Equivalence involving nondimensional quantity

[rmsrosa]

In water treatment, for example, it is common to treat mg/l as ppm. Currently, UnitfulEquivalences.jl does not allow equivalences involving nondimensional quantities such as concentration. I managed to do that with two dispatches of edconvert:

@equivalence DensityConcentration

function edconvert(d::dimtype(Unitful.Density), x::Unitful.Quantity{T,D,U}, e::DensityConcentration) where {T,D,U}
    D == NoDims ? x * 1u"kg/L" : throw(_eqconversion_error(d, D, e))
end

function edconvert(d::Unitful.Dimensions{()}, x::Unitful.Quantity{T,D,U}, e::DensityConcentration) where {T,D,U}
    D == Unitful.𝐌/Unitful.𝐋^3 ? x * 1u"L/kg" : throw(_eqconversion_error(d, D, e))
end

But since it is a direct relation, it would be nice to be able to define them at once with something like

@equivalence DensityConcentration
@eqrelation DensityConcentration Unitful.Density / Unitful.NoDims = 1u"kg/l"

I extended dimtype with dimtype(::Unitful.Dimensions{()}) = typeof(Unitful.NoDims) (despite the type piracy warning) so it could handle nondimensional quantities, but it was not enough since apparently edrelation also complains about Unitful.NoDims as the second argument. I think this is also related to the fact that I had to use Unitful.Quantity{T,D,U} in my first edconvert solution above.

As for the second edconvert, the following code also works for the conversion from Density to NoDims, without the need for Unitful.Quantity{T,D,U}, but for some reason that I do not understand it also gives me a warning of type piracy:

edconvert(d::Unitful.Dimensions{()}, x::Unitful.Density, e::DensityConcentration) = x * 1u"L/kg"

So I guess this is a "feature request" to allow one nondimensional quantity in @eqrelation.

[sostock]

I have never seen a “type piracy warning”, can you elaborate on that?

The first edconvert methods can be simplified as follows:

edconvert(::dimtype(Unitful.Density), x::DimensionlessQuantity, e::DensityConcentration) = x * 1u"kg/L"

But I agree that this should work with the macro. I will implement it so that @eqrelation can be used with DimensionlessQuantity.

[sostock]

#7 implements the necessary dimtype method.

One thing that is still missing for dimensionless quantities to work correctly is support for plain numbers: uconvert(u"kg/L", 5u"mm/m", DensityConcentration()) works, but uconvert(u"kg/L", 5, DensityConcentration()) does not. I will implement that as well. Note that the conversion to plain numbers already works: uconvert(NoUnits, 5u"kg/L", DensityConcentration()) Support for plain numbers is now included in #7 as well.

[rmsrosa]

Many thanks for implementing this!

And thanks for telling me about DimensionlessQuantity. I was looking for something like this to use as you mentioned, but somehow I missed it.

As for the type piracy, the warning I get is "An imported function has been extended without using module defined typed arguments.", which shows up in julia-vscode/StaticLint.jl as "TypePiracy".

I just learned about this so I will just point to two places that explain it better: julialang docs: Avoid type piracy and discourse: Clarification on type piracy.

So I agree that my extension to dimtype is a type piracy since both types I use are defined in other modules, but the following extensions use the type DensityConcentration, which is defined in my module, but they still give me this warning:

edconvert(d::dimtype(Unitful.Density), x::Unitful.DimensionlessQuantity, e::DensityConcentration) = x * 1u"kg/l"

edconvert(d::dimtype(NoDims), x::Unitful.Density, e::DensityConcentration) = x * 1u"L/kg"

Anyways, I will try out now with the direct relation that you implemented and get back here to let you know.

[rmsrosa]

Ah, I guess I need to wait till the pull-request is closed, no worries. And I like that you proposed to make the definition of DimensionlessQuantity in Unitful.jl consistent with the others.

[sostock]

Thanks, I know about type piracy but did not know that there are linters who warn about it.

However, the edconvert methods are not type piracy (since you own DensityConcentration), so I wonder why you get a warning for them. Maybe the linter doesn’t recognise that DensityConcentration is a type from your module, since its definition is obfuscated by the @equivalence macro? Is the warning fixed by defining the DensityConcentration struct yourself instead of using the macro?

[rmsrosa]

Is the warning fixed by defining the DensityConcentration struct yourself instead of using the macro?

If I define it directly then the warning disappears.

[sostock]

If I define it directly then the warning disappears.

One more reason to remove the @equivalence macro 😄