Struct ProjectDb 

pub struct ProjectDb { /* private fields */ }

Stateful incremental project database.

A thin, path-keyed shell around a salsa database: file texts are salsa inputs, and every derived artifact (parse tree, HIR, symbol index, resolutions, LIR, StoryData) is a memoized tracked query with real dependency tracking and early cutoff. Both the compiler (one-shot) and LSP/IDE (long-lived) use this as their project model; editor overlays are plain input writes.

Implementations

impl ProjectDb

pub fn new() -> ProjectDb

Create an empty project database.

pub fn with_id_base(id_base: u32) -> ProjectDb

Create an empty project database whose FileIds start counting from id_base instead of 0 (issue #1580).

A long-lived host that keeps multiple independent ProjectDb instances alive at once — brink-lsp’s per-native-project extent partitioning, one db per governing brink.toml — needs every instance’s FileIds to be mutually disjoint: each db mints its own FileIds starting at 0 internally, so two dbs each holding a first-registered file would otherwise both mint FileId(0), and a caller merging per-project data into one FileId-keyed map (as brink-lsp’s cross-project ProjectAnalyses does) would silently conflate two unrelated files. Callers are responsible for choosing non-overlapping id_base ranges (e.g. a fixed stride per project index) — this constructor only seeds the counter, it does not police collisions across instances it knows nothing about.

pub fn set_file(&mut self, path: &str, source: String) -> FileId

Add or replace a file. An existing file’s text is overwritten in place (an input write); derived queries recompute lazily on next read.

Path→FileId identity is durable (#536): re-adding a path that was previously remove_filed reinstates its original FileId and salsa input, so per-file memos are overwritten in place instead of accumulating under freshly-minted dead ids.

pub fn update_file(&mut self, path: &str, source: String) -> FileId

Incrementally update a file. Identical to set_file: salsa’s dependency tracking decides what recomputes.

pub fn remove_file(&mut self, path: &str)

Remove a file from the database.

The salsa input is tombstoned, not forgotten (#536): salsa can never reclaim an input or the memos keyed on it, so the SourceFile is parked in retired with its text cleared (releasing the source and invalidating stale derived memos) while dropping out of the project file list and every path/id map. From a consumer’s view the file is gone — enumeration, lookups, and INCLUDE resolution behave exactly as if it never existed; re-adding the path reuses its original FileId.

pub fn set_entry(&mut self, path: &str) -> Option<FileId>

Set the compile entry point (for the lir_product and story_data queries). The file must already be in the database.

pub fn entry(&self) -> Option<FileId>

The current compile entry point, if any.

pub fn set_dialect(&mut self, dialect: Option<DialogueDialect>)

Set the analysis options (host manifest + external-check severity) used by the analysis and downstream queries. Register (or clear) the screenplay dialect config (#3064 B1). UNGUARDED like set_analysis_options — the salsa write stamps the revision unconditionally, so callers guard against no-op writes (IdeSession::set_dialect does).

pub fn dialect_config(&self) -> Option<&DialogueDialect>

The registered dialect config, if any.

pub fn resolved_dialect(&self) -> Option<&Arc<ResolvedDialect>>

The compiled dialect (memoized — regexes compile once per config change), if one is registered and valid.

pub fn set_analysis_options(&mut self, options: AnalysisOptions)

pub fn analysis_options(&self) -> &AnalysisOptions

The analysis options currently registered with the database.

pub fn set_native_root(&mut self, root: Option<String>)

Register the directory native .brink file keys are root-relative to (issue #1572).

A native file’s module — and therefore every DefinitionId it qualifies — is a pure function of its root-relative key (decision-log 2026-07-22 “Native module identity”). brink-driver’s discover_native already registers such keys, so a compile leaves this None and nothing changes. A consumer that must key by some other prefix — the LSP keys by absolute OS path, because every path it holds round-trips through a file:// URI — declares that prefix here, and the identity it mints then matches a real compile of the same tree byte for byte instead of embedding the machine’s directory layout. Paths not under root are unaffected.

