Package org.xlsx4j.org.apache.poi.ss.util

Class NormalisedDecimal

java.lang.Object

org.xlsx4j.org.apache.poi.ss.util.NormalisedDecimal

final class NormalisedDecimal
extends Object

Represents a transformation of a 64 bit IEEE double quantity having a decimal exponent and a fixed point (15 decimal digit) significand. Some quirks of Excel's calculation behaviour are simpler to reproduce with numeric quantities in this format. This class is currently used to help:

1. Comparison operations
2. Conversions to text

This class does not handle negative numbers or zero.

The value of a NormalisedDecimal is given by
significand × 10^{decimalExponent}
where:
significand = wholePart + fractionalPart / 2²⁴

Field Summary

Fields

Modifier and Type	Field	Description
private final int	_fractionalPart	The fractional part of the significand.
private final int	_relativeDecimalExponent	The decimal exponent increased by one less than the digit count of _wholePart
private final long	_wholePart	The whole part of the significand (typically 15 digits).
private static final BigDecimal	BD_2_POW_24
private static final int	C_2_POW_19	219
private static final int	EXPONENT_OFFSET	Number of powers of ten contained in the significand
private static final int	FRAC_HALF	the value of _fractionalPart that represents 0.5
private static final int	LOG_BASE_10_OF_2_TIMES_2_POW_20	log10(2)×220
private static final long	MAX_REP_WHOLE_PART	1015

Constructor Summary

Constructors

Constructor	Description
NormalisedDecimal(long wholePart, int fracPart, int decimalExponent)

Method Summary

Modifier and Type	Method	Description
int	compareNormalised(NormalisedDecimal other)	assumes both this and other are normalised
(package private) BigInteger	composeFrac()
static NormalisedDecimal	create(BigInteger frac, int binaryExponent)
int	getDecimalExponent()
private String	getFractionalDigits()
BigDecimal	getFractionalPart()
String	getSignificantDecimalDigits()
String	getSignificantDecimalDigitsLastDigitRounded()	Rounds the first whole digit position (considers only units digit, not frational part).
ExpandedDouble	normaliseBaseTwo()	Convert to an equivalent ExpandedDouble representation (binary frac and exponent).
NormalisedDecimal	roundUnits()	Rounds at the digit with value 10decimalExponent
String	toString()

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Field Details

EXPONENT_OFFSET

private static final int EXPONENT_OFFSET

Number of powers of ten contained in the significand

See Also:

• Constant Field Values

BD_2_POW_24

private static final BigDecimal BD_2_POW_24

LOG_BASE_10_OF_2_TIMES_2_POW_20

private static final int LOG_BASE_10_OF_2_TIMES_2_POW_20

log₁₀(2)×2²⁰

See Also:

• Constant Field Values

C_2_POW_19

private static final int C_2_POW_19

2¹⁹

See Also:

• Constant Field Values

FRAC_HALF

private static final int FRAC_HALF

the value of _fractionalPart that represents 0.5

See Also:

• Constant Field Values

MAX_REP_WHOLE_PART

private static final long MAX_REP_WHOLE_PART

10¹⁵

See Also:

• Constant Field Values

_relativeDecimalExponent

private final int _relativeDecimalExponent

The decimal exponent increased by one less than the digit count of _wholePart

_wholePart

private final long _wholePart

The whole part of the significand (typically 15 digits). 47-50 bits long (MSB may be anywhere from bit 46 to 49) LSB is units bit.

_fractionalPart

private final int _fractionalPart

The fractional part of the significand. 24 bits (only top 14-17 bits significant): a value between 0x000000 and 0xFFFF80

Constructor Details

NormalisedDecimal

NormalisedDecimal(long wholePart,
 int fracPart,
 int decimalExponent)

Method Details

create

public static NormalisedDecimal create(BigInteger frac,
 int binaryExponent)

roundUnits

public NormalisedDecimal roundUnits()

Rounds at the digit with value 10^{decimalExponent}

normaliseBaseTwo

public ExpandedDouble normaliseBaseTwo()

Convert to an equivalent ExpandedDouble representation (binary frac and exponent). The resulting transformed object is easily converted to a 64 bit IEEE double:

• bits 2-53 of the #getSignificand() become the 52 bit 'fraction'.
• #getBinaryExponent() is biased by 1023 to give the 'exponent'.

The sign bit must be obtained from somewhere else.

Returns:

a new NormalisedDecimal normalised to base 2 representation.

composeFrac

BigInteger composeFrac()

Returns:

the significand as a fixed point number (with 24 fraction bits and 47-50 whole bits)

getSignificantDecimalDigits

public String getSignificantDecimalDigits()

getSignificantDecimalDigitsLastDigitRounded

public String getSignificantDecimalDigitsLastDigitRounded()

Rounds the first whole digit position (considers only units digit, not frational part). Caller should check total digit count of result to see whether the rounding operation caused a carry out of the most significant digit

getDecimalExponent

public int getDecimalExponent()

Returns:

the number of powers of 10 which have been extracted from the significand and binary exponent.

compareNormalised

public int compareNormalised(NormalisedDecimal other)

assumes both this and other are normalised

getFractionalPart

public BigDecimal getFractionalPart()

getFractionalDigits

private String getFractionalDigits()

toString

public String toString()

Overrides:

toString in class Object