From 15fa18bc5656adac26799dc5d8791414f68f6b0c Mon Sep 17 00:00:00 2001 From: John Fitzpatrick Date: Thu, 30 Mar 2000 00:17:54 +0000 Subject: [PATCH] initial checkin X-SVN-Rev: 1031 --- icu4j/src/com/ibm/icu/math/BigDecimal.java | 4357 +++++++++++++++++++ icu4j/src/com/ibm/icu/math/MathContext.java | 582 +++ icu4j/src/com/ibm/math/BigDecimal.java | 4357 +++++++++++++++++++ icu4j/src/com/ibm/math/MathContext.java | 582 +++ 4 files changed, 9878 insertions(+) create mode 100755 icu4j/src/com/ibm/icu/math/BigDecimal.java create mode 100755 icu4j/src/com/ibm/icu/math/MathContext.java create mode 100755 icu4j/src/com/ibm/math/BigDecimal.java create mode 100755 icu4j/src/com/ibm/math/MathContext.java diff --git a/icu4j/src/com/ibm/icu/math/BigDecimal.java b/icu4j/src/com/ibm/icu/math/BigDecimal.java new file mode 100755 index 00000000000..427616279f2 --- /dev/null +++ b/icu4j/src/com/ibm/icu/math/BigDecimal.java @@ -0,0 +1,4357 @@ +/* Generated from 'BigDecimal.nrx' 27 Mar 2000 22:38:05 [v1.162] */ +/* Options: Binary Comments Crossref Format Java Logo Strictargs Strictcase Trace2 Verbose3 */ +package com.ibm.math; +import java.math.BigInteger; + +/* ------------------------------------------------------------------ */ +/* BigDecimal -- Decimal arithmetic for Java */ +/* ------------------------------------------------------------------ */ +/* Copyright IBM Corporation, 1996, 2000. All Rights Reserved. */ +/* */ +/* The BigDecimal class provides immutable arbitrary-precision */ +/* floating point (including integer) decimal numbers. */ +/* */ +/* As the numbers are decimal, there is an exact correspondence */ +/* between an instance of a BigDecimal object and its String */ +/* representation; the BigDecimal class provides direct conversions */ +/* to and from String and character array objects, and well as */ +/* conversions to and from the Java primitive types (which may not */ +/* be exact). */ +/* ------------------------------------------------------------------ */ +/* Notes: */ +/* */ +/* 1. A BigDecimal object is never changed in value once constructed; */ +/* this avoids the need for locking. Note in particular that the */ +/* mantissa array may be shared between many BigDecimal objects, */ +/* so that once exposed it must not be altered. */ +/* */ +/* 2. This class looks at MathContext class fields directly (for */ +/* performance). It must not and does not change them. */ +/* */ +/* 3. Exponent checking is delayed until finish(), as we know */ +/* intermediate calculations cannot cause 31-bit overflow. */ +/* [This assertion depends on MAX_DIGITS in MathContext.] */ +/* */ +/* 4. Comments for the public API now follow the javadoc conventions. */ +/* The NetRexx -comments option is used to pass these comments */ +/* through to the generated Java code (with -format, if desired). */ +/* */ +/* 5. System.arraycopy is faster than explicit loop as follows */ +/* Mean length 4: equal */ +/* Mean length 8: x2 */ +/* Mean length 16: x3 */ +/* Mean length 24: x4 */ +/* From prior experience, we expect mean length a little below 8, */ +/* but arraycopy is still the one to use, in general, until later */ +/* measurements suggest otherwise. */ +/* */ +/* 6. 'DMSRCN' referred to below is the original (1981) IBM S/370 */ +/* assembler code implementation of the algorithms below; it is */ +/* now called IXXRCN and is available with the OS/390 and VM/ESA */ +/* operating systems. */ +/* ------------------------------------------------------------------ */ +/* Change History: */ +/* 1997.09.02 Initial version (derived from netrexx.lang classes) */ +/* 1997.09.12 Add lostDigits checking */ +/* 1997.10.06 Change mantissa to a byte array */ +/* 1997.11.22 Rework power [did not prepare arguments, etc.] */ +/* 1997.12.13 multiply did not prepare arguments */ +/* 1997.12.14 add did not prepare and align arguments correctly */ +/* 1998.05.02 0.07 packaging changes suggested by Sun and Oracle */ +/* 1998.05.21 adjust remainder operator finalization */ +/* 1998.06.04 rework to pass MathContext to finish() and round() */ +/* 1998.06.06 change format to use round(); support rounding modes */ +/* 1998.06.25 rename to BigDecimal and begin merge */ +/* zero can now have trailing zeros (i.e., exp\=0) */ +/* 1998.06.28 new methods: movePointXxxx, scale, toBigInteger */ +/* unscaledValue, valueof */ +/* 1998.07.01 improve byteaddsub to allow array reuse, etc. */ +/* 1998.07.01 make null testing explicit to avoid JIT bug [Win32] */ +/* 1998.07.07 scaled division [divide(BigDecimal, int, int)] */ +/* 1998.07.08 setScale, faster equals */ +/* 1998.07.11 allow 1E6 (no sign) ; new double/float conversion */ +/* 1998.10.12 change package to com.ibm.math */ +/* 1998.12.14 power operator no longer rounds RHS [to match ANSI] */ +/* add toBigDecimal() and BigDecimal(java.math.BigDecimal) */ +/* 1998.12.29 improve byteaddsub by using table lookup */ +/* 1999.02.04 lostdigits=0 behaviour rounds instead of digits+1 guard */ +/* 1999.02.05 cleaner code for BigDecimal(char[]) */ +/* 1999.02.06 add javadoc comments */ +/* 1999.02.11 format() changed from 7 to 2 method form */ +/* 1999.03.05 null pointer checking is no longer explicit */ +/* 1999.03.05 simplify; changes from discussion with J. Bloch: */ +/* null no longer permitted for MathContext; drop boolean, */ +/* byte, char, float, short constructor, deprecate double */ +/* constructor, no blanks in string constructor, add */ +/* offset and length version of char[] constructor; */ +/* add valueOf(double); drop booleanValue, charValue; */ +/* add ...Exact versions of remaining convertors */ +/* 1999.03.13 add toBigIntegerExact */ +/* 1999.03.13 1.00 release to IBM Centre for Java Technology */ +/* 1999.05.27 1.01 correct 0-0.2 bug under scaled arithmetic */ +/* 1999.06.29 1.02 constructors should not allow exponent > 9 digits */ +/* 1999.07.03 1.03 lost digits should not be checked if digits=0 */ +/* 1999.07.06 lost digits Exception message changed */ +/* 1999.07.10 1.04 more work on 0-0.2 (scaled arithmetic) */ +/* 1999.07.17 improve messages from pow method */ +/* 1999.08.08 performance tweaks */ +/* 1999.08.15 fastpath in multiply */ +/* 1999.11.05 1.05 fix problem in intValueExact [e.g., 5555555555] */ +/* 1999.12.22 1.06 remove multiply fastpath, and improve performance */ +/* 2000.01.01 copyright update [Y2K has arrived] */ +/* ------------------------------------------------------------------ */ + + + + + +/** + * The BigDecimal class implements immutable + * arbitrary-precision decimal numbers. The methods of the + * BigDecimal class provide operations for fixed and + * floating point arithmetic, comparison, format conversions, and + * hashing. + *