Ink (.ink) files never consult this: their module is their file stem, which no path prefix can change.

pub fn native_root(&self) -> Option<&str>

The registered native source root, if any — see set_native_root.

pub fn set_ink_root(&mut self, root: Option<String>)

Register the directory .ink file keys are root-relative to (issue #1696) — ink’s sibling of set_native_root, consulted by hir::root_content_scope_path’s qualifier rather than by module identity.

brink-compiler/src/driver.rs’s prepare_driver registers this for every ink compile, using brink_driver::native_source_root (the same root-discovery rule native compiles already use) fed the entry path. None — no caller has registered a root — is byte-identical to the pre-#1696 world: the qualifier stays the file’s raw registered path.

pub fn ink_root(&self) -> Option<&str>

The registered ink source root, if any — see set_ink_root.

pub fn segment_count(&self, id: FileId) -> Option<usize>

The number of per-knot segments a file splits into (#3084) — pulling this warms file_segments_query, so perf instrumentation can price the segmentation toll as its own stage. None for an unknown file id.

pub fn projection(&self, id: FileId) -> Option<Arc<Projection>>

The file’s assembled, identity-joined projection (#3064 B2) — the per-segment memoized replacement for IdeSession’s retired wipe-on-every-edit projection cache. None for an unknown id.

pub fn line_contexts(&self, id: FileId) -> Option<Arc<Vec<LineContext>>>

The file’s assembled per-line contexts (#3064 B3) — per-segment memoized for ink (an edit reclassifies the edited knot’s fragment only), whole-file for native. Dialect-classified when a dialect config is registered (set_dialect).

pub fn semantic_tokens(&self, id: FileId) -> Option<Arc<Vec<RawToken>>>

The file’s assembled semantic tokens (#3064 B4) — per-segment memoized for ink with a range-free resolution-kind seam, so both shift edits and unrelated-content edits leave untouched segments’ token memos validated. Whole-file for native.

pub fn segment_manifest(&self, id: FileId) -> Option<(Vec<(String, u32)>, u32)>

The outbound-delta segment manifest (#3064 option A, ruled 2026-08-24): one entry per segment — its VERSION KEY and the first line it owns — plus the file’s total line count. The version key is the salsa tracked-struct id as index:generation: stable across shift edits (only the tracked offset field moves), changed exactly when the segment’s content changes (a new identity), and ABA-safe (slot reuse and identity-hash collisions both bump the generation). A consumer caches per-segment slices under this key, re-fetches only keys it hasn’t seen, and drops keys that leave the manifest. None for an unknown file or a non-ink file (no segment road there).

pub fn segment_line_contexts_slice( &self, id: FileId, key: &str, ) -> Option<Vec<LineContext>>

One segment’s owned line-context slice by manifest version key (#3064 option A) — concatenating every manifest entry’s slice in order reproduces line_contexts exactly (parity-gated). None when the key no longer names a live segment (the consumer’s manifest is stale — re-fetch it).

pub fn segment_semantic_tokens_slice( &self, id: FileId, key: &str, ) -> Option<Vec<RawToken>>

One segment’s owned semantic-token slice by manifest version key, token lines RELATIVE to the segment’s owned start (#3064 option A) — cached slices survive shift edits; the consumer adds the manifest’s owned-from line back at assembly.

pub fn segment_semantic_tokens_slice_fast( &self, id: FileId, key: &str, ) -> Option<Vec<RawToken>>

segment_semantic_tokens_slice’s classifier-only sibling (#3064 micro): never pulls the symbol index or resolutions — the keystroke path’s source, refined by the deferred refresh.

pub fn file_id(&self, path: &str) -> Option<FileId>

Look up a file’s ID by path.

pub fn file_path(&self, id: FileId) -> Option<&str>

Look up a file’s path by ID.

pub fn file_ids(&self) -> impl Iterator<Item = FileId>

Iterate over all registered file IDs.

