruleset,
    1.54 +        const RuleBasedNumberFormat* formatter,
    1.55 +        const UnicodeString& description,
    1.56 +        UErrorCode& status);
    1.57 +    virtual ~SameValueSubstitution();
    1.58 +
    1.59 +    virtual int64_t transformNumber(int64_t number) const { return number; }
    1.60 +    virtual double transformNumber(double number) const { return number; }
    1.61 +    virtual double composeRuleValue(double newRuleValue, double /*oldRuleValue*/) const { return newRuleValue; }
    1.62 +    virtual double calcUpperBound(double oldUpperBound) const { return oldUpperBound; }
    1.63 +    virtual UChar tokenChar() const { return (UChar)0x003d; } // '='
    1.64 +
    1.65 +public:
    1.66 +    static UClassID getStaticClassID(void);
    1.67 +    virtual UClassID getDynamicClassID(void) const;
    1.68 +};
    1.69 +
    1.70 +SameValueSubstitution::~SameValueSubstitution() {}
    1.71 +
    1.72 +class MultiplierSubstitution : public NFSubstitution {
    1.73 +    double divisor;
    1.74 +    int64_t ldivisor;
    1.75 +
    1.76 +public:
    1.77 +    MultiplierSubstitution(int32_t _pos,
    1.78 +        double _divisor,
    1.79 +        const NFRuleSet* _ruleSet,
    1.80 +        const RuleBasedNumberFormat* formatter,
    1.81 +        const UnicodeString& description,
    1.82 +        UErrorCode& status)
    1.83 +        : NFSubstitution(_pos, _ruleSet, formatter, description, status), divisor(_divisor)
    1.84 +    {
    1.85 +        ldivisor = util64_fromDouble(divisor);
    1.86 +        if (divisor == 0) {
    1.87 +            status = U_PARSE_ERROR;
    1.88 +        }
    1.89 +    }
    1.90 +    virtual ~MultiplierSubstitution();
    1.91 +
    1.92 +    virtual void setDivisor(int32_t radix, int32_t exponent, UErrorCode& status) { 
    1.93 +        divisor = uprv_pow(radix, exponent);
    1.94 +        ldivisor = util64_fromDouble(divisor);
    1.95 +
    1.96 +        if(divisor == 0) {
    1.97 +            status = U_PARSE_ERROR;
    1.98 +        }
    1.99 +    }
   1.100 +
   1.101 +    virtual UBool operator==(const NFSubstitution& rhs) const;
   1.102 +
   1.103 +    virtual int64_t transformNumber(int64_t number) const {
   1.104 +        return number / ldivisor;
   1.105 +    }
   1.106 +
   1.107 +    virtual double transformNumber(double number) const {
   1.108 +        if (getRuleSet()) {
   1.109 +            return uprv_floor(number / divisor);
   1.110 +        } else {
   1.111 +            return number/divisor;
   1.112 +        }
   1.113 +    }
   1.114 +
   1.115 +    virtual double composeRuleValue(double newRuleValue, double /*oldRuleValue*/) const {
   1.116 +        return newRuleValue * divisor;
   1.117 +    }
   1.118 +
   1.119 +    virtual double calcUpperBound(double /*oldUpperBound*/) const { return divisor; }
   1.120 +
   1.121 +    virtual UChar tokenChar() const { return (UChar)0x003c; } // '<'
   1.122 +
   1.123 +public:
   1.124 +    static UClassID getStaticClassID(void);
   1.125 +    virtual UClassID getDynamicClassID(void) const;
   1.126 +};
   1.127 +
   1.128 +MultiplierSubstitution::~MultiplierSubstitution() {}
   1.129 +
   1.130 +class ModulusSubstitution : public NFSubstitution {
   1.131 +    double divisor;
   1.132 +    int64_t  ldivisor;
   1.133 +    const NFRule* ruleToUse;
   1.134 +public:
   1.135 +    ModulusSubstitution(int32_t pos,
   1.136 +        double _divisor,
   1.137 +        const NFRule* rulePredecessor,
   1.138 +        const NFRuleSet* ruleSet,
   1.139 +        const RuleBasedNumberFormat* formatter,
   1.140 +        const UnicodeString& description,
   1.141 +        UErrorCode& status);
   1.142 +    virtual ~ModulusSubstitution();
   1.143 +
   1.144 +    virtual void setDivisor(int32_t radix, int32_t exponent, UErrorCode& status) { 
   1.145 +        divisor = uprv_pow(radix, exponent);
   1.146 +        ldivisor = util64_fromDouble(divisor);
   1.147 +
   1.148 +        if (divisor == 0) {
   1.149 +            status = U_PARSE_ERROR;
   1.150 +        }
   1.151 +    }
   1.152 +
   1.153 +    virtual UBool operator==(const NFSubstitution& rhs) const;
   1.154 +
   1.155 +    virtual void doSubstitution(int64_t number, UnicodeString& toInsertInto, int32_t pos) const;
   1.156 +    virtual void doSubstitution(double number, UnicodeString& toInsertInto, int32_t pos) const;
   1.157 +
   1.158 +    virtual int64_t transformNumber(int64_t number) const { return number % ldivisor; }
   1.159 +    virtual double transformNumber(double number) const { return uprv_fmod(number, divisor); }
   1.160 +
   1.161 +    virtual UBool doParse(const UnicodeString& text, 
   1.162 +        ParsePosition& parsePosition,
   1.163 +        double baseValue,
   1.164 +        double upperBound,
   1.165 +        UBool lenientParse,
   1.166 +        Formattable& result) const;
   1.167 +
   1.168 +    virtual double composeRuleValue(double newRuleValue, double oldRuleValue) const {
   1.169 +        return oldRuleValue - uprv_fmod(oldRuleValue, divisor) + newRuleValue;
   1.170 +    }
   1.171 +
   1.172 +    virtual double calcUpperBound(double /*oldUpperBound*/) const { return divisor; }
   1.173 +
   1.174 +    virtual UBool isModulusSubstitution() const { return TRUE; }
   1.175 +
   1.176 +    virtual UChar tokenChar() const { return (UChar)0x003e; } // '>'
   1.177 +
   1.178 +	virtual void toString(UnicodeString& result) const;
   1.179 +
   1.180 +public:
   1.181 +    static UClassID getStaticClassID(void);
   1.182 +    virtual UClassID getDynamicClassID(void) const;
   1.183 +};
   1.184 +
   1.185 +ModulusSubstitution::~ModulusSubstitution() {}
   1.186 +
   1.187 +class IntegralPartSubstitution : public NFSubstitution {
   1.188 +public:
   1.189 +    IntegralPartSubstitution(int32_t _pos,
   1.190 +        const NFRuleSet* _ruleSet,
   1.191 +        const RuleBasedNumberFormat* formatter,
   1.192 +        const UnicodeString& description,
   1.193 +        UErrorCode& status)
   1.194 +        : NFSubstitution(_pos, _ruleSet, formatter, description, status) {}
   1.195 +    virtual ~IntegralPartSubstitution();
   1.196 +
   1.197 +    virtual int64_t transformNumber(int64_t number) const { return number; }
   1.198 +    virtual double transformNumber(double number) const { return uprv_floor(number); }
   1.199 +    virtual double composeRuleValue(double newRuleValue, double oldRuleValue) const { return newRuleValue + oldRuleValue; }
   1.200 +    virtual double calcUpperBound(double /*oldUpperBound*/) const { return DBL_MAX; }
   1.201 +    virtual UChar tokenChar() const { return (UChar)0x003c; } // '<'
   1.202 +
   1.203 +public:
   1.204 +    static UClassID getStaticClassID(void);
   1.205 +    virtual UClassID getDynamicClassID(void) const;
   1.206 +};
   1.207 +
   1.208 +IntegralPartSubstitution::~IntegralPartSubstitution() {}
   1.209 +
   1.210 +class FractionalPartSubstitution : public NFSubstitution {
   1.211 +    UBool byDigits;
   1.212 +    UBool useSpaces;
   1.213 +    enum { kMaxDecimalDigits = 8 };
   1.214 +public:
   1.215 +    FractionalPartSubstitution(int32_t pos,
   1.216 +        const NFRuleSet* ruleSet,
   1.217 +        const RuleBasedNumberFormat* formatter,
   1.218 +        const UnicodeString& description,
   1.219 +        UErrorCode& status);
   1.220 +    virtual ~FractionalPartSubstitution();
   1.221 +
   1.222 +    virtual UBool operator==(const NFSubstitution& rhs) const;
   1.223 +
   1.224 +    virtual void doSubstitution(double number, UnicodeString& toInsertInto, int32_t pos) const;
