001    /*
002     * Copyright (C) 2011 The Guava Authors
003     *
004     * Licensed under the Apache License, Version 2.0 (the "License");
005     * you may not use this file except in compliance with the License.
006     * You may obtain a copy of the License at
007     *
008     * http://www.apache.org/licenses/LICENSE-2.0
009     *
010     * Unless required by applicable law or agreed to in writing, software
011     * distributed under the License is distributed on an "AS IS" BASIS,
012     * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
013     * See the License for the specific language governing permissions and
014     * limitations under the License.
015     */
016    
017    package com.google.common.math;
018    
019    import static com.google.common.base.Preconditions.checkArgument;
020    import static com.google.common.math.DoubleUtils.IMPLICIT_BIT;
021    import static com.google.common.math.DoubleUtils.SIGNIFICAND_BITS;
022    import static com.google.common.math.DoubleUtils.getSignificand;
023    import static com.google.common.math.DoubleUtils.isFinite;
024    import static com.google.common.math.DoubleUtils.isNormal;
025    import static com.google.common.math.DoubleUtils.scaleNormalize;
026    import static com.google.common.math.MathPreconditions.checkInRange;
027    import static com.google.common.math.MathPreconditions.checkNonNegative;
028    import static com.google.common.math.MathPreconditions.checkRoundingUnnecessary;
029    import static java.lang.Math.abs;
030    import static java.lang.Math.copySign;
031    import static java.lang.Math.getExponent;
032    import static java.lang.Math.log;
033    import static java.lang.Math.rint;
034    
035    import com.google.common.annotations.Beta;
036    import com.google.common.annotations.VisibleForTesting;
037    import com.google.common.primitives.Booleans;
038    
039    import java.math.BigInteger;
040    import java.math.RoundingMode;
041    
042    /**
043     * A class for arithmetic on doubles that is not covered by {@link java.lang.Math}.
044     *
045     * @author Louis Wasserman
046     * @since 11.0
047     */
048    @Beta
049    public final class DoubleMath {
050      /*
051       * This method returns a value y such that rounding y DOWN (towards zero) gives the same result
052       * as rounding x according to the specified mode.
053       */
054      static double roundIntermediate(double x, RoundingMode mode) {
055        if (!isFinite(x)) {
056          throw new ArithmeticException("input is infinite or NaN");
057        }
058        switch (mode) {
059          case UNNECESSARY:
060            checkRoundingUnnecessary(isMathematicalInteger(x));
061            return x;
062    
063          case FLOOR:
064            if (x >= 0.0 || isMathematicalInteger(x)) {
065              return x;
066            } else {
067              return x - 1.0;
068            }
069    
070          case CEILING:
071            if (x <= 0.0 || isMathematicalInteger(x)) {
072              return x;
073            } else {
074              return x + 1.0;
075            }
076    
077          case DOWN:
078            return x;
079    
080          case UP:
081            if (isMathematicalInteger(x)) {
082              return x;
083            } else {
084              return x + Math.copySign(1.0, x);
085            }
086    
087          case HALF_EVEN:
088            return rint(x);
089    
090          case HALF_UP: {
091            double z = rint(x);
092            if (abs(x - z) == 0.5) {
093              return x + copySign(0.5, x);
094            } else {
095              return z;
096            }
097          }
098    
099          case HALF_DOWN: {
100            double z = rint(x);
101            if (abs(x - z) == 0.5) {
102              return x;
103            } else {
104              return z;
105            }
106          }
107    
108          default:
109            throw new AssertionError();
110        }
111      }
112    
113      /**
114       * Returns the {@code int} value that is equal to {@code x} rounded with the specified rounding
115       * mode, if possible.
116       *
117       * @throws ArithmeticException if
118       *         <ul>
119       *         <li>{@code x} is infinite or NaN
120       *         <li>{@code x}, after being rounded to a mathematical integer using the specified
121       *         rounding mode, is either less than {@code Integer.MIN_VALUE} or greater than {@code
122       *         Integer.MAX_VALUE}
123       *         <li>{@code x} is not a mathematical integer and {@code mode} is
124       *         {@link RoundingMode#UNNECESSARY}
125       *         </ul>
126       */
127      public static int roundToInt(double x, RoundingMode mode) {
128        double z = roundIntermediate(x, mode);
129        checkInRange(z > MIN_INT_AS_DOUBLE - 1.0 & z < MAX_INT_AS_DOUBLE + 1.0);
130        return (int) z;
131      }
132    
133      private static final double MIN_INT_AS_DOUBLE = -0x1p31;
134      private static final double MAX_INT_AS_DOUBLE = 0x1p31 - 1.0;
135    
136      /**
137       * Returns the {@code long} value that is equal to {@code x} rounded with the specified rounding
138       * mode, if possible.
