Decimal v1.5.0 Decimal View Source
Decimal arithmetic on arbitrary precision floating-point numbers.
A number is represented by a signed coefficient and exponent such that: sign
* coefficient * 10 ^ exponent
. All numbers are represented and calculated
exactly, but the result of an operation may be rounded depending on the
context the operation is performed with, see: Decimal.Context
. Trailing
zeros in the coefficient are never truncated to preserve the number of
significant digits unless explicitly done so.
There are also special values such as NaN (not a number) and ±Infinity.
-0 and +0 are two distinct values.
Some operation results are not defined and will return NaN.
This kind of NaN is quiet, any operation returning a number will return
NaN when given a quiet NaN (the NaN value will flow through all operations).
The other kind of NaN is signalling which is the value that can be reached
in result
field on Decimal.Error
when the result is NaN. Any operation
given a signalling NaN return will signal :invalid_operation
.
Exceptional conditions are grouped into signals, each signal has a flag and a
trap enabler in the context. Whenever a signal is triggered it’s flag is set
in the context and will be set until explicitly cleared. If the signal is trap
enabled Decimal.Error
will be raised.
Specifications
This implementation follows the above standards as closely as possible. But at some places the implementation diverges from the specification. The reasons are different for each case but may be that the specification doesn’t map to this environment, ease of implementation or that API will be nicer. Still, the implementation is close enough that the specifications can be seen as additional documentation that can be used when things are unclear.
The specification models the sign of the number as 1, for a negative number,
and 0 for a positive number. Internally this implementation models the sign as
1 or -1 such that the complete number will be sign * coefficient *
10 ^ exponent
and will refer to the sign in documentation as either positive
or negative.
There is currently no maximum or minimum values for the exponent. Because of that all numbers are “normal”. This means that when an operation should, according to the specification, return a number that “underflow” 0 is returned instead of Etiny. This may happen when dividing a number with infinity. Additionally, overflow, underflow and clamped may never be signalled.
Link to this section Summary
Types
The coefficient of the power of 10
. Non-negative because the sign is stored separately in sign
The exponent to which 10
is raised
1
for positive-1
for negative
This implementation models the sign
as 1
or -1
such that the complete number will be: sign * coef * 10 ^ exp
Functions
The absolute value of given number. Sets the number’s sign to positive
Adds two numbers together
Compares two numbers numerically. If the first number is greater than the second
:gt
is returned, if less than :lt
is returned, if both numbers are equal
:eq
is returned
Compares two numbers numerically. If the first number is greater than the second
#Decimal<1>
is returned, if less than #Decimal<-1>
is returned. Otherwise,
if both numbers are equal #Decimal<0>
is returned. If either number is a quiet
NaN, then that number is returned
Returns true
if argument is a decimal number, otherwise false
Divides two numbers
Divides two numbers and returns the integer part
Compares two numbers numerically and returns true
if they are equal,
otherwise false
Creates a new decimal number from a floating point number
Gets the process’ context
Returns true
if number is ±Infinity, otherwise false
Compares two values numerically and returns the maximum. Unlike most other
functions in Decimal
if a number is NaN the result will be the other number.
Only if both numbers are NaN will NaN be returned
Compares two values numerically and returns the minimum. Unlike most other
functions in Decimal
if a number is NaN the result will be the other number.
Only if both numbers are NaN will NaN be returned
Negates the given number
Multiplies two numbers
Returns true
if number is NaN, otherwise false
Check if given number is negative
Creates a new decimal number from an integer or a string representation
Creates a new decimal number from the sign, coefficient and exponent such that
the number will be: sign * coefficient * 10 ^ exponent
Parses a binary into a decimal
Applies the context to the given number rounding it to specified precision
Check if given number is positive
Reduces the given number. Removes trailing zeros from coefficient while keeping the number numerically equivalent by increasing the exponent
Remainder of integer division of two numbers. The result will have the sign of the first number
Rounds the given number to specified decimal places with the given strategy (default is to round to nearest one). If places is negative, at least that many digits to the left of the decimal point will be zero
Set the process’ context
Subtracts second number from the first. Equivalent to Decimal.add/2
when the
second number’s sign is negated
Returns the decimal converted to a float
Returns the decimal represented as an integer
Converts given number to its string representation
Update the process’ context
Runs function with given context
Link to this section Types
coefficient() :: non_neg_integer() | :qNaN | :sNaN | :inf
The coefficient of the power of 10
. Non-negative because the sign is stored separately in sign
.
non_neg_integer
- when thet
represents a number, instead of one of the special values below.:qNaN
- a quiet NaN was produced by a previous operation. Quiet NaNs propagate quietly, unlike signaling NaNs that return errors (based on theDecimal.Context
).:sNaN
- signalling NaN that indicated an error occurred that should stop the next operation with an error (based on theDecimal.Context
).:inf
- Infinity.
