Recursive type definitions are not parsed correctly

[vzarytovskii]

I hope it's fine if I will be creating issues I've found (and will be trying fixing them) here.
Recursive types definition are resulting in errors when parsing types declared with and

type Maybe<'T> = | Just B<'T> | Nothing
and 'T maybe = Maybe<'T>

results in

  type_definition [1, 0] - [2, 24]
    type_abbrev_defn [1, 5] - [2, 24]
      type_name [1, 5] - [1, 14]
        identifier [1, 5] - [1, 10]
        type_arguments [1, 10] - [1, 14]
          type_argument_defn [1, 11] - [1, 13]
            type_argument [1, 11] - [1, 13]
              identifier [1, 12] - [1, 13]
      ERROR [1, 15] - [2, 12]
        ERROR [1, 17] - [2, 3]
          union_type_cases [1, 17] - [1, 39]
            union_type_case [1, 19] - [1, 23]
              identifier [1, 19] - [1, 23]
            ERROR [1, 24] - [1, 29]
              ERROR [1, 24] - [1, 25]
              ERROR [1, 27] - [1, 28]
            union_type_case [1, 32] - [1, 39]
              identifier [1, 32] - [1, 39]
        type [2, 4] - [2, 12]
          type [2, 4] - [2, 6]
            type_argument [2, 4] - [2, 6]
              identifier [2, 5] - [2, 6]
          long_identifier [2, 7] - [2, 12]
            identifier [2, 7] - [2, 12]
      type [2, 15] - [2, 24]
        long_identifier [2, 15] - [2, 20]
          identifier [2, 15] - [2, 20]
        type_attributes [2, 21] - [2, 23]
          type_attribute [2, 21] - [2, 23]
            type [2, 21] - [2, 23]
              type_argument [2, 21] - [2, 23]
                identifier [2, 22] - [2, 23]

[Nsidorenco]

Very nice @vzarytovskii!
We should consider adding these failing examples the corpus to ensure we don't regress in the future.
I will try to get it cleaned up such that the current tests doesn't fail.

[vzarytovskii]

Very nice @vzarytovskii Vlad Zarytovskii FTE! We should consider adding these failing examples the corpus to ensure we don't regress in the future. I will try to get it cleaned up such that the current tests doesn't fail.

Sure, give me some time, to figure out how corpus works :)

[Nsidorenco]

Come to think of it, is type Maybe<'T> = | Just B<'T> | Nothing valid f#?
I've added the and keyword for types in 275f3ca, but the union cases does not match the rules in the language specification

[vzarytovskii]

Come to think of it, is type Maybe<'T> = | Just B<'T> | Nothing valid f#?

It's a typo, apologies. B not supposed to be there but 'of' instead.

https://sharplab.io/#v2:EYLgtghglgdgNAGxAMwM4B8AuBPADgUwAIBZCbYfAHgHIAVAPkIF5D1CApAV1U0IHtkhOq0IA5PpgAWsAOYBYAFAQYAEyG1CkckRaltNBkA=

I've added the and keyword for types in 275f3ca, but the union cases does not match the rules in the language specification

[vzarytovskii]

So, it's working for

type A<'T> = | B of 'T
and C<'T> = A<'T>

But not for something like

type A<'T> = | B of 'T
and 'T C = A<'T>

(notice the prefix style type paramter)

Let me try fix it.

Update: the following type definition doesn't work either:

type 'T X = | Y of 'T