Particles & Unitful

[aplavin]

I noticed that sometimes mixing units and MCM particles results in weird results – nesting two levels of Particles, resulting in 2000*2000 floats underneath:

julia> using MonteCarloMeasurements, Unitful

julia> x = Particles()
4.44089e-19 ± 1.0 Particles{Float64, 2000}


julia> y = 123u"m" * x
1.02318e-15 ± 123.0 Particles{Float64, 2000}
 m

julia> x * y
-2.8e-17 ± 2.0e-15 m ± 98.1 ± 74.0 m Particles{Quantity{Particles{Float64, 2000}, 𝐋, Unitful.FreeUnits{(m,), 𝐋, nothing}}, 2000}

I wonder how to avoid/fix that?

[baggepinnen]

It seems you're creating an object with the wrong type, see
https://baggepinnen.github.io/MonteCarloMeasurements.jl/stable/examples/#Unitful-interaction-1

[aplavin]

Thanks, I saw that part, just not sure how to apply it in my case.
I start with a non-unitful Particles that are passed to a (ideally, blackbox) function that does some unitful calculations in it but its input and output are non-unitful:

julia> function f(x)  # simple MWE for f
           y = 123u"m" * x
           ustrip(u"cm", x*y)
       end

julia> x = Particles()  # create an plain non-unitful particles
8.88178e-19 ± 1.0 Particles{Float64, 2000}

julia> f(x)  # apply f to x, expecting to get just Particles{Float64, 2000}:
-1.2e-16 ± 1.3e-13 ± 9810.0 ± 7400.0 Particles{Particles{Float64, 2000}, 2000}

[baggepinnen]

The issue is probably some missing methods in the unitful extension. The multiplication 123u"m" * x dispatches to

*(x::AbstractQuantity, y::Number)

in Unitful, instead of

function Base.$f(y::Quantity{S,D,U}, p::$PT{T,N}) where {S, D, U, T <: Quantity, N}
    QT = Base.promote_op($op, typeof(y), T)
    $PT{QT,N}($(op).(y, p.particles))
end

in MCM.jl due to the T <: Quantity. Changing this to T <: Number appears to resolve the issue.

Try out

• fix multiplication of particles and quantity #159

and see if that works out for you, if it does, I'll make a new release