Decimal numbers of arbitrary magnitude and arbitrary precision

Decimal fixed-point arithmetic.

The package decimalfp provides a Decimal number type which can represent decimal numbers of arbitrary magnitude and arbitrary precision, i.e. any number of fractional digits.

Usage

decimalfp.Decimal instances are created by giving a value (default: 0) and a precision (i.e the number of fractional digits, default: None).

>>> Decimal()
Decimal(0)

If precision is given, it must be of type int and >= 0.

>>> Decimal(5, 3)
Decimal(5, 3)
>>> Decimal(5555, -3)
ValueError: Precision must be >= 0.

If value is given, it must either be a string, an instance of numbers.Integral, number.Rational (for example fractions.Fraction), decimal.Decimal, a finite instance of numbers.Real (for example float) or be convertable to a float or an int.

The value is always adjusted to the given precision or the precision is calculated from the given value, if no precision is given.

>>> Decimal(b'12.345')
Decimal('12.345')
>>> Decimal(u'12.345')
Decimal('12.345')
>>> Decimal(u'12.345', 5)
Decimal('12.345', 5)
>>> Decimal(15, 4)
Decimal(15, 4)
>>> Decimal(10 ** 23)
Decimal(100000000000000000000000)
>>> Decimal(10 ** 48)
Decimal(1000000000000000000000000000000000000000000000000)
>>> Decimal(2 ** 48)
Decimal(281474976710656)
>>> Decimal(2 ** 98)
Decimal(316912650057057350374175801344)
>>> Decimal(Fraction(7, 56))
Decimal('0.125')
>>> Decimal(Fraction(8106479329266893, 4503599627370496), 7)
Decimal('1.8', 7)
>>> Decimal(1.8, 7)
Decimal('1.8', 7)
>>> Decimal(decimal.Decimal('-19.26', 5), 3)
Decimal('-19.26', 3)

When the given precision is lower than the precision of the given value, the result is rounded, according to the rounding mode of the current context held by the standard module decimal (which defaults to ROUND_HALF_EVEN).

>>> Decimal(u'12.345', 2)
Decimal('12.34')
>>> Decimal(u'12.3456', 3)
Decimal('12.346')
>>> Decimal(0.2, 3)
Decimal('0.2', 3)
>>> Decimal(0.2, 17)
Decimal('0.20000000000000001')
>>> Decimal(0.2, 55)
Decimal('0.200000000000000011102230246251565404236316680908203125', 55)

When no precision is given and the given value is a float or a numbers.Rational (but no Decimal), the Decimal constructor tries to convert value exactly. But, for performance reasons, this is done only up a fixed limit of fractional digits. This limit defaults to 32 and is accessible as decimalfp.LIMIT_PREC. If value can not be represented as a Decimal within this limit, an exception is raised.

>>> Decimal(0.2)
ValueError: Can't convert 0.2 exactly to Decimal.
>>> Decimal(Fraction(1, 7))
ValueError: Can't convert Fraction(1, 7) exactly to Decimal.

Decimal does not deal with infinity, division by 0 always raises a ZeroDivisionError. Likewise, infinite instances of type float or decimal.Decimal can not be converted to Decimal instances. The same is true for the ‘not a number’ instances of these types.

Computations

When importing decimalfp, its Decimal type is registered in Pythons numerical stack as number.Rational. It supports all operations defined for that base class and its instances can be mixed in computations with instances of all numeric types mentioned above.

All numerical operations give an exact result, i.e. they are not automatically constraint to the precision of the operands or to a number of significant digits (like the floating-point Decimal type from the standard module decimal). When the result can not exactly be represented by a Decimal instance within the limit given by decimalfp.LIMIT_PREC, an instance of fractions.Fraction is returned.

Addition and subtraction

Adding or subtracting Decimal instances results in a Decimal instance with a precision equal to the maximum of the precisions of the operands.

>>> Decimal('7.3') + Decimal('8.275')
Decimal('15.575')
>>> Decimal('-7.3', 4) + Decimal('8.275')
Decimal('0.975', 4)

In operations with other numerical types the precision of the result is at least equal to the precision of the involved Decimal instance, but may be greater, if neccessary. If the needed precision exceeds decimalfp.LIMIT_PREC, an instance of fractions.Fraction is returned.

>>> 0.25 + Decimal(3)
Decimal('3.25')
>>> 0.25 - Decimal(-3, 5)
Decimal('3.25', 5)
>>> 0.725 + Decimal('3')
Fraction(33551817223910195, 9007199254740992)
>>> Decimal('3') + Fraction(1, 7)
Fraction(22, 7)

Multiplication and division

Multiplying Decimal instances results in a Decimal instance with precision equal to the sum of the precisions of the operands.

>>> Decimal('5.000') * Decimal('2.5')
Decimal('12.5', 4)

Dividing Decimal instances results in a Decimal instance with precision at least equal to max(0, numerator.precision - denominator.precision), but may be greater, if needed.

