Automatically strip unitless values

[haberdashPI]

I would have expected this expression to evaluate to a unitless value:

julia> using Unitful: s, Hz
julia> x = 10s * 2Hz
20 Hz s

Instead, one has to do

julia> using Unitful: s, Hz
julia> x = uconvert(unit(1),10s * 2Hz)
20//1

In small expressions this is plenty readable, but in large expressions it gets somewhat ugly. How feasible is that to change? I'd like this to work because then it becomes fairly natural to work with unitful values in larger expressions of unitless values. If this is a reasonable change, I'd be happy to create a pull request if I can be pointed in the right direction.

[Petr-Hlavenka]

Hi,
much simpler for your case is:

julia> Float64(x)
20.0

Little more advance approach with full control of the conversion is following:
I'm using a custom overload of the divide function with a signature physical_value/Unit where both have the same physical dimension (thus the result is unitless):

import Base./
/{T, D, UA,  UU}(x::Unitful.Quantity{T,D,UA}, y::Unitful.Units{UU,D}) =
    float(x.val*Unitful.convfact(y, unit(x)))

for your case, works like this:

julia> x/unit(1)
20.0

But, I use it very often for generating dimensionless output in the units of my choice

julia> y = 1000Hz;
julia> println("Result is y = $(y/u"kHz") [kHz]")
Result is y = 1.0 [kHz]

[haberdashPI]

Thanks, that is much cleaner! My question was more whether it can be automatically detected when units cancel and units striped (and whether that was desirable to do). Obviously I could create a short function (or operator, as in your example) to make manual removal more succinct.

[Petr-Hlavenka]

See my attempts in #48.
Usually this is not a problem and the stripping off the units occurs somewhere where one converts it to a number - e.g. plotting etc. The reason the automatic conversion to number when unitless is unwanted is that the author wanted to keep some conversions "exact" - see the last paragraph in this manual section.
I personally dislike this behavior and therefore have added this one function that does it for the division with units.

[haberdashPI]

Ah, that's helpful, thanks for the pointers. I hadn't actually understood the last section of that section of the manual, I didn't realize that typeof(10.0s / s) = Float64. I would have assumed, based on the behavior above, that typeof(10.0s / s) <: DimensionlessQuantity.