All Downloads are FREE. Search and download functionalities are using the official Maven repository.

com.sun.electric.util.math.MutableInterval Maven / Gradle / Ivy

The newest version!
/* -*- tab-width: 4 -*-
 *
 * Electric(tm) VLSI Design System
 *
 * File: MutableInterval.java
 *
 * Copyright (c) 2004, Oracle and/or its affiliates. All rights reserved.
 *
 * Electric(tm) is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 3 of the License, or
 * (at your option) any later version.
 *
 * Electric(tm) 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 General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with Electric(tm); see the file COPYING.  If not, write to
 * the Free Software Foundation, Inc., 59 Temple Place, Suite 330,
 * Boston, Mass 02111-1307, USA.
 */

package com.sun.electric.util.math;

import java.math.BigDecimal;

/**
 * Mutable class for representation of intervals X = [a, b].
 * a and b are double precision floation point numbers with a <= b. 
 * 
 * Closed ("extended") interval system is implemented, which is the extension
 * of regular interval system.
 *
 * The elements of the regular interval system (IR) are intervals with regular 
 * IEEE floating point number bounds.
 *
 * In the closed ("extended") interval system (IR*) the set of regular 
 * intervals IR is extended by infinite intervals and the empty interval.
 * All arithmetic operations and mathematical functions are closed on IR*.
 * For details see the paper 
 * 
 * Interval Arithmetic Specification by Chiriaev and Walster. The 
 * implementation is following this specification.  
 * Three special intervals are supported in the extended system:
 * 
    *
  • [ EMPTY ], represented as [ NaN, NaN ] *
  • [ -INF ], represented as [ -INF, -MAX ] *
  • [ +INF ], represented as [ MAX, +INF] *
* where MAX is the largest regular floating point number. All interval * operations are guaranteed to produce results consistent with these * representations. */ public class MutableInterval { /* Fields */ private double inf; private double sup; /* Constants */ /** * A constant holding 3/4*ulp(1). * next(x) is round-to-nearest * of (x + x*ULP_EPS) for positive normalized doble numbers. */ private static final double ULP_EPS = 1.5/(1L << 53); // 0x1.8p-53; /** * A constant holding prev(1). * prev(x) is round-to-nearest * if (x*SCALE_DOWN) for positive normalized double numbers. */ private static final double SCALE_DOWN = 1.0 - 1.0/(1L << 53); // 0x1.fffffffffffffp-1; /** * A constant holding minimal positive normalized double number. */ private static final double MIN_NORMAL = Double.MIN_VALUE*(1L << 52); // 0x1.0p-1022; /** * A constant limiting those x for which x*ULP_EPS is noramlized. */ private static final double ULP_EPS_NORMAL = MIN_NORMAL / ULP_EPS; // 0x1.5555555555555p-970 /** * A constant holding ulp(Double.MAX_VALUE). */ private static final double MAX_ULP = Double.MAX_VALUE/(1L << 53)/SCALE_DOWN; // 0x1.0p971; /** * A constant holding range of long numbers exactly represented * by double numbers. All longs in [ -EXACT_LONG, EXACT_LONG] are * represented exactly. Some of other longs are not. */ private static final long EXACT_LONG = 1L << 53; // ----------------------------------------------------------------------- // Constructors // ----------------------------------------------------------------------- /** * Constructs a point interval [0,0]. */ public MutableInterval() { } /** * Constructs a point interval [x,x]. */ public MutableInterval(int x) { assign(x); } /** * Constructs sharpest interval containing x. */ public MutableInterval(long x) { assign(x); } /** * Constructs a point interval [x,x]. * * Special cases in the extended system: *
    *
  • if x == +INF then [ +INF ] is constructed *
  • if x == -INF then [ -INF ] is constructed *
  • if x == NaN then the entire interval [ -INF,+INF ] is constructed *
*/ public MutableInterval(double x) { assign(x); } /** * Constructs the interval [inf, sup]. * * Special cases in the extended system: *
    *
  • if inf < sup then the entire interval [ -INF,+INF ] is constructed *
* * @param inf The infimum of the interval to be constructed. * @param sup The supremum of the interval to be constructed. */ public MutableInterval(int inf, int sup) { assign(inf, sup); } /** * Constructs the interval [inf, sup]. * * Special cases in the extended system: *
    *
  • if inf < sup then the entire interval [ -INF,+INF ] is constructed *
* * @param inf The infimum of the interval to be constructed. * @param sup The supremum of the interval to be constructed. */ public MutableInterval(long inf, long sup) { assign(inf, sup); } /** * Constructs the interval [inf, sup]. * * Special cases in the extended system: *
    *
  • if inf == sup == -INF the interval [ -INF ] is constructed *
  • if inf == sup == +INF the interval [ +INF ] is constructed *
