Implementation plan and status

This RFD is a draft. Checked items represent design or implementation work which has landed together with tests and documentation.

Phase 1: Candidate-universe contract

• Define precisely which local and upstream crates belong to the visible impl universe.
• Define how builtin, auto, negative, and specialization candidates join user-defined impl discovery.
• Define candidate-source completeness in the presence of unavailable or unsupported metadata.
• Specify stable local and external impl identities.

Phase 2: Query and index design

• Define the mandatory trait-keyed candidate API.
• Define the conservative simplified-self-type key and fallback bucket.
• Select a stable tracked index with a private tracked map and keyed backdated lookup methods for local impl indexing.
• Extend TcxDb with trait signatures, impl signatures, and relevant-impl enumeration using owned metadata values.
• Define deterministic ordering and deduplication across local, external, and fallback sources.

Phase 3: Semantic implementation

• Index local impl signatures by trait without reading unrelated trait impls during lookup.
• Import represented external trait signatures, relevant explicit-impl identities, and impl headers through separate TcxDb operations.
• Expose local impls of external traits once defining predicates are complete and merge them with the reachable external candidate source.
• Add no-false-negative simplified-self-type filtering to external metadata discovery; local indexing remains open.
• Replace the solver's direct local_impls scan with the accepted candidate API.
• Preserve incomplete-source behavior for unavailable candidate sources and unsupported individual headers.

Phase 4: Required tests

• Expose a structured query recorder from the test harness while retaining Salsa WillExecute and typed TcxDb events.
• Normalize semantic keys and internal Salsa IDs for stable assertions.
• Add cold, warm-cache, relevant-edit, and unrelated-edit goal traces.
• Land every semantic discovery test listed in the RFD.
• Compare indexed and exhaustive results over generated fixtures.
• Land Salsa event tests proving unrelated-trait edits stop at an equal keyed lookup result.
• Land signature-versus-body invalidation tests.
• Land self-head bucket invalidation tests when that index is enabled.
• Land stable external-metadata identity and reuse tests.
• Update md/design/trait-solver.md, the Trait System architecture, and the MVP limitation notes after the indexed global path lands.