   1.225 +    virtual void doSubstitution(int64_t /*number*/, UnicodeString& /*toInsertInto*/, int32_t /*_pos*/) const {}
   1.226 +    virtual int64_t transformNumber(int64_t /*number*/) const { return 0; }
   1.227 +    virtual double transformNumber(double number) const { return number - uprv_floor(number); }
   1.228 +
   1.229 +    virtual UBool doParse(const UnicodeString& text,
   1.230 +        ParsePosition& parsePosition,
   1.231 +        double baseValue,
   1.232 +        double upperBound,
   1.233 +        UBool lenientParse,
   1.234 +        Formattable& result) const;
   1.235 +
   1.236 +    virtual double composeRuleValue(double newRuleValue, double oldRuleValue) const { return newRuleValue + oldRuleValue; }
   1.237 +    virtual double calcUpperBound(double /*oldUpperBound*/) const { return 0.0; }
   1.238 +    virtual UChar tokenChar() const { return (UChar)0x003e; } // '>'
   1.239 +
   1.240 +public:
   1.241 +    static UClassID getStaticClassID(void);
   1.242 +    virtual UClassID getDynamicClassID(void) const;
   1.243 +};
   1.244 +
   1.245 +FractionalPartSubstitution::~FractionalPartSubstitution() {}
   1.246 +
   1.247 +class AbsoluteValueSubstitution : public NFSubstitution {
   1.248 +public:
   1.249 +    AbsoluteValueSubstitution(int32_t _pos,
   1.250 +        const NFRuleSet* _ruleSet,
   1.251 +        const RuleBasedNumberFormat* formatter,
   1.252 +        const UnicodeString& description,
   1.253 +        UErrorCode& status)
   1.254 +        : NFSubstitution(_pos, _ruleSet, formatter, description, status) {}
   1.255 +    virtual ~AbsoluteValueSubstitution();
   1.256 +
   1.257 +    virtual int64_t transformNumber(int64_t number) const { return number >= 0 ? number : -number; }
   1.258 +    virtual double transformNumber(double number) const { return uprv_fabs(number); }
   1.259 +    virtual double composeRuleValue(double newRuleValue, double /*oldRuleValue*/) const { return -newRuleValue; }
   1.260 +    virtual double calcUpperBound(double /*oldUpperBound*/) const { return DBL_MAX; }
   1.261 +    virtual UChar tokenChar() const { return (UChar)0x003e; } // '>'
   1.262 +
   1.263 +public:
   1.264 +    static UClassID getStaticClassID(void);
   1.265 +    virtual UClassID getDynamicClassID(void) const;
   1.266 +};
   1.267 +
   1.268 +AbsoluteValueSubstitution::~AbsoluteValueSubstitution() {}
   1.269 +
   1.270 +class NumeratorSubstitution : public NFSubstitution {
   1.271 +    double denominator;
   1.272 +    int64_t ldenominator;
   1.273 +    UBool withZeros;
   1.274 +public:
   1.275 +    static inline UnicodeString fixdesc(const UnicodeString& desc) {
   1.276 +        if (desc.endsWith(LTLT, 2)) {
   1.277 +            UnicodeString result(desc, 0, desc.length()-1);
   1.278 +            return result;
   1.279 +        }
   1.280 +        return desc;
   1.281 +    }
   1.282 +    NumeratorSubstitution(int32_t _pos,
   1.283 +        double _denominator,
   1.284 +        const NFRuleSet* _ruleSet,
   1.285 +        const RuleBasedNumberFormat* formatter,
   1.286 +        const UnicodeString& description,
   1.287 +        UErrorCode& status)
   1.288 +        : NFSubstitution(_pos, _ruleSet, formatter, fixdesc(description), status), denominator(_denominator) 
   1.289 +    {
   1.290 +        ldenominator = util64_fromDouble(denominator);
   1.291 +        withZeros = description.endsWith(LTLT, 2);
   1.292 +    }
   1.293 +    virtual ~NumeratorSubstitution();
   1.294 +
   1.295 +    virtual UBool operator==(const NFSubstitution& rhs) const;
   1.296 +
   1.297 +    virtual int64_t transformNumber(int64_t number) const { return number * ldenominator; }
   1.298 +    virtual double transformNumber(double number) const { return uprv_round(number * denominator); }
   1.299 +
   1.300 +    virtual void doSubstitution(int64_t /*number*/, UnicodeString& /*toInsertInto*/, int32_t /*_pos*/) const {}
   1.301 +    virtual void doSubstitution(double number, UnicodeString& toInsertInto, int32_t pos) const;
   1.302 +    virtual UBool doParse(const UnicodeString& text, 
   1.303 +        ParsePosition& parsePosition,
   1.304 +        double baseValue,
   1.305 +        double upperBound,
   1.306 +        UBool /*lenientParse*/,
   1.307 +        Formattable& result) const;
   1.308 +
   1.309 +    virtual double composeRuleValue(double newRuleValue, double oldRuleValue) const { return newRuleValue / oldRuleValue; }
   1.310 +    virtual double calcUpperBound(double /*oldUpperBound*/) const { return denominator; }
   1.311 +    virtual UChar tokenChar() const { return (UChar)0x003c; } // '<'
   1.312 +private:
   1.313 +    static const UChar LTLT[2];
   1.314 +
   1.315 +public:
   1.316 +    static UClassID getStaticClassID(void);
   1.317 +    virtual UClassID getDynamicClassID(void) const;
   1.318 +};
   1.319 +
   1.320 +NumeratorSubstitution::~NumeratorSubstitution() {}
   1.321 +
   1.322 +class NullSubstitution : public NFSubstitution {
   1.323 +public:
   1.324 +    NullSubstitution(int32_t _pos,
   1.325 +        const NFRuleSet* _ruleSet,
   1.326 +        const RuleBasedNumberFormat* formatter,
   1.327 +        const UnicodeString& description,
   1.328 +        UErrorCode& status)
   1.329 +        : NFSubstitution(_pos, _ruleSet, formatter, description, status) {}
   1.330 +    virtual ~NullSubstitution();
   1.331 +
   1.332 +    virtual void toString(UnicodeString& /*result*/) const {}
   1.333 +    virtual void doSubstitution(double /*number*/, UnicodeString& /*toInsertInto*/, int32_t /*_pos*/) const {}
   1.334 +    virtual void doSubstitution(int64_t /*number*/, UnicodeString& /*toInsertInto*/, int32_t /*_pos*/) const {}
   1.335 +    virtual int64_t transformNumber(int64_t /*number*/) const { return 0; }
   1.336 +    virtual double transformNumber(double /*number*/) const { return 0; }
   1.337 +    virtual UBool doParse(const UnicodeString& /*text*/,
   1.338 +        ParsePosition& /*parsePosition*/, 
   1.339 +        double baseValue,
   1.340 +        double /*upperBound*/,
   1.341 +        UBool /*lenientParse*/,
   1.342 +        Formattable& result) const
   1.343 +    { result.setDouble(baseValue); return TRUE; }
   1.344 +    virtual double composeRuleValue(double /*newRuleValue*/, double /*oldRuleValue*/) const { return 0.0; } // never called
   1.345 +    virtual double calcUpperBound(double /*oldUpperBound*/) const { return 0; } // never called
   1.346 +    virtual UBool isNullSubstitution() const { return TRUE; }
   1.347 +    virtual UChar tokenChar() const { return (UChar)0x0020; } // ' ' never called
   1.348 +
   1.349 +public:
   1.350 +    static UClassID getStaticClassID(void);
   1.351 +    virtual UClassID getDynamicClassID(void) const;
   1.352 +};
   1.353 +
   1.354 +NullSubstitution::~NullSubstitution() {}
   1.355 +
   1.356 +NFSubstitution*
   1.357 +NFSubstitution::makeSubstitution(int32_t pos,
   1.358 +                                 const NFRule* rule,
   1.359 +                                 const NFRule* predecessor,
   1.360 +                                 const NFRuleSet* ruleSet,
   1.361 +                                 const RuleBasedNumberFormat* formatter,
   1.362 +                                 const UnicodeString& description,
   1.363 +                                 UErrorCode& status)
   1.364 +{
   1.365 +    // if the description is empty, return a NullSubstitution
   1.366 +    if (description.length() == 0) {
   1.367 +        return new NullSubstitution(pos, ruleSet, formatter, description, status);
   1.368 +    }
   1.369 +
   1.370 +    switch (description.charAt(0)) {
   1.371 +        // if the description begins with '<'...
