Skip to content

docs: clarify the reachability-sweep class-member root set (#191 follow-up) #196

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
jagguji merged 1 commit into from
Aug 18, 2026
Merged

docs: clarify the reachability-sweep class-member root set (#191 follow-up) #196

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 15 additions & 4 deletions docs/TYPE_MAPPING.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -694,15 +694,26 @@ metrics. Orphan removal belongs in a reachability sweep **after** traversal, not

**That sweep now exists (`sweepUnreachableEntries`, #191).** After every IR tree is final but BEFORE
`stabilizeNames`, it walks the emitted roots (component props + `baseSpreads`, function/const/context
signatures, class members) and drops any registered entry unreachable from them. It is **key-based**
signatures, and **every class member that emits a binding** — the `@new` ctor, `@send` methods, `@get`
getters, `@set` write-only setters, and `@scope` statics) and drops any registered entry unreachable
from them. It is **key-based**
(every module-mode ref is keyed via `refTo`), so a type reached only through a keyed ref whose home is
late-bound (#128) is never dropped; a deep walk (`collectAllRefKeys`) covers refs inside inline-record
fields and spreads that the shallow `collectRefKeys` skips. Running it before naming also lets a live
sibling reclaim a base name an orphan squatted (`jsonBoxed<T>(): BoxOf<T>` flags its return and strands
`boxOf<'a>`; the sweep drops it and `readBox(): BoxOf<string>` reclaims the clean `boxOf`). Removing an
orphan can also dissolve a spurious SCC merge it was wedged into, giving cleaner module homes. This is
the general mechanism #178 called for; the per-site method-path rollback stays as a cheap early exit.
Fixture: [`return-only-generic-orphan`](../test/golden/cases/return-only-generic-orphan). (#191)
orphan can also dissolve a spurious SCC merge it was wedged into (ReScript's cross-file module graph
must be **acyclic** — a dependency cycle between two `*Types.res` homes is a compile error, so it is
forced into one file, #35), giving cleaner module homes. This is the general mechanism #178 called for;
the per-site method-path rollback stays as a cheap early exit.

The class-member root set must stay **complete**: a type reached *only* through a write-only `@set`
setter or a `@scope` static (`@get`+`@set` normally pair up, so a getter usually hides it) has no other
path to a root, so dropping it emits an external referencing a `*Types.res` type that was never written
— a dangling reference ReScript rejects at compile time. So all class members that bind a type are
roots, not just ctor/methods/getters; the `class-setter-static-reachable` fixture guards that invariant.
Fixtures: [`return-only-generic-orphan`](../test/golden/cases/return-only-generic-orphan),
[`class-setter-static-reachable`](../test/golden/cases/class-setter-static-reachable). (#191)

`opaqueUnion` also gained the **structural dedup** the other builders always had (#180): it registered
unconditionally, so two `type.id`s widening to the same module each got their own — blend emitted
Expand Down
Loading