在Java中表示分数的最佳方法是什么？

Question

我正在尝试在Java中使用分数。

我想要实现算术函数。为此，我首先需要一种规范化函数的方法。我知道在没有通分之前无法将1/6和1/2相加。我将不得不添加1/6和3/6。一个天真的方法是将2/12和6/12相加，然后约分。如何以最小的性能惩罚实现共同分母？哪种算法最好？

---

第8版（感谢hstoerr）：

改进包括：

• equals()方法现在与compareTo()方法一致

final class Fraction extends Number {
    private int numerator;
    private int denominator;

    public Fraction(int numerator, int denominator) {
        if(denominator == 0) {
            throw new IllegalArgumentException("denominator is zero");
        }
        if(denominator < 0) {
            numerator *= -1;
            denominator *= -1;
        }
        this.numerator = numerator;
        this.denominator = denominator;
    }

    public Fraction(int numerator) {
        this.numerator = numerator;
        this.denominator = 1;
    }

    public int getNumerator() {
        return this.numerator;
    }

    public int getDenominator() {
        return this.denominator;
    }

    public byte byteValue() {
        return (byte) this.doubleValue();
    }

    public double doubleValue() {
        return ((double) numerator)/((double) denominator);
    }

    public float floatValue() {
        return (float) this.doubleValue();
    }

    public int intValue() {
        return (int) this.doubleValue();
    }

    public long longValue() {
        return (long) this.doubleValue();
    }

    public short shortValue() {
        return (short) this.doubleValue();
    }

    public boolean equals(Fraction frac) {
        return this.compareTo(frac) == 0;
    }

    public int compareTo(Fraction frac) {
        long t = this.getNumerator() * frac.getDenominator();
        long f = frac.getNumerator() * this.getDenominator();
        int result = 0;
        if(t>f) {
            result = 1;
        }
        else if(f>t) {
            result = -1;
        }
        return result;
    }
}

---

我已删除所有先前的版本，感谢以下人员：

• Dave Ray
• cletus
• duffymo
• James
• Milhous
• Oscar Reyes
• Jason S
• Francisco Canedo
• Outlaw Programmer
• Beska

- eleven81

---

放弃自己编写的代码，使用Apache Commons：） http://commons.apache.org/math/userguide/fraction.html - Patrick

如果Patrick的评论作为答案发布，它本应该得到+1的赞同。在大多数情况下，那就是正确的答案；就像《Effective Java》所说的那样，“了解并使用库”。原问题也很清晰有用。 - Jonik

当前版本实现了equals...但没有实现hashCode。如果您想将Fraction用作Map中的键，请实现hashCode。 - Xantix

还有一个bug：您想在compareTo中使用long以防止溢出，但乘法仅使用int。 - maaartinus

你的代码中没有 hashCode。仅定义 equals 可能会导致哈希映射中出现非常糟糕的行为（比两者都不定义还要糟糕）。请注意，定义 hashCode 需要某种规范化以保持与 equals 一致。 - maaartinus

Answer 1

恰好我不久前为Project Euler问题编写了一个BigFraction类。它保留了BigInteger分子和分母，因此永远不会溢出。但是对于许多你知道永远不会溢出的操作来说，它可能会慢一点...无论如何，如果您需要它，请使用它。我一直想以某种方式展示它 :)

编辑：包括单元测试在内的最新版本代码现在托管在GitHub上，也可以通过Maven Central获得。我将我的原始代码保留在这里，以便这个答案不只是一个链接...

---

import java.math.*;

/**
 * Arbitrary-precision fractions, utilizing BigIntegers for numerator and
 * denominator.  Fraction is always kept in lowest terms.  Fraction is
 * immutable, and guaranteed not to have a null numerator or denominator.
 * Denominator will always be positive (so sign is carried by numerator,
 * and a zero-denominator is impossible).
 */
public final class BigFraction extends Number implements Comparable<BigFraction>
{
  private static final long serialVersionUID = 1L; //because Number is Serializable
  private final BigInteger numerator;
  private final BigInteger denominator;

  public final static BigFraction ZERO = new BigFraction(BigInteger.ZERO, BigInteger.ONE, true);
  public final static BigFraction ONE = new BigFraction(BigInteger.ONE, BigInteger.ONE, true);

  /**
   * Constructs a BigFraction with given numerator and denominator.  Fraction
   * will be reduced to lowest terms.  If fraction is negative, negative sign will
   * be carried on numerator, regardless of how the values were passed in.
   */
  public BigFraction(BigInteger numerator, BigInteger denominator)
  {
    if(numerator == null)
      throw new IllegalArgumentException("Numerator is null");
    if(denominator == null)
      throw new IllegalArgumentException("Denominator is null");
    if(denominator.equals(BigInteger.ZERO))
      throw new ArithmeticException("Divide by zero.");

    //only numerator should be negative.
    if(denominator.signum() < 0)
    {
      numerator = numerator.negate();
      denominator = denominator.negate();
    }

    //create a reduced fraction
    BigInteger gcd = numerator.gcd(denominator);
    this.numerator = numerator.divide(gcd);
    this.denominator = denominator.divide(gcd);
  }

