types: #99 lib.d.ts loader + checker features — 250→256 Pass#1263
Open
nickna wants to merge 11 commits into
Open
types: #99 lib.d.ts loader + checker features — 250→256 Pass#1263nickna wants to merge 11 commits into
nickna wants to merge 11 commits into
Conversation
…tructor) First increment of #99 (lib.d.ts modeling), seam-first: a single TryResolveLibType switch, consulted by both ResolveTypeNameCore (for the type) and ReportUnknownTypeName (so the name isn't flagged TS2304), resolves TYPE-position lib names SharpTS doesn't yet parse from the vendored .d.ts files. Checked AFTER user declarations so a user interface of the same name wins. A later increment backs this switch with a real lib.d.ts loader. Covers the Symbol family: SymbolConstructor -> the existing WellKnownSymbolTypes shape; Symbol -> a new member-less wrapper interface (a symbol value is assignable to it, but it is not assignable back to the symbol primitive). Flips symbolProperty55, symbolType15.
…eInfo) The real-loader foundation for #99 increment 2. LibTypeLoader parses ambient lib declarations embedded as libdefs/*.d.ts resources (the vendored TypeScript lib is a dev-only submodule, so it must be embedded to work at runtime) and resolves them to modeled TypeInfo via a throwaway checker's new ExtractLibTypes (PreRegister into a fresh env, diagnostics suppressed, extracted by name). Built once, cached process-wide. TryResolveLibType now sources SymbolConstructor from the LOADED lib instead of a hand-modeled shape (Symbol stays hand-modeled — the wrapper needs a member-less shape a symbol can satisfy, which the loaded interface can't give until primitive apparent members are modeled). No baseline change (symbolProperty55/symbolType15 still pass); this is the pipeline that later increments fill from the real es5.d.ts + resolve the primitive-vs-interface tension to flip the remaining lib tests. Full xUnit 15452/0, conformance 252 Pass (0 regressions).
…ts/extends First test flipped by the lib loader. A wrapper name (String/Number/Boolean) resolves to the primitive in type position, so 'class C implements String' and 'interface I extends String' saw a non-interface and reported spurious TS2304/TS2312. Now the implements and interface-extends resolution consult LibTypeLoader when the name isn't a user interface, using the (currently member-less) lib wrapper interface as a valid implements/extends target. 'var s: String' still resolves to the primitive, so no assignability change there. Flips stringLiteralTypeIsSubtypeOfString. Full xUnit 15452/0, conformance 253 Pass.
…pes (TS2416)
'class C implements I' validation runs before the method-body pass, so an
un-annotated method (e.g. '[Symbol.toPrimitive]() { return "" }') had a
<inferred> (~any) return type that hid a mismatch with the interface. Capture the
resolved interfaces during that pass and re-check un-annotated method return types
against them at the end of class-checking (after inference is published), reporting
TS2416 at the method's own line. Only un-annotated methods are re-checked, so
annotated ones aren't double-reported.
Flips symbolProperty24. Full xUnit 15452/0, conformance 254 Pass.
…lap (TS2367) A symbol can never equal a boolean/number/string/bigint, so ==/!=/===/!== between them is unintentional (TS2367). Scoped to the case where exactly one operand is a symbol and the other is a concrete primitive — symbol vs symbol / vs any / vs a union that could hold a symbol is left untouched, so no broad no-overlap sweep. Flips symbolType9. Full xUnit 15452/0, conformance 255 Pass.
…nts (TS2353) Excess-property checking ran for assignments but not call arguments — a fresh object literal passed to a typed parameter with extra properties was accepted. Wire CheckExcessProperties into both the non-generic argument loop and the generic call path (which skipped argument validation entirely), scoped to direct object literals (tsc's fresh-literal rule). Also: CheckExcessProperties now flattens a generic interface argument (I<T>) to its concrete members, and attributes TS2353 to the offending property's own line (not the enclosing statement's) so a multi-line call argument reports correctly. Flips symbolProperty21. Full xUnit 15452/0, conformance 256 Pass.
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.
Stacked on #1262 (base is
worktree-epic-99; will retarget tomainonce #1262 merges). Continues the TypeScript conformance work (epic #80 / #99), taking the 296-test subset baseline 250 → 256 Pass / 29 → 23 Fail. Zero regressions at every step; full xUnit 15452/0 and conformance green on every commit.#99 — from deferred epic to a working lib loader
TryResolveLibTypeconsulted by bothResolveTypeNameCore(for the type) andReportUnknownTypeName(so the name isn't flagged TS2304). ResolvesSymbol/SymbolConstructorin type position. → flips symbolProperty55, symbolType15.LibTypeLoaderparses ambient lib declarations embedded aslibdefs/*.d.tsresources (the vendored TypeScript lib is a dev-only submodule, absent at runtime) and resolves them to realTypeInfovia a throwaway checker'sExtractLibTypes.SymbolConstructornow comes from the loaded lib. Built once, cached.implements/extendsresolution consults the loaded (member-less) wrapper interfaces forString/Number/Boolean, soclass C implements String/interface I extends Stringare valid targets.var s: Stringstill resolves to the primitive — zero assignability change. → flips stringLiteralTypeIsSubtypeOfString.Checker features (found while working the Fail bucket)
implementsnow checks against inferred method return types (the validation ran before body inference, when un-annotated returns were<inferred>≈any). Narrow after-inference re-check of un-annotated methods, reported at the method line. → symbolProperty24.symbol == boolean/number/string/bigintflagged as a no-overlap comparison, scoped to symbol-vs-concrete-primitive. → symbolType9.Remaining Fail bucket is the hard core: globalThis/Intl ambient types, lib-version awareness (the 6 lib-drift tests), and the accessor getter-return-inference feature.