You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
tdeutils/kcalc/knumber/knumber.cpp

694 lines
15 KiB

// -*- c-basic-offset: 2 -*-
/* This file is part of the KDE libraries
Copyright (c) 2005 Klaus Niederkrueger <kniederk@math.uni-koeln.de>
This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Library General Public
License as published by the Free Software Foundation; either
version 2 of the License, or (at your option) any later version.
This library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Library General Public License for more details.
You should have received a copy of the GNU Library General Public License
along with this library; see the file COPYING.LIB. If not, write to
the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
Boston, MA 02110-1301, USA.
*/
#include <math.h>
#include <config.h>
#include <tqregexp.h>
#include <tqstring.h>
#include "knumber.h"
KNumber const KNumber::Zero(0);
KNumber const KNumber::One(1);
KNumber const KNumber::MinusOne(-1);
KNumber const KNumber::Pi("3.141592653589793238462643383279502884197169"
"39937510582097494459230781640628620899862803"
"4825342117068");
KNumber const KNumber::Euler("2.718281828459045235360287471352662497757"
"24709369995957496696762772407663035354759"
"4571382178525166427");
KNumber const KNumber::NotDefined("nan");
bool KNumber::_float_output = false;
bool KNumber::_fraction_input = false;
bool KNumber::_splitoffinteger_output = false;
KNumber::KNumber(signed int num)
{
_num = new _knuminteger(num);
}
KNumber::KNumber(unsigned int num)
{
_num = new _knuminteger(num);
}
KNumber::KNumber(signed long int num)
{
_num = new _knuminteger(num);
}
KNumber::KNumber(unsigned long int num)
{
_num = new _knuminteger(num);
}
KNumber::KNumber(unsigned long long int num)
{
_num = new _knuminteger(num);
}
KNumber::KNumber(double num)
{
if ( isinf(num) ) _num = new _knumerror( _knumber::Infinity );
else if ( isnan(num) ) _num = new _knumerror( _knumber::UndefinedNumber );
else _num = new _knumfloat(num);
}
KNumber::KNumber(KNumber const & num)
{
switch(num.type()) {
case SpecialType:
_num = new _knumerror(*(num._num));
return;
case IntegerType:
_num = new _knuminteger(*(num._num));
return;
case FractionType:
_num = new _knumfraction(*(num._num));
return;
case FloatType:
_num = new _knumfloat(*(num._num));
return;
};
}
KNumber::KNumber(TQString const & num)
{
if (TQRegExp("^(inf|-inf|nan)$").exactMatch(num))
_num = new _knumerror(num);
else if (TQRegExp("^[+-]?\\d+$").exactMatch(num))
_num = new _knuminteger(num);
else if (TQRegExp("^[+-]?\\d+/\\d+$").exactMatch(num)) {
_num = new _knumfraction(num);
simplifyRational();
}
else if (TQRegExp("^[+-]?\\d+(\\.\\d*)?(e[+-]?\\d+)?$").exactMatch(num))
if (_fraction_input == true) {
_num = new _knumfraction(num);
simplifyRational();
} else
_num = new _knumfloat(num);
else
_num = new _knumerror("nan");
}
KNumber::NumType KNumber::type(void) const
{
if(dynamic_cast<_knumerror *>(_num))
return SpecialType;
if(dynamic_cast<_knuminteger *>(_num))
return IntegerType;
if(dynamic_cast<_knumfraction *>(_num))
return FractionType;
if(dynamic_cast<_knumfloat *>(_num))
return FloatType;
return SpecialType;
}
// This method converts a fraction to an integer, whenever possible,
// i.e. 5/1 --> 5
// This method should be called, whenever such a inproper fraction can occur,
// e.g. when adding 4/3 + 2/3....