  /**
   * Constructs a BigFraction from a whole number.
   */
  public BigFraction(BigInteger numerator)
  {
    this(numerator, BigInteger.ONE, true);
  }

  public BigFraction(long numerator, long denominator)
  {
    this(BigInteger.valueOf(numerator), BigInteger.valueOf(denominator));
  }

  public BigFraction(long numerator)
  {
    this(BigInteger.valueOf(numerator), BigInteger.ONE, true);
  }

  /**
   * Constructs a BigFraction from a floating-point number.
   * 
   * Warning: round-off error in IEEE floating point numbers can result
   * in answers that are unexpected.  For example, 
   *     System.out.println(new BigFraction(1.1))
   * will print:
   *     2476979795053773/2251799813685248
   * 
   * This is because 1.1 cannot be expressed exactly in binary form.  The
   * given fraction is exactly equal to the internal representation of
   * the double-precision floating-point number.  (Which, for 1.1, is:
   * (-1)^0 * 2^0 * (1 + 0x199999999999aL / 0x10000000000000L).)
   * 
   * NOTE: In many cases, BigFraction(Double.toString(d)) may give a result
   * closer to what the user expects.
   */
  public BigFraction(double d)
  {
    if(Double.isInfinite(d))
      throw new IllegalArgumentException("double val is infinite");
    if(Double.isNaN(d))
      throw new IllegalArgumentException("double val is NaN");

    //special case - math below won't work right for 0.0 or -0.0
    if(d == 0)
    {
      numerator = BigInteger.ZERO;
      denominator = BigInteger.ONE;
      return;
    }

    final long bits = Double.doubleToLongBits(d);
    final int sign = (int)(bits >> 63) & 0x1;
    final int exponent = ((int)(bits >> 52) & 0x7ff) - 0x3ff;
    final long mantissa = bits & 0xfffffffffffffL;

    //number is (-1)^sign * 2^(exponent) * 1.mantissa
    BigInteger tmpNumerator = BigInteger.valueOf(sign==0 ? 1 : -1);
    BigInteger tmpDenominator = BigInteger.ONE;

    //use shortcut: 2^x == 1 << x.  if x is negative, shift the denominator
    if(exponent >= 0)
      tmpNumerator = tmpNumerator.multiply(BigInteger.ONE.shiftLeft(exponent));
    else
      tmpDenominator = tmpDenominator.multiply(BigInteger.ONE.shiftLeft(-exponent));

    //1.mantissa == 1 + mantissa/2^52 == (2^52 + mantissa)/2^52
    tmpDenominator = tmpDenominator.multiply(BigInteger.valueOf(0x10000000000000L));
    tmpNumerator = tmpNumerator.multiply(BigInteger.valueOf(0x10000000000000L + mantissa));

    BigInteger gcd = tmpNumerator.gcd(tmpDenominator);
    numerator = tmpNumerator.divide(gcd);
    denominator = tmpDenominator.divide(gcd);
  }

  /**
   * Constructs a BigFraction from two floating-point numbers.
   * 
   * Warning: round-off error in IEEE floating point numbers can result
   * in answers that are unexpected.  See BigFraction(double) for more
   * information.
   * 
   * NOTE: In many cases, BigFraction(Double.toString(numerator) + "/" + Double.toString(denominator))
   * may give a result closer to what the user expects.
   */
  public BigFraction(double numerator, double denominator)
  {
    if(denominator == 0)
      throw new ArithmeticException("Divide by zero.");

    BigFraction tmp = new BigFraction(numerator).divide(new BigFraction(denominator));
    this.numerator = tmp.numerator;
    this.denominator = tmp.denominator;
  }

  /**
   * Constructs a new BigFraction from the given BigDecimal object.
   */
  public BigFraction(BigDecimal d)
  {
    this(d.scale() < 0 ? d.unscaledValue().multiply(BigInteger.TEN.pow(-d.scale())) : d.unscaledValue(),
         d.scale() < 0 ? BigInteger.ONE                                             : BigInteger.TEN.pow(d.scale()));
  }

  public BigFraction(BigDecimal numerator, BigDecimal denominator)
  {
    if(denominator.equals(BigDecimal.ZERO))
      throw new ArithmeticException("Divide by zero.");

    BigFraction tmp = new BigFraction(numerator).divide(new BigFraction(denominator));
    this.numerator = tmp.numerator;
    this.denominator = tmp.denominator;
  }

  /**
   * Constructs a BigFraction from a String.  Expected format is numerator/denominator,
   * but /denominator part is optional.  Either numerator or denominator may be a floating-
   * point decimal number, which in the same format as a parameter to the
   * <code>BigDecimal(String)</code> constructor.
   * 
   * @throws NumberFormatException  if the string cannot be properly parsed.
   */
  public BigFraction(String s)
  {
    int slashPos = s.indexOf('/');
    if(slashPos < 0)
    {
      BigFraction res = new BigFraction(new BigDecimal(s));
      this.numerator = res.numerator;
      this.denominator = res.denominator;
    }
    else
    {
      BigDecimal num = new BigDecimal(s.substring(0, slashPos));
      BigDecimal den = new BigDecimal(s.substring(slashPos+1, s.length()));
      BigFraction res = new BigFraction(num, den);
      this.numerator = res.numerator;
      this.denominator = res.denominator;
    }
  }

  /**
   * Returns this + f.
   */
  public BigFraction add(BigFraction f)
  {
    if(f == null)
      throw new IllegalArgumentException("Null argument");