   1.372 +    case gLessThan:
   1.373 +        // throw an exception if the rule is a negative number
   1.374 +        // rule
   1.375 +        if (rule->getBaseValue() == NFRule::kNegativeNumberRule) {
   1.376 +            // throw new IllegalArgumentException("<< not allowed in negative-number rule");
   1.377 +            status = U_PARSE_ERROR;
   1.378 +            return NULL;
   1.379 +        }
   1.380 +
   1.381 +        // if the rule is a fraction rule, return an
   1.382 +        // IntegralPartSubstitution
   1.383 +        else if (rule->getBaseValue() == NFRule::kImproperFractionRule
   1.384 +            || rule->getBaseValue() == NFRule::kProperFractionRule
   1.385 +            || rule->getBaseValue() == NFRule::kMasterRule) {
   1.386 +            return new IntegralPartSubstitution(pos, ruleSet, formatter, description, status);
   1.387 +        }
   1.388 +
   1.389 +        // if the rule set containing the rule is a fraction
   1.390 +        // rule set, return a NumeratorSubstitution
   1.391 +        else if (ruleSet->isFractionRuleSet()) {
   1.392 +            return new NumeratorSubstitution(pos, (double)rule->getBaseValue(),
   1.393 +                formatter->getDefaultRuleSet(), formatter, description, status);
   1.394 +        }
   1.395 +
   1.396 +        // otherwise, return a MultiplierSubstitution
   1.397 +        else {
   1.398 +            return new MultiplierSubstitution(pos, rule->getDivisor(), ruleSet,
   1.399 +                formatter, description, status);
   1.400 +        }
   1.401 +
   1.402 +        // if the description begins with '>'...
   1.403 +    case gGreaterThan:
   1.404 +        // if the rule is a negative-number rule, return
   1.405 +        // an AbsoluteValueSubstitution
   1.406 +        if (rule->getBaseValue() == NFRule::kNegativeNumberRule) {
   1.407 +            return new AbsoluteValueSubstitution(pos, ruleSet, formatter, description, status);
   1.408 +        }
   1.409 +
   1.410 +        // if the rule is a fraction rule, return a
   1.411 +        // FractionalPartSubstitution
   1.412 +        else if (rule->getBaseValue() == NFRule::kImproperFractionRule
   1.413 +            || rule->getBaseValue() == NFRule::kProperFractionRule
   1.414 +            || rule->getBaseValue() == NFRule::kMasterRule) {
   1.415 +            return new FractionalPartSubstitution(pos, ruleSet, formatter, description, status);
   1.416 +        }
   1.417 +
   1.418 +        // if the rule set owning the rule is a fraction rule set,
   1.419 +        // throw an exception
   1.420 +        else if (ruleSet->isFractionRuleSet()) {
   1.421 +            // throw new IllegalArgumentException(">> not allowed in fraction rule set");
   1.422 +            status = U_PARSE_ERROR;
   1.423 +            return NULL;
   1.424 +        }
   1.425 +
   1.426 +        // otherwise, return a ModulusSubstitution
   1.427 +        else {
   1.428 +            return new ModulusSubstitution(pos, rule->getDivisor(), predecessor,
   1.429 +                ruleSet, formatter, description, status);
   1.430 +        }
   1.431 +
   1.432 +        // if the description begins with '=', always return a
   1.433 +        // SameValueSubstitution
   1.434 +    case gEquals:
   1.435 +        return new SameValueSubstitution(pos, ruleSet, formatter, description, status);
   1.436 +
   1.437 +        // and if it's anything else, throw an exception
   1.438 +    default:
   1.439 +        // throw new IllegalArgumentException("Illegal substitution character");
   1.440 +        status = U_PARSE_ERROR;
   1.441 +    }
   1.442 +    return NULL;
   1.443 +}
   1.444 +
   1.445 +NFSubstitution::NFSubstitution(int32_t _pos,
   1.446 +                               const NFRuleSet* _ruleSet,
   1.447 +                               const RuleBasedNumberFormat* formatter,
   1.448 +                               const UnicodeString& description,
   1.449 +                               UErrorCode& status)
   1.450 +                               : pos(_pos), ruleSet(NULL), numberFormat(NULL)
   1.451 +{
   1.452 +    // the description should begin and end with the same character.
   1.453 +    // If it doesn't that's a syntax error.  Otherwise,
   1.454 +    // makeSubstitution() was the only thing that needed to know
   1.455 +    // about these characters, so strip them off
   1.456 +    UnicodeString workingDescription(description);
   1.457 +    if (description.length() >= 2
   1.458 +        && description.charAt(0) == description.charAt(description.length() - 1))
   1.459 +    {
   1.460 +        workingDescription.remove(description.length() - 1, 1);
   1.461 +        workingDescription.remove(0, 1);
   1.462 +    }
   1.463 +    else if (description.length() != 0) {
   1.464 +        // throw new IllegalArgumentException("Illegal substitution syntax");
   1.465 +        status = U_PARSE_ERROR;
   1.466 +        return;
   1.467 +    }
   1.468 +
   1.469 +    // if the description was just two paired token characters
   1.470 +    // (i.e., "<<" or ">>"), it uses the rule set it belongs to to
   1.471 +    // format its result
   1.472 +    if (workingDescription.length() == 0) {
   1.473 +        this->ruleSet = _ruleSet;
   1.474 +    }
   1.475 +    // if the description contains a rule set name, that's the rule
   1.476 +    // set we use to format the result: get a reference to the
   1.477 +    // names rule set
   1.478 +    else if (workingDescription.charAt(0) == gPercent) {
   1.479 +        this->ruleSet = formatter->findRuleSet(workingDescription, status);
   1.480 +    }
   1.481 +    // if the description begins with 0 or #, treat it as a
   1.482 +    // DecimalFormat pattern, and initialize a DecimalFormat with
   1.483 +    // that pattern (then set it to use the DecimalFormatSymbols
   1.484 +    // belonging to our formatter)
   1.485 +    else if (workingDescription.charAt(0) == gPound || workingDescription.charAt(0) ==gZero) {
   1.486 +        DecimalFormatSymbols* sym = formatter->getDecimalFormatSymbols();
   1.487 +        if (!sym) {
   1.488 +            status = U_MISSING_RESOURCE_ERROR;
   1.489 +            return;
   1.490 +        }
   1.491 +        this->numberFormat = new DecimalFormat(workingDescription, *sym, status);
   1.492 +        /* test for NULL */
   1.493 +        if (this->numberFormat == 0) {
   1.494 +            status = U_MEMORY_ALLOCATION_ERROR;
   1.495 +            return;
   1.496 +        }
   1.497 +        if (U_FAILURE(status)) {
   1.498 +            delete (DecimalFormat*)this->numberFormat;
   1.499 +            this->numberFormat = NULL;
   1.500 +            return;
   1.501 +        }
   1.502 +        // this->numberFormat->setDecimalFormatSymbols(formatter->getDecimalFormatSymbols());
   1.503 +    }
   1.504 +    // if the description is ">>>", this substitution bypasses the
   1.505 +    // usual rule-search process and always uses the rule that precedes
   1.506 +    // it in its own rule set's rule list (this is used for place-value
   1.507 +    // notations: formats where you want to see a particular part of
   1.508 +    // a number even when it's 0)
   1.509 +    else if (workingDescription.charAt(0) == gGreaterThan) {
   1.510 +        // this causes problems when >>> is used in a frationalPartSubstitution
   1.511 +        // this->ruleSet = NULL;
   1.512 +        this->ruleSet = _ruleSet;
   1.513 +        this->numberFormat = NULL;
   1.514 +    }
   1.515 +    // and of the description is none of these things, it's a syntax error
   1.516 +    else {
   1.517 +        // throw new IllegalArgumentException("Illegal substitution syntax");
   1.518 +        status = U_PARSE_ERROR;
   1.519 +    }
   1.520 +}
   1.521 +
   1.522 +NFSubstitution::~NFSubstitution()
   1.523 +{
   1.524 +  // cast away const
   1.525 +  delete (NumberFormat*)numberFormat; numberFormat = NULL;
   1.526 +}
   1.527 +
   1.528 +/**
   1.529 + * Set's the substitution's divisor.  Used by NFRule.setBaseValue().