void KNumber::simplifyRational(void)
{
if (type() != FractionType)
return;
_knumfraction *tmp_num = dynamic_cast<_knumfraction *>(_num);
if (tmp_num->isInteger()) {
_knumber *tmp_num2 = tmp_num->intPart();
delete tmp_num;
_num = tmp_num2;
}
}
KNumber const & KNumber::operator=(KNumber const & num)
{
if (this == & num)
return *this;
delete _num;
switch(num.type()) {
case SpecialType:
_num = new _knumerror();
break;
case IntegerType:
_num = new _knuminteger();
break;
case FractionType:
_num = new _knumfraction();
break;
case FloatType:
_num = new _knumfloat();
break;
};
_num->copy(*(num._num));
return *this;
}
KNumber & KNumber::operator +=(KNumber const &arg)
{
KNumber tmp_num = *this + arg;
delete _num;
switch(tmp_num.type()) {
case SpecialType:
_num = new _knumerror();
break;
case IntegerType:
_num = new _knuminteger();
break;
case FractionType:
_num = new _knumfraction();
break;
case FloatType:
_num = new _knumfloat();
break;
};
_num->copy(*(tmp_num._num));
return *this;
}
KNumber & KNumber::operator -=(KNumber const &arg)
{
KNumber tmp_num = *this - arg;
delete _num;
switch(tmp_num.type()) {
case SpecialType:
_num = new _knumerror();
break;
case IntegerType:
_num = new _knuminteger();
break;
case FractionType:
_num = new _knumfraction();
break;
case FloatType:
_num = new _knumfloat();
break;
};
_num->copy(*(tmp_num._num));
return *this;
}
// increase the digit at 'position' by one
static void _inc_by_one(TQString &str, int position)
{
for (int i = position; i >= 0; i--)
{
char last_char = str[i].latin1();
switch(last_char)
{
case '0':
str[i] = '1';
break;
case '1':
str[i] = '2';
break;
case '2':
str[i] = '3';
break;
case '3':
str[i] = '4';
break;
case '4':
str[i] = '5';
break;
case '5':
str[i] = '6';
break;
case '6':
str[i] = '7';
break;
case '7':
str[i] = '8';
break;
case '8':
str[i] = '9';
break;
case '9':
str[i] = '0';
if (i == 0) str.prepend('1');
continue;
case '.':
continue;
}
break;
}
}
// Cut off if more digits in fractional part than 'precision'
static void _round(TQString &str, int precision)
{
int decimalSymbolPos = str.find('.');
if (decimalSymbolPos == -1)
if (precision == 0) return;
else if (precision > 0) // add dot if missing (and needed)
{
str.append('.');
decimalSymbolPos = str.length() - 1;
}
// fill up with more than enough zeroes (in case fractional part too short)
str.append(TQString().fill('0', precision));
// Now decide whether to round up or down
char last_char = str[decimalSymbolPos + precision + 1].latin1();
switch (last_char)
{
case '0':
case '1':
case '2':
case '3':
case '4':
// nothing to do, rounding down
break;
case '5':
case '6':
case '7':
case '8':
case '9':
// rounding up
_inc_by_one(str, decimalSymbolPos + precision);
break;
default:
break;
}
decimalSymbolPos = str.find('.');
str.truncate(decimalSymbolPos + precision + 1);
// if precision == 0 delete also '.'
if (precision == 0) str = str.section('.', 0, 0);
}
static TQString roundNumber(const TQString &numStr, int precision)
{
TQString tmpString = numStr;
if (precision < 0 ||
! TQRegExp("^[+-]?\\d+(\\.\\d+)*(e[+-]?\\d+)?$").exactMatch(tmpString))
return numStr;
// Skip the sign (for now)
bool neg = (tmpString[0] == '-');
if (neg || tmpString[0] == '+') tmpString.remove(0, 1);
// Split off exponential part (including 'e'-symbol)
TQString mantString = tmpString.section('e', 0, 0,
TQString::SectionCaseInsensitiveSeps);
TQString expString = tmpString.section('e', 1, 1,
TQString::SectionCaseInsensitiveSeps |
TQString::SectionIncludeLeadingSep);
if (expString.length() == 1) expString = TQString();
_round(mantString, precision);
