bigdecimal 3.1.8 → 4.1.2

data/lib/bigdecimal/newton.rb

@@ -2,6 +2,8 @@
 require "bigdecimal/ludcmp"
 require "bigdecimal/jacobian"
 
+warn "'bigdecimal/newton' is deprecated and will be removed in a future release."
+
 #
 # newton.rb
 #

data/lib/bigdecimal/util.rb

@@ -1,4 +1,4 @@
-# frozen_string_literal: false
+# frozen_string_literal: true
 #
 #--
 # bigdecimal/util extends various native classes to provide the #to_d method,
@@ -119,8 +119,11 @@ class Rational < Numeric
   #
   # Returns the value as a BigDecimal.
   #
-  # The required +precision+ parameter is used to determine the number of
-  # significant digits for the result.
+  # The +precision+ parameter is used to determine the number of
+  # significant digits for the result. When +precision+ is set to +0+,
+  # the number of digits to represent the float being converted is determined
+  # automatically.
+  # The default +precision+ is +0+.
   #
   #     require 'bigdecimal'
   #     require 'bigdecimal/util'
@@ -129,7 +132,7 @@ class Rational < Numeric
   #
   # See also Kernel.BigDecimal.
   #
-  def to_d(precision)
+  def to_d(precision=0)
     BigDecimal(self, precision)
   end
 end
@@ -141,29 +144,27 @@ class Complex < Numeric
   #     cmp.to_d(precision)  -> bigdecimal
   #
   # Returns the value as a BigDecimal.
+  # If the imaginary part is not +0+, an error is raised
   #
-  # The +precision+ parameter is required for a rational complex number.
-  # This parameter is used to determine the number of significant digits
-  # for the result.
+  # The +precision+ parameter is used to determine the number of
+  # significant digits for the result. When +precision+ is set to +0+,
+  # the number of digits to represent the float being converted is determined
+  # automatically.
+  # The default +precision+ is +0+.
   #
   #     require 'bigdecimal'
   #     require 'bigdecimal/util'
   #
   #     Complex(0.1234567, 0).to_d(4)   # => 0.1235e0
   #     Complex(Rational(22, 7), 0).to_d(3)   # => 0.314e1
+  #     Complex(1, 1).to_d   # raises ArgumentError
   #
   # See also Kernel.BigDecimal.
   #
-  def to_d(*args)
-    BigDecimal(self) unless self.imag.zero? # to raise eerror
+  def to_d(precision=0)
+    BigDecimal(self) unless self.imag.zero? # to raise error
 
-    if args.length == 0
-      case self.real
-      when Rational
-        BigDecimal(self.real) # to raise error
-      end
-    end
-    self.real.to_d(*args)
+    BigDecimal(self.real, precision)
   end
 end
 

