Reject non-primitive string interpolation at the typechecker - #813
Merged
Conversation
Conversion to String must be named in the source. An interpolated part now has to be an Int, a Float, a Bool or a String — exactly the set the __to_str lowering renders on every backend. Any other embed is a type error that names the offending type and asks for a named function returning String; no standard-library renderer is offered, because none exists. Reverts the wasm-gc List<Int> interpolation stringifier: with the checker rule in place, no typechecked program can reach the compound arm of the interpolation emitter. That arm now raises a codegen error naming the function and the embed type instead of bailing out to a silent unreachable trap stub, so an internal pipeline that skips the typecheck gets a diagnostic rather than a module that traps at runtime. The remaining unsupported-shape bails in the MIR body emitter still produce a trap stub; AVER_WASMGC_REQUIRE_MIR=1 enumerates them. Migrates the self-hosted evaluator's two Expr diagnostics onto a named exprLabel function and regenerates the self-host output. The two wasm-gc tests that covered the old behaviour are rewritten rather than dropped: one pins the new type error, the other keeps the three-way packed-field projection differential by building its string through an explicit local render function. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
An unresolved type variable was accepted alongside `Type::Invalid` on the
theory that it always follows an earlier diagnostic. It does not. Matching
on a bare `Option.None` with an arm `Option.Some(x) -> "{x}"` binds `x` to
the `T` of an `Option<T>` nothing in the program fixes; the file otherwise
type-checks clean, so the embed passed the interpolation rule and reached
the backends with no renderable type. A bare `[]` subject is the same hole.
A bare `Type::Var` is now its own verdict with its own diagnostic: the type
could not be determined, pin it and convert through a named function
returning String. `Type::Invalid` keeps its acceptance — that one really is
downstream of an error already reported.
The verdict is decided where the embed is inferred rather than in a later
sweep, because there is no later chance to pin it: the checker holds no
ambient substitution (every `subst` map lives and dies inside one
`compatible` / `match_with` call), `Spanned::set_ty` writes a `OnceLock` so
the type read at the rule is the final stamp every backend will see, and
the one top-down channel fires only for the fixed recogniser list, which
never descends into an interpolation part.
Records the argument for the emitter's "typechecked source cannot reach
this arm" claim at the wasm-gc error site, now that the checker's partition
of `Type` has no remainder: the four accepted primitives are exactly the
four the dispatch handles, `Invalid` implies the compile is already gated,
and everything else is rejected by name. Pins the reported line number for
both interpolation diagnostics on a line past the first, since an embed's
own line is 1-based inside the `{...}` fragment and only the outer node's
line is source-accurate.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Interpolation renders primitives only —
Int,Float,Bool,String, exactly the set the__to_strlowering implements on every backend. Any other embed is a type error naming the offending type and asking for a named function returningString; no standard-library renderer is added.List<Int>interpolation stringifier: no typechecked program can reach the compound arm any more.unreachabletrap stub, so pipelines that skip the typecheck get a diagnostic. The other unsupported-shape bails in the MIR emitter still trap-stub;AVER_WASMGC_REQUIRE_MIR=1lists them.Exprdiagnostics to a namedexprLabelfunction (self-host regenerated). The rest of the tree already used named render functions.🤖 Generated with Claude Code