Instantiate this in non-super property/element access expressions
#29468
Conversation
|
@ahejlsberg per your offline review, this now does some more convoluted things with making a this-type type reference and passing that around with a refined |
| @@ -6685,6 +6688,23 @@ namespace ts { | |||
| } | |||
|
|
|||
| function resolveTypeReferenceMembers(type: TypeReference): void { | |||
|
|
|||
| @@ -6631,6 +6631,9 @@ namespace ts { | |||
| else if (type.flags & TypeFlags.Intersection) { | |||
| return getIntersectionType(map((<IntersectionType>type).types, t => getTypeWithThisArgument(t, thisArgument, needApparentType))); | |||
| } | |||
| else if (type.flags & TypeFlags.TypeParameter && thisArgument && (type as TypeParameter).isThisType) { | |||
| return getTypeWithThisArgument(createTypeReference(globalThisType, [type]), needApparentType ? instantiateType(thisArgument, createTypeMapper([type], [getApparentType(type)])) : thisArgument, needApparentType); | |||
There was a problem hiding this comment.
- can you format the conditional one more than one line? It's not readable as-is. (I would extract the middle argument to a temp.)
- Why create a reference to
globalThisType?
| @@ -6685,6 +6688,23 @@ namespace ts { | |||
| } | |||
|
|
|||
| function resolveTypeReferenceMembers(type: TypeReference): void { | |||
|
|
|||
| if (type.target === globalThisType && type.typeArguments && type.typeArguments[0].flags & TypeFlags.TypeParameter && | |||
There was a problem hiding this comment.
I also don't understand what globalThisType is doing here (unless I'm completely mistaken and it's a marker type rather than the type of globalThis.)
There was a problem hiding this comment.
It's the ThisType<T> marker type.
There was a problem hiding this comment.
This fix looks ok now that I have been reminded that globalThisType is ThisType<T> not typeof globalThis. It's pretty confusing to reuse ThisType like this, though, so I think you should think again whether the fix is worth it before merging.
@ahejlsberg, can you take a look at this PR since @weswigham changed the way it is implemented?
|
Unfortunately, we never finished reviewing this PR. It is pretty old now, so I'm going to close it to reduce the number of open PRs. |
Fixes #28159