  • if inf < sup then the entire interval [ -INF,+INF ] is constructed *
  • if inf == NaN or sup == NaN then the entire interval [ -INF,+INF ] is constructed *
* * @param inf The infimum of the interval to be constructed. * @param sup The supremum of the interval to be constructed. */ public MutableInterval(double inf, double sup) { assign(inf, sup); } /** * Constructs the interval same as x. */ public MutableInterval(MutableInterval x) { assign(x); } /** * Constructs the interval from string. */ public MutableInterval(String s) { assign(s); } /** * Constructs the interval from character array. */ public MutableInterval(char[] b) { assign(b); } // ----------------------------------------------------------------------- // Assigns // ----------------------------------------------------------------------- /** * Assigns a point interval [x,x]. */ public MutableInterval assign(int x) { inf = sup = (double)x; return this; } /** * Assigns sharpest interval containing x. */ public MutableInterval assign(long x) { double xd = (double)x; inf = sup = xd; if (Math.abs(x) > EXACT_LONG) { long xx = (long)xd; if (xx > x || x == Long.MAX_VALUE) inf = prev(xd); else if (xx < x) sup = next(xd); } return this; } /** * Assigns a point interval. * * Special cases in the extended system: *
    *
  • if x == +INF then [ +INF ] is constructed *
  • if x == -INF then [ -INF ] is constructed *
  • if x == NaN then the entire interval [ -INF,+INF ] is constructed *
*/ public MutableInterval assign(double x) { inf = sup = x; if (x == Double.POSITIVE_INFINITY) inf = Double.MAX_VALUE; else if (x == Double.NEGATIVE_INFINITY) sup = -Double.MAX_VALUE; else if (x != x) assignEntire(); return this; } /** * Assigns the interval [inf, sup]. * * Special cases in the extended system: *
    *
  • if inf < sup then the entire interval [ -INF,+INF ] is constructed *
* * @param inf The infimum of the interval to be constructed. * @param sup The supremum of the interval to be constructed. */ public MutableInterval assign(int inf, int sup) { if (inf <= sup) { this.inf = (double)inf; this.sup = (double)sup; } else assignEntire(); return this; } /** * Assigns the interval [inf, sup]. * * Special cases in the extended system: *
    *
  • if inf < sup then the the entire interval [ -INF,+INF ] is constructed *
* * @param inf The infimum of the interval to be constructed. * @param sup The supremum of the interval to be constructed. */ public MutableInterval assign(long inf, long sup) { if (inf <= sup) { this.inf = (double)inf; this.sup = (double)sup; if (inf < -EXACT_LONG || sup > EXACT_LONG) { if ((long)this.inf > inf || inf == Long.MAX_VALUE) this.inf = prev(this.inf); if ((long)this.sup < sup) this.sup = next(this.sup); } } else assignEntire(); return this; } /** * Assigns the interval [inf, sup]. * * Special cases in the extended system: *
    *
  • if inf == sup == -INF the interval [ -INF ] is constructed *
  • if inf == sup == +INF the interval [ +INF ] is constructed *
  • if inf < sup then the entire interval [ -INF,+INF ] is constructed *
  • if inf == NaN or sup == NaN then the entire [ -INF,+INF ] is constructed *
* * @param inf The infimum of the interval to be constructed. * @param sup The supremum of the interval to be constructed. */ public MutableInterval assign(double inf, double sup) { if (inf <= sup) { this.inf = (inf == Double.POSITIVE_INFINITY ? Double.MAX_VALUE : inf); this.sup = (sup == Double.NEGATIVE_INFINITY ? -Double.MAX_VALUE : sup); } else assignEntire(); return this; } /** * Assigns interval same as x. */ public MutableInterval assign(MutableInterval x) { this.inf = x.inf; this.sup = x.sup; return this; } /** * Assigns the interval from string. */ public MutableInterval assign(String s) { parse(s); return this; } /** * Assigns the interval from character array. */ public MutableInterval assign(char[] b) { parse(new String(b)); return this; } /** * Creates and returns a copy of this MutableInterval. */ public MutableInterval clon() { try { return (MutableInterval)clone(); } catch (CloneNotSupportedException e) { return new MutableInterval(inf, sup); } } /** * Assigns entire interval [ -INF, +INF ]. */ public MutableInterval assignEntire() { inf = Double.NEGATIVE_INFINITY; sup = Double.POSITIVE_INFINITY; return this; } /** * Assigns empty interval [-EMPTY ]. */ public MutableInterval assignEmpty() { inf = sup = Double.NaN; return this; } // ----------------------------------------------------------------------- // Interval bounds // ----------------------------------------------------------------------- /** * Returns the infimum of this interval. * * Special cases: *
  • x.inf() == NaN for x == [ EMPTY ]. *
  • x.inf() == MAX_DOUBLE for x == [ +INF ].
*/ public double inf() { return inf; } /** * Returns the supremum of this interval. * * Special cases: *
  • x.sup() == NaN for x == [ EMPTY ]. *
  • x.sup() == MAX_DOUBLE for x == [ +INF ].