139       *
140       * @throws ArithmeticException if
141       *         <ul>
142       *         <li>{@code x} is infinite or NaN
143       *         <li>{@code x}, after being rounded to a mathematical integer using the specified
144       *         rounding mode, is either less than {@code Long.MIN_VALUE} or greater than {@code
145       *         Long.MAX_VALUE}
146       *         <li>{@code x} is not a mathematical integer and {@code mode} is
147       *         {@link RoundingMode#UNNECESSARY}
148       *         </ul>
149       */
150      public static long roundToLong(double x, RoundingMode mode) {
151        double z = roundIntermediate(x, mode);
152        checkInRange(MIN_LONG_AS_DOUBLE - z < 1.0 & z < MAX_LONG_AS_DOUBLE_PLUS_ONE);
153        return (long) z;
154      }
155    
156      private static final double MIN_LONG_AS_DOUBLE = -0x1p63;
157      /*
158       * We cannot store Long.MAX_VALUE as a double without losing precision.  Instead, we store
159       * Long.MAX_VALUE + 1 == -Long.MIN_VALUE, and then offset all comparisons by 1.
160       */
161      private static final double MAX_LONG_AS_DOUBLE_PLUS_ONE = 0x1p63;
162    
163      /**
164       * Returns the {@code BigInteger} value that is equal to {@code x} rounded with the specified
165       * rounding mode, if possible.
166       *
167       * @throws ArithmeticException if
168       *         <ul>
169       *         <li>{@code x} is infinite or NaN
170       *         <li>{@code x} is not a mathematical integer and {@code mode} is
171       *         {@link RoundingMode#UNNECESSARY}
172       *         </ul>
173       */
174      public static BigInteger roundToBigInteger(double x, RoundingMode mode) {
175        x = roundIntermediate(x, mode);
176        if (MIN_LONG_AS_DOUBLE - x < 1.0 & x < MAX_LONG_AS_DOUBLE_PLUS_ONE) {
177          return BigInteger.valueOf((long) x);
178        }
179        int exponent = getExponent(x);
180        long significand = getSignificand(x);
181        BigInteger result = BigInteger.valueOf(significand).shiftLeft(exponent - SIGNIFICAND_BITS);
182        return (x < 0) ? result.negate() : result;
183      }
184    
185      /**
186       * Returns {@code true} if {@code x} is exactly equal to {@code 2^k} for some finite integer
187       * {@code k}.
188       */
189      public static boolean isPowerOfTwo(double x) {
190        return x > 0.0 && isFinite(x) && LongMath.isPowerOfTwo(getSignificand(x));
191      }
192    
193      /**
194       * Returns the base 2 logarithm of a double value.
195       *
196       * <p>Special cases:
197       * <ul>
198       * <li>If {@code x} is NaN or less than zero, the result is NaN.
199       * <li>If {@code x} is positive infinity, the result is positive infinity.
200       * <li>If {@code x} is positive or negative zero, the result is negative infinity.
201       * </ul>
202       *
203       * <p>The computed result is within 1 ulp of the exact result.
204       *
205       * <p>If the result of this method will be immediately rounded to an {@code int},
206       * {@link #log2(double, RoundingMode)} is faster.
207       */
208      public static double log2(double x) {
209        return log(x) / LN_2; // surprisingly within 1 ulp according to tests
210      }
211    
212      private static final double LN_2 = log(2);
213    
214      /**
215       * Returns the base 2 logarithm of a double value, rounded with the specified rounding mode to an
216       * {@code int}.
217       *
218       * <p>Regardless of the rounding mode, this is faster than {@code (int) log2(x)}.
219       *
220       * @throws IllegalArgumentException if {@code x <= 0.0}, {@code x} is NaN, or {@code x} is
221       *         infinite
222       */
223      @SuppressWarnings("fallthrough")
224      public static int log2(double x, RoundingMode mode) {
225        checkArgument(x > 0.0 && isFinite(x), "x must be positive and finite");
226        int exponent = getExponent(x);
227        if (!isNormal(x)) {
228          return log2(x * IMPLICIT_BIT, mode) - SIGNIFICAND_BITS;
229          // Do the calculation on a normal value.
230        }
231        // x is positive, finite, and normal
232        boolean increment;
233        switch (mode) {
234          case UNNECESSARY:
235            checkRoundingUnnecessary(isPowerOfTwo(x));
236            // fall through
237          case FLOOR:
238            increment = false;
239            break;
240          case CEILING:
241            increment = !isPowerOfTwo(x);
242            break;
243          case DOWN:
244            increment = exponent < 0 & !isPowerOfTwo(x);
245            break;
246          case UP:
247            increment = exponent >= 0 & !isPowerOfTwo(x);
248            break;
249          case HALF_DOWN:
250          case HALF_EVEN:
251          case HALF_UP:
252            double xScaled = scaleNormalize(x);
253            // sqrt(2) is irrational, and the spec is relative to the "exact numerical result,"
254            // so log2(x) is never exactly exponent + 0.5.
255            increment = (xScaled * xScaled) > 2.0;
256            break;
257          default:
258            throw new AssertionError();
259        }
260        return increment ? exponent + 1 : exponent;
261      }
262    
263      /**
264       * Returns {@code true} if {@code x} represents a mathematical integer.
265       *
266       * <p>This is equivalent to, but not necessarily implemented as, the expression {@code
267       * !Double.isNaN(x) && !Double.isInfinite(x) && x == Math.rint(x)}.