   1.530 + * A no-op for all substitutions except multiplier and modulus
   1.531 + * substitutions.
   1.532 + * @param radix The radix of the divisor
   1.533 + * @param exponent The exponent of the divisor
   1.534 + */
   1.535 +void
   1.536 +NFSubstitution::setDivisor(int32_t /*radix*/, int32_t /*exponent*/, UErrorCode& /*status*/) {
   1.537 +  // a no-op for all substitutions except multiplier and modulus substitutions
   1.538 +}
   1.539 +
   1.540 +
   1.541 +//-----------------------------------------------------------------------
   1.542 +// boilerplate
   1.543 +//-----------------------------------------------------------------------
   1.544 +
   1.545 +UOBJECT_DEFINE_RTTI_IMPLEMENTATION(NFSubstitution)
   1.546 +
   1.547 +/**
   1.548 + * Compares two substitutions for equality
   1.549 + * @param The substitution to compare this one to
   1.550 + * @return true if the two substitutions are functionally equivalent
   1.551 + */
   1.552 +UBool
   1.553 +NFSubstitution::operator==(const NFSubstitution& rhs) const
   1.554 +{
   1.555 +  // compare class and all of the fields all substitutions have
   1.556 +  // in common
   1.557 +  // this should be called by subclasses before their own equality tests
   1.558 +  return typeid(*this) == typeid(rhs)
   1.559 +  && pos == rhs.pos
   1.560 +  && (ruleSet == NULL) == (rhs.ruleSet == NULL)
   1.561 +  // && ruleSet == rhs.ruleSet causes circularity, other checks to make instead?
   1.562 +  && (numberFormat == NULL
   1.563 +      ? (rhs.numberFormat == NULL)
   1.564 +      : (*numberFormat == *rhs.numberFormat));
   1.565 +}
   1.566 +
   1.567 +/**
   1.568 + * Returns a textual description of the substitution
   1.569 + * @return A textual description of the substitution.  This might
   1.570 + * not be identical to the description it was created from, but
   1.571 + * it'll produce the same result.
   1.572 + */
   1.573 +void
   1.574 +NFSubstitution::toString(UnicodeString& text) const
   1.575 +{
   1.576 +  // use tokenChar() to get the character at the beginning and
   1.577 +  // end of the substitutin token.  In between them will go
   1.578 +  // either the name of the rule set it uses, or the pattern of
   1.579 +  // the DecimalFormat it uses
   1.580 +  text.remove();
   1.581 +  text.append(tokenChar());
   1.582 +
   1.583 +  UnicodeString temp;
   1.584 +  if (ruleSet != NULL) {
   1.585 +    ruleSet->getName(temp);
   1.586 +  } else if (numberFormat != NULL) {
   1.587 +    numberFormat->toPattern(temp);
   1.588 +  }
   1.589 +  text.append(temp);
   1.590 +  text.append(tokenChar());
   1.591 +}
   1.592 +
   1.593 +//-----------------------------------------------------------------------
   1.594 +// formatting
   1.595 +//-----------------------------------------------------------------------
   1.596 +
   1.597 +/**
   1.598 + * Performs a mathematical operation on the number, formats it using
   1.599 + * either ruleSet or decimalFormat, and inserts the result into
   1.600 + * toInsertInto.
   1.601 + * @param number The number being formatted.
   1.602 + * @param toInsertInto The string we insert the result into
   1.603 + * @param pos The position in toInsertInto where the owning rule's
   1.604 + * rule text begins (this value is added to this substitution's
   1.605 + * position to determine exactly where to insert the new text)
   1.606 + */
   1.607 +void
   1.608 +NFSubstitution::doSubstitution(int64_t number, UnicodeString& toInsertInto, int32_t _pos) const
   1.609 +{
   1.610 +    if (ruleSet != NULL) {
   1.611 +        // perform a transformation on the number that is dependent
   1.612 +        // on the type of substitution this is, then just call its
   1.613 +        // rule set's format() method to format the result
   1.614 +        ruleSet->format(transformNumber(number), toInsertInto, _pos + this->pos);
   1.615 +    } else if (numberFormat != NULL) {
   1.616 +        // or perform the transformation on the number (preserving
   1.617 +        // the result's fractional part if the formatter it set
   1.618 +        // to show it), then use that formatter's format() method
   1.619 +        // to format the result
   1.620 +        double numberToFormat = transformNumber((double)number);
   1.621 +        if (numberFormat->getMaximumFractionDigits() == 0) {
   1.622 +            numberToFormat = uprv_floor(numberToFormat);
   1.623 +        }
   1.624 +
   1.625 +        UnicodeString temp;
   1.626 +        numberFormat->format(numberToFormat, temp);
   1.627 +        toInsertInto.insert(_pos + this->pos, temp);
   1.628 +    }
   1.629 +}
   1.630 +
   1.631 +/**
   1.632 + * Performs a mathematical operation on the number, formats it using
   1.633 + * either ruleSet or decimalFormat, and inserts the result into
   1.634 + * toInsertInto.
   1.635 + * @param number The number being formatted.
   1.636 + * @param toInsertInto The string we insert the result into
   1.637 + * @param pos The position in toInsertInto where the owning rule's
   1.638 + * rule text begins (this value is added to this substitution's
   1.639 + * position to determine exactly where to insert the new text)
   1.640 + */
   1.641 +void
   1.642 +NFSubstitution::doSubstitution(double number, UnicodeString& toInsertInto, int32_t _pos) const {
   1.643 +    // perform a transformation on the number being formatted that
   1.644 +    // is dependent on the type of substitution this is
   1.645 +    double numberToFormat = transformNumber(number);
   1.646 +
   1.647 +    // if the result is an integer, from here on out we work in integer
   1.648 +    // space (saving time and memory and preserving accuracy)
   1.649 +    if (numberToFormat == uprv_floor(numberToFormat) && ruleSet != NULL) {
   1.650 +        ruleSet->format(util64_fromDouble(numberToFormat), toInsertInto, _pos + this->pos);
   1.651 +
   1.652 +        // if the result isn't an integer, then call either our rule set's
   1.653 +        // format() method or our DecimalFormat's format() method to
   1.654 +        // format the result
   1.655 +    } else {
   1.656 +        if (ruleSet != NULL) {
   1.657 +            ruleSet->format(numberToFormat, toInsertInto, _pos + this->pos);
   1.658 +        } else if (numberFormat != NULL) {
   1.659 +            UnicodeString temp;
   1.660 +            numberFormat->format(numberToFormat, temp);
   1.661 +            toInsertInto.insert(_pos + this->pos, temp);
   1.662 +        }
   1.663 +    }
   1.664 +}
   1.665 +
   1.666 +
   1.667 +    //-----------------------------------------------------------------------
   1.668 +    // parsing
   1.669 +    //-----------------------------------------------------------------------
   1.670 +
   1.671 +#ifdef RBNF_DEBUG
   1.672 +#include <stdio.h>
   1.673 +#endif
   1.674 +
   1.675 +/**
   1.676 + * Parses a string using the rule set or DecimalFormat belonging
   1.677 + * to this substitution.  If there's a match, a mathematical
   1.678 + * operation (the inverse of the one used in formatting) is
   1.679 + * performed on the result of the parse and the value passed in
   1.680 + * and returned as the result.  The parse position is updated to
   1.681 + * point to the first unmatched character in the string.
   1.682 + * @param text The string to parse
   1.683 + * @param parsePosition On entry, ignored, but assumed to be 0.
   1.684 + * On exit, this is updated to point to the first unmatched
   1.685 + * character (or 0 if the substitution didn't match)
   1.686 + * @param baseValue A partial parse result that should be
   1.687 + * combined with the result of this parse
   1.688 + * @param upperBound When searching the rule set for a rule
   1.689 + * matching the string passed in, only rules with base values
   1.690 + * lower than this are considered
   1.691 + * @param lenientParse If true and matching against rules fails,
   1.692 + * the substitution will also try matching the text against
   1.693 + * numerals using a default-costructed NumberFormat.  If false,
   1.694 + * no extra work is done.  (This value is false whenever the
   1.695 + * formatter isn't in lenient-parse mode, but is also false
   1.696 + * under some conditions even when the formatter _is_ in
   1.697 + * lenient-parse mode.)
   1.698 + * @return If there's a match, this is the result of composing
   1.699 + * baseValue with whatever was returned from matching the
   1.700 + * characters.  This will be either a Long or a Double.  If there's
   1.701 + * no match this is new Long(0) (not null), and parsePosition
   1.702 + * is left unchanged.
