-
Notifications
You must be signed in to change notification settings - Fork 381
Adding a New Unit
So you want to add a quantity or unit that is not yet part of Units.NET?
- Great, but before you start!
- Requirements
- Quick Summary of Steps
- Detailed steps
- Logarithmic Units
- Code Style
Sometimes we just have to say no, sorry! 😅 We simply want to avoid bloating the library.
- Is widely used, preferably across domains.
- Has multiple units to convert between (e.g.
Length
has kilometer, feet, nanometer etc.). - Can convert to other quantities (e.g.
Length x Length = Area
). - Can be represented by a
double
numeric value, integer values are not well supported and may suffer from precision errors. - Is not dimensionless/unitless (consider using
Ratio
).
- Is widely used.
- Can be converted to other units of the same quantity.
- The conversion function is well established without ambiguous competing standards.
We have made some exceptions to these "rules" so start a discussion with us if you still think it belongs in the library.
Ok, enough of that. Let's move on! 😃
In order to build and run tests locally you need to have some tools installed.
-
.NET 6 SDK
to generate and build code
Units.NET uses CodeGen, a C# command line app that reads JSON files with quantity and unit definitions and generates C# code.
To add a quantity or a unit:
- Add or change a quantity JSON file.
- Run
generate-code.bat
file. - Specify test values for the new units in the generated test code.
Easy peasy. Below are the detailed steps.
- Place in Common/UnitDefinitions
- See Length.json as an example.
- Use reliable references, such as UN/ECE Recommendation No. 21, Google, Wolfram Alpha or online converters.
- If the conversion is a simple
*
or/
operation, then prefer*
inFromBaseToUnit
and/
inFromUnitToBase
. As an example,Length.Centimeter
is defined as"FromBaseToUnit": "{x} * 100"
and"FromUnitToBase": "{x} / 100"
, instead of{x} / 0.001
and{x} * 0.001
. - Prefer
1e3
and1e-5
notation instead of1000
and0.00001
- Prefer a constant if the conversion factor is finite and well known (
Inch FromUnitToBase: {x} * 2.54e-2
) - Prefer a calculation if the conversion factor is infinite (
PrinterPoint FromUnitToBase: ({x} / 72.27) * 2.54e-2
). If the calculation is not available, specify the most precise constant you can.double
numeric type can represent 15-17 significant decimal digits
Prefer the most widely used abbreviation in the domain, but try to adapt to our conventions.
- Use superscript (
cm²
,m³
) instead ofcm^2
,m^3
- Use
∆
for delta (not▲
) - Use
·
for products (N·m
instead ofNm
,N*m
orN.m
) - Use
/
over⁻¹
, such askm/h
andJ/(mol·K)
- Use
h
for hours,min
for minutes ands
for seconds (m
is ambiguous with meters) - Use abbreviations defined by SI Unit System, such as
l
instead ofL
for liters - Use suffixes to distinguish variants of similar units, such as
gal (U.S.)
vsgal (imp.)
for gallons-
(U.S.)
for United States -
(imp.)
for imperial / British units
-
- Add other popular variants to be able to parse those too, e.g.
[ "tsp", "t", "ts", "tspn", "t.", "ts.", "tsp.", "tspn.", "teaspoon" ]
forVolumeUnit.MetricTeaspoon
wheretsp
is used by default inToString()
The base unit dimensions of the quantity, such as "L": 1
for Length
and "L": 2
for Area
(Length*Length
).
The 7 SI base units are:
-
L
- Length -
M
- Mass -
T
- Time -
I
- ElectricCurrent -
Θ
- Temperature -
N
- AmountOfSubstance -
J
- LuminousIntensity
when converting from one unit to another with FromUnitToBaseFunc
and FromBaseToUnitFunc
conversion functions. It is typically chosen as an SI derived unit (Meter
, Newtonmeter
, Squaremeter
etc). This choice affects the precision of conversions for much bigger/smaller units than BaseUnit
.
BaseUnits
(optional) - the SI base units of a unit
Don't confuse this with the quantity's BaseUnit
, which is discussed to be renamed.
If specified, you can create quantities with consistent units for a given unit system:
new Length(1, UnitSystem.SI).ToString() // "1 m"
new Length(1, myBritishEngineeringUnitSystem).ToString() // "1 ft"
-
LengthUnit.Inch
has{ "L": "Inch" }
(L=1) -
AreaUnit.SquareCentimeter
has{ "L": "Centimeter" }
, because we ignore dimensions (L=2) -
VolumeUnit.Cubicfeet
has{ "L": "Foot" }
, because we ignore dimensions (L=3) -
ForceUnit.Newton
has{ "L": "Meter", "M": "Kilogram", "T": "Second" }
, becauseN = 1 kg * 1 m / s² = Kilogram * Meter / Second²
and we ignore the dimensions -
ForceUnit.PoundForce
has{ "L": "Foot", "M": "Pound", "T": "Second" }
, becauseN = 1 lbm * 1 ft / s² = Pound * Foot / Second²
and we ignore the dimensions -
MassConcentrationUnit.GramPerDeciliter
has{ "L": "Centimeter", "M": "Gram" }
, becauseDeciliter = 1 cm * 1cm * 1cm = Centimeter³
and we ignore the dimensions
The only consequence of not specifying BaseUnits
is that you cannot construct these units by passing a UnitSystem
to the quantity constructor as in the example above.
