F# Unit of Measure, Casting without losing the measure type

Question

Is there built in version of the type casting functions that preserves units and if not how would I make them? So for example with this code how would I cast intWithSecondsMeasure to a float without losing the measure or multiplying by 1.0<s>?

[<Measure>] type s
let intWithSecondsMeasure = 1<s>
let justAFloat = float intWithSecondsMeasure 

Answer 1

The answer provided by @kvb certainly works, but I'd prefer not to use the unbox operator for this conversion. There's a better, built in way that I think should be compiled as a NOP to IL (I haven't checked but unbox is probably going to end up as a unbox instruction in IL and thus adds a runtime type check).

The preferred way to do unit conversions in F# is LanguagePrimitives.TypeWithMeasure (MSDN).

let inline float32toFloat (x:float32<'u>) : float<'u> = 
    x |> float |> LanguagePrimitives.FloatWithMeasure

Answer 2

I don't think that there's a built-in way to do it, but you can easily define your own unit-preserving conversion function:

let float_unit (x:int<'u>) : float<'u> = unbox float x
let floatWithSecondsMeasure = float_unit intWithSecondsMeasure

Answer 3

I compiled the code from kvb and Johannes answers.

Johannes answer

let float32toFloat (x:int<'u>) : float<'u> = 
    x |> float |> LanguagePrimitives.FloatWithMeasure

.method public static float64  float32toFloat(int32 x) cil managed
{
  // Code size       3 (0x3)
  .maxstack  8
  IL_0000:  ldarg.0
  IL_0001:  conv.r8
  IL_0002:  ret
} // end of method Program::float32toFloat

kvb answer with parentheses added.

let float_unit (x:int<'u>) : float<'u> = unbox (float x)

.method public static float64  float_unit(int32 x) cil managed
{
  // Code size       13 (0xd)
  .maxstack  8
  IL_0000:  ldarg.0
  IL_0001:  conv.r8
  IL_0002:  box        [mscorlib]System.Double
  IL_0007:  unbox.any  [mscorlib]System.Double
  IL_000c:  ret
} // end of method Program::float_unit

kvb answer

let float_unit (x:int<'u>) : float<'u> = unbox float x

.method public static float64  float_unit(int32 x) cil managed
{
  // Code size       19 (0x13)
  .maxstack  8
  IL_0000:  newobj     instance void Program/float_unit@4::.ctor()
  IL_0005:  call       !!0 [FSharp.Core]Microsoft.FSharp.Core.LanguagePrimitives/IntrinsicFunctions::UnboxGeneric<class [FSharp.Core]Microsoft.FSharp.Core.FSharpFunc`2<int32,float64>>(object)
  IL_000a:  ldarg.0
  IL_000b:  tail.
  IL_000d:  callvirt   instance !1 class [FSharp.Core]Microsoft.FSharp.Core.FSharpFunc`2<int32,float64>::Invoke(!0)
  IL_0012:  ret
} // end of method Program::float_unit

Answer 4

See my answer to this question:

Unit-safe square roots

which suggests this today:

[<Measure>] 
type s
let intWithSecondsMeasure = 1<s>

let intUtoFloatU< [<Measure>] 'u>( x : int<'u> ) : float<'u> = //'
    let i = int x       //  drop the units
    let f = float i     //  cast
    box f :?> float<'u> //' restore the units

let floatWithS = intUtoFloatU intWithSecondsMeasure