   1.703 + */
   1.704 +UBool
   1.705 +NFSubstitution::doParse(const UnicodeString& text,
   1.706 +                        ParsePosition& parsePosition,
   1.707 +                        double baseValue,
   1.708 +                        double upperBound,
   1.709 +                        UBool lenientParse,
   1.710 +                        Formattable& result) const
   1.711 +{
   1.712 +#ifdef RBNF_DEBUG
   1.713 +    fprintf(stderr, "<nfsubs> %x bv: %g ub: %g\n", this, baseValue, upperBound);
   1.714 +#endif
   1.715 +    // figure out the highest base value a rule can have and match
   1.716 +    // the text being parsed (this varies according to the type of
   1.717 +    // substitutions: multiplier, modulus, and numerator substitutions
   1.718 +    // restrict the search to rules with base values lower than their
   1.719 +    // own; same-value substitutions leave the upper bound wherever
   1.720 +    // it was, and the others allow any rule to match
   1.721 +    upperBound = calcUpperBound(upperBound);
   1.722 +
   1.723 +    // use our rule set to parse the text.  If that fails and
   1.724 +    // lenient parsing is enabled (this is always false if the
   1.725 +    // formatter's lenient-parsing mode is off, but it may also
   1.726 +    // be false even when the formatter's lenient-parse mode is
   1.727 +    // on), then also try parsing the text using a default-
   1.728 +    // constructed NumberFormat
   1.729 +    if (ruleSet != NULL) {
   1.730 +        ruleSet->parse(text, parsePosition, upperBound, result);
   1.731 +        if (lenientParse && !ruleSet->isFractionRuleSet() && parsePosition.getIndex() == 0) {
   1.732 +            UErrorCode status = U_ZERO_ERROR;
   1.733 +            NumberFormat* fmt = NumberFormat::createInstance(status);
   1.734 +            if (U_SUCCESS(status)) {
   1.735 +                fmt->parse(text, result, parsePosition);
   1.736 +            }
   1.737 +            delete fmt;
   1.738 +        }
   1.739 +
   1.740 +        // ...or use our DecimalFormat to parse the text
   1.741 +    } else if (numberFormat != NULL) {
   1.742 +        numberFormat->parse(text, result, parsePosition);
   1.743 +    }
   1.744 +
   1.745 +    // if the parse was successful, we've already advanced the caller's
   1.746 +    // parse position (this is the one function that doesn't have one
   1.747 +    // of its own).  Derive a parse result and return it as a Long,
   1.748 +    // if possible, or a Double
   1.749 +    if (parsePosition.getIndex() != 0) {
   1.750 +        UErrorCode status = U_ZERO_ERROR;
   1.751 +        double tempResult = result.getDouble(status);
   1.752 +
   1.753 +        // composeRuleValue() produces a full parse result from
   1.754 +        // the partial parse result passed to this function from
   1.755 +        // the caller (this is either the owning rule's base value
   1.756 +        // or the partial result obtained from composing the
   1.757 +        // owning rule's base value with its other substitution's
   1.758 +        // parse result) and the partial parse result obtained by
   1.759 +        // matching the substitution (which will be the same value
   1.760 +        // the caller would get by parsing just this part of the
   1.761 +        // text with RuleBasedNumberFormat.parse() ).  How the two
   1.762 +        // values are used to derive the full parse result depends
   1.763 +        // on the types of substitutions: For a regular rule, the
   1.764 +        // ultimate result is its multiplier substitution's result
   1.765 +        // times the rule's divisor (or the rule's base value) plus
   1.766 +        // the modulus substitution's result (which will actually
   1.767 +        // supersede part of the rule's base value).  For a negative-
   1.768 +        // number rule, the result is the negative of its substitution's
   1.769 +        // result.  For a fraction rule, it's the sum of its two
   1.770 +        // substitution results.  For a rule in a fraction rule set,
   1.771 +        // it's the numerator substitution's result divided by
   1.772 +        // the rule's base value.  Results from same-value substitutions
   1.773 +        // propagate back upard, and null substitutions don't affect
   1.774 +        // the result.
   1.775 +        tempResult = composeRuleValue(tempResult, baseValue);
   1.776 +        result.setDouble(tempResult);
   1.777 +        return TRUE;
   1.778 +        // if the parse was UNsuccessful, return 0
   1.779 +    } else {
   1.780 +        result.setLong(0);
   1.781 +        return FALSE;
   1.782 +    }
   1.783 +}
   1.784 +
   1.785 +UBool
   1.786 +NFSubstitution::isNullSubstitution() const {
   1.787 +    return FALSE;
   1.788 +}
   1.789 +
   1.790 +    /**
   1.791 +     * Returns true if this is a modulus substitution.  (We didn't do this
   1.792 +     * with instanceof partially because it causes source files to
   1.793 +     * proliferate and partially because we have to port this to C++.)
   1.794 +     * @return true if this object is an instance of ModulusSubstitution
   1.795 +     */
   1.796 +UBool
   1.797 +NFSubstitution::isModulusSubstitution() const {
   1.798 +    return FALSE;
   1.799 +}
   1.800 +
   1.801 +//===================================================================
   1.802 +// SameValueSubstitution
   1.803 +//===================================================================
   1.804 +
   1.805 +/**
   1.806 + * A substitution that passes the value passed to it through unchanged.
   1.807 + * Represented by == in rule descriptions.
   1.808 + */
   1.809 +SameValueSubstitution::SameValueSubstitution(int32_t _pos,
   1.810 +                        const NFRuleSet* _ruleSet,
   1.811 +                        const RuleBasedNumberFormat* formatter,
   1.812 +                        const UnicodeString& description,
   1.813 +                        UErrorCode& status)
   1.814 +: NFSubstitution(_pos, _ruleSet, formatter, description, status)
   1.815 +{
   1.816 +    if (0 == description.compare(gEqualsEquals, 2)) {
   1.817 +        // throw new IllegalArgumentException("== is not a legal token");
   1.818 +        status = U_PARSE_ERROR;
   1.819 +    }
   1.820 +}
   1.821 +
   1.822 +UOBJECT_DEFINE_RTTI_IMPLEMENTATION(SameValueSubstitution)
   1.823 +
   1.824 +//===================================================================
   1.825 +// MultiplierSubstitution
   1.826 +//===================================================================
   1.827 +
   1.828 +UOBJECT_DEFINE_RTTI_IMPLEMENTATION(MultiplierSubstitution)
   1.829 +
   1.830 +UBool MultiplierSubstitution::operator==(const NFSubstitution& rhs) const
   1.831 +{
   1.832 +    return NFSubstitution::operator==(rhs) &&
   1.833 +        divisor == ((const MultiplierSubstitution*)&rhs)->divisor;
   1.834 +}
   1.835 +
   1.836 +
   1.837 +//===================================================================
   1.838 +// ModulusSubstitution
   1.839 +//===================================================================
   1.840 +
   1.841 +/**
   1.842 + * A substitution that divides the number being formatted by the its rule's
   1.843 + * divisor and formats the remainder.  Represented by "&gt;&gt;" in a
   1.844 + * regular rule.
