Module db 

Salsa-based incremental computation layer.

After the mir-analyzer 0.22 migration, semantic analysis (Pass 2, class issues, references) is owned by AnalysisSession in DocumentStore. Salsa here is now responsible for the LSP-side compact representations only: parsed AST cache (parse), per-file declaration index (index), workspace aggregation (workspace_index). These power cross-file LSP features (workspace symbols, document symbols, find-implementations, hover from index) that don’t require the analyzer’s full type system.

Modules

index

file_index salsa query — derives a compact FileIndex from a parsed document. Depends on parsed_doc, so editing a file reparses once and the index re-extracts from the new AST.

mir_queries

php-lsp salsa queries on mir’s shared database: keyed on mir’s SourceFile input and run over &dyn MirDatabase, so they live in the single converged db. The Arc-wrapper types they return (ParsedArc, IndexArc, SymbolMapArc) live in parse.rs / index.rs / symbol_map.rs.

parse

The parsed_doc salsa query: parses a SourceFile into an Arc<ParsedDoc> under salsa memoization. Downstream queries (file_index, symbol_map, semantic diagnostics) depend on this one, so each file is parsed at most once per revision.

symbol_map

symbol_map salsa query — derives a SymbolMap from a parsed document.

workspace_index

workspace_index salsa query — aggregates every file’s FileIndex into a single structure with pre-built reverse maps.