*/ public double sup() { return sup; } // ----------------------------------------------------------------------- // Info functions // ----------------------------------------------------------------------- /** * Returns true iff this interval is a point interval. * i.e. inf() == sup(). * * Special cases in the extended system: *
    *
  • x.isPoint() == false for x == [ EMPTY ] *
*/ public boolean isPoint() { return inf == sup; } /** * Same as isPoint(). */ public boolean isDegenerate() { return inf == sup; } /** * Returns true if this == [ EMPTY ]. */ public boolean isEmpty() { return inf != inf; } /** * Return true if either x.inf() or x.sup() is infinite. */ public boolean isInfinite() { return inf == Double.NEGATIVE_INFINITY || sup == Double.POSITIVE_INFINITY; } /** * Returns true iff this interval has an ulp accuracy of n. * I.e. x.inf() and x.sup() have a distance of at most n machine numbers. * * Special cases in the extended system: *
    *
  • x.hasUlpAcc(n) == false for x == [ EMPTY ] or any infinite interval *
*/ public boolean hasUlpAcc(int n) { if (isInfinite()) return false; double x = inf; int i = 0; while (i++ < n && x < sup) x = next(x); return x == sup; } /** * Returns true iff this interval is an entire interval [ -INF, +INF ]. * * Special cases in the extended system: *
    *
  • x.isEntire(n) == false for x == [ EMPTY ] *
*/ public boolean isEntire() { return -inf == sup && sup == Double.POSITIVE_INFINITY; } /** * Always returns true indicating that the extended system is currently used. */ public static boolean isExtended() { return true; } /** * Always returns false, indicating that this is not native implementation. */ public static boolean isNative() { return false; } // ----------------------------------------------------------------------- // Utility functions // ----------------------------------------------------------------------- /** * Returns double number nearest to the midpoint of this interval, i.e. * * x.mid == (x.inf() + x.sup()) / 2. * * Special cases in the extended system: *
    *
  • x.mid() == NaN for x == [ EMPTY ] *
  • x.mid() == 0.0 for x == [ ENTIRE ] *
  • x.mid() == +INF for x == [ +INF ] or x = [ a, +INF ] *
  • x.mid() == -INF for x == [ -INF ] or x = [ -INF, a] *
*/ public double mid() { double mid = 0.5*(inf + sup); if (mid > Double.NEGATIVE_INFINITY && mid < Double.POSITIVE_INFINITY) return mid; return -inf == sup ? 0 : 0.5*inf + 0.5*sup; } /** * Returns an upper bound for the width (diameter) of this interval, i.e. * * x.wid() == x.sup()-x.inf() * * Special cases in the extended system: *
    *
  • x.wid() == NaN for x == [ EMPTY ] *
  • x.wid() == +INF for any infinite interval *
*/ public double wid() { return addPosUp(sup, -inf); } /** * Returns an upper bound for the radius of this interval, i.e. * * x.mid() - x.rad() <= x.inf() <= x.sup <= x.mid() + x.rad() * * Special cases in the extended system: *
    *
  • x.rad() == NaN for x == [ EMPTY ] *
  • x.rad() == +INF for any infinite interval *
*/ public double rad() { double mid = (inf + sup) * 0.5; if (!(mid > Double.NEGATIVE_INFINITY && mid < Double.POSITIVE_INFINITY)) { if (inf == Double.NEGATIVE_INFINITY || sup == Double.POSITIVE_INFINITY) return Double.POSITIVE_INFINITY; mid = 0.5*inf + 0.5*sup; } return Math.max(addPosUp(sup, -mid), addPosUp(mid, -inf)); } /** * Returns the mignitude of this interval, i.e. * * x.mig() == min{abs(y) : y in x } * * Special cases in the extended system: *
    *
  • x.mig() == NaN for x == [ EMPTY ] *
*/ public double mig() { return inf <= 0 && sup >= 0 ? 0 : inf < 0 ? -sup : inf; } /** * Returns the magnitude of this interval, i.e. * * x.mag() == max{abs(y) : y in x } * * Special cases in the extended system: *
    *
  • x.mag() == NaN for x == [ EMPTY ] *
  • x.mag() == +INF for any infinite interval *
*/ public double mag() { return -inf > sup ? -inf : sup; } /** * Returns the interval of absolute values of this interval, i.e. * * x.abs() == [ x.mig(), x.mag() ] * * Special cases in the extended system: *
    *
  • x.abs() == [ EMPTY ] for x == [ EMPTY ] *
  • x.abs() == [ +INF ] for x == [ +/- INF ] *
*/ public MutableInterval abs() { if (sup <= 0) { double h = sup; this.sup = -inf; this.inf = -h; } else if (inf < 0) { if (-inf > sup) this.sup = -inf; // else sup = this.sup; inf = 0; } // else { inf = this.inf; sup = this.sup; } return this; } /** * Returns an enclosure for the range of minima of this interval and the * interval x, i.e. * * x.min(y) == { z : z == min(a, b) : a in x, b in y } * * Special cases in the extended system: *
    *
  • x.min(y) == [ EMPTY ] for x == [ EMPTY ] and y == [ EMPTY ] *
*/ public MutableInterval min(MutableInterval y) { if (this.inf != this.inf) // inf = this.inf; sup = this.sup; return this; inf = (this.inf < y.inf ? this.inf : y.inf); sup = (this.sup < y.sup ? this.sup : y.sup); return this; } /** * Returns an enclosure for the range of maxima of this interval and the * interval x, i.e. * * x.max(y) == { z : z == max(a, b) : a in x, b in y } * * Special cases in the extended system: *
    *
  • x.max(y) == [ EMPTY ] for x == [ EMPTY ] and y == [ EMPTY ] *
