[ABI] Initial draft of canonicalization and minimization of generic signatures #12740
Conversation
docs/ABI/GenericSignature.md
Outdated
| 2. Constraints follow, ordered first by the | ||
| [type parameter ordering](#type-parameter-ordering) of the left-hand | ||
| operand and then by constraint kind. The left-hand side of a constraint | ||
| is always a type parameter `T`, which can be a generic parameter of a |
There was a problem hiding this comment.
"a generic parameter or a"?
docs/ABI/GenericSignature.md
Outdated
| nested type thereof (e.g., `T.SubSequence.Iterator.Element`). | ||
| Constraints are ordered as follows: | ||
| 1. A superclass constraint `T: C`, where `C` is a class. | ||
| 2. A layout constraints (e.g., `T: some-layout`), where the right-hand |
docs/ABI/GenericSignature.md
Outdated
|
|
||
| ### Same-type constaints | ||
|
|
||
| The canonical form of superclass, layout, and conformance constraints are trivially canonicalized using the anchor of the appropriate equivalence class. Same-type constraints, on the other hand, are responsible for forming those equivalence classes. Let's expand our running example to include a third `Collection`: |
There was a problem hiding this comment.
What about superclass constraints where the right hand side is a generic class?
There was a problem hiding this comment.
Yeah, I should call that out. The RHS needs to be a canonical concrete type.
|
It's not clear to me how we canonicalize concrete structural types that contain type parameters inside of them, eg |
|
Canonicalizing concrete structural types containing type parameters needs to be called out. I'll push an update shortly that does so, thanks! |
Also address Slava's feedback on the first draft.
|
I think this is in "good enough" shape to merge. Additional comments are certainly welcome! |
This is a rough, initial draft describing the canonicalization and minimization of generic signatures for the purposes of the ABI. Comments greatly appreciated!
Addresses the documentation part of SR-3733 and rdar://problem/31412994.