data/lib/bigdecimal.rb

@@ -1,5 +1,404 @@
 if RUBY_ENGINE == 'jruby'
   JRuby::Util.load_ext("org.jruby.ext.bigdecimal.BigDecimalLibrary")
+
+  class BigDecimal
+    def _decimal_shift(i) # :nodoc:
+      to_java.move_point_right(i).to_d
+    end
+  end
 else
   require 'bigdecimal.so'
 end
+
+class BigDecimal
+  module Internal # :nodoc:
+    # Default extra precision for intermediate calculations
+    # This value is currently the same as BigDecimal.double_fig, but defined separately for future changes.
+    EXTRA_PREC = 16
+
+    # Coerce x to BigDecimal with the specified precision.
+    # TODO: some methods (example: BigMath.exp) require more precision than specified to coerce.
+    def self.coerce_to_bigdecimal(x, prec, method_name) # :nodoc:
+      case x
+      when BigDecimal
+        return x
+      when Integer, Float
+        return BigDecimal(x, 0)
+      when Rational
+        return BigDecimal(x, [prec, 2 * BigDecimal.double_fig].max)
+      end
+      raise ArgumentError, "#{x.inspect} can't be coerced into BigDecimal"
+    end
+
+    def self.coerce_validate_prec(prec, method_name, accept_zero: false) # :nodoc:
+      unless Integer === prec
+        original = prec
+        # Emulate Integer.try_convert for ruby < 3.1
+        if prec.respond_to?(:to_int)
+          prec = prec.to_int
+        else
+          raise TypeError, "no implicit conversion of #{original.class} into Integer"
+        end
+        raise TypeError, "can't convert #{original.class} to Integer" unless Integer === prec
+      end
+
+      if accept_zero
+        raise ArgumentError, "Negative precision for #{method_name}" if prec < 0
+      else
+        raise ArgumentError, "Zero or negative precision for #{method_name}" if prec <= 0
+      end
+      prec
+    end
+
+    def self.infinity_computation_result # :nodoc:
+      if BigDecimal.mode(BigDecimal::EXCEPTION_ALL).anybits?(BigDecimal::EXCEPTION_INFINITY)
+        raise FloatDomainError, "Computation results in 'Infinity'"
+      end
+      BigDecimal::INFINITY
+    end
+
+    def self.underflow_computation_result # :nodoc:
+      if BigDecimal.mode(BigDecimal::EXCEPTION_ALL).anybits?(BigDecimal::EXCEPTION_UNDERFLOW)
+        raise FloatDomainError, 'Exponent underflow'
+      end
+      BigDecimal(0)
+    end
+
+    def self.nan_computation_result # :nodoc:
+      if BigDecimal.mode(BigDecimal::EXCEPTION_ALL).anybits?(BigDecimal::EXCEPTION_NaN)
+        raise FloatDomainError, "Computation results to 'NaN'"
+      end
+      BigDecimal::NAN
+    end
+
+    # Iteration for Newton's method with increasing precision
+    def self.newton_loop(prec, initial_precision: BigDecimal.double_fig / 2, safe_margin: 2) # :nodoc:
+      precs = []
+      while prec > initial_precision
+        precs << prec
+        prec = (precs.last + 1) / 2 + safe_margin
+      end
+      precs.reverse_each do |p|
+        yield p
+      end
+    end
+
+    # Calculates Math.log(x.to_f) considering large or small exponent
+    def self.float_log(x) # :nodoc:
+      Math.log(x._decimal_shift(-x.exponent).to_f) + x.exponent * Math.log(10)
+    end
+
+    # Calculating Taylor series sum using binary splitting method
+    # Calculates f(x) = (x/d0)*(1+(x/d1)*(1+(x/d2)*(1+(x/d3)*(1+...))))
+    # x.n_significant_digits or ds.size must be small to be performant.
+    def self.taylor_sum_binary_splitting(x, ds, prec) # :nodoc:
+      fs = ds.map {|d| [0, BigDecimal(d)] }
+      # fs = [[a0, a1], [b0, b1], [c0, c1], ...]