+ * As the numbers are decimal, there is an exact correspondence between + * an instance of a BigDecimal object and its + * String representation; the BigDecimal class + * provides direct conversions to and from String and + * character array (char[]) objects, as well as conversions + * to and from the Java primitive types (which may not be exact) and + * BigInteger. + *

+ * In the descriptions of constructors and methods in this documentation, + * the value of a BigDecimal number object is shown as the + * result of invoking the toString() method on the object. + * The internal representation of a decimal number is neither defined + * nor exposed, and is not permitted to affect the result of any + * operation. + *

+ * The floating point arithmetic provided by this class is defined by + * the ANSI X3.274-1996 standard, and is also documented at + * http://www2.hursley.ibm.com/decimal + *
[This URL will change.] + * + *

Operator methods

+ *

+ * Operations on BigDecimal numbers are controlled by a + * {@link MathContext} object, which provides the context (precision and + * other information) for the operation. Methods that can take a + * MathContext parameter implement the standard arithmetic + * operators for BigDecimal objects and are known as + * operator methods. The default settings provided by the + * constant {@link MathContext#DEFAULT} (digits=9, + * form=SCIENTIFIC, lostDigits=false, roundingMode=ROUND_HALF_UP) + * perform general-purpose floating point arithmetic to nine digits of + * precision. The MathContext parameter must not be + * null. + *

+ * Each operator method also has a version provided which does + * not take a MathContext parameter. For this version of + * each method, the context settings used are digits=0, + * form=PLAIN, lostDigits=false, roundingMode=ROUND_HALF_UP; + * these settings perform fixed point arithmetic with unlimited + * precision, as defined for the original BigDecimal class in Java 1.1 + * and Java 1.2. + *

+ * For monadic operators, only the optional MathContext + * parameter is present; the operation acts upon the current object. + *

+ * For dyadic operators, a BigDecimal parameter is always + * present; it must not be null. + * The operation acts with the current object being the left-hand operand + * and the BigDecimal parameter being the right-hand operand. + *

+ * For example, adding two BigDecimal objects referred to + * by the names award and extra could be + * written as any of: + *

+ * award.add(extra) + *
award.add(extra, MathContext.DEFAULT) + *
award.add(extra, acontext) + * + *

+ * (where acontext is a MathContext object), + * which would return a BigDecimal object whose value is + * the result of adding award and extra under + * the appropriate context settings. + *

+ * When a BigDecimal operator method is used, a set of + * rules define what the result will be (and, by implication, how the + * result would be represented as a character string). + * These rules are defined in the BigDecimal arithmetic documentation + * (see the URL above), but in summary: + *