# Trait alias work-plan and architecture overview

## RFC #3437

[Link to RFC](https://github.com/Jules-Bertholet/rfcs/blob/implementable-trait-alias/text/3437-implementable-trait-alias.md)

The following work summary is extracted and predicted from the RFC text. The items are ordered by descending priority and potential for a quick prototype

- Syntactical support from `rustc_ast`, `rustc_hir` and `rustc_ast_lowering`
- [ ] Change the parser rule to try the `path` NV first

- Name resolution
- ~~`rustc_resolve` needs a new phase in the late resolver, because it needs to wait for `SupertraitItem` to register before it can resolve `type SubtraitItem = Vec<Self::SupertraitItem>`.~~
- *This is a shared work* with Pre-RFCing `impl Subtrait`
- Two steps for this quest
- [x] Handle the simple case: simple `Ident` as item names, resolve them against the
- [ ] :construction: Handle the slightly harder case: item names as paths, resolve them while skipping the generics, e.g. `fn Qualified::<A, B>::name() ...` should only be resolved against `Qualified<...>`

- HIR book-keeping
- This can start *independently* to avoid the syntax bikeshedding
- For each conceptual trait alias items, we will record the supertraits, the associated items and method bodies
- Modify the queries
- `associated_item_def_ids` because it assumes that all associated items in a `trait`-def block are owned by the principal trait definition, and we are going to break this assumption with implementable trait alias
- was noch?

- Crate level book-keeping
- [x] Propagate supertrait `DefId`s across crate boundary
- [x] Propagate supertrait bounds with generics across crate boundary

- Trait solver
- This is the heavy lifting, and I presume that the foundation is already there.
- [ ] Handle the simple case: items without generics
- [ ] :construction: Modify the crate-level trait-impl collector to re-route the supertrait items into their right trait bins to simulate `impl`
- [ ] :construction: Modify the coherence check to enforce the rule in which either no impl items or all impl items should be present for any non-marker supertraits
- [ ] :dragon:s ? surprises? secret side-quests?

## `impl Subtrait` for Supertrait items

- First one needs to promote the Pre-RFC text to a formal RFC and @dingxf will take care of it
- ... so that we can have a RFC number to refer to at least

## PRs and Issues

- [ ] [#143527](https://github.com/rust-lang/rust/issues/143527) resolver, maybe the necessary HIR changes so that we have a minimal UI test case to check into the compiler