converting to and from value/quantity point with *out* including mp_units

[Cazadorro]

I have values with specified units coming across the wire. They have a different byte order than my host system. I've created a set of routines to take even non integer values (floats, enums, new type patterned types, and units) and correcty byte swap them to my hosts systems byte order.

This has even worked with with mp_units for the most part. I check using concepts if a type "looks like a mp_unit type" then byte_swap(x.numerical_value_in(T::reference)) * T:: reference;. The type that comes out is a byte swapped copy of the type that comes in. This is doubly great because I do not even have to include mp_units to have this template defined.

The problem comes when I encounter a quantity point type. There's no numerical value in, and there's no non deprecated way of using T::reference. I thought maybe I could use ADL and absolute<T:: reference>(byte_swap(x.quanity_from_zero(). numerical_value_in(T:: reference))) but .. absolute is not a function at all.

Now I'm stuck between a rock and a hard place. My code based pre-dates some of these changes, so there's lots of places where this kind of thing used to work, but no longer does. I can't include mp units in the byte swap header, many downstream users do not use mp_units, and it causes compilation issues for their targets.

Is there a way to avoid including mp_units here? I tried to use T::reference * x + T:: point_origin as well , but I guess that wasn't tested on clang or something because it gives incomprehensible errors that don't make any sense.

[mpusz]

T::reference * x + T:: point_origin will not work as we do not allow typing si::metre * 3, but x * T::reference + T:: point_origin should work just fine. Check https://godbolt.org/z/xfa1x4zx4.

[mpusz]

I unmarked because this is still deprecated, is this supposed to be? I get a million deprecation warnings.

@Cazadorro, could you please tell me what is deprecated? There should be no deprecation in the code that I sent you. Maybe you still use absolute helper instead of point?

[chiphogg]

Hmm... seems like it would be nice if + were commutative, as a general principle to apply to anything we call "+".

@chiphogg, + for point creation is commutative. However, * for quantity creation is not.

Ah, I misread which operation was not commutative. It's a lot easier to see * being non-commutative than +. Thanks!

[Cazadorro]

I unmarked because this is still deprecated, is this supposed to be? I get a million deprecation warnings.

@Cazadorro, could you please tell me what is deprecated? There should be no deprecation in the code that I sent you. Maybe you still use absolute helper instead of point?

x * T::reference is deprecated, and it says because it's multiplying Celsius/kelvin directly with a value instead of using absolute/delta. I do use mp_units::absolute<K>(x) to initialize, am I not supposed to use Delta and absolute anymore? Those were only just introduced in 2.3? And they are the only easily searchable things that come up even when typing "temperature" in the docs

[mpusz]

Right, in case of temperatures, things are getting harder. After an escalation from one of the users, we decided to disallow the multiply syntax for temperature quantities. In that case, we must explicitly use the point (previously called absolute) and delta construction helpers. Unfortunately, I do not see an easy workaround for your issue here 😢

[Cazadorro]

Right, in case of temperatures, things are getting harder. After an escalation from one of the users, we decided to disallow the multiply syntax for temperature quantities. In that case, we must explicitly use the point (previously called absolute) and delta construction helpers. Unfortunately, I do not see an easy workaround for your issue here 😢

Okay, for now I have to use absolute because VCPKG is supposed to stay on the latest stable release, I'm on 2.4. I'm getting around this by manually creating the type again using return T(typename T:: quantity_type(byte_swapped_value, T:: reference));.

If there was an official way to handle this it would need to be attached to the class some how or ADLed. But the biggest issue wasn't the way I had to do this, that's not a big deal, it was the confusion and lack of resouces and time investment in finding out how to do this at all. This needs to be documented somewhere.