Tracking Issue for Field Projections

[BennoLossin]

This is a tracking issue for work on Field Projections.
The feature gate for the issue is #![feature(field_projections)].

This is a lang experiment at the moment and the feature is subject to change substantially over time.

We have an active project goal for this effort: rust-lang/rust-project-goals#390

About tracking issues

Tracking issues are used to record the overall progress of implementation.
They are also used as hubs connecting to other relevant issues, e.g., bugs or open design questions.
A tracking issue is however not meant for large scale discussion, questions, or bug reports about a feature.
Instead, open a dedicated issue for the specific matter and add the relevant feature gate label.
Discussion comments will get marked as off-topic or deleted.
Repeated discussions on the tracking issue may lead to the tracking issue getting locked.

Steps

• Approve as lang experiment

  • We accepted this experiment in the 2025-08-13 lang design meeting.

• Hold a second design, notes from 2026-04-01

• a-mir-formality: create a formal model of the borrow checker changes

  • familiarization & experimentation
    • do small experiments using the traits from the borrow checker
    • add our traits to the model
  • add lang-item crate enable creation of lang item crate a-mir-formality#258
  • wait for Niko's new expression based syntax
  • Copy traits to new syntax field projection lang items a-mir-formality#306
  • Integrate HasPlace with place expressions and projections
  • Integrate PlaceRead into the borrow checker (for Copy targets only)
  • Integrate PlaceWrite, PlaceDrop, DropHusk and PlaceRead (also for !Copy) into the borrow checker
  • Integrate PlaceBorrow into the borrow checker
  • Refine & Explore Model
    • Comprehensive test-suite
    • Ensure functionality of common patterns
  • Author document for explaining the model

• Implementation: implement a compiler experiment

  • add field reflection via field representing types (FRTs)
    • add FRTs & Field trait: add field representing types #152730
    • improve FRT documentation improve documentation of field representing types #154927
    • fix FRTs not being invariant make field representing types invariant over the base type #154778
    • fix privacy violation add field representing types #152730 (comment)
    • fix compiler hang Compiler hang due to field-representing types defining a struct in terms of itself. #153256
    • support unsized base types (but only sized fields)
    • support for unions & enums via extra trait
    • figure out packed structs: direct support vs Unaligned<T>
    • add field_of! in expression position to improve ergonomics (reducing need for turbofish)
    • figure out if we actually want & need FRTs for projections
  • Add operation and projection traits to core add lang items for field projections via virtual places #154940
  • add manual projection desugaring macros
  • syntax (require unsafe to access)
  • borrow checker integration
    • depends on a-mir-formality model
  • relax unsafe requirement on syntax

• Experimentation: stress-test the experiment in real code

  • Questions to resolve by this experiment:
    • How much new syntax do we even need?
    • How bad is the ambiguity when making @place be inferred?
    • What effects does this have on SemVer in std & in general?
  • Linux Kernel
    • replace Gary's pointer projection infrastructure
    • Untrusted<T>
    • SeqLockRef<'_, T>
    • MutexGuard<'_ ,T>
    • Mutex<T> -> Rcu<U>
    • Arc<T>
    • Box<T>
    • ArcBorrow<'_, T>
  • Standard Library
    • MappedMutexGuard<'_, T>
    • UnsafeCell<T>
    • Cell<T>
    • ManuallyDrop<T>
    • MaybeUninit<T>
    • *const T & *mut T
    • NonNull<T>
    • Box<T>
    • Arc<T>
    • Rc<T>
    • &[mut] T
    • coordinate with t-libs-api on the creation of:
      • ArcRef<T> & UniqueArcRef<T>
      • ArcBorrow<'_, T>
      • ArcBox<T> (?)
      • WrappingPtr<T>
      • MutexGuardRef<'_, T>
      • MutexRef<'_, T>
      • Own<'_, T> (aka &own T)
    • ArcRef<T> & UniqueArcRef<T>
    • ArcBorrow<'_, T>
    • ArcBox<T> (?)
    • WrappingPtr<T>
    • cell::Ref[Mut]<'_, T>
    • MutexGuardRef<'_, T>
    • MutexRef<'_, T>
    • {btree_map,hash_map}::Entry @joshtriplett
  • pyo3 @davidhewitt
    • PyRef[Mut]<'_, T>
  • crubit @tmandry

• User Research

  • What is the best name for HasPlace? Some good candidates are Place and Derefable.
  • How should the borrowing syntax look like? What's the best for the 95% use-case and what do users understand best?

Unresolved Questions

• Should FRTs be provably (un)inhabited?
  Making them inhabited would allow APIs to take their value to reduce the amount of turbo-fish syntax in code my_fun(field_of!(Struct, field)) instead of my_fun::<field_of!(Struct, field)>(). There currently is no known advantage for provably uninhabited FRTs.
  • Resolved: they should be inhabited ZSTs to allow reducing the amount of turbofish syntax required to use them.

Implementation history

• Implement more traits for FRTs #155588
• add lang items for field projections via virtual places #154940
• improve documentation of field representing types #154927
• make field representing types invariant over the base type #154778
• field projection lang items a-mir-formality#306
• enable creation of lang item crate a-mir-formality#258
• rustfmt: add test for field representing type builtin syntax #153229
• add field representing types #152730
• Add field representing types #146307

[BennoLossin]

@rustbot labels +T-lang +B-experimental +F-field_projection

[rustbot]

Error: Label B-experimental can only be set by Rust team members

Please file an issue on GitHub at triagebot if there's a problem with this bot, or reach out on #triagebot on Zulip.