268       */
269      public static boolean isMathematicalInteger(double x) {
270        return isFinite(x)
271            && (x == 0.0 ||
272                SIGNIFICAND_BITS - Long.numberOfTrailingZeros(getSignificand(x)) <= getExponent(x));
273      }
274    
275      /**
276       * Returns {@code n!}, that is, the product of the first {@code n} positive
277       * integers, {@code 1} if {@code n == 0}, or e n!}, or
278       * {@link Double#POSITIVE_INFINITY} if {@code n! > Double.MAX_VALUE}.
279       *
280       * <p>The result is within 1 ulp of the true value.
281       *
282       * @throws IllegalArgumentException if {@code n < 0}
283       */
284      public static double factorial(int n) {
285        checkNonNegative("n", n);
286        if (n > MAX_FACTORIAL) {
287          return Double.POSITIVE_INFINITY;
288        } else {
289          // Multiplying the last (n & 0xf) values into their own accumulator gives a more accurate
290          // result than multiplying by EVERY_SIXTEENTH_FACTORIAL[n >> 4] directly.
291          double accum = 1.0;
292          for (int i = 1 + (n & ~0xf); i <= n; i++) {
293            accum *= i;
294          }
295          return accum * EVERY_SIXTEENTH_FACTORIAL[n >> 4];
296        }
297      }
298    
299      @VisibleForTesting
300      static final int MAX_FACTORIAL = 170;
301    
302      @VisibleForTesting
303      static final double[] EVERY_SIXTEENTH_FACTORIAL = {
304          0x1.0p0,
305          0x1.30777758p44,
306          0x1.956ad0aae33a4p117,
307          0x1.ee69a78d72cb6p202,
308          0x1.fe478ee34844ap295,
309          0x1.c619094edabffp394,
310          0x1.3638dd7bd6347p498,
311          0x1.7cac197cfe503p605,
312          0x1.1e5dfc140e1e5p716,
313          0x1.8ce85fadb707ep829,
314          0x1.95d5f3d928edep945};
315    
316      /**
317       * Returns {@code true} if {@code a} and {@code b} are within {@code tolerance} of each other.
318       *
319       * <p>Technically speaking, this is equivalent to
320       * {@code Math.abs(a - b) <= tolerance || Double.valueOf(a).equals(Double.valueOf(b))}.
321       *
322       * <p>Notable special cases include:
323       * <ul>
324       * <li>All NaNs are fuzzily equal.
325       * <li>If {@code a == b}, then {@code a} and {@code b} are always fuzzily equal.
326       * <li>Positive and negative zero are always fuzzily equal.
327       * <li>If {@code tolerance} is zero, and neither {@code a} nor {@code b} is NaN, then
328       * {@code a} and {@code b} are fuzzily equal if and only if {@code a == b}.
329       * <li>With {@link Double#POSITIVE_INFINITY} tolerance, all non-NaN values are fuzzily equal.
330       * <li>With finite tolerance, {@code Double.POSITIVE_INFINITY} and {@code
331       * Double.NEGATIVE_INFINITY} are fuzzily equal only to themselves.
332       * </li>
333       *
334       * <p>This is reflexive and symmetric, but <em>not</em> transitive, so it is <em>not</em> an
335       * equivalence relation and <em>not</em> suitable for use in {@link Object#equals}
336       * implementations.
337       *
338       * @throws IllegalArgumentException if {@code tolerance} is {@code < 0} or NaN
339       * @since 13.0
340       */
341      @Beta
342      public static boolean fuzzyEquals(double a, double b, double tolerance) {
343        MathPreconditions.checkNonNegative("tolerance", tolerance);
344        return
345              Math.copySign(a - b, 1.0) <= tolerance
346               // copySign(x, 1.0) is a branch-free version of abs(x), but with different NaN semantics
347              || (a == b) // needed to ensure that infinities equal themselves
348              || ((a != a) && (b != b)); // x != x is equivalent to Double.isNaN(x), but faster
349      }
350    
351      /**
352       * Compares {@code a} and {@code b} "fuzzily," with a tolerance for nearly-equal values.
353       *
354       * <p>This method is equivalent to
355       * {@code fuzzyEquals(a, b, tolerance) ? 0 : Double.compare(a, b)}. In particular, like
356       * {@link Double#compare(double, double)}, it treats all NaN values as equal and greater than all
357       * other values (including {@link Double#POSITIVE_INFINITY}).
358       *
359       * <p>This is <em>not</em> a total ordering and is <em>not</em> suitable for use in
360       * {@link Comparable#compareTo} implementations.  In particular, it is not transitive.
361       *
362       * @throws IllegalArgumentException if {@code tolerance} is {@code < 0} or NaN
363       * @since 13.0
364       */
365      @Beta
366      public static int fuzzyCompare(double a, double b, double tolerance) {
367        if (fuzzyEquals(a, b, tolerance)) {
368          return 0;
369        } else if (a < b) {
370          return -1;
371        } else if (a > b) {
372          return 1;
373        } else {
374          return Booleans.compare(Double.isNaN(a), Double.isNaN(b));
375        }
376      }
377    
378      private DoubleMath() {}
379    }