    //n1/d1 + n2/d2 = (n1*d2 + d1*n2)/(d1*d2) 
    return new BigFraction(numerator.multiply(f.denominator).add(denominator.multiply(f.numerator)),
                           denominator.multiply(f.denominator));
  }

  /**
   * Returns this + b.
   */
  public BigFraction add(BigInteger b)
  {
    if(b == null)
      throw new IllegalArgumentException("Null argument");

    //n1/d1 + n2 = (n1 + d1*n2)/d1
    return new BigFraction(numerator.add(denominator.multiply(b)),
                           denominator, true);
  }

  /**
   * Returns this + n.
   */
  public BigFraction add(long n)
  {
    return add(BigInteger.valueOf(n));
  }

  /**
   * Returns this - f.
   */
  public BigFraction subtract(BigFraction f)
  {
    if(f == null)
      throw new IllegalArgumentException("Null argument");

    return new BigFraction(numerator.multiply(f.denominator).subtract(denominator.multiply(f.numerator)),
                           denominator.multiply(f.denominator));
  }

  /**
   * Returns this - b.
   */
  public BigFraction subtract(BigInteger b)
  {
    if(b == null)
      throw new IllegalArgumentException("Null argument");

    return new BigFraction(numerator.subtract(denominator.multiply(b)),
                           denominator, true);
  }

  /**
   * Returns this - n.
   */
  public BigFraction subtract(long n)
  {
    return subtract(BigInteger.valueOf(n));
  }

  /**
   * Returns this * f.
   */
  public BigFraction multiply(BigFraction f)
  {
    if(f == null)
      throw new IllegalArgumentException("Null argument");

    return new BigFraction(numerator.multiply(f.numerator), denominator.multiply(f.denominator));
  }

  /**
   * Returns this * b.
   */
  public BigFraction multiply(BigInteger b)
  {
    if(b == null)
      throw new IllegalArgumentException("Null argument");

    return new BigFraction(numerator.multiply(b), denominator);
  }

  /**
   * Returns this * n.
   */
  public BigFraction multiply(long n)
  {
    return multiply(BigInteger.valueOf(n));
  }

  /**
   * Returns this / f.
   */
  public BigFraction divide(BigFraction f)
  {
    if(f == null)
      throw new IllegalArgumentException("Null argument");

    if(f.numerator.equals(BigInteger.ZERO))
      throw new ArithmeticException("Divide by zero");

    return new BigFraction(numerator.multiply(f.denominator), denominator.multiply(f.numerator));
  }

  /**
   * Returns this / b.
   */
  public BigFraction divide(BigInteger b)
  {
    if(b == null)
      throw new IllegalArgumentException("Null argument");

    if(b.equals(BigInteger.ZERO))
      throw new ArithmeticException("Divide by zero");

    return new BigFraction(numerator, denominator.multiply(b));
  }

  /**
   * Returns this / n.
   */
  public BigFraction divide(long n)
  {
    return divide(BigInteger.valueOf(n));
  }

  /**
   * Returns this^exponent.
   */
  public BigFraction pow(int exponent)
  {
    if(exponent == 0)
      return BigFraction.ONE;
    else if (exponent == 1)
      return this;
    else if (exponent < 0)
      return new BigFraction(denominator.pow(-exponent), numerator.pow(-exponent), true);
    else
      return new BigFraction(numerator.pow(exponent), denominator.pow(exponent), true);
  }

  /**
   * Returns 1/this.
   */
  public BigFraction reciprocal()
  {
    if(this.numerator.equals(BigInteger.ZERO))
      throw new ArithmeticException("Divide by zero");

    return new BigFraction(denominator, numerator, true);
  }

  /**
   * Returns the complement of this fraction, which is equal to 1 - this.
   * Useful for probabilities/statistics.

   */
  public BigFraction complement()
  {
    return new BigFraction(denominator.subtract(numerator), denominator, true);
  }

  /**
   * Returns -this.
   */
  public BigFraction negate()
  {
    return new BigFraction(numerator.negate(), denominator, true);
  }

  /**
   * Returns -1, 0, or 1, representing the sign of this fraction.
   */
  public int signum()
  {
    return numerator.signum();
  }

  /**
   * Returns the absolute value of this.
   */
  public BigFraction abs()
  {
    return (signum() < 0 ? negate() : this);
  }

  /**
   * Returns a string representation of this, in the form
   * numerator/denominator.
   */
  public String toString()
  {
    return numerator.toString() + "/" + denominator.toString();
  }

  /**
   * Returns if this object is equal to another object.
   */
  public boolean equals(Object o)
  {
    if(!(o instanceof BigFraction))
      return false;

    BigFraction f = (BigFraction)o;
    return numerator.equals(f.numerator) && denominator.equals(f.denominator);
  }

  /**
   * Returns a hash code for this object.
   */
  public int hashCode()
  {
    //using the method generated by Eclipse, but streamlined a bit..
    return (31 + numerator.hashCode())*31 + denominator.hashCode();
  }

  /**
   * Returns a negative, zero, or positive number, indicating if this object
   * is less than, equal to, or greater than f, respectively.
   */
  public int compareTo(BigFraction f)
  {
    if(f == null)
      throw new IllegalArgumentException("Null argument");

    //easy case: this and f have different signs
    if(signum() != f.signum())
      return signum() - f.signum();