if(neg) mantString.prepend('-');
return mantString + expString;
}
TQString const KNumber::toTQString(int width, int prec) const
{
TQString tmp_str;
if (*this == Zero) // important to avoid infinite loops below
return "0";
switch (type()) {
case IntegerType:
if (width > 0) { //result needs to be cut-off
bool tmp_bool = _fraction_input; // stupid work-around
_fraction_input = false;
tmp_str = (KNumber("1.0")*(*this)).toTQString(width, -1);
_fraction_input = tmp_bool;
} else
tmp_str = TQString(_num->ascii());
break;
case FractionType:
if (_float_output) {
bool tmp_bool = _fraction_input; // stupid work-around
_fraction_input = false;
tmp_str = (KNumber("1.0")*(*this)).toTQString(width, -1);
_fraction_input = tmp_bool;
} else { // _float_output == false
if(_splitoffinteger_output) {
// split off integer part
KNumber int_part = this->integerPart();
if (int_part == Zero)
tmp_str = TQString(_num->ascii());
else if (int_part < Zero)
tmp_str = int_part.toTQString() + " " + (int_part - *this)._num->ascii();
else
tmp_str = int_part.toTQString() + " " + (*this - int_part)._num->ascii();
} else
tmp_str = TQString(_num->ascii());
if (width > 0 && tmp_str.length() > width) {
//result needs to be cut-off
bool tmp_bool = _fraction_input; // stupid work-around
_fraction_input = false;
tmp_str = (KNumber("1.0")*(*this)).toTQString(width, -1);
_fraction_input = tmp_bool;
}
}
break;
case FloatType:
if (width > 0)
tmp_str = TQString(_num->ascii(width));
else
// rough estimate for maximal decimal precision (10^3 = 2^10)
tmp_str = TQString(_num->ascii(3*mpf_get_default_prec()/10));
break;
default:
return TQString(_num->ascii());
}
if (prec >= 0)
return roundNumber(tmp_str, prec);
else
return tmp_str;
}
void KNumber::setDefaultFloatOutput(bool flag)
{
_float_output = flag;
}
void KNumber::setDefaultFractionalInput(bool flag)
{
_fraction_input = flag;
}
void KNumber::setSplitoffIntegerForFractionOutput(bool flag)
{
_splitoffinteger_output = flag;
}
void KNumber::setDefaultFloatPrecision(unsigned int prec)
{
// Need to transform decimal digits into binary digits
unsigned long int bin_prec = static_cast<unsigned long int>
(double(prec) * M_LN10 / M_LN2 + 1);
mpf_set_default_prec(bin_prec);
}
KNumber const KNumber::abs(void) const
{
KNumber tmp_num;
delete tmp_num._num;
tmp_num._num = _num->abs();
return tmp_num;
}
KNumber const KNumber::cbrt(void) const
{
KNumber tmp_num;
delete tmp_num._num;
tmp_num._num = _num->cbrt();
return tmp_num;
}
KNumber const KNumber::sqrt(void) const
{
KNumber tmp_num;
delete tmp_num._num;
tmp_num._num = _num->sqrt();
return tmp_num;
}
KNumber const KNumber::integerPart(void) const
{
KNumber tmp_num;
delete tmp_num._num;
tmp_num._num = _num->intPart();
return tmp_num;
}
KNumber const KNumber::power(KNumber const &exp) const
{
if (*this == Zero) {
if(exp == Zero)
return KNumber("nan"); // 0^0 not defined
else if (exp < Zero)
return KNumber("inf");
else
return KNumber(0);
}
if (exp == Zero) {
if (*this != Zero)
return One;
else
return KNumber("nan");
}
else if (exp < Zero) {
KNumber tmp_num;
KNumber tmp_num2 = -exp;
delete tmp_num._num;
tmp_num._num = _num->power(*(tmp_num2._num));
return One/tmp_num;
}
else {
KNumber tmp_num;
delete tmp_num._num;
tmp_num._num = _num->power(*(exp._num));
return tmp_num;
}
}
KNumber const KNumber::operator-(void) const
{
KNumber tmp_num;
delete tmp_num._num;
tmp_num._num = _num->change_sign();
return tmp_num;
}
KNumber const KNumber::operator+(KNumber const & arg2) const
{
KNumber tmp_num;
delete tmp_num._num;
tmp_num._num = _num->add(*arg2._num);
tmp_num.simplifyRational();
return tmp_num;
}
