Fix presentation of HKT and TypeLambdas in Scala3doc

[BarkingBad]

No description provided.

[abgruszecki]

This seems to improve some things, so I think we can merge it on that basis. However, we still display type lambdas instead of HKTs, see all the exnteds clauses here: https://scala3doc.virtuslab.com/pr-missing-wildcard-in-hkts/scala3/api/scala/collection/immutable.html. That should be considered a bug, and isn't fixed here.

[romanowski]

We cannot simply strip anything after $ in type names

[BarkingBad]

@abgruszecki I am not sure if it is correct behaviour, but look at this example, given code like this:

import scala.collection.immutable.ArraySeq trait SomeTraitWithHKTs[A[_], B[_], C[_], D[_], E[_]] abstract class Foo[Z] trait SomeTrait extends SomeTraitWithHKTs[Foo, scala.collection.immutable.ArraySeq, ArraySeq, List, [B] =>> List[B]]

we get following AppliedType's parameter list:

Full output

List(HKTypeLambda(List(Z), List(TypeBounds(TypeRef(TermRef(ThisType(TypeRef(NoPrefix,module class <root>)),object scala),Nothing),TypeRef(TermRef(ThisType(TypeRef(NoPrefix,module class <root>)),object scala),Any))), AppliedType(TypeRef(TermRef(ThisType(TypeRef(NoPrefix,module class tests)),object hkts),Foo),List(TypeParamRef(Z)))), TypeRef(TermRef(TermRef(TermRef(ThisType(TypeRef(NoPrefix,module class <root>)),object scala),object collection),object immutable),class ArraySeq), HKTypeLambda(List(A), List(TypeBounds(TypeRef(TermRef(ThisType(TypeRef(NoPrefix,module class <root>)),object scala),Nothing),TypeRef(TermRef(ThisType(TypeRef(NoPrefix,module class <root>)),object scala),Any))), AppliedType(TypeRef(TermRef(ThisType(TypeRef(NoPrefix,module class collection)),object immutable),ArraySeq),List(TypeParamRef(A)))), TypeRef(TermRef(TermRef(ThisType(TypeRef(NoPrefix,module class <root>)),object scala),package),List), HKTypeLambda(List(B), List(TypeBounds(TypeRef(TermRef(ThisType(TypeRef(NoPrefix,module class <root>)),object scala),Nothing),TypeRef(TermRe

which is actually List of [HKTypeLambda, TypeRef, HKTypeLambda, TypeRef, HKTypeLambda] and the output is:

1. There is different behaviour to ArraySeq declared as an import or as fully qualified name making it just ArraySeq or type lambda of [A] =>> ArraySeq[A].
2. If i would like to collapse the Foo type lamda, I could not disambiguate it with actual type lambdas, like [B] =>> List[B].

What do you think about these two things?

[abgruszecki]

Ah, I see, in some cases we do display List as simply List. Do we have what you wrote as a test case? If not, we should.

At some point we should find out what exactly makes the types in stdlib display type lambdas instead of HKTs (variance annotations, perhaps?), but it's less of a problem than I thought, I'd say.

[BarkingBad]

I am still not sure why this thing is happening, but I have shown the example where import of fullly qualified name makes difference for ArraySeq.
@romanowski the memberInfo was the unfinished work. I implemented parseTypeDef to use memberInfo property, however, there come out another weird problem, for given code:

trait A[X] { type SomeType[X] def x: SomeType[X] } trait B[D] extends A[D]

the expected behaviour to see SomeType from B class perspective would be with applied D symbolic name, not X. However, we get following, because memberInfo method from dotty.tools.dotc.core.Types returns such name though it is run in context of B class symbol.

[BarkingBad]

Unfortunately I am enforced to rollback all memberInfo changes for TypeDefs as it returns HKTypeLambdas which lose variance