    //next easy case: this and f have the same denominator
    if(denominator.equals(f.denominator))
      return numerator.compareTo(f.numerator);

    //not an easy case, so first make the denominators equal then compare the numerators 
    return numerator.multiply(f.denominator).compareTo(denominator.multiply(f.numerator));
  }

  /**
   * Returns the smaller of this and f.
   */
  public BigFraction min(BigFraction f)
  {
    if(f == null)
      throw new IllegalArgumentException("Null argument");

    return (this.compareTo(f) <= 0 ? this : f);
  }

  /**
   * Returns the maximum of this and f.
   */
  public BigFraction max(BigFraction f)
  {
    if(f == null)
      throw new IllegalArgumentException("Null argument");

    return (this.compareTo(f) >= 0 ? this : f);
  }

  /**
   * Returns a positive BigFraction, greater than or equal to zero, and less than one.
   */
  public static BigFraction random()
  {
    return new BigFraction(Math.random());
  }

  public final BigInteger getNumerator() { return numerator; }
  public final BigInteger getDenominator() { return denominator; }

  //implementation of Number class.  may cause overflow.
  public byte   byteValue()   { return (byte) Math.max(Byte.MIN_VALUE,    Math.min(Byte.MAX_VALUE,    longValue())); }
  public short  shortValue()  { return (short)Math.max(Short.MIN_VALUE,   Math.min(Short.MAX_VALUE,   longValue())); }
  public int    intValue()    { return (int)  Math.max(Integer.MIN_VALUE, Math.min(Integer.MAX_VALUE, longValue())); }
  public long   longValue()   { return Math.round(doubleValue()); }
  public float  floatValue()  { return (float)doubleValue(); }
  public double doubleValue() { return toBigDecimal(18).doubleValue(); }

  /**
   * Returns a BigDecimal representation of this fraction.  If possible, the
   * returned value will be exactly equal to the fraction.  If not, the BigDecimal
   * will have a scale large enough to hold the same number of significant figures
   * as both numerator and denominator, or the equivalent of a double-precision
   * number, whichever is more.
   */
  public BigDecimal toBigDecimal()
  {
    //Implementation note:  A fraction can be represented exactly in base-10 iff its
    //denominator is of the form 2^a * 5^b, where a and b are nonnegative integers.
    //(In other words, if there are no prime factors of the denominator except for
    //2 and 5, or if the denominator is 1).  So to determine if this denominator is
    //of this form, continually divide by 2 to get the number of 2's, and then
    //continually divide by 5 to get the number of 5's.  Afterward, if the denominator
    //is 1 then there are no other prime factors.

    //Note: number of 2's is given by the number of trailing 0 bits in the number
    int twos = denominator.getLowestSetBit();
    BigInteger tmpDen = denominator.shiftRight(twos); // x / 2^n === x >> n

    final BigInteger FIVE = BigInteger.valueOf(5);
    int fives = 0;
    BigInteger[] divMod = null;

    //while(tmpDen % 5 == 0) { fives++; tmpDen /= 5; }
    while(BigInteger.ZERO.equals((divMod = tmpDen.divideAndRemainder(FIVE))[1]))
    {
      fives++;
      tmpDen = divMod[0];
    }

    if(BigInteger.ONE.equals(tmpDen))
    {
      //This fraction will terminate in base 10, so it can be represented exactly as
      //a BigDecimal.  We would now like to make the fraction of the form
      //unscaled / 10^scale.  We know that 2^x * 5^x = 10^x, and our denominator is
      //in the form 2^twos * 5^fives.  So use max(twos, fives) as the scale, and
      //multiply the numerator and deminator by the appropriate number of 2's or 5's
      //such that the denominator is of the form 2^scale * 5^scale.  (Of course, we
      //only have to actually multiply the numerator, since all we need for the
      //BigDecimal constructor is the scale.
      BigInteger unscaled = numerator;
      int scale = Math.max(twos, fives);

      if(twos < fives)
        unscaled = unscaled.shiftLeft(fives - twos); //x * 2^n === x << n
      else if (fives < twos)
        unscaled = unscaled.multiply(FIVE.pow(twos - fives));

      return new BigDecimal(unscaled, scale);
    }

    //else: this number will repeat infinitely in base-10.  So try to figure out
    //a good number of significant digits.  Start with the number of digits required
    //to represent the numerator and denominator in base-10, which is given by
    //bitLength / log[2](10).  (bitLenth is the number of digits in base-2).
    final double LG10 = 3.321928094887362; //Precomputed ln(10)/ln(2), a.k.a. log[2](10)
    int precision = Math.max(numerator.bitLength(), denominator.bitLength());
    precision = (int)Math.ceil(precision / LG10);

    //If the precision is less than 18 digits, use 18 digits so that the number
    //will be at least as accurate as a cast to a double.  For example, with
    //the fraction 1/3, precision will be 1, giving a result of 0.3.  This is
    //quite a bit different from what a user would expect.
    if(precision < 18)
      precision = 18;

    return toBigDecimal(precision);
  }

  /**
   * Returns a BigDecimal representation of this fraction, with a given precision.
   * @param precision  the number of significant figures to be used in the result.
   */
  public BigDecimal toBigDecimal(int precision)
  {
    return new BigDecimal(numerator).divide(new BigDecimal(denominator), new MathContext(precision, RoundingMode.HALF_EVEN));
  }