KNumber const KNumber::operator-(KNumber const & arg2) const
{
return *this + (-arg2);
}
KNumber const KNumber::operator*(KNumber const & arg2) const
{
KNumber tmp_num;
delete tmp_num._num;
tmp_num._num = _num->multiply(*arg2._num);
tmp_num.simplifyRational();
return tmp_num;
}
KNumber const KNumber::operator/(KNumber const & arg2) const
{
KNumber tmp_num;
delete tmp_num._num;
tmp_num._num = _num->divide(*arg2._num);
tmp_num.simplifyRational();
return tmp_num;
}
KNumber const KNumber::operator%(KNumber const & arg2) const
{
if (type() != IntegerType || arg2.type() != IntegerType)
return Zero;
KNumber tmp_num;
delete tmp_num._num;
_knuminteger const *tmp_arg1 = dynamic_cast<_knuminteger const *>(_num);
_knuminteger const *tmp_arg2 = dynamic_cast<_knuminteger const *>(arg2._num);
tmp_num._num = tmp_arg1->mod(*tmp_arg2);
return tmp_num;
}
KNumber const KNumber::operator&(KNumber const & arg2) const
{
if (type() != IntegerType || arg2.type() != IntegerType)
return Zero;
KNumber tmp_num;
delete tmp_num._num;
_knuminteger const *tmp_arg1 = dynamic_cast<_knuminteger const *>(_num);
_knuminteger const *tmp_arg2 = dynamic_cast<_knuminteger const *>(arg2._num);
tmp_num._num = tmp_arg1->intAnd(*tmp_arg2);
return tmp_num;
}
KNumber const KNumber::operator|(KNumber const & arg2) const
{
if (type() != IntegerType || arg2.type() != IntegerType)
return Zero;
KNumber tmp_num;
delete tmp_num._num;
_knuminteger const *tmp_arg1 = dynamic_cast<_knuminteger const *>(_num);
_knuminteger const *tmp_arg2 = dynamic_cast<_knuminteger const *>(arg2._num);
tmp_num._num = tmp_arg1->intOr(*tmp_arg2);
return tmp_num;
}
KNumber const KNumber::operator<<(KNumber const & arg2) const
{
if (type() != IntegerType || arg2.type() != IntegerType)
return KNumber("nan");
_knuminteger const *tmp_arg1 = dynamic_cast<_knuminteger const *>(_num);
_knuminteger const *tmp_arg2 = dynamic_cast<_knuminteger const *>(arg2._num);
KNumber tmp_num;
delete tmp_num._num;
tmp_num._num = tmp_arg1->shift(*tmp_arg2);
return tmp_num;
}
KNumber const KNumber::operator>>(KNumber const & arg2) const
{
if (type() != IntegerType || arg2.type() != IntegerType)
return KNumber("nan");
KNumber tmp_num = -arg2;
_knuminteger const *tmp_arg1 = dynamic_cast<_knuminteger const *>(_num);
_knuminteger const *tmp_arg2 = dynamic_cast<_knuminteger const *>(tmp_num._num);
KNumber tmp_num2;
delete tmp_num2._num;
tmp_num2._num = tmp_arg1->shift(*tmp_arg2);
return tmp_num2;
}
KNumber::operator bool(void) const
{
if (*this == Zero)
return false;
return true;
}
KNumber::operator signed long int(void) const
{
return static_cast<signed long int>(*_num);
}
KNumber::operator unsigned long int(void) const
{
return static_cast<unsigned long int>(*_num);
}
KNumber::operator unsigned long long int(void) const
{
#if SIZEOF_UNSIGNED_LONG == 8
return static_cast<unsigned long int>(*this);
#elif SIZEOF_UNSIGNED_LONG == 4
KNumber tmp_num1 = this->abs().integerPart();
unsigned long long int tmp_num2 = static_cast<unsigned long int>(tmp_num1) +
(static_cast<unsigned long long int>(
static_cast<unsigned long int>(tmp_num1 >> KNumber("32"))) << 32) ;
#warning the cast operator from KNumber to unsigned long long int is probably buggy, when a sign is involved
if (*this > KNumber(0))
return tmp_num2;
else
return static_cast<unsigned long long int> (- static_cast<signed long long int>(tmp_num2));
#else
#error "SIZEOF_UNSIGNED_LONG is a unhandled case"
#endif
}
KNumber::operator double(void) const
{
return static_cast<double>(*_num);
}
int const KNumber::compare(KNumber const & arg2) const
{
return _num->compare(*arg2._num);
}