The exponent to which 10
is raised.
rounding() :: :down | :half_up | :half_even | :ceiling | :floor | :half_down | :up
1
for positive-1
for negative
signal() :: :invalid_operation | :division_by_zero | :rounded | :inexact
t() :: %Decimal{coef: coefficient(), exp: exponent(), sign: sign()}
This implementation models the sign
as 1
or -1
such that the complete number will be: sign * coef * 10 ^ exp
.
coef
- the coefficient of the power of10
.exp
- the exponent of the power of10
.sign
-1
for positive,-1
for negative.
Link to this section Functions
The absolute value of given number. Sets the number’s sign to positive.
Adds two numbers together.
Exceptional conditions
- If one number is -Infinity and the other +Infinity
:invalid_operation
will be signalled.
Examples
iex> Decimal.add(1, "1.1")
#Decimal<2.1>
iex> Decimal.add(1, "Inf")
#Decimal<Infinity>
Compares two numbers numerically. If the first number is greater than the second
:gt
is returned, if less than :lt
is returned, if both numbers are equal
:eq
is returned.
Neither number can be a NaN. If you need to handle quiet NaNs, use compare/2
.
Examples
iex> Decimal.cmp("1.0", 1)
:eq
iex> Decimal.cmp("Inf", -1)
:gt
Compares two numbers numerically. If the first number is greater than the second
#Decimal<1>
is returned, if less than #Decimal<-1>
is returned. Otherwise,
if both numbers are equal #Decimal<0>
is returned. If either number is a quiet
NaN, then that number is returned.
Examples
iex> Decimal.compare("1.0", 1)
#Decimal<0>
iex> Decimal.compare("Inf", -1)
#Decimal<1>
Returns true
if argument is a decimal number, otherwise false
.
Divides two numbers.
Exceptional conditions
- If both numbers are ±Infinity
:invalid_operation
is signalled. - If both numbers are ±0
:invalid_operation
is signalled. - If second number (denominator) is ±0
:division_by_zero
is signalled.
Examples
iex> Decimal.div(3, 4)
#Decimal<0.75>
iex> Decimal.div("Inf", -1)
#Decimal<-Infinity>
Divides two numbers and returns the integer part.
Exceptional conditions
- If both numbers are ±Infinity
:invalid_operation
is signalled. - If both numbers are ±0
:invalid_operation
is signalled. - If second number (denominator) is ±0
:division_by_zero
is signalled.
Examples
iex> Decimal.div_int(5, 2)
#Decimal<2>
iex> Decimal.div_int("Inf", -1)
#Decimal<-Infinity>
Integer division of two numbers and the remainder. Should be used when both
div_int/2
and rem/2
is needed. Equivalent to: {Decimal.div_int(x, y),
Decimal.rem(x, y)}
.
Exceptional conditions
- If both numbers are ±Infinity
:invalid_operation
is signalled. - If both numbers are ±0
:invalid_operation
is signalled. - If second number (denominator) is ±0
:division_by_zero
is signalled.
Examples
iex> Decimal.div_rem(5, 2)
{Decimal.new(2), Decimal.new(1)}
Compares two numbers numerically and returns true
if they are equal,
otherwise false
.
Examples
iex> Decimal.equal?("1.0", 1)
true
iex> Decimal.equal?(1, -1)
false
Creates a new decimal number from a floating point number.
Floating point numbers will be converted to decimal numbers with
:io_lib_format.fwrite_g/1
. Since this conversion is not exact it is
recommended to give an integer or a string via new/1
instead.
Examples
iex> Decimal.from_float(3.14)
#Decimal<3.14>
Gets the process’ context.
Returns true
if number is ±Infinity, otherwise false
.
Compares two values numerically and returns the maximum. Unlike most other
functions in Decimal
if a number is NaN the result will be the other number.
Only if both numbers are NaN will NaN be returned.