  //--------------------------------------------------------------------------
  //  PRIVATE FUNCTIONS
  //--------------------------------------------------------------------------

  /**
   * Private constructor, used when you can be certain that the fraction is already in
   * lowest terms.  No check is done to reduce numerator/denominator.  A check is still
   * done to maintain a positive denominator.
   * 
   * @param throwaway  unused variable, only here to signal to the compiler that this
   *                   constructor should be used.
   */
  private BigFraction(BigInteger numerator, BigInteger denominator, boolean throwaway)
  {
    if(denominator.signum() < 0)
    {
      this.numerator = numerator.negate();
      this.denominator = denominator.negate();
    }
    else
    {
      this.numerator = numerator;
      this.denominator = denominator;
    }
  }

}

- Kip

---

如果参数为null，则抛出NullPointerException。实际上，代码将自动执行此操作，因此您的检查（以及替换为IllegalArgumentException）是不必要的代码膨胀。 - cletus

我不同意；如果另一个用户在没有查看我的源代码的情况下使用此类，并且出现NullPointerException，他会认为是 我的代码有bug。但是IllegalArgumentException表明他违反了javado c所暗示的契约（即使我没有明确说明）。 - Kip

https://dev59.com/HHVD5IYBdhLWcg3wXaYd - cletus

只是一个问题，Commons Math 中的 Fraction 和 BigFraction 有什么问题？ - Mortimer

@Mortimer：不确定，我从未看过它。 - Kip

Answer 2

• 将其变成不可变对象；
• 将其变成规范形式，也就是将6/4变成3/2（最大公约数算法对此很有用）;
• 称之为Rational，因为您所表示的是一个有理数;
• 您可以使用BigInteger来存储任意精度的值。如果不使用，则使用long，它具有更简单的实现方式；
• 始终将分母设为正数。符号应由分子携带；
• 扩展Number；
• 实现Comparable<T>；
• 实现equals()和hashCode()；
• 为由String表示的数字添加工厂方法；
• 添加一些便捷的工厂方法；
• 添加一个toString()；并且
• 将其设为Serializable。

实际上，试试这个大小是否合适。它可以运行，但可能存在一些问题：

public class BigRational extends Number implements Comparable<BigRational>, Serializable {
    public final static BigRational ZERO = new BigRational(BigInteger.ZERO, BigInteger.ONE);
    private final static long serialVersionUID = 1099377265582986378L;

    private final BigInteger numerator, denominator;

    private BigRational(BigInteger numerator, BigInteger denominator) {
        this.numerator = numerator;
        this.denominator = denominator;
    }

    private static BigRational canonical(BigInteger numerator, BigInteger denominator, boolean checkGcd) {
        if (denominator.signum() == 0) {
            throw new IllegalArgumentException("denominator is zero");
        }
        if (numerator.signum() == 0) {
            return ZERO;
        }
        if (denominator.signum() < 0) {
            numerator = numerator.negate();
            denominator = denominator.negate();
        }
        if (checkGcd) {
            BigInteger gcd = numerator.gcd(denominator);
            if (!gcd.equals(BigInteger.ONE)) {
                numerator = numerator.divide(gcd);
                denominator = denominator.divide(gcd);
            }
        }
        return new BigRational(numerator, denominator);
    }

    public static BigRational getInstance(BigInteger numerator, BigInteger denominator) {
        return canonical(numerator, denominator, true);
    }

    public static BigRational getInstance(long numerator, long denominator) {
        return canonical(new BigInteger("" + numerator), new BigInteger("" + denominator), true);
    }

    public static BigRational getInstance(String numerator, String denominator) {
        return canonical(new BigInteger(numerator), new BigInteger(denominator), true);
    }

    public static BigRational valueOf(String s) {
        Pattern p = Pattern.compile("(-?\\d+)(?:.(\\d+)?)?0*(?:e(-?\\d+))?");
        Matcher m = p.matcher(s);
        if (!m.matches()) {
            throw new IllegalArgumentException("Unknown format '" + s + "'");
        }

        // this translates 23.123e5 to 25,123 / 1000 * 10^5 = 2,512,300 / 1 (GCD)
        String whole = m.group(1);
        String decimal = m.group(2);
        String exponent = m.group(3);
        String n = whole;

        // 23.123 => 23123
        if (decimal != null) {
            n += decimal;
        }
        BigInteger numerator = new BigInteger(n);

        // exponent is an int because BigInteger.pow() takes an int argument
        // it gets more difficult if exponent needs to be outside {-2 billion,2 billion}
        int exp = exponent == null ? 0 : Integer.valueOf(exponent);
        int decimalPlaces = decimal == null ? 0 : decimal.length();
        exp -= decimalPlaces;
        BigInteger denominator;
        if (exp < 0) {
            denominator = BigInteger.TEN.pow(-exp);
        } else {
            numerator = numerator.multiply(BigInteger.TEN.pow(exp));
            denominator = BigInteger.ONE;
        }

