Allow redefining None to 0 for @[Flags] enum

[bew]

Followup on #4395 (comment)

#4395 now prevents All & None in @[Flags] enums.

But as @oprypin pointed out / demonstrated, redefining None & All can be useful for documentation purpose.

To this end:

it should be possible to redefine None and All but only if their value matches what they would have been anyway.

None should always be 0, and All should always be all the values OR-ed together.

So something like this should work:

@[Flags]
enum Foo
  None = 0
  A
  B
  C
  All = A | B | C
end

It's easy to handle None, just need to check it's 0.
About All, first it's very tedious to repeat all members, second it's harder to detect if there are more members after it. To simplify All and still be able to documente it, there could be a special value like All = :all that will skip this member but retain its doc somehow...

---

As handling None is easy I implemented it in this PR.
Also, I don't think All is common when using enum flags, so I think it's ok to not be able to document it.

WDYT?

[jhass]

In general I'm +1 to this, not sure I like :all though. Can we allow a self-reference here? Something like All = All or All = Foo::All

[oprypin]

Mentioning :all was a bad call because it's not even part of this PR 🐱.

I would be very happy to have just None merged.

[bew]

Yeah I know @oprypin I just wanted to add some context, and I needed to mention All for completeness 😃

[bew]

I just added a better empty enum check that ignores None & All in flags enums, now that they are allowed

[straight-shoota]

What's nb_members supposed to mean? I'd suggest to use a more descriptive name.

[bew]

it's the number of members, maybe number_of_members then? it just felt too long ^^

[straight-shoota]

There is no "too long" for names, only "too short" 😄 num_members would be fine, tough. Or members_size, enum_size.

[bew]

num_members is good 😃 ty

[bcardiff]

i think there is a spec missing regarding how All is ignored. From the PR initial message it wasn't clear if the user needs to A | B | C or not

[oprypin]

There is no change in behavior of All.

[bew]

Yeah sorry if it wasn't clear @bcardiff, the title only mention the possibility to redefine None, not All. The PR inital msg is providing context and explanations, maybe it should have gone in an issue.

[bew]

Thank you!

[straight-shoota]

No, thank you, @bew =)