Rename, clarify, and document code for "erasing" query values #151565

Zalathar · 2026-01-24T02:47:15Z

In order to reduce compile times and code size for the compiler itself, the query system has a mechanism for “erasing” and ”restoring” query values in certain contexts. See #109333 for the original implementation.

Unfortunately, the erasure system has very little documentation, and involves a dizzying assortment of similarly-named types, traits, and functions.

This PR therefore renames several parts of the erasure API and implementation to hopefully be clearer, and adds comments to better explain the purpose and mechanism behind value erasure.

Summary of renames:

fn erase → erase_val (avoiding ambiguity with module erase)
fn restore → restore_val
type Erase<T> → Erased<T> (for actual erased values of T)
trait EraseType → Erasable (for types that can be erased and restored)
associated type EraseType::Result → Erasable::Storage
implementation-detail struct Erased<T> → ErasedData<Storage>

There should be no change to compiler behaviour.

rustbot · 2026-01-24T02:47:20Z

r? @SparrowLii

rustbot has assigned @SparrowLii.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Zalathar · 2026-01-24T02:48:45Z

compiler/rustc_middle/src/query/erase.rs

+/// Trait for types that can be erased into [`Erased<Self>`].
+///
+/// Erasing and unerasing values is performed by [`erase_val`] and [`restore_val`].
+pub trait Erasable: Copy {


Erasable goes against the notional convention of naming traits after verbs rather than adjectives, but in this case I find erasable so much clearer that I think it's justified.

certainly can't be worse the status quo, since it wasn't Erase before. what does "EraseType" mean? "you can erase this type"? yeah it implements the trait, doesn't it?

( type Erase simply doesn't make much sense to me. )

compiler/rustc_middle/src/query/erase.rs

rustbot · 2026-01-25T13:49:20Z

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

petrochenkov · 2026-01-26T11:52:40Z

r? @petrochenkov

petrochenkov · 2026-01-26T12:36:43Z

compiler/rustc_middle/src/query/erase.rs

    const {
-        if size_of::<T>() != size_of::<T::Result>() {
+        if size_of::<T>() != size_of::<T::Storage>() {
            panic!("size of T must match erased type T::Result")


Suggested change

panic!("size of T must match erased type T::Result")

panic!("size of T must match erased type T::Storage")

petrochenkov · 2026-01-26T12:36:57Z

compiler/rustc_middle/src/query/erase.rs

-    Erased::<<T as EraseType>::Result> {
+    ErasedData::<<T as Erasable>::Storage> {
        // `transmute_unchecked` is needed here because it does not have `transmute`'s size check
        // (and thus allows to transmute between `T` and `MaybeUninit<T::Result>`) (we do the size


Suggested change

// (and thus allows to transmute between `T` and `MaybeUninit<T::Result>`) (we do the size

// (and thus allows to transmute between `T` and `MaybeUninit<T::Storage>`) (we do the size

petrochenkov · 2026-01-26T12:47:09Z

compiler/rustc_middle/src/query/erase.rs

        $(
-            impl<'tcx> EraseType for $($fake_path)::+<'tcx> {
-                type Result = [u8; size_of::<$($fake_path)::+<'static>>()];
+            impl<'tcx> Erasable for $($fake_path)::+<'tcx> {


Suggested change

impl<'tcx> Erasable for $($fake_path)::+<'tcx> {

impl Erasable for $($fake_path)::+<'_> {

I think this will work for any paths with one lifetime argument at the end, not just 'tcx.
A number of explicit impls from the above can probably be merged into this macro too.

petrochenkov · 2026-01-26T12:51:19Z

compiler/rustc_middle/src/query/inner.rs

    let key = key.into_query_param();
    if let Some(res) = try_get_cached(tcx, query_cache, &key) {
-        erase::restore(res).map(drop)
+        erase::restore_val(res).map(drop)


Hmm, why drop?
Erasable types are Copy, they cannot have any nontrivial drop.

petrochenkov · 2026-01-26T12:59:28Z

compiler/rustc_middle/src/query/plumbing.rs

                queries::$name::Key<'tcx>,
                QueryMode,
-            ) -> Option<Erase<$V>>,)*
+            ) -> Option<::rustc_middle::query::erase::Erased<$V>>,)*


Suggested change

) -> Option<::rustc_middle::query::erase::Erased<$V>>,)*

) -> Option<$crate::query::erase::Erased<$V>>,)*

And two more instances above.

