Rollup merge of rust-lang#152730 - BennoLossin:field-projections-lang-item, r=oli-obk

add field representing types

*[View all comments](https://triagebot.infra.rust-lang.org/gh-comments/rust-lang/rust/pull/152730)*

> [!NOTE]
> This is a rewrite of rust-lang#146307 by using a lang item instead of a custom `TyKind`. We still need a `hir::TyKind::FieldOf` variant, because resolving the field name cannot be done before HIR construction. The advantage of doing it this way is that we don't need to make any changes to types after HIR (including symbol mangling). At the very beginning of this feature implementation, I tried to do it using a lang item, but then quickly abandoned the approach, because at that time I was still intending to support nested fields.

Here is a [range-diff](https://triagebot.infra.rust-lang.org/gh-range-diff/rust-lang/rust/605f49b27444a738ea4032cb77e3bdc4eb811bab..d15f5052095b3549111854a2555dd7026b0a729e/605f49b27444a738ea4032cb77e3bdc4eb811bab..f5f42d1e03495dbaa23671c46b15fccddeb3492f) between the two PRs

---

# Add Field Representing Types (FRTs)

This PR implements the first step of the field projection lang experiment (Tracking Issue: rust-lang#145383). Field representing types (FRTs) are a new kind of type. They can be named through the use of the `field_of!` macro with the first argument being the type and the second the name of the field (or variant and field in the case of an enum). No nested fields are supported.

FRTs natively implement the `Field` trait that's also added in this PR. It exposes information about the field such as the type of the field, the type of the base (i.e. the type that contains the field) and the offset within that base type. Only fields of non-packed structs are supported, fields of enums an unions have unique types for each field, but those do not implement the `Field` trait.

This PR was created in collaboration with @dingxiangfei2009, it wouldn't have been possible without him, so huge thanks for mentoring me!

I updated my library solution for field projections to use the FRTs from `core` instead of creating my own using the hash of the name of the field. See the [Rust-for-Linux/field-projection `lang-experiment` branch](https://github.com/Rust-for-Linux/field-projection/tree/lang-experiment).

## API added to `core::field`

```rust
pub unsafe trait Field {
    type Base;

    type Type;

    const OFFSET: usize;
}

pub macro field_of($Container:ty, $($fields:expr)+ $(,)?);
```

Along with a perma-unstable type that the compiler uses in the expansion of the macro:

```rust
#[unstable(feature = "field_representing_type_raw", issue = "none")]
pub struct FieldRepresentingType<T: ?Sized, const VARIANT: u32, const FIELD: u32> {
    _phantom: PhantomData<T>,
}
```

## Explanation of Field Representing Types (FRTs)

FRTs are used for compile-time & trait-level reflection for fields of structs & tuples. Each struct & tuple has a unique compiler-generated type nameable through the `field_of!` macro. This type natively contains information about the field such as the outermost container, type of the field and its offset. Users may implement additional traits on these types in order to record custom information (for example a crate may define a [`PinnableField` trait](https://github.com/Rust-for-Linux/field-projection/blob/lang-experiment/src/marker.rs#L9-L23) that records whether the field is structurally pinned).

They are the foundation of field projections, a general operation that's generic over the fields of a struct. This genericism needs to be expressible in the trait system. FRTs make this possible, since an operation generic over fields can just be a function with a generic parameter `F: Field`.

> [!NOTE]
> The approach of field projections has changed considerably since this PR was opened. In the end we might not need FRTs, so this API is highly experimental.

FRTs should act as though they were defined as `struct MyStruct_my_field<StructGenerics>;` next to the struct. So it should be local to the crate defining the struct so that one can implement any trait for the FRT from that crate. The `Field` traits should be implemented by the compiler & populated with correct information (`unsafe` code needs to be able to rely on them being correct).

## TODOs

There are some `FIXME(FRTs)` scattered around the code:
- Diagnostics for `field_of!` can be improved
  - `tests/ui/field_representing_types/nonexistent.rs`
  - `tests/ui/field_representing_types/non-struct.rs`
  - `tests/ui/field_representing_types/offset.rs`
  - `tests/ui/field_representing_types/not-field-if-packed.rs`
  - `tests/ui/field_representing_types/invalid.rs`
- Simple type alias already seem to work, but might need some extra work in `compiler/rustc_hir_analysis/src/hir_ty_lowering/mod.rs`

r? @oli-obk

Author: JonathanBrouwer

Cargo.lock

@@ -4706,6 +4706,7 @@ dependencies = [
  "ena",
  "indexmap",
  "rustc-hash 2.1.1",
+ "rustc_abi",
  "rustc_ast_ir",
  "rustc_data_structures",
  "rustc_error_messages",

compiler/rustc_abi/src/layout.rs

@@ -21,7 +21,64 @@ mod simple;
 mod ty;
 