+      # f(x) = a0/a1+(x/a1)*(1+b0/b1+(x/b1)*(1+c0/c1+(x/c1)*(1+d0/d1+(x/d1)*(1+...))))
+      while fs.size > 1
+        # Merge two adjacent fractions
+        # from: (1 + a0/a1 + x/a1 * (1 + b0/b1 + x/b1 * rest))
+        # to:   (1 + (a0*b1+x*(b0+b1))/(a1*b1) + (x*x)/(a1*b1) * rest)
+        xn = xn ? xn.mult(xn, prec) : x
+        fs = fs.each_slice(2).map do |(a, b)|
+          b ||= [0, BigDecimal(1)._decimal_shift([xn.exponent, 0].max + 2)]
+          [
+            (a[0] * b[1]).add(xn * (b[0] + b[1]), prec),
+            a[1].mult(b[1], prec)
+          ]
+        end
+      end
+      BigDecimal(fs[0][0]).div(fs[0][1], prec)
+    end
+  end
+
+  #  call-seq:
+  #    self ** other -> bigdecimal
+  #
+  #  Returns the \BigDecimal value of +self+ raised to power +other+:
+  #
+  #    b = BigDecimal('3.14')
+  #    b ** 2              # => 0.98596e1
+  #    b ** 2.0            # => 0.98596e1
+  #    b ** Rational(2, 1) # => 0.98596e1
+  #
+  #  Related: BigDecimal#power.
+  #
+  def **(y)
+    case y
+    when BigDecimal, Integer, Float, Rational
+      power(y)
+    when nil
+      raise TypeError, 'wrong argument type NilClass'
+    else
+      x, y = y.coerce(self)
+      x**y
+    end
+  end
+
+  # call-seq:
+  #   power(n)
+  #   power(n, prec)
+  #
+  # Returns the value raised to the power of n.
+  #
+  # Also available as the operator **.
+  #
+  def power(y, prec = 0)
+    prec = Internal.coerce_validate_prec(prec, :power, accept_zero: true)
+    x = self
+    y = Internal.coerce_to_bigdecimal(y, prec.nonzero? || n_significant_digits, :power)
+
+    return Internal.nan_computation_result if x.nan? || y.nan?
+    return BigDecimal(1) if y.zero?
+
+    if y.infinite?
+      if x < 0
+        return BigDecimal(0) if x < -1 && y.negative?
+        return BigDecimal(0) if x > -1 && y.positive?
+        raise Math::DomainError, 'Result undefined for negative base raised to infinite power'
+      elsif x < 1
+        return y.positive? ? BigDecimal(0) : BigDecimal::Internal.infinity_computation_result
+      elsif x == 1
+        return BigDecimal(1)
+      else
+        return y.positive? ? BigDecimal::Internal.infinity_computation_result : BigDecimal(0)
+      end
+    end
+
+    if x.infinite? && y < 0
+      # Computation result will be +0 or -0. Avoid overflow.
+      neg = x < 0 && y.frac.zero? && y % 2 == 1
+      return neg ? -BigDecimal(0) : BigDecimal(0)
+    end
+
+    if x.zero?
+      return BigDecimal(1) if y.zero?
+      return BigDecimal(0) if y > 0
+      if y.frac.zero? && y % 2 == 1 && x.sign == -1
+        return -BigDecimal::Internal.infinity_computation_result
+      else
+        return BigDecimal::Internal.infinity_computation_result
+      end
+    elsif x < 0
+      if y.frac.zero?
+        if y % 2 == 0
+          return (-x).power(y, prec)
+        else
+          return -(-x).power(y, prec)
+        end
+      else
+        raise Math::DomainError, 'Computation results in complex number'
+      end
+    elsif x == 1
+      return BigDecimal(1)
+    end
+
+    limit = BigDecimal.limit
+    frac_part = y.frac
+
+    if frac_part.zero? && prec.zero? && limit.zero?
+      # Infinite precision calculation for `x ** int` and `x.power(int)`
+      int_part = y.fix.to_i
+      int_part = -int_part if (neg = int_part < 0)
+      ans = BigDecimal(1)
+      n = 1
+      xn = x
+      while true
+        ans *= xn if int_part.allbits?(n)
+        n <<= 1