   1.845 + */
   1.846 +ModulusSubstitution::ModulusSubstitution(int32_t _pos,
   1.847 +                                         double _divisor,
   1.848 +                                         const NFRule* predecessor,
   1.849 +                                         const NFRuleSet* _ruleSet,
   1.850 +                                         const RuleBasedNumberFormat* formatter,
   1.851 +                                         const UnicodeString& description,
   1.852 +                                         UErrorCode& status)
   1.853 + : NFSubstitution(_pos, _ruleSet, formatter, description, status)
   1.854 + , divisor(_divisor)
   1.855 + , ruleToUse(NULL)
   1.856 +{
   1.857 +  ldivisor = util64_fromDouble(_divisor);
   1.858 +
   1.859 +  // the owning rule's divisor controls the behavior of this
   1.860 +  // substitution: rather than keeping a backpointer to the rule,
   1.861 +  // we keep a copy of the divisor
   1.862 +
   1.863 +  if (ldivisor == 0) {
   1.864 +      status = U_PARSE_ERROR;
   1.865 +  }
   1.866 +
   1.867 +  if (0 == description.compare(gGreaterGreaterGreaterThan, 3)) {
   1.868 +    // the >>> token doesn't alter how this substituion calculates the
   1.869 +    // values it uses for formatting and parsing, but it changes
   1.870 +    // what's done with that value after it's obtained: >>> short-
   1.871 +    // circuits the rule-search process and goes straight to the
   1.872 +    // specified rule to format the substitution value
   1.873 +    ruleToUse = predecessor;
   1.874 +  }
   1.875 +}
   1.876 +
   1.877 +UOBJECT_DEFINE_RTTI_IMPLEMENTATION(ModulusSubstitution)
   1.878 +
   1.879 +UBool ModulusSubstitution::operator==(const NFSubstitution& rhs) const
   1.880 +{
   1.881 +  return NFSubstitution::operator==(rhs) &&
   1.882 +  divisor == ((const ModulusSubstitution*)&rhs)->divisor &&
   1.883 +  ruleToUse == ((const ModulusSubstitution*)&rhs)->ruleToUse;
   1.884 +}
   1.885 +
   1.886 +//-----------------------------------------------------------------------
   1.887 +// formatting
   1.888 +//-----------------------------------------------------------------------
   1.889 +
   1.890 +
   1.891 +/**
   1.892 + * If this is a &gt;&gt;&gt; substitution, use ruleToUse to fill in
   1.893 + * the substitution.  Otherwise, just use the superclass function.
   1.894 + * @param number The number being formatted
   1.895 + * @toInsertInto The string to insert the result of this substitution
   1.896 + * into
   1.897 + * @param pos The position of the rule text in toInsertInto
   1.898 + */
   1.899 +void
   1.900 +ModulusSubstitution::doSubstitution(int64_t number, UnicodeString& toInsertInto, int32_t _pos) const
   1.901 +{
   1.902 +    // if this isn't a >>> substitution, just use the inherited version
   1.903 +    // of this function (which uses either a rule set or a DecimalFormat
   1.904 +    // to format its substitution value)
   1.905 +    if (ruleToUse == NULL) {
   1.906 +        NFSubstitution::doSubstitution(number, toInsertInto, _pos);
   1.907 +
   1.908 +        // a >>> substitution goes straight to a particular rule to
   1.909 +        // format the substitution value
   1.910 +    } else {
   1.911 +        int64_t numberToFormat = transformNumber(number);
   1.912 +        ruleToUse->doFormat(numberToFormat, toInsertInto, _pos + getPos());
   1.913 +    }
   1.914 +}
   1.915 +
   1.916 +/**
   1.917 +* If this is a &gt;&gt;&gt; substitution, use ruleToUse to fill in
   1.918 +* the substitution.  Otherwise, just use the superclass function.
   1.919 +* @param number The number being formatted
   1.920 +* @toInsertInto The string to insert the result of this substitution
   1.921 +* into
   1.922 +* @param pos The position of the rule text in toInsertInto
   1.923 +*/
   1.924 +void
   1.925 +ModulusSubstitution::doSubstitution(double number, UnicodeString& toInsertInto, int32_t _pos) const
   1.926 +{
   1.927 +    // if this isn't a >>> substitution, just use the inherited version
   1.928 +    // of this function (which uses either a rule set or a DecimalFormat
   1.929 +    // to format its substitution value)
   1.930 +    if (ruleToUse == NULL) {
   1.931 +        NFSubstitution::doSubstitution(number, toInsertInto, _pos);
   1.932 +
   1.933 +        // a >>> substitution goes straight to a particular rule to
   1.934 +        // format the substitution value
   1.935 +    } else {
   1.936 +        double numberToFormat = transformNumber(number);
   1.937 +
   1.938 +        ruleToUse->doFormat(numberToFormat, toInsertInto, _pos + getPos());
   1.939 +    }
   1.940 +}
   1.941 +
   1.942 +//-----------------------------------------------------------------------
   1.943 +// parsing
   1.944 +//-----------------------------------------------------------------------
   1.945 +
   1.946 +/**
   1.947 + * If this is a &gt;&gt;&gt; substitution, match only against ruleToUse.
   1.948 + * Otherwise, use the superclass function.
   1.949 + * @param text The string to parse
   1.950 + * @param parsePosition Ignored on entry, updated on exit to point to
   1.951 + * the first unmatched character.
   1.952 + * @param baseValue The partial parse result prior to calling this
   1.953 + * routine.
   1.954 + */
   1.955 +UBool
   1.956 +ModulusSubstitution::doParse(const UnicodeString& text,
   1.957 +                             ParsePosition& parsePosition,
   1.958 +                             double baseValue,
   1.959 +                             double upperBound,
   1.960 +                             UBool lenientParse,
   1.961 +                             Formattable& result) const
   1.962 +{
   1.963 +    // if this isn't a >>> substitution, we can just use the
   1.964 +    // inherited parse() routine to do the parsing
   1.965 +    if (ruleToUse == NULL) {
   1.966 +        return NFSubstitution::doParse(text, parsePosition, baseValue, upperBound, lenientParse, result);
   1.967 +
   1.968 +        // but if it IS a >>> substitution, we have to do it here: we
   1.969 +        // use the specific rule's doParse() method, and then we have to
   1.970 +        // do some of the other work of NFRuleSet.parse()
   1.971 +    } else {
   1.972 +        ruleToUse->doParse(text, parsePosition, FALSE, upperBound, result);
   1.973 +
   1.974 +        if (parsePosition.getIndex() != 0) {
   1.975 +            UErrorCode status = U_ZERO_ERROR;
   1.976 +            double tempResult = result.getDouble(status);
   1.977 +            tempResult = composeRuleValue(tempResult, baseValue);
   1.978 +            result.setDouble(tempResult);
   1.979 +        }
   1.980 +
   1.981 +        return TRUE;
   1.982 +    }
   1.983 +}
   1.984 +/**
   1.985 + * Returns a textual description of the substitution
   1.986 + * @return A textual description of the substitution.  This might
   1.987 + * not be identical to the description it was created from, but
   1.988 + * it'll produce the same result.
   1.989 + */
   1.990 +void
   1.991 +ModulusSubstitution::toString(UnicodeString& text) const
   1.992 +{
   1.993 +  // use tokenChar() to get the character at the beginning and
   1.994 +  // end of the substitutin token.  In between them will go
   1.995 +  // either the name of the rule set it uses, or the pattern of
   1.996 +  // the DecimalFormat it uses
   1.997 +
   1.998 +  if ( ruleToUse != NULL ) { // Must have been a >>> substitution.
   1.999 +      text.remove();
  1.1000 +      text.append(tokenChar());
  1.1001 +      text.append(tokenChar());
  1.1002 +      text.append(tokenChar());
  1.1003 +  } else { // Otherwise just use the super-class function.
  1.1004 +	  NFSubstitution::toString(text);
  1.1005 +  }
  1.1006 +}
  1.1007 +//===================================================================
  1.1008 +// IntegralPartSubstitution
  1.1009 +//===================================================================
  1.1010 +
  1.1011 +UOBJECT_DEFINE_RTTI_IMPLEMENTATION(IntegralPartSubstitution)
  1.1012 +
  1.1013 +
  1.1014 +//===================================================================
  1.1015 +// FractionalPartSubstitution
  1.1016 +//===================================================================
  1.1017 +
  1.1018 +
  1.1019 +    /**
  1.1020 +     * Constructs a FractionalPartSubstitution.  This object keeps a flag
  1.1021 +     * telling whether it should format by digits or not.  In addition,
  1.1022 +     * it marks the rule set it calls (if any) as a fraction rule set.
  1.1023 +     */
  1.1024 +FractionalPartSubstitution::FractionalPartSubstitution(int32_t _pos,
  1.1025 +                             const NFRuleSet* _ruleSet,
  1.1026 +                             const RuleBasedNumberFormat* formatter,
  1.1027 +                             const UnicodeString& description,
  1.1028 +                             UErrorCode& status)
  1.1029 + : NFSubstitution(_pos, _ruleSet, formatter, description, status)
  1.1030 + , byDigits(FALSE)
  1.1031 + , useSpaces(TRUE)
  1.1032 +
  1.1033 +{
  1.1034 +    // akk, ruleSet can change in superclass constructor
  1.1035 +    if (0 == description.compare(gGreaterGreaterThan, 2) ||
  1.1036 +        0 == description.compare(gGreaterGreaterGreaterThan, 3) ||
  1.1037 +        _ruleSet == getRuleSet()) {
  1.1038 +        byDigits = TRUE;
  1.1039 +        if (0 == description.compare(gGreaterGreaterGreaterThan, 3)) {
  1.1040 +            useSpaces = FALSE;
  1.1041 +        }
  1.1042 +    } else {
  1.1043 +        // cast away const
  1.1044 +        ((NFRuleSet*)getRuleSet())->makeIntoFractionRuleSet();
  1.1045 +    }
  1.1046 +}
  1.1047 +
  1.1048 +//-----------------------------------------------------------------------
  1.1049 +// formatting
  1.1050 +//-----------------------------------------------------------------------
  1.1051 +
  1.1052 +/**
  1.1053 + * If in "by digits" mode, fills in the substitution one decimal digit
  1.1054 + * at a time using the rule set containing this substitution.
