net.sf.saxon.value

Class FloatingPointConverter

java.lang.Object

net.sf.saxon.value.FloatingPointConverter

public class FloatingPointConverter
extends java.lang.Object

This is a utility class that handles formatting of numbers as strings.

The algorithm for converting a floating point number to a string is taken from Guy L. Steele and Jon L. White, How to Print Floating-Point Numbers Accurately, ACM SIGPLAN 1990. It is algorithm (FPP)² from that paper. There are three separate implementations of the algorithm:

• One using long arithmetic and generating non-exponential output representations
• One using BigInteger arithmetic and generating non-exponential output representation
• One using BigInteger arithmetic and generating exponential output representations

The choice of method depends on the value of the number being formatted.

The module contains some residual code (mainly the routine for formatting integers) from the class AppenderHelper by Jack Shirazi in the O'Reilly book Java Performance Tuning. The floating point routines in that module were found to be unsuitable, since they used floating point arithmetic which introduces rounding errors.

There are several reasons for doing this conversion within Saxon, rather than leaving it all to Java. Firstly, there are differences in the required output format, notably the absence of ".0" when formatting whole numbers, and the different rules for the range of numbers where exponential notation is used. Secondly, there are bugs in some Java implementations, for example JDK outputs 0.001 as 0.0010, and IKVM/GNU gets things very wrong sometimes. Finally, this implementation is faster for "everyday" numbers, though it is slower for more extreme numbers. It would probably be reasonable to hand over formatting to the Java platform (at least when running the Sun JDK) for exponents outside the range -7 to +7.

Field Summary

Fields

Modifier and Type	Field and Description
static long	DOUBLE_SIGN_MASK
static int	FLOAT_SIGN_MASK
static FloatingPointConverter	THE_INSTANCE

Method Summary

Modifier and Type	Method and Description
static FastStringBuffer	appendDouble(FastStringBuffer s, double d, boolean forceExponential) Append a string representation of a double value to a string buffer
static FastStringBuffer	appendFloat(FastStringBuffer s, float f, boolean forceExponential) Append a string representation of a float value to a string buffer
static FastStringBuffer	appendInt(FastStringBuffer s, int i) Format an integer, appending the string representation of the integer to a string buffer

Methods inherited from class java.lang.Object

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

Field Detail

THE_INSTANCE

public static FloatingPointConverter THE_INSTANCE

DOUBLE_SIGN_MASK

public static final long DOUBLE_SIGN_MASK

See Also:

Constant Field Values

FLOAT_SIGN_MASK

public static final int FLOAT_SIGN_MASK

See Also:

Constant Field Values

Method Detail

appendInt

public static FastStringBuffer appendInt(FastStringBuffer s,
                                         int i)

Format an integer, appending the string representation of the integer to a string buffer

Parameters:

s - the string buffer

i - the integer to be formatted

Returns:

the supplied string buffer, containing the appended integer

appendDouble

public static FastStringBuffer appendDouble(FastStringBuffer s,
                                            double d,
                                            boolean forceExponential)

Append a string representation of a double value to a string buffer

Parameters:

s - the string buffer to which the result will be appended

d - the double to be formatted

Returns:

the original string buffer, now containing the string representation of the supplied double

appendFloat

public static FastStringBuffer appendFloat(FastStringBuffer s,
                                           float f,
                                           boolean forceExponential)

Append a string representation of a float value to a string buffer

Parameters:

s - the string buffer to which the result will be appended

f - the float to be formatted

forceExponential - forces exponential notation if set (if not set, exponential notation is used only for values outside the range 1e-6 to 1e+6)

Returns:

the original string buffer, now containing the string representation of the supplied float