+        break if n > int_part
+        xn *= xn
+        # Detect overflow/underflow before consuming infinite memory
+        if (xn.exponent.abs - 1) * int_part / n >= 0x7FFFFFFFFFFFFFFF
+          return ((xn.exponent > 0) ^ neg ? BigDecimal::Internal.infinity_computation_result : BigDecimal(0)) * (int_part.even? || x > 0 ? 1 : -1)
+        end
+      end
+      return neg ? BigDecimal(1) / ans : ans
+    end
+
+    result_prec = prec.nonzero? || [x.n_significant_digits, y.n_significant_digits, BigDecimal.double_fig].max + BigDecimal.double_fig
+    result_prec = [result_prec, limit].min if prec.zero? && limit.nonzero?
+
+    prec2 = result_prec + BigDecimal::Internal::EXTRA_PREC
+
+    if y < 0
+      inv = x.power(-y, prec2)
+      return BigDecimal(0) if inv.infinite?
+      return BigDecimal::Internal.infinity_computation_result if inv.zero?
+      return BigDecimal(1).div(inv, result_prec)
+    end
+
+    if frac_part.zero? && y.exponent < Math.log(result_prec) * 5 + 20
+      # Use exponentiation by squaring if y is an integer and not too large
+      pow_prec = prec2 + y.exponent
+      n = 1
+      xn = x
+      ans = BigDecimal(1)
+      int_part = y.fix.to_i
+      while true
+        ans = ans.mult(xn, pow_prec) if int_part.allbits?(n)
+        n <<= 1
+        break if n > int_part
+        xn = xn.mult(xn, pow_prec)
+      end
+      ans.mult(1, result_prec)
+    else
+      if x > 1 && x.finite?
+        # To calculate exp(z, prec), z needs prec+max(z.exponent, 0) precision if z > 0.
+        # Estimate (y*log(x)).exponent
+        logx_exponent = x < 2 ? (x - 1).exponent : Math.log10(x.exponent).round
+        ylogx_exponent = y.exponent + logx_exponent
+        prec2 += [ylogx_exponent, 0].max
+      end
+      BigMath.exp(BigMath.log(x, prec2).mult(y, prec2), result_prec)
+    end
+  end
+
+  # Returns the square root of the value.
+  #
+  # Result has at least prec significant digits.
+  #
+  def sqrt(prec)
+    prec = Internal.coerce_validate_prec(prec, :sqrt, accept_zero: true)
+    return Internal.infinity_computation_result if infinite? == 1
+
+    raise FloatDomainError, 'sqrt of negative value' if self < 0
+    raise FloatDomainError, "sqrt of 'NaN'(Not a Number)" if nan?
+    return self if zero?
+
+    if prec == 0
+      limit = BigDecimal.limit
+      prec = n_significant_digits + BigDecimal.double_fig
+      prec = [limit, prec].min if limit.nonzero?
+    end
+
+    ex = exponent / 2
+    x = _decimal_shift(-2 * ex)
+    y = BigDecimal(Math.sqrt(x.to_f), 0)
+    Internal.newton_loop(prec + BigDecimal::Internal::EXTRA_PREC) do |p|
+      y = y.add(x.div(y, p), p).div(2, p)
+    end
+    y._decimal_shift(ex).mult(1, prec)
+  end
+end
+
+# Core BigMath methods for BigDecimal (log, exp) are defined here.
+# Other methods (sin, cos, atan) are defined in 'bigdecimal/math.rb'.
+module BigMath
+  module_function
+
+  # call-seq:
+  #   BigMath.log(decimal, numeric)    -> BigDecimal
+  #
+  # Computes the natural logarithm of +decimal+ to the specified number of
+  # digits of precision, +numeric+.
+  #
+  # If +decimal+ is zero or negative, raises Math::DomainError.
+  #
+  # If +decimal+ is positive infinity, returns Infinity.
+  #
+  # If +decimal+ is NaN, returns NaN.
+  #
+  def log(x, prec)
+    prec = BigDecimal::Internal.coerce_validate_prec(prec, :log)
+    raise Math::DomainError, 'Complex argument for BigMath.log' if Complex === x
+
+    x = BigDecimal::Internal.coerce_to_bigdecimal(x, prec, :log)