        // done
        return canonical(numerator, denominator, true);
    }

    // Comparable
    public int compareTo(BigRational o) {
        // note: this is a bit of cheat, relying on BigInteger.compareTo() returning
        // -1, 0 or 1.  For the more general contract of compareTo(), you'd need to do
        // more checking
        if (numerator.signum() != o.numerator.signum()) {
            return numerator.signum() - o.numerator.signum();
        } else {
            // oddly BigInteger has gcd() but no lcm()
            BigInteger i1 = numerator.multiply(o.denominator);
            BigInteger i2 = o.numerator.multiply(denominator);
            return i1.compareTo(i2); // expensive!
        }
    }

    public BigRational add(BigRational o) {
        if (o.numerator.signum() == 0) {
            return this;
        } else if (numerator.signum() == 0) {
            return o;
        } else if (denominator.equals(o.denominator)) {
            return new BigRational(numerator.add(o.numerator), denominator);
        } else {
            return canonical(numerator.multiply(o.denominator).add(o.numerator.multiply(denominator)), denominator.multiply(o.denominator), true);
        }
    }


    public BigRational multiply(BigRational o) {
        if (numerator.signum() == 0 || o.numerator.signum( )== 0) {
            return ZERO;
        } else if (numerator.equals(o.denominator)) {
            return canonical(o.numerator, denominator, true);
        } else if (o.numerator.equals(denominator)) {
            return canonical(numerator, o.denominator, true);
        } else if (numerator.negate().equals(o.denominator)) {
            return canonical(o.numerator.negate(), denominator, true);
        } else if (o.numerator.negate().equals(denominator)) {
            return canonical(numerator.negate(), o.denominator, true);
        } else {
            return canonical(numerator.multiply(o.numerator), denominator.multiply(o.denominator), true);
        }
    }

    public BigInteger getNumerator() { return numerator; }
    public BigInteger getDenominator() { return denominator; }
    public boolean isInteger() { return numerator.signum() == 0 || denominator.equals(BigInteger.ONE); }
    public BigRational negate() { return new BigRational(numerator.negate(), denominator); }
    public BigRational invert() { return canonical(denominator, numerator, false); }
    public BigRational abs() { return numerator.signum() < 0 ? negate() : this; }
    public BigRational pow(int exp) { return canonical(numerator.pow(exp), denominator.pow(exp), true); }
    public BigRational subtract(BigRational o) { return add(o.negate()); }
    public BigRational divide(BigRational o) { return multiply(o.invert()); }
    public BigRational min(BigRational o) { return compareTo(o) <= 0 ? this : o; }
    public BigRational max(BigRational o) { return compareTo(o) >= 0 ? this : o; }

    public BigDecimal toBigDecimal(int scale, RoundingMode roundingMode) {
        return isInteger() ? new BigDecimal(numerator) : new BigDecimal(numerator).divide(new BigDecimal(denominator), scale, roundingMode);
    }

    // Number
    public int intValue() { return isInteger() ? numerator.intValue() : numerator.divide(denominator).intValue(); }
    public long longValue() { return isInteger() ? numerator.longValue() : numerator.divide(denominator).longValue(); }
    public float floatValue() { return (float)doubleValue(); }
    public double doubleValue() { return isInteger() ? numerator.doubleValue() : numerator.doubleValue() / denominator.doubleValue(); }

    @Override
    public String toString() { return isInteger() ? String.format("%,d", numerator) : String.format("%,d / %,d", numerator, denominator); }

    @Override
    public boolean equals(Object o) {
        if (this == o) return true;
        if (o == null || getClass() != o.getClass()) return false;

        BigRational that = (BigRational) o;

        if (denominator != null ? !denominator.equals(that.denominator) : that.denominator != null) return false;
        if (numerator != null ? !numerator.equals(that.numerator) : that.numerator != null) return false;

        return true;
    }

    @Override
    public int hashCode() {
        int result = numerator != null ? numerator.hashCode() : 0;
        result = 31 * result + (denominator != null ? denominator.hashCode() : 0);
        return result;
    }

    public static void main(String args[]) {
        BigRational r1 = BigRational.valueOf("3.14e4");
        BigRational r2 = BigRational.getInstance(111, 7);
        dump("r1", r1);
        dump("r2", r2);
        dump("r1 + r2", r1.add(r2));
        dump("r1 - r2", r1.subtract(r2));
        dump("r1 * r2", r1.multiply(r2));
        dump("r1 / r2", r1.divide(r2));
        dump("r2 ^ 2", r2.pow(2));
    }

    public static void dump(String name, BigRational r) {
        System.out.printf("%s = %s%n", name, r);
        System.out.printf("%s.negate() = %s%n", name, r.negate());
        System.out.printf("%s.invert() = %s%n", name, r.invert());
        System.out.printf("%s.intValue() = %,d%n", name, r.intValue());
        System.out.printf("%s.longValue() = %,d%n", name, r.longValue());
        System.out.printf("%s.floatValue() = %,f%n", name, r.floatValue());
        System.out.printf("%s.doubleValue() = %,f%n", name, r.doubleValue());
        System.out.println();
    }
}

