RFC: Add Quantity::as_ and Quantity::cast for changing the underlying storage type.

[adamreichold]

A first go at something helping with #293.

The restriction to $units<V> instead of U: Units<V> seems a bit weird, but it was the only way I saw to enforce that the same set of base units would be used for the conversion. (The more general version fails with trait bound errors as e.g. the f32 base units do not implement Units<i32> and without generic associated types there is no way to ask a Units<V> implementor for its equivalent Units<W> implementor AFAIU.)

Finally, while I added minimal doc tests, I am not sure where actual unit tests should end up as this involves at least two storage types by definition.

[adamreichold]

without generic associated types there is no way to ask a Units implementor for its equivalent Units implementor AFAIU

I think we could add a separate trait

trait ChangeStorage<Ul, Ur>: Units<Ul> {
    type Units: Unit<Ur>;
}

impl<Ul, Ur> ChangeStorage<Ul, Ur> for $units<Ul> {
    type Units = $units<Ur>;
}

but I am not sure how much this would add w.r.t. generality as it would still cover only the $units<_> case by default.

[iliekturtles]

Thanks for the PR! I'll take a look this coming week and see if I can figure out anything with the traits.

[iliekturtles]

I don't think using $units is needed/desired.

The following diff compiles so it works... right?

I went a slightly different way and defined a QuantityCast trait although I suppose no trait or even From/Into (too implicit?) could be used. This is a very rough implementation and essentially delegates the conversion down to num::cast::NumCast which in turn relies on num::cast::ToPrimitive.

I'll do some thinking about usability over the weekend as I think the ability to do type conversions is proven, it's just a matter of determining what the interface should be.

diff --git a/src/lib.rs b/src/lib.rs
index be430b4..32b73cf 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -280,6 +280,10 @@ pub mod num {
     #[cfg(feature = "bigint-support")]
     pub use num_rational::BigRational;
 
+    pub mod cast {
+        pub use num_traits::cast::*;
+    }
+
     #[cfg(any(feature = "rational-support", feature = "bigint-support"))]
     pub mod rational {
         pub use num_rational::*;
diff --git a/src/system.rs b/src/system.rs
index 6324553..e9aac7c 100644
--- a/src/system.rs
+++ b/src/system.rs
@@ -590,6 +590,34 @@ macro_rules! system {
         impl_ops!(Sub, sub, -, SubAssign, sub_assign, -=, Diff,
             Div, div, /, DivAssign, div_assign, /=, sub_div);
 
+        trait QuantityCast<D, U, V>: Sized
+        where
+            D: Dimension + ?Sized,
+            U: Units<V> + ?Sized,
+            V: $crate::num::Num + $crate::num::cast::ToPrimitive + $crate::Conversion<V>,
+        {
+            fn from(q: Quantity<D, U, V>) -> Option<Self>;
+        }
+
+        impl<D, Ur, Ul, Vr, Vl> QuantityCast<D, Ul, Vl> for Quantity<D, Ur, Vr>
+        where
+            D: Dimension + ?Sized,
+            Ur: Units<Vr> + ?Sized,
+            Ul: Units<Vl> + Units<Vr> + ?Sized,
+            Vr: $crate::num::Num + $crate::num::cast::NumCast + $crate::Conversion<Vr>,
+            Vl: $crate::num::Num + $crate::num::cast::ToPrimitive + $crate::Conversion<Vl>,
+        {
+            fn from(q: Quantity<D, Ul, Vl>) -> Option<Self>
+            {
+                <Vr as $crate::num::cast::NumCast>::from(q.value)
+                    .map(|v| Quantity {
+                        dimension: $crate::lib::marker::PhantomData,
+                        units: $crate::lib::marker::PhantomData,
+                        value: change_base::<D, Ul, Ur, Vr>(&v),
+                    })
+            }
+        }
+
         impl<D, U, V> Quantity<D, U, V>
         where
             D: Dimension + ?Sized,

[adamreichold]

The trait bound which did not work out for me was

Ul: Units<Vl> + Units<Vr> + ?Sized,

as e.g. SI<f32> would only implement Units<f32> but not say Units<i32>.

[dzmitry-lahoda]

consider impl https://github.com/kas-gui/easy-cast traits may be

[dzmitry-lahoda]

casts are good, for example #529 i would like casts because inputs are 64 bit, but I lift system to 128 bit as needed for aggregations