>>> Decimal('5.2000') / Decimal('2.5')
Decimal('2.08', 3)
>>> Decimal('5.2003') / Decimal('2.5')
Decimal('2.08012')

In operations with other numerical types the precision of the result is at least equal to the precision of the involved Decimal instance, but may be greater, if neccessary. If the needed precision exceeds decimalfp.LIMIT_PREC, an instance of fractions.Fraction is returned.

>>> 3 * Decimal('7.5')
Decimal('22.5')
>>> Decimal(5) * 0.25
Decimal('1.25')
>>> Decimal('3') * Fraction(1, 7)
Fraction(3, 7)

Rounding

Decimal supports rounding via the built-in function round.

Note

In Python 3.x the function round uses the rounding mode ROUND_HALF_EVEN and returns an int when called with one argument, otherwise the same type as the number to be rounded.

>>> round(Decimal('12.345'))
12
>>> round(Decimal('12.345'), 2)
Decimal('12.34')
>>> round(Decimal('1234.5'), -2)
Decimal(1200)

In addition, via the method adjusted() a Decimal with a different precision can be derived, supporting all rounding modes defined by the standard library module decimal.

The rounding modes defined in decimal are wrapped into the Enum ROUNDING.

>>> d = Decimal('12.345')
>>> d.adjusted(2)           # default rounding mode is ROUND_HALF_EVEN !
Decimal('12.34')
>>> d.adjusted(2, ROUNDING.ROUND_HALF_UP)
Decimal('12.35')
>>> d.adjusted(1, ROUNDING.ROUND_UP)
Decimal('12.4')

For the details of the different rounding modes see the documentation of the standard library module decimal.

round and adjusted only allow to round to a quantum that’s a power to 10. The method quantize() can be used to round to any quantum and it does also support all rounding modes mentioned above.

>>> d = Decimal('12.345')
>>># equivalent to round(d, 2) or d.adjusted(2)
>>># (default rounding mode ROUNDING.ROUND_HALF_EVEN):
>>> d.quantize(Decimal('0.01'))
Decimal('12.34')
>>> d.quantize(Decimal('0.05'))
Decimal('12.35')
>>> d.quantize('0.6')
Decimal('12.6')
>>> d.quantize(4)
Decimal('12')

Class Decimal

class decimalfp.Decimal

Decimal number with a given number of fractional digits.

Parameters
  • value (see below) – numerical value (default: None)

  • precision (numbers.Integral) – number of fractional digits (default: None)

If value is given, it must either be a string, an instance of numbers.Integral, number.Rational (for example fractions.Fraction), decimal.Decimal, a finite instance of numbers.Real (for example float) or be convertable to a float or an int.

If a string is given as value, it must be a string in one of two formats:

  • [+|-]<int>[.<frac>][<e|E>[+|-]<exp>] or

  • [+|-].<frac>[<e|E>[+|-]<exp>].

If given value is None, Decimal(0) is returned.

Returns

Decimal instance derived from value according

to precision

The value is always adjusted to the given precision or the precision is calculated from the given value, if no precision is given. For performance reasons, in the latter case the conversion of a numbers.Rational (like fractions.Fraction) or a float tries to give an exact result as a Decimal only up to a fixed limit of fractional digits (decimalfp.LIMIT_PREC).

Raises
  • TypeErrorprecision is given, but not of type Integral.

  • TypeErrorvalue is not an instance of the types listed above and not convertable to float or int.

  • ValueErrorprecision is given, but not >= 0.

  • ValueErrorvalue can not be converted to a Decimal (with a number of fractional digits <= LIMIT_PREC if no precision is given).

Decimal instances are immutable.

Create and return new Decimal instance.

classmethod from_float(f)

Convert a finite float (or int) to a Decimal.

Parameters

f (float or int) – number to be converted to a Decimal

Returns

Decimal instance derived from f

Raises
  • TypeErrorf is neither a float nor an int.

  • ValueErrorf can not be converted to a Decimal with a precision <= LIMIT_PREC.

Beware that Decimal.from_float(0.3) != Decimal(‘0.3’).

classmethod from_decimal(d)

Convert a finite decimal number to a Decimal.

Parameters

d (see below) – decimal number to be converted to a Decimal

d can be of type Decimal, numbers.Integral or decimal.Decimal.

Returns

Decimal instance derived from d

Raises
  • TypeErrord is not an instance of the types listed above.

  • ValueErrord can not be converted to a Decimal.

classmethod from_real(r, exact=True)

Convert a Real number to a Decimal.

Parameters
  • r (numbers.Real) – number to be converted to a Decimal

  • exact (bool) – True if r shall exactly be represented by the resulting Decimal

Returns

Decimal instance derived from r

Raises
  • TypeErrorr is not an instance of numbers.Real.

  • ValueErrorexact is True and r can not exactly be converted to a Decimal with a precision <= LIMIT_PREC.