  1.1055 + * Otherwise, uses the superclass function.
  1.1056 + * @param number The number being formatted
  1.1057 + * @param toInsertInto The string to insert the result of formatting
  1.1058 + * the substitution into
  1.1059 + * @param pos The position of the owning rule's rule text in
  1.1060 + * toInsertInto
  1.1061 + */
  1.1062 +void
  1.1063 +FractionalPartSubstitution::doSubstitution(double number, UnicodeString& toInsertInto, int32_t _pos) const
  1.1064 +{
  1.1065 +  // if we're not in "byDigits" mode, just use the inherited
  1.1066 +  // doSubstitution() routine
  1.1067 +  if (!byDigits) {
  1.1068 +    NFSubstitution::doSubstitution(number, toInsertInto, _pos);
  1.1069 +
  1.1070 +    // if we're in "byDigits" mode, transform the value into an integer
  1.1071 +    // by moving the decimal point eight places to the right and
  1.1072 +    // pulling digits off the right one at a time, formatting each digit
  1.1073 +    // as an integer using this substitution's owning rule set
  1.1074 +    // (this is slower, but more accurate, than doing it from the
  1.1075 +    // other end)
  1.1076 +  } else {
  1.1077 +    //          int32_t numberToFormat = (int32_t)uprv_round(transformNumber(number) * uprv_pow(10, kMaxDecimalDigits));
  1.1078 +    //          // this flag keeps us from formatting trailing zeros.  It starts
  1.1079 +    //          // out false because we're pulling from the right, and switches
  1.1080 +    //          // to true the first time we encounter a non-zero digit
  1.1081 +    //          UBool doZeros = FALSE;
  1.1082 +    //          for (int32_t i = 0; i < kMaxDecimalDigits; i++) {
  1.1083 +    //              int64_t digit = numberToFormat % 10;
  1.1084 +    //              if (digit != 0 || doZeros) {
  1.1085 +    //                  if (doZeros && useSpaces) {
  1.1086 +    //                      toInsertInto.insert(_pos + getPos(), gSpace);
  1.1087 +    //                  }
  1.1088 +    //                  doZeros = TRUE;
  1.1089 +    //                  getRuleSet()->format(digit, toInsertInto, _pos + getPos());
  1.1090 +    //              }
  1.1091 +    //              numberToFormat /= 10;
  1.1092 +    //          }
  1.1093 +
  1.1094 +    DigitList dl;
  1.1095 +    dl.set(number);
  1.1096 +    dl.roundFixedPoint(20);     // round to 20 fraction digits.
  1.1097 +    dl.reduce();                // Removes any trailing zeros.
  1.1098 +    
  1.1099 +    UBool pad = FALSE;
  1.1100 +    for (int32_t didx = dl.getCount()-1; didx>=dl.getDecimalAt(); didx--) {
  1.1101 +      // Loop iterates over fraction digits, starting with the LSD.
  1.1102 +      //   include both real digits from the number, and zeros
  1.1103 +      //   to the left of the MSD but to the right of the decimal point.
  1.1104 +      if (pad && useSpaces) {
  1.1105 +        toInsertInto.insert(_pos + getPos(), gSpace);
  1.1106 +      } else {
  1.1107 +        pad = TRUE;
  1.1108 +      }
  1.1109 +      int64_t digit = didx>=0 ? dl.getDigit(didx) - '0' : 0;
  1.1110 +      getRuleSet()->format(digit, toInsertInto, _pos + getPos());
  1.1111 +    }
  1.1112 +
  1.1113 +    if (!pad) {
  1.1114 +      // hack around lack of precision in digitlist. if we would end up with
  1.1115 +      // "foo point" make sure we add a " zero" to the end.
  1.1116 +      getRuleSet()->format((int64_t)0, toInsertInto, _pos + getPos());
  1.1117 +    }
  1.1118 +  }
  1.1119 +}
  1.1120 +
  1.1121 +//-----------------------------------------------------------------------
  1.1122 +// parsing
  1.1123 +//-----------------------------------------------------------------------
  1.1124 +
  1.1125 +/**
  1.1126 + * If in "by digits" mode, parses the string as if it were a string
  1.1127 + * of individual digits; otherwise, uses the superclass function.
  1.1128 + * @param text The string to parse
  1.1129 + * @param parsePosition Ignored on entry, but updated on exit to point
  1.1130 + * to the first unmatched character
  1.1131 + * @param baseValue The partial parse result prior to entering this
  1.1132 + * function
  1.1133 + * @param upperBound Only consider rules with base values lower than
  1.1134 + * this when filling in the substitution
  1.1135 + * @param lenientParse If true, try matching the text as numerals if
  1.1136 + * matching as words doesn't work
  1.1137 + * @return If the match was successful, the current partial parse
  1.1138 + * result; otherwise new Long(0).  The result is either a Long or
  1.1139 + * a Double.
  1.1140 + */
  1.1141 +
  1.1142 +UBool
  1.1143 +FractionalPartSubstitution::doParse(const UnicodeString& text,
  1.1144 +                ParsePosition& parsePosition,
  1.1145 +                double baseValue,
  1.1146 +                double /*upperBound*/,
  1.1147 +                UBool lenientParse,
  1.1148 +                Formattable& resVal) const
  1.1149 +{
  1.1150 +    // if we're not in byDigits mode, we can just use the inherited
  1.1151 +    // doParse()
  1.1152 +    if (!byDigits) {
  1.1153 +        return NFSubstitution::doParse(text, parsePosition, baseValue, 0, lenientParse, resVal);
  1.1154 +
  1.1155 +        // if we ARE in byDigits mode, parse the text one digit at a time
  1.1156 +        // using this substitution's owning rule set (we do this by setting
  1.1157 +        // upperBound to 10 when calling doParse() ) until we reach
  1.1158 +        // nonmatching text
  1.1159 +    } else {
  1.1160 +        UnicodeString workText(text);
  1.1161 +        ParsePosition workPos(1);
  1.1162 +        double result = 0;
  1.1163 +        int32_t digit;
  1.1164 +//          double p10 = 0.1;
  1.1165 +
  1.1166 +        DigitList dl;
  1.1167 +        NumberFormat* fmt = NULL;
  1.1168 +        while (workText.length() > 0 && workPos.getIndex() != 0) {
  1.1169 +            workPos.setIndex(0);
  1.1170 +            Formattable temp;
  1.1171 +            getRuleSet()->parse(workText, workPos, 10, temp);
  1.1172 +            UErrorCode status = U_ZERO_ERROR;
  1.1173 +            digit = temp.getLong(status);
  1.1174 +//            digit = temp.getType() == Formattable::kLong ?