*/ public MutableInterval max(MutableInterval y) { if (this.inf != this.inf) // inf = this.inf; sup = this.sup; return this; inf = (this.inf > y.inf ? this.inf : y.inf); sup = (this.sup > y.sup ? this.sup : y.sup); return this; } /** * Returns an upper bound for the Hausdorff distance of this interval * and the interval x, i.e. * * x.dist(y) == max{ abs(x.inf()-y.inf()), abs(x.sup() - y.sup()) } * * Special cases in the extended system: *
    *
  • x.dist(y) == NaN for x == [ EMPTY ] or y == [ EMPTY ] *
*/ double dist(MutableInterval x) { if (isEmpty() || x.isEmpty()) return Double.NaN; if ( this.inf == x.inf && this.sup == x.sup) return 0; if (isInfinite() || x.isInfinite()) return Double.POSITIVE_INFINITY;; double dinf = this.inf > x.inf ? addPosUp(this.inf, -x.inf) : addPosUp(-this.inf, x.inf); double dsup = this.sup > x.sup ? addPosUp(this.sup, -x.sup) : addPosUp(-this.sup, x.sup); return Math.max(dinf, dsup); } // ----------------------------------------------------------------------- // Set operations // ----------------------------------------------------------------------- /** * Returns the intersection of this interval and the interval x. * * Special cases in the extended system: *
    *
  • x.intersect(y) == [ EMPTY ] iff x and y are disjoint *
*/ public MutableInterval intersect(MutableInterval y) { double l, u; if (isEmpty() || y.isEmpty()) { this.inf = Double.NaN; this.sup = Double.NaN; return this; } l = (this.inf > y.inf ? this.inf : y.inf); u = (this.sup < y.sup ? this.sup : y.sup); if (l > u) { this.inf = Double.NaN; this.sup = Double.NaN; return this; } this.inf = l; this.sup = u; return this; } /** * Same as x.intersect(y) */ public MutableInterval ix(MutableInterval y) { return intersect(y); } /** * Returns the convex hull of this interval and the interval x. * * Special cases in the extended system: *
    *
  • x.interval_hull(y) == [ EMPTY ] for x == y == [ EMPTY ] *
*/ public MutableInterval interval_hull(MutableInterval y) { if (isEmpty()) { this.inf = y.inf; this.sup = y.sup; return this; } if (y.isEmpty()) return this; inf = (this.inf < y.inf ? this.inf : y.inf); sup = (this.sup > y.sup ? this.sup : y.sup); return this; } /** * Returns the convex hull of this interval and the double x. * * Special cases in the extended system: *
    *
  • x.interval_hull(y) == [ -INF,+INF ] for y == NaN *
*/ public MutableInterval interval_hull(double x) { if (isEmpty()) return assign(x); if (x != x) return assignEntire(); inf = (this.inf < x ? this.inf : x); sup = (this.sup > x ? this.sup : x); return this; } /** * Same as x.interval_hull(y) */ public MutableInterval ih(MutableInterval y) { return interval_hull(y); } // ----------------------------------------------------------------------- // Set relations // ----------------------------------------------------------------------- /** * Returns true iff this interval and interval x are disjoint., i.e. * * Special cases in the extended system: *
    *
  • x.disjoint(y) == true for x or y == [ EMPTY ] *
*/ public boolean disjoint(MutableInterval y) { return !(this.inf <= y.sup && y.inf <= this.sup); } /** * Same as x.disjoint(y) */ public boolean dj(MutableInterval y) { return !(this.inf <= y.sup && y.inf <= this.sup); } /** * Returns true iff double number x is contained in this interval. * * Special cases in the extended system: *
    *
  • x.contains(y) == false for x == [ EMPTY ] or y == NaN *
*/ public boolean in(long x) { double xd = (double)x; if (xd >= -EXACT_LONG && xd <= EXACT_LONG) return inf <= xd && xd <= sup; long xx = (long)xd; return (xx > x || x == Long.MAX_VALUE ? inf < xd : inf <= xd) && (xx < x ? xd < sup : xd <= sup); } /** * Returns true iff double number x is contained in this interval. * * Special cases in the extended system: *
    *
  • x.contains(y) == false for x == [ EMPTY ] or y == NaN *
*/ public boolean in(double y) { return y >= this.inf && y <= this.sup; } /** * Returns true iff this interval is in interior of interval x. * * Special cases in the extended system: *
    *
  • x.in_interior(y) == true for x == [ EMPTY ] *
*/ public boolean in_interior(MutableInterval y) { return (inf > y.inf && sup < y.sup) || isEmpty(); } /** * Same as x.interior(y) */ public boolean interior(MutableInterval y) { return in_interior(y); } /** * Returns true iff this interval is a proper subset of interval x. * * Special cases in the extended system: *
    *
  • x.proper_subset(y) == true for x == [ EMPTY ] and y != [ EMPTY ] *
*/ public boolean proper_subset(MutableInterval y) { return (inf >= y.inf && sup <= y.sup && (inf > y.inf || sup < y.sup)) || (isEmpty() && ! y.isEmpty()); } /** * Returns true iff this interval is a subset of interval x. * Special cases in the extended system: * *
    *
  • x.subset(y) == true for x == [ EMPTY ] *
*/ public boolean subset(MutableInterval y) { return y.inf <= inf && sup <= y.sup || isEmpty(); } /** * Returns true iff this interval is a proper superset of interval x. * * Special cases in the extended system: *
    *
  • x.proper_superset(y) == true for x != [ EMPTY ] and y == [ EMPTY ] *
