Clarify relation between Kind and PointKind

[burnpanck]

From what I understand, the "kind of a quantity" is some higher-level categorisation of parameters that ought not to be compatible, despite them bing described using seemingly compatible units in the unit system of choice. That is, it further divides dimensions into sub-dimensions which share the same units. This becomes particularly useful when working in systems with Natural Units, where most quantities have a dimension of 1.

With this interpretation, why is there a need to represent the kind of dimension differently when it is used within a ? quantity_point(_kind) instead of within a quantity(_kind)? After all, they also share the same dimensions. IMHO, this seems a needless complication to me. It seems to me that, in fact, one should be able to user kinds and dimensions interchangeably, as they seem to "divide" the same categorical space. Am I missing something here?

[JohelEGP]

quantity_kind builds on quantity. It's a matter of separation of concerns. Kinds doesn't subdivide dimensions, they wrap them. PointKind exists to name Kinds for quantity points. They all might as well be units::point<name-of-the-kind>, and be declared in user-code as using my-point-kind = units::point<name-of-the-kind>;. But I chose to follow the quantity precedent for better error messages.

[burnpanck]

Yeah, if there is a justification for quantity_kind with respect to quantity, then there is one for quantity_point_kind with respect to quantity_point. But still, why is there a need for PointKind to name the Kind for quantity points? Why not supply the Kind to the quantity point directly? After all, we have no PointDimension nor PointUnit either.

[JohelEGP]

[...] why is there a need for PointKind to name the Kind for quantity points? Why not supply the Kind to the quantity point directly?

For better error messages. You can't supply it directly, but you can approximate that. It's done like quantity_kind_point<downcast_point_kind<some-kind>, some-unit> and looks like quantity_kind_point<detail::_point_kind_base<some-kind>, some-unit>. If we could pass them directly, that'd benefit #211.

After all, we have no PointDimension nor PointUnit either.

Yes, because all about quantity is wrapped. The only introduced concept are kinds and their particularity of only providing operations between related kinds.