输出结果为：

r1 = 31,400
r1.negate() = -31,400
r1.invert() = 1 / 31,400
r1.intValue() = 31,400
r1.longValue() = 31,400
r1.floatValue() = 31,400.000000
r1.doubleValue() = 31,400.000000

r2 = 111 / 7
r2.negate() = -111 / 7
r2.invert() = 7 / 111
r2.intValue() = 15
r2.longValue() = 15
r2.floatValue() = 15.857142
r2.doubleValue() = 15.857143

r1 + r2 = 219,911 / 7
r1 + r2.negate() = -219,911 / 7
r1 + r2.invert() = 7 / 219,911
r1 + r2.intValue() = 31,415
r1 + r2.longValue() = 31,415
r1 + r2.floatValue() = 31,415.857422
r1 + r2.doubleValue() = 31,415.857143

r1 - r2 = 219,689 / 7
r1 - r2.negate() = -219,689 / 7
r1 - r2.invert() = 7 / 219,689
r1 - r2.intValue() = 31,384
r1 - r2.longValue() = 31,384
r1 - r2.floatValue() = 31,384.142578
r1 - r2.doubleValue() = 31,384.142857

r1 * r2 = 3,485,400 / 7
r1 * r2.negate() = -3,485,400 / 7
r1 * r2.invert() = 7 / 3,485,400
r1 * r2.intValue() = 497,914
r1 * r2.longValue() = 497,914
r1 * r2.floatValue() = 497,914.281250
r1 * r2.doubleValue() = 497,914.285714

r1 / r2 = 219,800 / 111
r1 / r2.negate() = -219,800 / 111
r1 / r2.invert() = 111 / 219,800
r1 / r2.intValue() = 1,980
r1 / r2.longValue() = 1,980
r1 / r2.floatValue() = 1,980.180176
r1 / r2.doubleValue() = 1,980.180180

r2 ^ 2 = 12,321 / 49
r2 ^ 2.negate() = -12,321 / 49
r2 ^ 2.invert() = 49 / 12,321
r2 ^ 2.intValue() = 251
r2 ^ 2.longValue() = 251
r2 ^ 2.floatValue() = 251.448975
r2 ^ 2.doubleValue() = 251.448980

- cletus

Answer 3

我正在尝试在Java中使用真分数。

Apache Commons Math 已经有一个 Fraction 类相当长的时间了。大多数时候，对于“我希望Java有像核心库中的X这样的东西！”的答案可以在 Apache Commons library 的范围内找到。

- yawmark

---

我告诉你为什么这个库的评分这么低，因为Apache Commons库不适合新手使用。首先，在该页面上没有直接下载链接（它隐藏在侧边栏菜单中），其次没有使用说明（如何将jar添加到构建路径），第三，即使我把所有东西都添加进去了，还是会出现classDefNotFound错误。所以，对于我们这些只会复制和粘贴的人来说，你得不到赞成票。 - Noumenon

@Noumenon，使用任何构建管理器（例如Maven），然后在POM中添加依赖项怎么样？ - eugene.polschikov

我希望能看到一个针对新手的小提示，“如何在你的项目中使用它”。可以将这个建议加入其中。话虽如此，我确实想出了如何做，并在我的工厂应用程序中使用了它，因为我需要显示英寸的分数，但我没有回来给你点赞。所以感谢你，这里是我迟来的点赞。 - Noumenon

这是公正的反馈。这里也是我迟来的感谢！ :) - yawmark

这个很容易使用。 - Eric

Answer 4

请将其变成不可变类型！一个分数的值不会改变——例如，一半不会变成三分之一。你可以使用withDenominator替代setDenominator，它返回一个新的分数，该分数具有相同的分子但指定了分母。

使用不可变类型可以使生活变得轻松得多。

重写equals和hashcode也是明智的选择，这样它就可以在映射和集合中使用。Outlaw Programmer关于算术运算符和字符串格式化的观点也很好。

作为一个一般性的指南，请查看BigInteger和BigDecimal。它们没有做同样的事情，但它们足够相似，可以给您提供很好的思路。

- Jon Skeet

---

请将其设置为不可变类型！ 分数的值不会改变-例如，一半不会变成三分之一。列表/元组/向量（1、2、3、4）也不会变成值（4、3、2、1），但似乎大多数人并不介意列表会改变状态。不是说我不赞同分数的不可变性，但它需要更好的论据。 分数感觉更像是一个值而不是一堆状态。 程序员的期望是否是正确的指导原因？ 我不是100％确定，但它听起来像个好主意。 - Jonas Kölker

在现实生活中，列表确实会改变：你如何编写购物清单？你从一张空白纸开始，然后在上面写下东西。即使进行到一半，你仍然称之为“购物清单”。话虽如此，函数式编程确实努力使列表不可变... - Jon Skeet

Answer 5

首先，我会除去设置器并使分数不可变。

你可能还想要添加加、减等方法，并可能需要一些获取不同字符串格式的表示方法。

编辑：我可能会将字段标记为“final”以表明我的意图，但我想这并不是大问题...

- Outlaw Programmer

---

我想知道我们最终会得到多少个“使其不可变”的答案 :) - Jon Skeet

Answer 6

• 如果没有像add()和multiply()等算术方法，它就有点毫无意义了。
• 你一定要重写equals()和hashCode()方法。
• 你应该添加一个将分数标准化的方法，或者自动执行标准化。思考一下是否希望1/2和2/4被视为相同 - 这会影响equals()、hashCode()和compareTo()方法。

- Michael Borgwardt

Answer 7

我需要将它们从小到大排序，因此最终我也需要将它们表示为double类型

这并非必需。实际上，如果您想正确处理相等性，不要依赖double正常工作。如果b*d为正数，则当ad < bc时，a/b < c/d。如果涉及负整数，则可以适当处理...