*/ public boolean proper_superset(MutableInterval y) { return (inf <= y.inf && y.sup <= sup && (inf < y.inf || y.sup < sup)) || (y.isEmpty() && ! isEmpty()); } /** * Returns true iff this interval is a superset of interval x. * * Special cases in the extended system: *
    *
  • x.superset(y) == true for y == [ EMPTY ] *
*/ public boolean superset(MutableInterval y) { return inf <= y.inf && y.sup <= sup || y.isEmpty(); } /** * Returns true iff this interval is set-equal to interval x. * * Special cases in the extended system: *
    *
  • x.seq(y) == true for x == y == [ EMPTY ] *
*/ public boolean seq(MutableInterval y) { return (inf == y.inf && sup == y.sup) || isEmpty() && y.isEmpty(); } /** * Returns true iff this interval is set-not-equal to interval x. */ public boolean sne(MutableInterval y) { return !seq(y); } /** * Returns true iff this interval is set-greater-or-equal to * interval x. * * Special cases in the extended system: *
    *
  • x.sge(y) == true for x == y == [ EMPTY ] *
*/ public boolean sge(MutableInterval y) { return inf >= y.inf && sup >= y.sup || isEmpty() && y.isEmpty(); } /** * Returns true iff this interval is set-greater than interval x. * * Special cases in the extended system: *
    *
  • x.sgt(y) == false for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public boolean sgt(MutableInterval y) { return inf > y.inf && sup > y.sup; } /** * Returns true iff this interval is set-less-or-equal to * interval x. * * Special cases in the extended system: *
    *
  • x.sle(y) == true for x == y == [ EMPTY ] *
*/ public boolean sle(MutableInterval y) { return inf <= y.inf && sup <= y.sup || isEmpty() && y.isEmpty(); } /** * Returns true iff this interval is set-less than interval x. * * Special cases in the extended system: *
    *
  • x.slt(y) == false for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public boolean slt(MutableInterval y) { return inf < y.inf && sup < y.sup; } // ----------------------------------------------------------------------- // Certainly relations // ----------------------------------------------------------------------- /** * Returns true iff this interval is certainly-equal to interval x. * * Special cases in the extended system: *
    *
  • x.ceq(y) == false for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public boolean ceq(MutableInterval y) { return sup <= y.inf && inf >= y.sup; } /** * Returns true iff this interval is certainly-not-equal to interval x. * * Special cases in the extended system: *
    *
  • x.cne(y) == true for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public boolean cne(MutableInterval y) { return !(inf <= y.sup && y.inf <= sup); } /** * Returns true iff this interval is certainly-greater-or-equal to * interval x. * * Special cases in the extended system: *
    *
  • x.cge(y) == false for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public boolean cge(MutableInterval y) { return inf >= y.sup; } /** * Returns true iff this interval is certainly-greater than interval x. * * Special cases in the extended system: *
    *
  • x.cgt(y) == false for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public boolean cgt(MutableInterval y) { return inf > y.sup; } /** * Returns true iff this interval is certainly-less-or-equal to * interval x. * * Special cases in the extended system: *
    *
  • x.cle(y) == false for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public boolean cle(MutableInterval y) { return sup <= y.inf; } /** * Returns true iff this interval is certainly-less than interval x. * * Special cases in the extended system: *
    *
  • x.clt(y) == false for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public boolean clt(MutableInterval y) { return sup < y.inf; } // ----------------------------------------------------------------------- // Possibly relations // ----------------------------------------------------------------------- /** * Returns true iff this interval is possibly-equal to interval x. * * Special cases in the extended system: *
    *
  • x.peq(y) == false for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public boolean peq(MutableInterval y) { return inf <= y.sup && sup >= y.inf; } /** * Returns true iff this interval is possibly-not-equal to interval x. * * Special cases in the extended system: *
    *
  • x.pne(y) == true for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public boolean pne(MutableInterval y) { return !(sup <= y.inf && inf >= y.sup); } /** * Returns true iff this interval is possibly-greater-or-equal to * interval x. * * Special cases in the extended system: *
    *
  • x.pge(y) == false for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public boolean pge(MutableInterval y) { return sup >= y.inf; } /** * Returns true iff this interval is possibly-greater than interval x. * * Special cases in the extended system: *
    *
  • x.pgt(y) == false for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public boolean pgt(MutableInterval y) { return sup > y.inf; } /** * Returns true iff this interval is possibly-less-or-equal to * interval x. * * Special cases in the extended system: *
