An Unbounded Decimal Floating-Point Type

Doc No: P2159R0
Date: 2020-04-23
Author:  Bill Seymour
Reply to:
Audience: LEWG-I, SG6 (Numerics)

Is there any interest in an unbounded decimal floating-point type for the numbers TS proposed in P1889?

(The author intends to implement an open-source version for his own use in any event. Should he write the standardese?)

This paper uses from P1889R1 enum class rounding (§3.3), integer (§9.3), and decimal128 (IEEE 754 decimal floating-point, future §8.5?).

The “scale” is the number of digits to the right of the decimal point. The rounding mode defaults to tie_to_even but can be changed at any time.

Note the arguments in P2010 for std::format, std::to_chars and std::from_chars which we might want to add. (P2010’s arguments against the I/O operators probably don’t apply since those operators could be written to use the numpunct facet from the stream’s imbued locale, and they could be optional in freestanding environments.)

The basic idea:

(std:: qualifiers omitted for simplicity)

class decimal
  // As a practical matter, we’ll need to limit the number of fractional digits
  // to something finite.
    static constexpr int max_scale = implementation-defined;

  // Constructors, destructor:
    constexpr decimal() noexcept;

    decimal(long double); // NB: implicit

    decimal(const integer& unscaled_value, int scale);
    decimal(integer&& unscaled_value, int scale) noexcept;

    // as if in "C" locale:
    template<class Ch, class Tr, class A>
      explicit decimal(const basic_string<Ch,Tr,A>&, int radix = 10);

    ~decimal() noexcept;

  // Copyable, noexcept-moveable, noexcept-swappable:
    decimal(const decimal&);
    decimal(decimal&&) noexcept;

    decimal& operator=(const decimal&);
    decimal& operator=(decimal&&) noexcept;

    void swap(decimal&) noexcept;

  // Capacity in units of decimal digits:
    size_t size() const noexcept;
    size_t capacity() const noexcept;
    void reserve(size_t digits);
    void shrink_to_fit();

  // Conversions:
    explicit operator bool() const noexcept;
    explicit operator integer() const;
    explicit operator long double() const noexcept;
    explicit operator decimal128() const noexcept;

    // as if in "C" locale:
    string to_string(int scale = -1, // < 0 means exact
                     int radix = 10) const;

  // Observers:
    const integer& unscaled_value() const noexcept;
    int scale() const noexcept;
    rounding rounding_mode() const noexcept;

    bool is_zero() const noexcept;
    bool is_neg() const noexcept;

    decimal integer_part() const;    // rounding::all_to_zero
    decimal fractional_part() const; // *this - integer_part()
    decimal modf(decimal*) const;    // after modf(double,double*)

  // Modifiers:
    void rescale(int new_scale) noexcept; // shifts decimal point in O(1) time
    void reset_rounding(rounding = rounding::tie_to_even) noexcept;

  // Comparisons:
    int compare(const decimal&) const noexcept;

  // Member operators:
    decimal& operator++();
    decimal& operator--();

    decimal  operator++(int);
    decimal  operator--(int);

    decimal& operator+=(const decimal&);
    decimal& operator-=(const decimal&);
    decimal& operator*=(const decimal&);
    decimal& operator/=(const decimal&);

  // Other arithmetic operations:
    decimal& abs() noexcept;
    decimal& negate() noexcept;
    decimal& sqrt();
    decimal& pow(const decimal& exponent);
    decimal& remainder(const decimal& divisor);
void swap(decimal&, decimal&) noexcept;

// Non-member operators:
decimal operator+(const decimal&);
decimal operator+(decimal&&) noexcept;

decimal operator-(const decimal&);
decimal operator-(decimal&&) noexcept;

decimal operator+(const decimal&, const decimal&);
decimal operator-(const decimal&, const decimal&);
decimal operator*(const decimal&, const decimal&);
decimal operator/(const decimal&, const decimal&);

// Other non-member arithmetic operations:
decimal abs(const decimal&);
decimal abs(decimal&&) noexcept;

decimal sqrt(const decimal&);

decimal pow(const decimal& base, const decimal& exponent);

decimal remainder(const decimal& dividend, const decimal& divisor);

// I/O:
template<class Ch, class Tr>
  basic_ostream<Ch,Tr>& operator<<(basic_ostream<Ch,Tr>&, const decimal&);

template<class Ch, class Tr>
  basic_istream<Ch,Tr>& operator>>(basic_istream<Ch,Tr>&, decimal&);

All corrections and suggestions will be welcome. All flames will be amusing.
Mail to: