StandardUnits live in Units, but prefixes in enumerations.h

[MichaelClerx]

Hi all!

What's the rationale for this?

(Another example is InterfaceType, which lives inside Variable

[agarny]

Fair point. @hsorby, @dladd?

[MichaelClerx]

SpecificationRule lives in its own file, specificationrules.h

(should be specificationrule.h for consistency, but preferably shouldn't be a separate file at all!)

[dladd]

Good catch- I think this may be a case of legacy inconsistencies that stick out now that things are better fleshed out. Although I have a vague memory of someone pointing out that there may be use-cases for SI prefixes outside of the context of units (maybe @hsorby?). That aside, my original preference was to hierarchically specify when possible- for example I prefer the StandardUnit vs. the Prefix specification below:

u->addUnit(libcellml::Units::StandardUnit::AMPERE, libcellml::Prefix::MICRO);

@MichaelClerx , it sounds like you prefer bringing all the enums into enumerations.h? Would a flatter structure for these make things more straightforward for the bindings? I'd agree that consistency is the more important consideration here.

[MichaelClerx]

@dladd For Python bindings it would be easiest if the enums were all defined constants instead (but this means you lose some type-safety on the C++ side I think)!

Organisation-wise, I don't like the "enumerations" file because it's hard for the user to guess which enumerations are and are not inside it. So I'd much prefer if Prefix, Format, and SpecificationRule either (1) Each lived in their own private header file, or (2) Were all part of some suitable class

[hsorby]

I would prefer to have all enumerations that aren't attached to a class to have either there own header file or live in the class header they are used by. For example the InterfaceType is only for variables and must be part of the class. Whilst the Prefix and StandardUnit are associated with the Units class as concerns libCellML I could make a case for; separate headers; in the class header but not the class declaration; in the class declaration. It just depends on which object model philosophy we choose to follow. I think as it stands we aren't following a single object model philosophy.

For me with regards to the bindings I think we would be better off bringing the prefix enumerations into the units class and saying as far as the libCellML object model is concerned prefixes and standard units are part of Units. I hope @nickerso has something to say if prefixes or standard units can be used outside of Units in the libCellML domain.