    *
  • x.ple(y) == false for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public boolean ple(MutableInterval y) { return inf <= y.sup; } /** * Returns true iff this interval is possibly-less than interval x. * * Special cases in the extended system: *
    *
  • x.plt(y) == false for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public boolean plt(MutableInterval y) { return inf < y.sup; } // ----------------------------------------------------------------------- // Arithmetic operations // ----------------------------------------------------------------------- /** * Unary operator -. Myltiplies this interval by -1 and returns the product. * * Special cases in the extended system: *
    *
  • -x == [ EMPTY ] for x == [ EMPTY ] *
*/ public MutableInterval negate() { double l = inf; inf = -sup; sup = -l; return this; } /** * Adds interval y to this interval and returns the sum. * * Special cases in the extended system: *
    *
  • x += y == [ EMPTY ] for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public MutableInterval add(MutableInterval y) { double l = this.inf + y.inf; if (l - this.inf > y.inf || l - y.inf > this.inf) { assert Math.abs(l) >= MIN_NORMAL*2; l = (l < 0 ? (l < -ULP_EPS_NORMAL ? l + l*ULP_EPS : l/SCALE_DOWN) : l <= Double.MAX_VALUE ? l*SCALE_DOWN : Double.MAX_VALUE); } double h = this.sup + y.sup; if (h - this.sup < y.sup || h - y.sup < this.sup) { assert Math.abs(h) >= MIN_NORMAL*2; h = (h > 0 ? (h > ULP_EPS_NORMAL ? h + h*ULP_EPS : h/SCALE_DOWN) : h >= -Double.MAX_VALUE ? h*SCALE_DOWN : -Double.MAX_VALUE); } inf = l; sup = h; return this; } /** * Adds double number y to this interval and returns the sum. * * Special cases in the extended system: *
    *
  • x += y == [ EMPTY ] for x == [ EMPTY ] or y == NaN *
*/ public MutableInterval add(double y) { double l = this.inf + y; if (l - this.inf > y || l - y > this.inf) { assert Math.abs(l) >= MIN_NORMAL*2; l = (l < 0 ? (l < -ULP_EPS_NORMAL ? l + l*ULP_EPS : l/SCALE_DOWN) : l <= Double.MAX_VALUE ? l*SCALE_DOWN : Double.MAX_VALUE); } else if (!(l < Double.POSITIVE_INFINITY)) { if (y >= this.sup) // x is not [EMPTY] and y is not NaN l = Double.MAX_VALUE; } double h = this.sup + y; if (h - this.sup < y || h - y < this.sup) { assert Math.abs(h) >= MIN_NORMAL*2; h = (h > 0 ? (h > ULP_EPS_NORMAL ? h + h*ULP_EPS : h/SCALE_DOWN) : h >= -Double.MAX_VALUE ? h*SCALE_DOWN : -Double.MAX_VALUE); } else if (!(h > Double.NEGATIVE_INFINITY)) { if (y <= this.inf) // x is not [EMPTY] and y is not NaN h = -Double.MAX_VALUE; } inf = l; sup = h; return this; } /** * Subtracts interval y from this interval and returns the difference. * * Special cases in the extended system: *
    *
  • x -= y == [ EMPTY ] for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public MutableInterval sub(MutableInterval y) { double l = this.inf - y.sup; if (this.inf - l < y.sup || l + y.sup > this.inf) { assert Math.abs(l) >= MIN_NORMAL*2; l = (l < 0 ? (l < -ULP_EPS_NORMAL ? l + l*ULP_EPS : l/SCALE_DOWN) : l <= Double.MAX_VALUE ? l*SCALE_DOWN : Double.MAX_VALUE); } double h = this.sup - y.inf; if (this.sup - h > y.inf || h + y.inf < this.sup) { assert Math.abs(h) >= MIN_NORMAL*2; h = (h > 0 ? (h > ULP_EPS_NORMAL ? h + h*ULP_EPS : h/SCALE_DOWN) : h >= -Double.MAX_VALUE ? h*SCALE_DOWN : -Double.MAX_VALUE); } inf = l; sup = h; return this; } /** * Subtracts double number y from this interval and returns the difference. * * Special cases in the extended system: *
    *
  • x -= y == [ EMPTY ] for x == [ EMPTY ] or y == NaN *
*/ public MutableInterval sub(double y) { double l = this.inf - y; if (this.inf - l < y || l + y > this.inf) { assert Math.abs(l) >= MIN_NORMAL*2; l = (l < 0 ? (l < -ULP_EPS_NORMAL ? l + l*ULP_EPS : l/SCALE_DOWN) : l <= Double.MAX_VALUE ? l*SCALE_DOWN : Double.MAX_VALUE); } else if (!(l < Double.POSITIVE_INFINITY)) { if (y >= this.sup) // x is not [EMPTY] and y is not NaN l = Double.MAX_VALUE; } double h = this.sup - y; if (this.sup - h > y || h + y < this.sup) { assert Math.abs(h) >= MIN_NORMAL*2; h = (h > 0 ? (h > ULP_EPS_NORMAL ? h + h*ULP_EPS : h/SCALE_DOWN) : h >= -Double.MAX_VALUE ? h*SCALE_DOWN : -Double.MAX_VALUE); } else if (!(h > Double.NEGATIVE_INFINITY)) { if (y <= this.inf) // x is not [EMPTY] and y is not NaN h = -Double.MAX_VALUE; } inf = l; sup = h; return this; } /** * Multiplies this interval by interval y and returns the product. * * * Special cases in the extended system: *
    *
  • x *= y == [ EMPTY ] for x == [ EMPTY ] or y == [ EMPTY ] *
