Extend Generic math to enums

[shadow-cs]

Regarding the recent availability of Generic math preview, it would make sense we also add something along the lines of IEnum which would implement IConvertible<long> (which itself was described as a possible future improvement of generic math in the announcement) as well as I{Span}Parseable (and possibly others) and implement IEnum in Enum base type.

This would further enhance the #104.

[foxesknow]

Why IConvertible<long> ? This wouldn't handle something like enum Foo : ulong

[shadow-cs]

Good point, I was thinking about it being generic. I'm not sure what the possibilities are to make sure it doesn't break anything so I went with the most compatible version I could find. It would be cool if we could separate Flags from non-flags with some constraint as well.

[jnm2]

I asked at dotnet/runtime#112819 about having the runtime automatically implement IBinaryInteger<self> on all enums. This would provide capabilities such as ulong.CreateTruncating(anyEnumValue), TEnum.ReadLittleEndian and tEnumValue.WriteLittleEndian, tEnumValue.GetByteCount(), and general usability with any generic numerics logic.

[tannergooding]

An issue is that not all enums are integers. They can be bool and some runtimes even allow floating-point

[jnm2]

Added a line:

This would only apply to enums whose underlying type implements IBinaryInteger<T>.

[ds5678]

After reading the comments in #8789 about the problems with automatic implementation at run time and compile time, I feel like this would be best addressed with:

• Members on enums (Champion "permit methods in enum declarations" #8987)
• Partial enums (Feature request: Partial enums #2669)
• Enums implementing interfaces

From there, a source generator could let developers opt-in to these interfaces with an attribute.