📏 Dimension based safety in elixir
Switch branches/tags
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
bin add docker aliases for local version testing Aug 14, 2016
config
lib fix credo readability issues Sep 12, 2018
test
.credo.exs disable another credo rule Sep 12, 2018
.gitignore indent docs properly Aug 26, 2016
.travis.yml latest credo only supports elixir 1.4 and above. Sep 12, 2018
LICENCE.md add licence Nov 27, 2015
README.md Add validated facts for units (#3) Jul 21, 2017
mix.exs bump ex_doc package Sep 12, 2018
mix.lock bump ex_doc package Sep 12, 2018

README.md

UnitFun

Build Status

Attempt to add units to numbers in elixir to give some added type saftey when dealing with numeric quantities.

Why?

One good example: pounds(dollars) should never be accidentally added to pence(cents) without conversion. Both are numeric. Wrapping the numeric data in a tuple with unit information seems like a good idea. This library gives a neat way of expressing this.

Example - Basic

First define some units:

defmodule Pounds do
  use UnitFun.Unit
end

defmodule Pence do
  use UnitFun.Unit
end

Then do something with them:

use UnitFun.MathsOperators
import UnitFun.UnitTypes

item_cost = 5 <~ Pounds # UnitFun.with_units(5, Pounds)
item_tax = 100 <~ Pence # UnitFun.with_units(100, Pence)

# The following will throw an error as the units mismatch:
item_cost + item_tax # UnitFun.add(item_cost, item_tax)

Example - Conversions

Conversions can be defined:

defimpl UnitFun.Convertor, for: Pence do
  def convert(_, Pounds, value), do: (value * 100)
end

defimpl UnitFun.Convertor, for: Pounds do
  #Note: please don't actually do a divison for any financial maths
  # You're going to lose data and have a bad time.
  def convert(_, Pence, value), do: (value / 100)
end

And now the following:

# returns: %UnitFun.Value{value: 6, units: Pounds}
total = item_cost + item_tax # UnitFun.add(item_cost, item_tax)

# returns: %UnitFun.Value{value: 600, units: Pence}
total_in_pence = total <~ Pence # UnitFun.with_units(total, Pence)

Example - Assertion

Errors can be raised if units aren't what they are expected to be:

UnitFun.assert_units(total_money, Miles)

Example - Composite units

New units can also be composed by multiplying existing units together:

  use UnitFun.MathsOperators
  import UnitFun.UnitTypes

  km_squared = Kilometers * Kilometers # UnitFun.multiply(Kilometers, Kilometers)

These newly defined units can then be used as with all previous examples

  edge = 4 <~ Kilometers # UnitFun.with_units(4, Kilometers)

  area = edge * edge # UnitFun.multiply(edge, edge)

  expected_area = 16 <~ km_squared # UnitFun.with_units(16, km_squared)
  assert area == expected_area # UnitFun.equals(area, expected_area)

Dividing/multiplying by united types returns values with new types so correctness can be asserted on.

  miles_per_hour = Miles / Hours # UnitFun.divide(Miles, Hours)

  speed = 40 <~ miles_per_hour # UnitFun.with_units(40, miles_per_hour)
  time_spent_travelling = 2 <~ Hours # UnitFun.with_units(2, hours)

  #the distance will be in Miles as the hours cancel out
  distance_travelled_in_two_hours = time_spent_travelling * speed  # UnitFun.multiply(time_spent_travelling, speed)

  assert distance_travelled_in_two_hours == 80 <~ Miles # UnitFun.with_units(80, Miles)

Example - Composite unit conversions

If there's a single unit way of representing some composite units this conversion can also be defined (N.B. there's currently no way of defining a conversion from a simple unit to a composite one):

defimpl UnitFun.Convertor, for: UnitFun.ConvertorComplexTest.Pascals do
  alias UnitFun.ConvertorComplexTest.Meters
  alias UnitFun.ConvertorComplexTest.Newtons
  alias UnitFun.Units.CompositeUnit

  def convert(_, %CompositeUnit{numerators: [%Newtons{}], denominators: [%Meters{}]}, value) do
    value
  end
end

Example - Custom mathematic functions.

All the maths is controlled by protocols. So for example if you decided pence should only be handled as integers (so rounding isn't an issue) the following protocol could be defined:

defimpl UnitFun.Maths.AddSubtractMaths, for: UnitFun.ExampleTest.Pence do
  def add(_, left, right) when is_integer(left) and is_integer(right) do
     left + right
  end
  def subtract(_, left, right) when is_integer(left) and is_integer(right) do
    left - right
  end
end

Now any addition using non integer quantities will raise a FunctionClauseError. For convinience if nothing is defined then the kernel +-/* are used.

Example - Facts about units

It's possible to define units with facts that must always hold true. This is handled by defining a list of functions that return true or false.

defmodule UnitFun.Examples.PositiveUnit do
  @moduledoc false
  use UnitFun.Unit

  defp greater_than_zero(x), do: x >= 0

  facts [
    &greater_than_zero/1
  ]

end

Now whenever a PositiveUnit value is constructed the greater_than_zero callback will be executed. If this returns False then an InvalidValueError will be raised.