 #[cfg(feature = "nightly")]
-pub use ty::{FIRST_VARIANT, FieldIdx, Layout, TyAbiInterface, TyAndLayout, VariantIdx};
+pub use ty::{Layout, TyAbiInterface, TyAndLayout};
+
+rustc_index::newtype_index! {
+    /// The *source-order* index of a field in a variant.
+    ///
+    /// This is how most code after type checking refers to fields, rather than
+    /// using names (as names have hygiene complications and more complex lookup).
+    ///
+    /// Particularly for `repr(Rust)` types, this may not be the same as *layout* order.
+    /// (It is for `repr(C)` `struct`s, however.)
+    ///
+    /// For example, in the following types,
+    /// ```rust
+    /// # enum Never {}
+    /// # #[repr(u16)]
+    /// enum Demo1 {
+    ///    Variant0 { a: Never, b: i32 } = 100,
+    ///    Variant1 { c: u8, d: u64 } = 10,
+    /// }
+    /// struct Demo2 { e: u8, f: u16, g: u8 }
+    /// ```
+    /// `b` is `FieldIdx(1)` in `VariantIdx(0)`,
+    /// `d` is `FieldIdx(1)` in `VariantIdx(1)`, and
+    /// `f` is `FieldIdx(1)` in `VariantIdx(0)`.
+    #[cfg_attr(feature = "nightly", derive(rustc_macros::HashStable_Generic))]
+    #[encodable]
+    #[orderable]
+    #[gate_rustc_only]
+    pub struct FieldIdx {}
+}
+
+impl FieldIdx {
+    /// The second field, at index 1.
+    ///
+    /// For use alongside [`FieldIdx::ZERO`], particularly with scalar pairs.
+    pub const ONE: FieldIdx = FieldIdx::from_u32(1);
+}
+
+rustc_index::newtype_index! {
+    /// The *source-order* index of a variant in a type.
+    ///
+    /// For enums, these are always `0..variant_count`, regardless of any
+    /// custom discriminants that may have been defined, and including any
+    /// variants that may end up uninhabited due to field types.  (Some of the
+    /// variants may not be present in a monomorphized ABI [`Variants`], but
+    /// those skipped variants are always counted when determining the *index*.)
+    ///
+    /// `struct`s, `tuples`, and `unions`s are considered to have a single variant
+    /// with variant index zero, aka [`FIRST_VARIANT`].
+    #[cfg_attr(feature = "nightly", derive(rustc_macros::HashStable_Generic))]
+    #[encodable]
+    #[orderable]
+    #[gate_rustc_only]
+    pub struct VariantIdx {
+        /// Equivalent to `VariantIdx(0)`.
+        const FIRST_VARIANT = 0;
+    }
+}
 
 // A variant is absent if it's uninhabited and only has ZST fields.
 // Present uninhabited variants only require space for their fields,

compiler/rustc_abi/src/layout/ty.rs

@@ -4,67 +4,14 @@ use std::ops::Deref;
 use rustc_data_structures::intern::Interned;
 use rustc_macros::HashStable_Generic;
 
+use crate::layout::{FieldIdx, VariantIdx};
 use crate::{
     AbiAlign, Align, BackendRepr, FieldsShape, Float, HasDataLayout, LayoutData, Niche,
     PointeeInfo, Primitive, Size, Variants,
 };
 
 // Explicitly import `Float` to avoid ambiguity with `Primitive::Float`.
 