Examples
iex> Decimal.max(1, "2.0")
#Decimal<2.0>
iex> Decimal.max(1, "NaN")
#Decimal<1>
iex> Decimal.max("NaN", "NaN")
#Decimal<NaN>
Compares two values numerically and returns the minimum. Unlike most other
functions in Decimal
if a number is NaN the result will be the other number.
Only if both numbers are NaN will NaN be returned.
Examples
iex> Decimal.min(1, "2.0")
#Decimal<1>
iex> Decimal.min(1, "NaN")
#Decimal<1>
iex> Decimal.min("NaN", "NaN")
#Decimal<NaN>
Negates the given number.
Examples
iex> Decimal.minus(1)
#Decimal<-1>
iex> Decimal.minus("-Inf")
#Decimal<Infinity>
Multiplies two numbers.
Exceptional conditions
- If one number is ±0 and the other is ±Infinity
:invalid_operation
is signalled.
Examples
iex> Decimal.mult("0.5", 3)
#Decimal<1.5>
iex> Decimal.mult("Inf", -1)
#Decimal<-Infinity>
Returns true
if number is NaN, otherwise false
.
Check if given number is negative
Creates a new decimal number from an integer or a string representation.
A decimal number will always be created exactly as specified with all digits kept - it will not be rounded with the context.
Backus–Naur form
sign ::= "+" | "-"
digit ::= "0" | "1" | "2" | "3" | "4" | "5" | "6" | "7" | "8" | "9"
indicator ::= "e" | "E"
digits ::= digit [digit]...
decimal-part ::= digits "." [digits] | ["."] digits
exponent-part ::= indicator [sign] digits
infinity ::= "Infinity" | "Inf"
nan ::= "NaN" [digits] | "sNaN" [digits]
numeric-value ::= decimal-part [exponent-part] | infinity
numeric-string ::= [sign] numeric-value | [sign] nan
Examples
iex> Decimal.new(1)
#Decimal<1>
iex> Decimal.new("3.14")
#Decimal<3.14>
new(1 | -1, non_neg_integer() | :qNaN | :sNaN | :inf, integer()) :: t()
Creates a new decimal number from the sign, coefficient and exponent such that
the number will be: sign * coefficient * 10 ^ exponent
.
A decimal number will always be created exactly as specified with all digits kept - it will not be rounded with the context.
Parses a binary into a decimal.
If successful, returns a tuple in the form of {:ok, decimal}
,
otherwise :error
.
Examples
iex> Decimal.parse("3.14")
{:ok, %Decimal{coef: 314, exp: -2, sign: 1}}
iex> Decimal.parse("-1.1e3")
{:ok, %Decimal{coef: 11, exp: 2, sign: -1}}
iex> Decimal.parse("bad")
:error
Applies the context to the given number rounding it to specified precision.
Check if given number is positive
Reduces the given number. Removes trailing zeros from coefficient while keeping the number numerically equivalent by increasing the exponent.
Remainder of integer division of two numbers. The result will have the sign of the first number.
Exceptional conditions
- If both numbers are ±Infinity
:invalid_operation
is signalled. - If both numbers are ±0
:invalid_operation
is signalled. - If second number (denominator) is ±0
:division_by_zero
is signalled.
Examples
iex> Decimal.rem(5, 2)
#Decimal<1>
Rounds the given number to specified decimal places with the given strategy (default is to round to nearest one). If places is negative, at least that many digits to the left of the decimal point will be zero.
Examples
iex> Decimal.round("1.234")
#Decimal<1>
iex> Decimal.round("1.234", 1)
#Decimal<1.2>
Set the process’ context.
Subtracts second number from the first. Equivalent to Decimal.add/2
when the
second number’s sign is negated.
Exceptional conditions
- If one number is -Infinity and the other +Infinity
:invalid_operation
will be signalled.
Examples
iex> Decimal.sub(1, "0.1")
#Decimal<0.9>
iex> Decimal.sub(1, "Inf")
#Decimal<-Infinity>
Returns the decimal converted to a float.
The returned float may have lower precision than the decimal. Fails if the decimal cannot be converted to a float.
Returns the decimal represented as an integer.
Fails when loss of precision will occur.
Converts given number to its string representation.
Options
:scientific
- number converted to scientific notation.:normal
- number converted without a exponent.:raw
- number converted to its raw, internal format.
update_context((Decimal.Context.t() -> Decimal.Context.t())) :: :ok
Update the process’ context.
with_context(Decimal.Context.t(), (() -> x)) :: x when x: var
Runs function with given context.