-
VolumeUnit.ImperialGallon
has noBaseUnits
, becauseVolume = Length³
and there is no length unit that when multiplied three times would result in imperial gallon. -
RatioUnit.DecimalFraction
has noBaseUnits
, because dimensionless units are not made up by any SI base units.
2. Run generate-code.bat
to generate unit classes, unit enumerations and base class for tests.
This step might no longer be necessary, I think Visual Studio 2017 and the new .csproj format automatically loads new files automatically.
- Override the missing abstract properties in the unit test class (ex: LengthTests.cs)
- Specify value as a constant, not a calculation, with preferably at least 7 significant figures where possible.
- Triple-check the number you write here, this is the most important piece as it verifies your conversion function in the .JSON file
- If possible, add a comment next to the value with an URL to where the value can be verified.
- Example:
InchesInOneMeter
in LengthTests.cs- I find the conversion factor to be
39.37007874
from an online unit conversion calculator, it has 10 significant figures so that is plenty - I override the
InchesInOneMeter
property (see below snippet) - I Google to double-check:
Inches In 1 Meter
and it tells me1 Meter = 39.3701 Inches
(Google typically has fewer significant figures) - If Google can't help me, I find a second source to confirm the conversion factor (another conversion website, wikipedia, Wolfram Alpha etc)
- I check again by intuition, is there really around 40 inches in a meter? Yes, sounds about right.
- I find the conversion factor to be
Example code snippet:
/// <summary>https://link-to-where-i-found-the-value.com</summary>
protected override double InchesInOneMeter => 39.37007874;
Make sure all the tests pass. Either run build.bat or run the tests from within Visual Studio with ReSharper or the built-in test runner.
Please see GitHub: Creating a pull request. If you still have any questions, you can reach out in Discussion.
There are many ways to do this, here is one way:
- Click the Fork button to get a copy of this repo on https://github.com/your_user/UnitsNet
- Go to your fork and clone it to a directory on your PC (see instructions in the Code button)
- Create a branch, f.ex.
add-somenewunit
- Do your work, including generating code, commit and push to your fork where you have full write-access
- Visit your fork on github.com with a browser, then create a pull request from your branch to the angularsen/UnitsNet repo.
For one-offs, this is enough.
If you need to create multiple pull requests based on the latest master
branch, or simply keep your fork's branches up to date with the main repo's master
branch, then you need to add the angularsen/UnitsNet repo as a remote to your git clone and fetch it.
# Add main repo as a remote named 'angularsen'
git add remote angularsen https://github.com/angularsen/UnitsNet
# Fetch branches/tags from all your remotes
git fetch
# Create and checkout new branch based on latest master
git checkout -b add-another-unit angularsen/master
# With multiple remotes, you need to tell it what remote a branch should push/pull to, assuming 'origin' is your fork.
git branch --set-upstream-to=origin/add-another-unit
# Do your work, stage changes, commit and push to your fork.
git add -A
git commit -m "My commit message"
git push
# If you need to keep your branch up to date, merge in angularsen's master branch to the current branch you are on. Also push this to your fork.
git merge angularsen/master
git push
# Then visit your fork at https://github.com/your_user/UnitsNet to create a pull request.
Units.NET supports logarithmic units by adding Logarithmic
and LogarithmicScalingFactor
(optional) properties.
-
LogarithmicScalingFactor
is used to provide a scaling factor in the logarithmic conversion. For example, a scaling factor of2
is required when implementing the ratio of the squares of two field amplitude quantities such as voltage. In most casesLogarithmicScalingFactor
will be1
.
To create a logarithmic unit, follow the same steps from the previous section making the following adjustments:
Step 1. Add property "Logarithmic": "True"
to the JSON file, just after BaseUnit
. LogarithmicScalingFactor
defaults to 1
if not defined.
Step 4. Provide custom implementations for logarithmic addition and subtraction unit tests. See LevelTests.cs for an example.
Refer to Level.json as an example implementation of logarithmic units.
- If you have the ReSharper plugin installed, there are code formatting settings checked into the repository that will take effect automatically.
- If you don't use ReSharper, at least follow the same conventions as in the existing code.