-rustc_index::newtype_index! {
-    /// The *source-order* index of a field in a variant.
-    ///
-    /// This is how most code after type checking refers to fields, rather than
-    /// using names (as names have hygiene complications and more complex lookup).
-    ///
-    /// Particularly for `repr(Rust)` types, this may not be the same as *layout* order.
-    /// (It is for `repr(C)` `struct`s, however.)
-    ///
-    /// For example, in the following types,
-    /// ```rust
-    /// # enum Never {}
-    /// # #[repr(u16)]
-    /// enum Demo1 {
-    ///    Variant0 { a: Never, b: i32 } = 100,
-    ///    Variant1 { c: u8, d: u64 } = 10,
-    /// }
-    /// struct Demo2 { e: u8, f: u16, g: u8 }
-    /// ```
-    /// `b` is `FieldIdx(1)` in `VariantIdx(0)`,
-    /// `d` is `FieldIdx(1)` in `VariantIdx(1)`, and
-    /// `f` is `FieldIdx(1)` in `VariantIdx(0)`.
-    #[derive(HashStable_Generic)]
-    #[encodable]
-    #[orderable]
-    pub struct FieldIdx {}
-}
-
-impl FieldIdx {
-    /// The second field, at index 1.
-    ///
-    /// For use alongside [`FieldIdx::ZERO`], particularly with scalar pairs.
-    pub const ONE: FieldIdx = FieldIdx::from_u32(1);
-}
-
-rustc_index::newtype_index! {
-    /// The *source-order* index of a variant in a type.
-    ///
-    /// For enums, these are always `0..variant_count`, regardless of any
-    /// custom discriminants that may have been defined, and including any
-    /// variants that may end up uninhabited due to field types.  (Some of the
-    /// variants may not be present in a monomorphized ABI [`Variants`], but
-    /// those skipped variants are always counted when determining the *index*.)
-    ///
-    /// `struct`s, `tuples`, and `unions`s are considered to have a single variant
-    /// with variant index zero, aka [`FIRST_VARIANT`].
-    #[derive(HashStable_Generic)]
-    #[encodable]
-    #[orderable]
-    pub struct VariantIdx {
-        /// Equivalent to `VariantIdx(0)`.
-        const FIRST_VARIANT = 0;
-    }
-}
 #[derive(Copy, Clone, PartialEq, Eq, Hash, HashStable_Generic)]
 #[rustc_pass_by_value]
 pub struct Layout<'a>(pub Interned<'a, LayoutData<FieldIdx, VariantIdx>>);

compiler/rustc_abi/src/lib.rs

@@ -64,9 +64,9 @@ pub use canon_abi::{ArmCall, CanonAbi, InterruptKind, X86Call};
 #[cfg(feature = "nightly")]
 pub use extern_abi::CVariadicStatus;
 pub use extern_abi::{ExternAbi, all_names};
+pub use layout::{FIRST_VARIANT, FieldIdx, LayoutCalculator, LayoutCalculatorError, VariantIdx};
 #[cfg(feature = "nightly")]
-pub use layout::{FIRST_VARIANT, FieldIdx, Layout, TyAbiInterface, TyAndLayout, VariantIdx};
-pub use layout::{LayoutCalculator, LayoutCalculatorError};
+pub use layout::{Layout, TyAbiInterface, TyAndLayout};
 
 /// Requirements for a `StableHashingContext` to be used in this crate.
 /// This is a hack to allow using the `HashStable_Generic` derive macro

compiler/rustc_ast/src/ast.rs

@@ -2553,6 +2553,11 @@ pub enum TyKind {
     /// Pattern types like `pattern_type!(u32 is 1..=)`, which is the same as `NonZero<u32>`,
     /// just as part of the type system.
     Pat(Box<Ty>, Box<TyPat>),
+    /// A `field_of` expression (e.g., `builtin # field_of(Struct, field)`).
+    ///
+    /// Usually not written directly in user code but indirectly via the macro
+    /// `core::field::field_of!(...)`.
+    FieldOf(Box<Ty>, Option<Ident>, Ident),
     /// Sometimes we need a dummy value when no error has occurred.
     Dummy,
     /// Placeholder for a kind that has failed to be defined.

compiler/rustc_ast/src/util/classify.rs

@@ -298,6 +298,7 @@ fn type_trailing_braced_mac_call(mut ty: &ast::Ty) -> Option<&ast::MacCall> {
             | ast::TyKind::ImplicitSelf
             | ast::TyKind::CVarArgs
             | ast::TyKind::Pat(..)
+            | ast::TyKind::FieldOf(..)
             | ast::TyKind::Dummy
             | ast::TyKind::Err(..) => break None,
         }

compiler/rustc_ast_lowering/src/lib.rs

@@ -1496,6 +1496,13 @@ impl<'a, 'hir> LoweringContext<'a, 'hir> {
             TyKind::Pat(ty, pat) => {
                 hir::TyKind::Pat(self.lower_ty_alloc(ty, itctx), self.lower_ty_pat(pat, ty.span))
             }
+            TyKind::FieldOf(ty, variant, field) => hir::TyKind::FieldOf(
+                self.lower_ty_alloc(ty, itctx),
+                self.arena.alloc(hir::TyFieldPath {
+                    variant: variant.map(|variant| self.lower_ident(variant)),
+                    field: self.lower_ident(*field),
+                }),
+            ),
             TyKind::MacCall(_) => {
                 span_bug!(t.span, "`TyKind::MacCall` should have been expanded by now")
             }

compiler/rustc_ast_pretty/src/pprust/state.rs

@@ -1377,6 +1377,23 @@ impl<'a> State<'a> {
                 self.word(" is ");
                 self.print_ty_pat(pat);
             }
+            ast::TyKind::FieldOf(ty, variant, field) => {
+                self.word("builtin # field_of");
+                self.popen();
+                let ib = self.ibox(0);
+                self.print_type(ty);
+                self.word(",");
+                self.space();
+
+                if let Some(variant) = variant {
+                    self.print_ident(*variant);
+                    self.word(".");
+                }
+                self.print_ident(*field);
+
+                self.end(ib);
+                self.pclose();
+            }
         }
         self.end(ib);
     }

compiler/rustc_const_eval/src/const_eval/machine.rs

@@ -12,8 +12,8 @@ use rustc_middle::mir::AssertMessage;
 use rustc_middle::mir::interpret::ReportedErrorInfo;
 use rustc_middle::query::TyCtxtAt;
 use rustc_middle::ty::layout::{HasTypingEnv, TyAndLayout, ValidityRequirement};
-use rustc_middle::ty::{self, Ty, TyCtxt};
-use rustc_middle::{bug, mir};
+use rustc_middle::ty::{self, FieldInfo, Ty, TyCtxt};
+use rustc_middle::{bug, mir, span_bug};
 use rustc_span::{Span, Symbol, sym};
 use rustc_target::callconv::FnAbi;
 use tracing::debug;
@@ -23,8 +23,9 @@ use crate::errors::{LongRunning, LongRunningWarn};
 use crate::interpret::{
     self, AllocId, AllocInit, AllocRange, ConstAllocation, CtfeProvenance, FnArg, Frame,
     GlobalAlloc, ImmTy, InterpCx, InterpResult, OpTy, PlaceTy, Pointer, RangeSet, Scalar,
-    compile_time_machine, err_inval, interp_ok, throw_exhaust, throw_inval, throw_ub,
-    throw_ub_custom, throw_unsup, throw_unsup_format, type_implements_dyn_trait,
+    compile_time_machine, ensure_monomorphic_enough, err_inval, interp_ok, throw_exhaust,
+    throw_inval, throw_ub, throw_ub_custom, throw_unsup, throw_unsup_format,
+    type_implements_dyn_trait,
 };
 
 /// When hitting this many interpreted terminators we emit a deny by default lint
@@ -619,6 +620,27 @@ impl<'tcx> interpret::Machine<'tcx> for CompileTimeMachine<'tcx> {
                 ecx.write_type_info(ty, dest)?;
             }
 
+            sym::field_offset => {
+                let frt_ty = instance.args.type_at(0);
+                ensure_monomorphic_enough(ecx.tcx.tcx, frt_ty)?;
+
+                let (ty, variant, field) = if let ty::Adt(def, args) = frt_ty.kind()
+                    && let Some(FieldInfo { base, variant_idx, field_idx, .. }) =
+                        def.field_representing_type_info(ecx.tcx.tcx, args)
+                {
+                    (base, variant_idx, field_idx)
+                } else {
+                    span_bug!(ecx.cur_span(), "expected field representing type, got {frt_ty}")
+                };
+                let layout = ecx.layout_of(ty)?;
+                let cx = ty::layout::LayoutCx::new(ecx.tcx.tcx, ecx.typing_env());
+
+                let layout = layout.for_variant(&cx, variant);
+                let offset = layout.fields.offset(field.index()).bytes();
+
+                ecx.write_scalar(Scalar::from_target_usize(offset, ecx), dest)?;
+            }
+
             _ => {
                 // We haven't handled the intrinsic, let's see if we can use a fallback body.
                 if ecx.tcx.intrinsic(instance.def_id()).unwrap().must_be_overridden {

compiler/rustc_const_eval/src/interpret/mod.rs

@@ -38,6 +38,8 @@ use self::place::{MemPlace, Place};
 pub use self::projection::{OffsetMode, Projectable};
 pub use self::stack::{Frame, FrameInfo, LocalState, ReturnContinuation};
 pub use self::util::EnteredTraceSpan;
-pub(crate) use self::util::{create_static_alloc, type_implements_dyn_trait};
+pub(crate) use self::util::{
+    create_static_alloc, ensure_monomorphic_enough, type_implements_dyn_trait,
+};
 pub use self::validity::{CtfeValidationMode, RangeSet, RefTracking};
 pub use self::visitor::ValueVisitor;