  1.1175 +//               temp.getLong() :
  1.1176 +//            (int32_t)temp.getDouble();
  1.1177 +
  1.1178 +            if (lenientParse && workPos.getIndex() == 0) {
  1.1179 +                if (!fmt) {
  1.1180 +                    status = U_ZERO_ERROR;
  1.1181 +                    fmt = NumberFormat::createInstance(status);
  1.1182 +                    if (U_FAILURE(status)) {
  1.1183 +                        delete fmt;
  1.1184 +                        fmt = NULL;
  1.1185 +                    }
  1.1186 +                }
  1.1187 +                if (fmt) {
  1.1188 +                    fmt->parse(workText, temp, workPos);
  1.1189 +                    digit = temp.getLong(status);
  1.1190 +                }
  1.1191 +            }
  1.1192 +
  1.1193 +            if (workPos.getIndex() != 0) {
  1.1194 +                dl.append((char)('0' + digit));
  1.1195 +//                  result += digit * p10;
  1.1196 +//                  p10 /= 10;
  1.1197 +                parsePosition.setIndex(parsePosition.getIndex() + workPos.getIndex());
  1.1198 +                workText.removeBetween(0, workPos.getIndex());
  1.1199 +                while (workText.length() > 0 && workText.charAt(0) == gSpace) {
  1.1200 +                    workText.removeBetween(0, 1);
  1.1201 +                    parsePosition.setIndex(parsePosition.getIndex() + 1);
  1.1202 +                }
  1.1203 +            }
  1.1204 +        }
  1.1205 +        delete fmt;
  1.1206 +
  1.1207 +        result = dl.getCount() == 0 ? 0 : dl.getDouble();
  1.1208 +        result = composeRuleValue(result, baseValue);
  1.1209 +        resVal.setDouble(result);
  1.1210 +        return TRUE;
  1.1211 +    }
  1.1212 +}
  1.1213 +
  1.1214 +UBool
  1.1215 +FractionalPartSubstitution::operator==(const NFSubstitution& rhs) const
  1.1216 +{
  1.1217 +  return NFSubstitution::operator==(rhs) &&
  1.1218 +  ((const FractionalPartSubstitution*)&rhs)->byDigits == byDigits;
  1.1219 +}
  1.1220 +
  1.1221 +UOBJECT_DEFINE_RTTI_IMPLEMENTATION(FractionalPartSubstitution)
  1.1222 +
  1.1223 +
  1.1224 +//===================================================================
  1.1225 +// AbsoluteValueSubstitution
  1.1226 +//===================================================================
  1.1227 +
  1.1228 +UOBJECT_DEFINE_RTTI_IMPLEMENTATION(AbsoluteValueSubstitution)
  1.1229 +
  1.1230 +//===================================================================
  1.1231 +// NumeratorSubstitution
  1.1232 +//===================================================================
  1.1233 +
  1.1234 +void
  1.1235 +NumeratorSubstitution::doSubstitution(double number, UnicodeString& toInsertInto, int32_t apos) const {
  1.1236 +    // perform a transformation on the number being formatted that
  1.1237 +    // is dependent on the type of substitution this is
  1.1238 +
  1.1239 +    double numberToFormat = transformNumber(number);
  1.1240 +    int64_t longNF = util64_fromDouble(numberToFormat);
  1.1241 +
  1.1242 +    const NFRuleSet* aruleSet = getRuleSet();
  1.1243 +    if (withZeros && aruleSet != NULL) {
  1.1244 +        // if there are leading zeros in the decimal expansion then emit them
  1.1245 +        int64_t nf =longNF;
  1.1246 +        int32_t len = toInsertInto.length();
  1.1247 +        while ((nf *= 10) < denominator) {
  1.1248 +            toInsertInto.insert(apos + getPos(), gSpace);
  1.1249 +            aruleSet->format((int64_t)0, toInsertInto, apos + getPos());
  1.1250 +        }
  1.1251 +        apos += toInsertInto.length() - len;
  1.1252 +    }
  1.1253 +
  1.1254 +    // if the result is an integer, from here on out we work in integer
  1.1255 +    // space (saving time and memory and preserving accuracy)
  1.1256 +    if (numberToFormat == longNF && aruleSet != NULL) {
  1.1257 +        aruleSet->format(longNF, toInsertInto, apos + getPos());
  1.1258 +
  1.1259 +        // if the result isn't an integer, then call either our rule set's
  1.1260 +        // format() method or our DecimalFormat's format() method to
  1.1261 +        // format the result
  1.1262 +    } else {
  1.1263 +        if (aruleSet != NULL) {
  1.1264 +            aruleSet->format(numberToFormat, toInsertInto, apos + getPos());
  1.1265 +        } else {
  1.1266 +            UErrorCode status = U_ZERO_ERROR;
  1.1267 +            UnicodeString temp;
  1.1268 +            getNumberFormat()->format(numberToFormat, temp, status);
  1.1269 +            toInsertInto.insert(apos + getPos(), temp);
  1.1270 +        }
  1.1271 +    }
  1.1272 +}
  1.1273 +
  1.1274 +UBool 
  1.1275 +NumeratorSubstitution::doParse(const UnicodeString& text, 
  1.1276 +                               ParsePosition& parsePosition,
  1.1277 +                               double baseValue,
  1.1278 +                               double upperBound,
  1.1279 +                               UBool /*lenientParse*/,
  1.1280 +                               Formattable& result) const
  1.1281 +{
  1.1282 +    // we don't have to do anything special to do the parsing here,
  1.1283 +    // but we have to turn lenient parsing off-- if we leave it on,
  1.1284 +    // it SERIOUSLY messes up the algorithm
  1.1285 +
  1.1286 +    // if withZeros is true, we need to count the zeros
  1.1287 +    // and use that to adjust the parse result
  1.1288 +    UErrorCode status = U_ZERO_ERROR;
  1.1289 +    int32_t zeroCount = 0;
  1.1290 +    UnicodeString workText(text);
  1.1291 +
  1.1292 +    if (withZeros) {
  1.1293 +        ParsePosition workPos(1);
  1.1294 +        Formattable temp;
  1.1295 +
  1.1296 +        while (workText.length() > 0 && workPos.getIndex() != 0) {
  1.1297 +            workPos.setIndex(0);
  1.1298 +            getRuleSet()->parse(workText, workPos, 1, temp); // parse zero or nothing at all
  1.1299 +            if (workPos.getIndex() == 0) {
  1.1300 +                // we failed, either there were no more zeros, or the number was formatted with digits
  1.1301 +                // either way, we're done
  1.1302 +                break;
  1.1303 +            }
  1.1304 +
  1.1305 +            ++zeroCount;
  1.1306 +            parsePosition.setIndex(parsePosition.getIndex() + workPos.getIndex());
  1.1307 +            workText.remove(0, workPos.getIndex());
  1.1308 +            while (workText.length() > 0 && workText.charAt(0) == gSpace) {
  1.1309 +                workText.remove(0, 1);
  1.1310 +                parsePosition.setIndex(parsePosition.getIndex() + 1);
  1.1311 +            }
  1.1312 +        }
  1.1313 +
  1.1314 +        workText = text;
  1.1315 +        workText.remove(0, (int32_t)parsePosition.getIndex());
  1.1316 +        parsePosition.setIndex(0);
  1.1317 +    }
  1.1318 +
  1.1319 +    // we've parsed off the zeros, now let's parse the rest from our current position
  1.1320 +    NFSubstitution::doParse(workText, parsePosition, withZeros ? 1 : baseValue, upperBound, FALSE, result);
  1.1321 +
  1.1322 +    if (withZeros) {
  1.1323 +        // any base value will do in this case.  is there a way to
  1.1324 +        // force this to not bother trying all the base values?
  1.1325 +
  1.1326 +        // compute the 'effective' base and prescale the value down
  1.1327 +        int64_t n = result.getLong(status); // force conversion!
  1.1328 +        int64_t d = 1;
  1.1329 +        int32_t pow = 0;
  1.1330 +        while (d <= n) {
  1.1331 +            d *= 10;
  1.1332 +            ++pow;
  1.1333 +        }
  1.1334 +        // now add the zeros
  1.1335 +        while (zeroCount > 0) {
  1.1336 +            d *= 10;
  1.1337 +            --zeroCount;
  1.1338 +        }
  1.1339 +        // d is now our true denominator
  1.1340 +        result.setDouble((double)n/(double)d);
  1.1341 +    }
  1.1342 +
  1.1343 +    return TRUE;
  1.1344 +}
  1.1345 +
  1.1346 +UBool
  1.1347 +NumeratorSubstitution::operator==(const NFSubstitution& rhs) const
  1.1348 +{
  1.1349 +    return NFSubstitution::operator==(rhs) &&
  1.1350 +        denominator == ((const NumeratorSubstitution*)&rhs)->denominator;
  1.1351 +}
  1.1352 +
  1.1353 +UOBJECT_DEFINE_RTTI_IMPLEMENTATION(NumeratorSubstitution)
  1.1354 +
  1.1355 +const UChar NumeratorSubstitution::LTLT[] = { 0x003c, 0x003c };
  1.1356 +        
  1.1357 +//===================================================================
  1.1358 +// NullSubstitution
  1.1359 +//===================================================================
  1.1360 +
  1.1361 +UOBJECT_DEFINE_RTTI_IMPLEMENTATION(NullSubstitution)
  1.1362 +
  1.1363 +U_NAMESPACE_END
  1.1364 +
  1.1365 +/* U_HAVE_RBNF */
  1.1366 +#endif
  1.1367 +

mercurial