+    return BigDecimal::Internal.nan_computation_result if x.nan?
+    raise Math::DomainError, 'Negative argument for log' if x < 0
+    return -BigDecimal::Internal.infinity_computation_result if x.zero?
+    return BigDecimal::Internal.infinity_computation_result if x.infinite?
+    return BigDecimal(0) if x == 1
+
+    prec2 = prec + BigDecimal::Internal::EXTRA_PREC
+
+    # Reduce x to near 1
+    if x > 1.01 || x < 0.99
+      # log(x) = log(x/exp(logx_approx)) + logx_approx
+      logx_approx = BigDecimal(BigDecimal::Internal.float_log(x), 0)
+      x = x.div(exp(logx_approx, prec2), prec2)
+    else
+      logx_approx = BigDecimal(0)
+    end
+
+    # Solve exp(y) - x = 0 with Newton's method
+    # Repeat: y -= (exp(y) - x) / exp(y)
+    y = BigDecimal(BigDecimal::Internal.float_log(x), 0)
+    exp_additional_prec = [-(x - 1).exponent, 0].max
+    BigDecimal::Internal.newton_loop(prec2) do |p|
+      expy = exp(y, p + exp_additional_prec)
+      y = y.sub(expy.sub(x, p).div(expy, p), p)
+    end
+    y.add(logx_approx, prec)
+  end
+
+  private_class_method def _exp_binary_splitting(x, prec) # :nodoc:
+    return BigDecimal(1) if x.zero?
+    # Find k that satisfies x**k / k! < 10**(-prec)
+    log10 = Math.log(10)
+    logx = BigDecimal::Internal.float_log(x.abs)
+    step = (1..).bsearch { |k| Math.lgamma(k + 1)[0] - k * logx > prec * log10 }
+    # exp(x)-1 = x*(1+x/2*(1+x/3*(1+x/4*(1+x/5*(1+...)))))
+    1 + BigDecimal::Internal.taylor_sum_binary_splitting(x, [*1..step], prec)
+  end
+
+  # call-seq:
+  #   BigMath.exp(decimal, numeric)    -> BigDecimal
+  #
+  # Computes the value of e (the base of natural logarithms) raised to the
+  # power of +decimal+, to the specified number of digits of precision.
+  #
+  # If +decimal+ is infinity, returns Infinity.
+  #
+  # If +decimal+ is NaN, returns NaN.
+  #
+  def exp(x, prec)
+    prec = BigDecimal::Internal.coerce_validate_prec(prec, :exp)
+    x = BigDecimal::Internal.coerce_to_bigdecimal(x, prec, :exp)
+    return BigDecimal::Internal.nan_computation_result if x.nan?
+    if x.infinite? || x.exponent >= 21 # exp(10**20) and exp(-10**20) overflows/underflows 64-bit exponent
+      if x.positive?
+        return BigDecimal::Internal.infinity_computation_result
+      elsif x.infinite?
+        # exp(-Infinity) is +0 by definition, this is not an underflow.
+        return BigDecimal(0)
+      else
+        return BigDecimal::Internal.underflow_computation_result
+      end
+    end
+
+    return BigDecimal(1) if x.zero?
+
+    # exp(x * 10**cnt) = exp(x)**(10**cnt)
+    cnt = x < -1 || x > 1 ? x.exponent : 0
+    prec2 = prec + BigDecimal::Internal::EXTRA_PREC + cnt
+    x = x._decimal_shift(-cnt)
+
+    # Decimal form of bit-burst algorithm
+    # Calculate exp(x.xxxxxxxxxxxxxxxx) as
+    # exp(x.xx) * exp(0.00xx) * exp(0.0000xxxx) * exp(0.00000000xxxxxxxx)
+    x = x.mult(1, prec2)
+    n = 2
+    y = BigDecimal(1)
+    BigDecimal.save_limit do
+      BigDecimal.limit(0)
+      while x != 0 do
+        partial_x = x.truncate(n)
+        x -= partial_x
+        y = y.mult(_exp_binary_splitting(partial_x, prec2), prec2)
+        n *= 2
+      end
+    end
+
+    # calculate exp(x * 10**cnt) from exp(x)
+    # exp(x * 10**k) = exp(x * 10**(k - 1)) ** 10
+    cnt.times do
+      y2 = y.mult(y, prec2)
+      y5 = y2.mult(y2, prec2).mult(y, prec2)
+      y = y5.mult(y5, prec2)
+    end
+
+    y.mult(1, prec)
+  end
+end