If exact is False and r can not exactly be represented by a Decimal with a precision <= LIMIT_PREC, the result is rounded to a precision = LIMIT_PREC.

precision

Return precision of self.

magnitude

Return magnitude of self in terms of power to 10.

I.e. the largest integer exp so that 10 ** exp <= self.

numerator

Return the normalized numerator of self.

I. e. the numerator from the pair of integers with the smallest positive denominator, whose ratio is equal to self.

denominator

Return the normalized denominator of ‘self’.

I. e. the smallest positive denominator from the pairs of integers, whose ratio is equal to self.

real

Return real part of self.

Returns self (Real numbers are their real component).

imag

Return imaginary part of self.

Returns 0 (Real numbers have no imaginary component).

adjusted(precision=None, rounding=None)

Return adjusted copy of self.

Parameters
  • precision (numbers.Integral) – number of fractional digits (default: None)

  • rounding (ROUNDING) – rounding mode (default: None)

Returns

Decimal instance derived from self, adjusted

to the given precision, using the given rounding mode

If no precision is given, the result is adjusted to the minimum precision preserving x == x.adjusted().

If no rounding mode is given, the default mode from the current context (from module decimal) is used.

If the given precision is less than the precision of self, the result is rounded and thus information may be lost.

quantize(quant, rounding=None)

Return integer multiple of quant closest to self.

Parameters
  • quant (Rational) – quantum to get a multiple from

  • rounding (ROUNDING) – rounding mode (default: None)

A string can be given for quant as long as it is convertable to a Decimal.

If no rounding mode is given, the default mode from the current context (from module decimal) is used.

Returns

Decimal instance that is the integer multiple of quant

closest to self (according to rounding mode); if result can not be represented as Decimal, an instance of Fraction is returned

Raises

TypeErrorquant is not a Rational number or can not be converted to a Decimal

as_tuple()

Return a tuple (sign, coeff, exp) equivalent to self.

self == (-1) ** sign * coeff * 10 ** exp.

as_fraction()

Return an instance of Fraction equal to self.

Returns the Fraction with the smallest positive denominator, whose ratio is equal to self.

as_integer_ratio()

Return a pair of integers whose ratio is equal to self.

Returns the pair of numerator and denominator with the smallest positive denominator, whose ratio is equal to self.

__repr__()

repr(self)

__str__()

str(self)

__format__(fmt_spec)

Return self converted to a string according to fmt_spec.

Parameters

fmt_spec (str) – a standard format specifier for a number

Returns

self converted to a string according to fmt_spec

Return type

str

__eq__(other)

self == other

__lt__(other)

self < other

__le__(other)

self <= other

__gt__(other)

self > other

__ge__(other)

self >= other

__hash__()

hash(self)

__trunc__()

math.trunc(self)

__pos__()

+self

__neg__()

-self

__abs__()

abs(self)

__add__(other)

self + other

__radd__(other)

self + other

__sub__(other)

self - other

__rsub__(other)

other - self

__mul__(other)

self * other

__rmul__(other)

self * other

__div__(other)

self / other

__rdiv__(other)

other / self

__truediv__(other)

self / other

__rtruediv__(other)

other / self

__pow__(other, mod=None)

self ** other

If other is an integer (or a Rational with denominator = 1), the result will be a Decimal. Otherwise, the result will be a float or complex since roots are generally irrational.

mod must always be None (otherwise a TypeError is raised).

__floor__()

math.floor(self)

__ceil__()

math.ceil(self)

__round__(precision=None)

round(self [, n_digits])

Round self to a given precision in decimal digits (default 0). n_digits may be negative.

This method is called by the built-in round function. It returns an int when called with one argument, otherwise a Decimal.

Rounding modes

Decimal supports all rounding modes defined by the standard library module decimal: ROUND_DOWN, ROUND_UP, ROUND_HALF_DOWN, ROUND_HALF_UP, ROUND_HALF_EVEN, ROUND_CEILING, ROUND_FLOOR and ROUND_05UP. Unless explicitely given, it uses the rounding mode set in the current context set in that module.

The rounding modes defined in decimal are wrapped into the Enum ROUNDING.

class decimalfp.ROUNDING

Enumeration of rounding modes.

ROUND_05UP = 1

Round away from zero if last digit after rounding towards zero would have been 0 or 5; otherwise round towards zero.

ROUND_CEILING = 2

Round towards Infinity.

ROUND_DOWN = 3

Round towards zero.

ROUND_FLOOR = 4

Round towards -Infinity.

ROUND_HALF_DOWN = 5

Round to nearest with ties going towards zero.

ROUND_HALF_EVEN = 6

Round to nearest with ties going to nearest even integer.

ROUND_HALF_UP = 7

Round to nearest with ties going away from zero.

ROUND_UP = 8

Round away from zero.

As shortcut to get or set the rounding mode, the package decimalfp provides the following two functions:

decimalfp.get_rounding()

Return rounding mode from current context.

decimalfp.set_rounding(rounding)

Set rounding mode in current context.

Parameters

rounding (ROUNDING) – rounding mode to be set