We are in rustc_middle, so this looks a bit weird in declarative macros, but it's a matter of style. Other code above uses $crate though.

zetanumbers

I don't find these changes to be useful or clarifying.

View changes since this review

zetanumbers · 2026-01-26T13:30:37Z

compiler/rustc_middle/src/query/erase.rs

+//! To improve compile times and code size for the compiler itself, query
+//! values are "erased" in some contexts (e.g. inside in-memory cache types),
+//! to reduce the number of generic instantiations created during codegen.


Source of this statement needs clarification. Benchmarks on the original PR (#109333 (comment)) do not seem to have introduced improvements.

Yeah, it would be good to check if it's still necessary.
#108638 says "The compile time of rustc_query_impl is reduced by 27%" which by itself doesn't sound like a convincing reason to complicate the query system, but pictures like #65031 (comment) make it more convincing.

rustbot assigned SparrowLii Jan 24, 2026

Zalathar commented Jan 24, 2026

View reviewed changes

deltragon reviewed Jan 24, 2026

View reviewed changes

compiler/rustc_middle/src/query/erase.rs Outdated Show resolved Hide resolved

Zalathar force-pushed the erase branch from 8a38f03 to e8b3211 Compare January 24, 2026 11:58

This comment has been minimized.

Sign in to view

Rename, clarify, and document code for "erasing" query values

e3b524a

Zalathar force-pushed the erase branch from e8b3211 to e3b524a Compare January 25, 2026 13:49

rustbot assigned petrochenkov and unassigned SparrowLii Jan 26, 2026

petrochenkov reviewed Jan 26, 2026

View reviewed changes

petrochenkov added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Jan 26, 2026

zetanumbers suggested changes Jan 26, 2026

View reviewed changes

	panic!("size of T must match erased type T::Result")
	panic!("size of T must match erased type T::Storage")

	// (and thus allows to transmute between `T` and `MaybeUninit<T::Result>`) (we do the size
	// (and thus allows to transmute between `T` and `MaybeUninit<T::Storage>`) (we do the size

	impl<'tcx> Erasable for $($fake_path)::+<'tcx> {
	impl Erasable for $($fake_path)::+<'_> {

	) -> Option<::rustc_middle::query::erase::Erased<$V>>,)*
	) -> Option<$crate::query::erase::Erased<$V>>,)*

Uh oh!

Rename, clarify, and document code for "erasing" query values #151565

Are you sure you want to change the base?

Rename, clarify, and document code for "erasing" query values #151565

Conversation

Zalathar commented Jan 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rustbot commented Jan 24, 2026

Uh oh!

Zalathar Jan 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

workingjubilee Jan 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

This comment has been minimized.

rustbot commented Jan 25, 2026

Uh oh!

petrochenkov commented Jan 26, 2026

Uh oh!

petrochenkov Jan 26, 2026

Choose a reason for hiding this comment

Uh oh!

petrochenkov Jan 26, 2026

Choose a reason for hiding this comment

Uh oh!

petrochenkov Jan 26, 2026

Choose a reason for hiding this comment

Uh oh!

petrochenkov Jan 26, 2026

Choose a reason for hiding this comment

Uh oh!

petrochenkov Jan 26, 2026

Choose a reason for hiding this comment

Uh oh!

zetanumbers left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

zetanumbers Jan 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

petrochenkov Jan 26, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

Zalathar commented Jan 24, 2026 •

edited

Loading

Zalathar Jan 24, 2026 •

edited

Loading

workingjubilee Jan 24, 2026 •

edited

Loading

zetanumbers left a comment •

edited

Loading

zetanumbers Jan 26, 2026 •

edited

Loading