*/ public MutableInterval mul(MutableInterval y) { double l, h; if (y.inf > 0) { if (this.inf > 0) { // x > 0, y > 0 l = this.inf*y.inf; h = this.sup*y.sup; } else if (this.sup < 0) { // x < 0, y > 0 l = this.inf*y.sup; h = this.sup*y.inf; } else { // 0 in x or x is [NaN,NaN], y > 0 l = this.inf*y.sup; h = this.sup*y.sup; } } else if (y.sup < 0) { if (this.inf > 0) { // x > 0, y < 0 l = this.sup*y.inf; h = this.inf*y.sup; } else if (this.sup < 0) { // x < 0, y < 0 l = this.sup*y.sup; h = this.inf*y.inf; } else { // 0 in x or x is [NaN,NaN], y < 0 l = this.sup*y.inf; h = this.inf*y.inf; } } else { if (this.inf > 0) { // x > 0, 0 in y or y is [NaN,NaN] l = this.sup*y.inf; h = this.sup*y.sup; } else if (this.sup < 0) { // x < 0, 0 in y or y is [NaN,NaN] l = this.inf*y.sup; h = this.inf*y.inf; } else { // 0 in x or x is [NaN,NaN], 0 in y or y is [NaN,NaN] l = this.sup*y.inf; double lo = this.inf*y.sup; if (l > lo || lo != lo) l = lo; h = this.inf*y.inf; double ho = this.sup*y.sup; if (h < ho || ho != ho) h = ho; } } if (l > 0) l = prevPos(l); else if (l < 0) l = prevNeg(l); else if (l == 0) l = (this.inf >= 0 && y.inf >= 0 || this.sup <= 0 && y.sup <= 0) ? 0 : -Double.MIN_VALUE; else if (this.inf == this.inf && y.inf == y.inf) return assignEntire(); if (h > 0) h = nextPos(h); else if (h < 0) h = nextNeg(h); else if (h == 0) h = (this.inf >= 0 && y.sup <= 0 || this.sup <= 0 && y.inf >= 0) ? 0 : Double.MIN_VALUE; else if (this.inf == this.inf && y.inf == y.inf) return assignEntire(); inf = l; sup = h; return this; } /** * Divides this interval by interval y and returns the quotient. * * Special cases in the extended system: *
    *
  • x /= y == [ EMPTY ] for x == [ EMPTY ] or y == [ EMPTY ] *
  • x /= y == [ ENTIRE ] if y contains 0 *
*/ public MutableInterval div(MutableInterval y) { double l, h; if (y.inf > 0) { if (this.inf > 0) { // x > 0, y > 0 l = this.inf/y.sup; h = this.sup/y.inf; } else if (this.sup < 0) { // x < 0, y > 0 l = this.inf/y.inf; h = this.sup/y.sup; } else { // 0 in x or x is [NaN,NaN], y > 0 l = this.inf/y.inf; h = this.sup/y.inf; } } else if (y.sup < 0) { if (this.inf > 0) { // x > 0, y < 0 l = this.sup/y.sup; h = this.inf/y.inf; } else if (this.sup < 0) { // x < 0, y < 0 l = this.sup/y.inf; h = this.inf/y.sup; } else { // 0 in x or x is [NaN,NaN], y < 0 l = this.sup/y.sup; h = this.inf/y.sup; } } else { // 0 in y or y is [NaN,NaN] l = h = Double.NaN; } if (l > 0) l = prevPos(l); else if (l < 0) l = prevNeg(l); else if (l == 0) l = (this.inf >= 0 && y.inf > 0 || this.sup <= 0 && y.sup < 0) ? 0 : -Double.MIN_VALUE; else if (this.inf == this.inf && y.inf == y.inf) return assignEntire(); if (h > 0) h = nextPos(h); else if (h < 0) h = nextNeg(h); else if (h == 0) h = (this.inf >= 0 && y.sup < 0 || this.sup <= 0 && y.inf > 0) ? 0 : Double.MIN_VALUE; else if (this.inf == this.inf && y.inf == y.inf) return assignEntire(); inf = l; sup = h; return this; } // ----------------------------------------------------------------------- // Elementary functions // ----------------------------------------------------------------------- /** * Replaces this interval by an interval enclosure of its exponential. */ public MutableInterval exp() { double l = Math.exp(this.inf); if (l > 0) l = prevPos(l); double h = Math.exp(this.sup); if (h > 0) h = nextPos(h); else if (h == 0) h = Double.MIN_VALUE; inf = l; sup = h; return this; } /** * Replaces this interval by an interval enclosure of its natural logarithm. */ public MutableInterval log() { double l = Math.log(this.inf); double h = Math.log(this.sup); if (l > 0) l = prevPos(l); else if (l < 0) l = prevNeg(l); else if (l != l && h == h) l = Double.NEGATIVE_INFINITY; if (h > 0) h = nextPos(h); else if (h < 0) h = nextNeg(h); inf = l; sup = h; return this; } // ----------------------------------------------------------------------- // I/O // ----------------------------------------------------------------------- /** * Return string representation. */ public String toString() { if (isEmpty()) return "[EMPTY ]"; StringBuffer buf = new StringBuffer(49); buf.append('['); append(buf, inf, false); buf.append(','); append(buf, sup, true); buf.append(']'); return buf.toString(); } private static void append(StringBuffer buf, double x, boolean isSup) { final int d = 16; // digits after floating point if (x == Double.NEGATIVE_INFINITY) buf.append(" -Infinity"); else if (x == Double.POSITIVE_INFINITY) buf.append(" Infinity"); else if (x == 0 && !isSup) buf.append("-.0000000000000000E+000"); else if (x == 0 && isSup) buf.append("0.0000000000000000E+000"); else { BigDecimal bx = new BigDecimal(x); String s = bx.unscaledValue().abs().toString(); int drop = 0; if (s.length() != d) { drop = s.length() - d; bx = bx.setScale(bx.scale() - drop, isSup ? BigDecimal.ROUND_CEILING : BigDecimal.ROUND_FLOOR); } s = bx.unscaledValue().abs().toString(); if (s.length() > d) { assert s.length() == d + 1 && s.charAt(s.length() - 1) == '0'; drop++; bx = bx.setScale(bx.scale() - 1, BigDecimal.ROUND_UNNECESSARY); s = bx.unscaledValue().abs().toString(); } buf.append(bx.signum() < 0 ? '-' : '0'); buf.append('.'); buf.append(s); buf.append('E'); int exp = d - bx.scale(); if (exp >= 0) buf.append('+'); else { buf.append('-'); exp = -exp; } assert (exp < 1000); buf.append((char)('0' + exp/100)); exp %= 100; buf.append((char)('0' + exp/10)); exp %= 10; buf.append((char)('0' + exp)); } } private void parse(String s) { s = s.trim(); if (s.length() < 2 || s.charAt(0) != '[' || s.charAt(s.length() - 1) != ']') throw new NumberFormatException(); int comma = s.indexOf(','); String ls = s.substring(1, comma < 0 ? s.length() - 1 : comma).trim(); BigDecimal lb = null; if (ls.equals("NaN") || ls.equals("+NaN") || ls.equals("-NaN")) inf = Double.NaN; else if (ls.equals("Infinity") || ls.equals("+Infinity")) inf = Double.POSITIVE_INFINITY; else if (ls.equals("-Infinity")) inf = Double.NEGATIVE_INFINITY; else if (ls.equals("EMPTY")) { if (comma >= 0) throw new NumberFormatException(); assignEmpty(); return; } else { lb = new BigDecimal(ls); inf = lb.doubleValue(); } if (comma >= 0) { String rs = s.substring(comma + 1, s.length() - 1).trim(); BigDecimal rb = null; if (rs.equals("NaN") || rs.equals("+NaN") || rs.equals("-NaN")) sup = Double.NaN; else if (rs.equals("Infinity") || rs.equals("+Infinity")) sup = Double.POSITIVE_INFINITY; else if (rs.equals("-Infinity")) sup = Double.NEGATIVE_INFINITY; else { rb = new BigDecimal(rs); sup = rb.doubleValue(); if (inf >= sup && (inf > sup || lb.compareTo(rb) > 0)) { assignEntire(); return; } sup = correct(rb, sup, true); } } else { sup = correct(lb, inf, true); } inf = correct(lb, inf, false); if (inf != inf || sup != sup || inf > sup) { assignEntire(); return; } if (inf == Double.POSITIVE_INFINITY) inf = Double.MAX_VALUE; if (sup == Double.NEGATIVE_INFINITY) sup = -Double.MAX_VALUE; } private static double correct(BigDecimal b, double d, boolean isSup) { if (b == null || Double.isInfinite(d)) return d; int diff = b.compareTo(new BigDecimal(d)); if (isSup) { if (diff > 0) d = next(d); } else { if (diff < 0) d = prev(d); } return d; } // ----------------------------------------------------------------------- // predecessor and successor of a number // ----------------------------------------------------------------------- /** * Returns the size of an ulp of the argument. An ulp of a * double value is the positive distance between this * floating-point value and the double value next * larger in magnitude. Note that for non-NaN x, * ulp(-x) == ulp(x). * *

Special Cases: *

    *
  • If the argument is NaN, then the result is NaN. *
  • If the argument is positive or negative infinity, then the * result is positive infinity. *
  • If the argument is positive or negative zero, then the result is * Double.MIN_VALUE. *
  • If the argument is ±Double.MAX_VALUE, then * the result is equal to 2971. *
* * @param d the floating-point value whose ulp is to be returned * @return the size of an ulp of the argument */ public static double ulp(double d) { if (d < 0) d = -d; if (d < Double.MAX_VALUE) { if (d > ULP_EPS_NORMAL) return (d + d*ULP_EPS) - d; if (d >= MIN_NORMAL*2) return d/SCALE_DOWN - d; return Double.MIN_VALUE; } if (d == Double.MAX_VALUE) return MAX_ULP; return d; } /** * Returns previous floating point double number. * * Special cases in the extended system: *
    *
  • prev(NaN) == NaN *
  • prev(+INF) == Double.MAX_VALUE *
  • prev(-INF) == -INF *
  • prev(0) == -Double.MIN_VALUE *
*/ public static double prev(double x) { if (x <= MIN_NORMAL) { if (x < -ULP_EPS_NORMAL) return x + x*ULP_EPS; else if (x <= -MIN_NORMAL) return x/SCALE_DOWN; else if (x == 0) return -Double.MIN_VALUE; else return Double.longBitsToDouble(Double.doubleToLongBits(x) + (x > 0 ? -1 : 1)); } else if (x > Double.MAX_VALUE) return Double.MAX_VALUE; else return x*SCALE_DOWN; } /** * Returns next floating point double number. * * Special cases: *
    *
  • next(NaN) == NaN *
  • next(+INF) == +INF *
  • next(-INF) == -Double.MAX_VALUE *
  • next(0) == Double.MIN_VALUE *
*/ public static double next(double x) { if (x >= -MIN_NORMAL) { if (x > ULP_EPS_NORMAL) return x + x*ULP_EPS; else if (x >= MIN_NORMAL) return x/SCALE_DOWN; else if (x == 0) return Double.MIN_VALUE; else return Double.longBitsToDouble(Double.doubleToLongBits(x) + (x > 0 ? 1 : -1)); } else if (x < -Double.MAX_VALUE) return -Double.MAX_VALUE; else return x*SCALE_DOWN; } private static double nextPos(double x) { assert x > 0; return x > ULP_EPS_NORMAL ? x + x*ULP_EPS : x >= MIN_NORMAL ? x/SCALE_DOWN : Double.longBitsToDouble(Double.doubleToLongBits(x) + 1); } private static double nextNeg(double x) { assert x < 0; return x <= -MIN_NORMAL ? (x >= -Double.MAX_VALUE ? x*SCALE_DOWN : -Double.MAX_VALUE) : Double.longBitsToDouble(Double.doubleToLongBits(x) - 1); } private static double prevPos(double x) { assert x > 0; return x >= MIN_NORMAL ? (x <= Double.MAX_VALUE ? x*SCALE_DOWN : Double.MAX_VALUE) : Double.longBitsToDouble(Double.doubleToLongBits(x) - 1); } private static double prevNeg(double x) { assert x < 0; return x < -ULP_EPS_NORMAL ? x + x*ULP_EPS : x <= -MIN_NORMAL ? x/SCALE_DOWN : Double.longBitsToDouble(Double.doubleToLongBits(x) + 1); } public static double addUp(double x, double y) { double z = x + y; if (z - x < y || z - y < x) { assert Math.abs(z) >= MIN_NORMAL*2; return (z > 0 ? (z > ULP_EPS_NORMAL ? z + z*ULP_EPS : z/SCALE_DOWN) : z >= -Double.MAX_VALUE ? z*SCALE_DOWN : -Double.MAX_VALUE); } return (z == z || x != x || y != y ? z : Double.POSITIVE_INFINITY); } public static double addPosUp(double x, double y) { double z = x + y; assert z >= 0 || z != z; if (z - x < y || z - y < x) { assert z >= MIN_NORMAL*2; return (z > ULP_EPS_NORMAL ? z + z*ULP_EPS : z/SCALE_DOWN); } return z; } }




© 2015 - 2024 Weber Informatics LLC | Privacy Policy