我可能会重写为：

public int compareTo(Fraction frac)
{
    // we are comparing this=a/b with frac=c/d 
    // by multiplying both sides by bd.
    // If bd is positive, then a/b < c/d <=> ad < bc.
    // If bd is negative, then a/b < c/d <=> ad > bc.
    // If bd is 0, then you've got other problems (either b=0 or d=0)
    int d = frac.getDenominator();
    long ad = (long)this.numerator * d;
    long bc = (long)this.denominator * frac.getNumerator();
    long diff = ((long)d*this.denominator > 0) ? (ad-bc) : (bc-ad);
    return (diff > 0 ? 1 : (diff < 0 ? -1 : 0));
}

使用long是为了确保两个大的int相乘时不会溢出。如果您可以保证分母始终为非负数（如果为负数，只需取反分子和分母），那么可以摆脱检查b*d是否为正数并节省一些步骤。不确定零分母的行为是什么。

不确定使用双倍比较的性能如何。 （也就是说，如果您非常关心性能）这是我用来检查的测试方法。（似乎工作正常。）

public static void main(String[] args)
{
    int a = Integer.parseInt(args[0]);
    int b = Integer.parseInt(args[1]);
    int c = Integer.parseInt(args[2]);
    int d = Integer.parseInt(args[3]);
    Fraction f1 = new Fraction(a,b); 
    Fraction f2 = new Fraction(c,d);
    int rel = f1.compareTo(f2);
    String relstr = "<=>";
    System.out.println(a+"/"+b+" "+relstr.charAt(rel+1)+" "+c+"/"+d);
}

（附注：您可能考虑重构以实现 Comparable 或 Comparator 用于您的类。）

- Jason S

---

如果例如a = 1，b = 3，c = -2，d = -3，则这并不是真的。 如果b和d为正数，则当且仅当ad < bc时，a / b < c / d才为真。 - Luke Woodward

啊，我把资格条件搞错了。(谢谢!) 条件应该是如果bd>0。 - Jason S

真的。更准确地说，当bd>0时，a/b < c/d <=> ac < bd成立。如果bd<0，则反之成立。（如果bd = 0，则你有一个糟糕的分数。 :-)） - Paul Brinkley

对不起，你的理解有误。当我们说 a/b < c/d 时，实际上是指 ad < bc，前提是 bd>0。 （我在代码注释中第一次就理解正确了！） - Jason S

Answer 8

一个非常微小的改进可能是保存你计算的双精度浮点数，这样你只需要在第一次访问时计算它。除非你经常访问此数字，否则这并不会带来很大的收益，但这也不是太难做到。

另外一个要点可能是你在分母中进行的错误检查...你自动将0更改为1。我不确定这是否适用于你的特定应用程序，但通常情况下，如果有人试图除以0，那么肯定出了什么问题。我会让它抛出一个异常（如果你觉得需要的话可以使用专门的异常），而不是以一种似乎是任意的方式更改值，这种方式用户不知道。

与其他评论相比，关于添加加减等方法...既然你没有提到需要它们，我就认为你不需要。除非你正在构建一个真正会被许多地方或其他人使用的库，否则遵循YAGNI原则（你不需要它，所以它不应该存在）。

- Beska

---

他有 getNumerator() 和 getDenominator() 方法，这让我相信他正在创建此类之外的新分数。如果存在这种逻辑，那么它可能应该在这里。 - Tim Frey

+1 在分母中悄悄地将0变为1是一场灾难的配方。 - maaartinus

Answer 9

有几种方法可以改善这个或任何值类型：

• 将您的类设置为不可变，包括使分子和分母成为final
• 自动将分数转换为规范形式，例如2/4 -> 1/2
• 实现toString()
• 实现“public static Fraction valueOf(String s)”以从字符串转换为分数。实现类似的工厂方法以从int、double等转换。
• 实现加法、乘法等
• 添加从整数构造函数
• 覆盖equals/hashCode
• 考虑将Fraction作为接口，并提供在必要时切换到BigInteger的实现
• 考虑子类化Number
• 考虑包含常用值的命名常量，如0和1
• 考虑使其可序列化
• 测试除以零的情况
• 记录您的API

基本上，查看其他值类的API，例如Double、Integer等，并按照它们的方式进行操作 :)

- Dave Ray

Answer 10

如果你把一个分数的分子和另一个分数的分母相乘，再把另一个分数的分子和第一个分数的分母相乘，就会得到两个分数（它们仍然是相同的值），这两个分数有相同的分母，你可以直接比较它们的分子。因此，你不需要计算双倍值：

public int compareTo(Fraction frac) {
    int t = this.numerator * frac.getDenominator();
    int f = frac.getNumerator() * this.denominator;
    if(t>f) return 1;
    if(f>t) return -1;
    return 0;
}

- Francisco Canedo

---

如果frac.getDenominator()和this.denominator有相反的符号，则此操作将失败（请参见我的帖子）。另外，您必须注意乘法可能会溢出的事实。 - Jason S

啊，是的，那是真的。但在这种情况下，我更喜欢Kip的实现，至少我能理解它。;) - Francisco Canedo

我想指出，在我的实现中，只有分子可以为负数。我还使用了BigIntegers，因此永远不会发生溢出（当然，这会牺牲一些性能）。 - Kip