data/sample/linear.rb

@@ -1,6 +1,3 @@
-#!/usr/local/bin/ruby
-# frozen_string_literal: false
-
 #
 # linear.rb
 #
@@ -13,62 +10,101 @@
 
 # :stopdoc:
 require "bigdecimal"
-require "bigdecimal/ludcmp"
 
-#
-# NOTE:
-#   Change following BigDecimal.limit() if needed.
-BigDecimal.limit(100)
-#
+# Requires gem matrix
+require "matrix"
+
+class PrecisionSpecifiedValue
+  # NOTE:
+  #   Change following PREC if needed.
+
+  attr_reader :value
+  def initialize(value, prec)
+    @value = BigDecimal(value)
+    @prec = prec
+  end
+
+  def unwrap(value)
+    PrecisionSpecifiedValue === value ? value.value : value
+  end
+
+  def coerce(other)
+    [self.class.new(unwrap(other), @prec), self]
+  end
+
+  def abs
+    self.class.new(@value.abs, @prec)
+  end
+
+  def >(other)
+    @value > unwrap(other)
+  end
+
+  def <(other)
+    @value < unwrap(other)
+  end
+
+  def -(other)
+    self.class.new(@value.sub(unwrap(other), @prec), @prec)
+  end
+
+  def +(other)
+    self.class.new(@value.add(unwrap(other), @prec), @prec)
+  end
+
+  def *(other)
+    self.class.new(@value.mult(unwrap(other), @prec), @prec)
+  end
+
+  def quo(other)
+    self.class.new(@value.div(unwrap(other), @prec), @prec)
+  end
+end
+
+return if __FILE__ != $0
 
-include LUSolve
 def rd_order(na)
-   printf("Number of equations ?") if(na <= 0)
-   n = ARGF.gets().to_i
+  printf("Number of equations ?") if(na <= 0)
+  ARGF.gets().to_i
 end
 
-na   = ARGV.size
-zero = BigDecimal("0.0")
-one  = BigDecimal("1.0")
+na = ARGV.size
 
 while (n=rd_order(na))>0
   a = []
-  as= []
   b = []
   if na <= 0
      # Read data from console.
      printf("\nEnter coefficient matrix element A[i,j]\n")
      for i in 0...n do
-       for j in 0...n do
+       a << n.times.map do |j|
          printf("A[%d,%d]? ",i,j); s = ARGF.gets
-         a  << BigDecimal(s)
-         as << BigDecimal(s)
+         BigDecimal(s)
        end
        printf("Contatant vector element b[%d] ? ",i)
        b << BigDecimal(ARGF.gets)
      end
   else
-     # Read data from specified file.
-     printf("Coefficient matrix and constant vector.\n")
-     for i in 0...n do
-       s = ARGF.gets
-       printf("%d) %s",i,s)
-       s = s.split
-       for j in 0...n do
-         a  << BigDecimal(s[j])
-         as << BigDecimal(s[j])
-       end
-       b << BigDecimal(s[n])
-     end
+    # Read data from specified file.
+    printf("Coefficient matrix and constant vector.\n")
+    for i in 0...n do
+      s = ARGF.gets
+      printf("%d) %s",i,s)
+      s = s.split
+      a << n.times.map {|j| BigDecimal(s[j]) }
+      b << BigDecimal(s[n])
+    end
   end
-  x = lusolve(a,b,ludecomp(a,n,zero,one),zero)
+
+  prec = 100
+  matrix = Matrix[*a.map {|row| row.map {|v| PrecisionSpecifiedValue.new(v, prec) } }]
+  vector = b.map {|v| PrecisionSpecifiedValue.new(v, prec) }
+  x = matrix.lup.solve(vector).map(&:value)
+
   printf("Answer(x[i] & (A*x-b)[i]) follows\n")
   for i in 0...n do
      printf("x[%d]=%s ",i,x[i].to_s)
-     s = zero
-     for j in 0...n do
-       s = s + as[i*n+j]*x[j]
-     end
-     printf(" & %s\n",(s-b[i]).to_s)
+     diff = a[i].zip(x).sum {|aij, xj| aij*xj }.sub(b[i], 10)
+     printf(" & %s\n", diff.to_s)
   end
 end