pub fn file_ids_topo(&self, entry: FileId) -> Vec<FileId>

Return file IDs in topological include order (included files before the files that include them), matching ink’s INCLUDE paste semantics. Only entry and files it transitively INCLUDEs are returned — see [IncludeGraph::topological_order] (issue #815).

pub fn compilation_closure(&self) -> Vec<FileId>

The current compile closure — the exact file set codegen builds from (compilation_closure_files): an ink entry’s transitive INCLUDE closure in topological order, or every discovered .brink module for a native entry. Empty when no entry is set. Issue #3017 reads this through brink-ide/brink-web to mark files that are on disk but not in the story — absent diagnostics on such a file look identical to clean diagnostics, so the editor says so instead.

pub fn parse(&self, id: FileId) -> Option<&Parse>

Get the parse tree for a file.

pub fn parse_native(&self, id: FileId) -> Option<&Parse>

Get the native (.brink) parse tree for a file (B0.10a, the native compile seam, issue #1106). The native-frontend sibling of parse — a distinct nominal Parse type. This runs the native parser regardless of the file’s extension; the extension-based frontend dispatch that decides which parser lowering uses lives in lowered_query, so parse() stays ink-typed and untouched for the LSP/IDE ink path.

pub fn hir(&self, id: FileId) -> Option<&HirFile>

Get the HIR for a file. None for an unknown file id, or for a tracked file [is_source_file] excludes (issue #2329 review finding): this per-file accessor reads lowered_query directly, so without this gate a brink.toml/.md/.json document would still return its bogus ink-lowered HIR.

pub fn manifest(&self, id: FileId) -> Option<&SymbolManifest>

Get the symbol manifest for a file. None for an unknown file id, or for a tracked file [is_source_file] excludes — see Self::hir’s doc (issue #2329 review finding).

pub fn source(&self, id: FileId) -> Option<&str>

Get the source text for a file.

pub fn file_diagnostics(&self, id: FileId) -> Option<&[Diagnostic]>

Get per-file diagnostics (parse + lowering). None for an unknown file id, or for a tracked file [is_source_file] excludes — see Self::hir’s doc (issue #2329 review finding).

pub fn admission_diagnostics(&self, id: FileId) -> Option<&[Diagnostic]>

Get the B0.3 HIR admission validator’s output for a file (docs/hir-admission-contract.md §4.2, issue #1172) — kept separate from Self::file_diagnostics because it is non-suppressible (never routed through apply_suppressions). None for an unknown file id, or for a tracked file [is_source_file] excludes — see Self::hir’s doc (issue #2329 review finding).

pub fn suppressions(&self, id: FileId) -> Option<&Suppressions>

Get suppression directives for a file — the text-scanned brink-disable/brink-expect comments merged with the file’s HIR-derived @[allow(…)] scopes (issue #1161), i.e. parsed ∪ HIR-derived, not parsed alone.

pub fn rebuild_include_graph(&mut self)

Rebuild include graph edges for all files.

No-op since the salsa migration: the include graph is a tracked query over the full file set and is always complete. Kept so batch-loading call sites need no change.

pub fn find_cycle(&self) -> Option<Vec<FileId>>

Detect cycles in the include graph.

Returns the first cycle found as an ordered path of file IDs.

pub fn compute_projects(&self) -> Vec<(FileId, Vec<FileId>)>

Compute independent projects — the unit every editor surface scopes itself to (the LSP analyzes one project at a time, and navigation only ever sees the files of the project the cursor’s file belongs to).

Returns (root, members) pairs sorted by root FileId; each project’s members are sorted by FileId.

One rule per frontend:

• Ink groups by INCLUDE reachability — a root file plus its transitive INCLUDE closure (see IncludeGraph::compute_projects). Unchanged.
• Native .brink files are one project, all of them. Issue #1562: .brink has no INCLUDE (the module system replaced it), so running them through the ink rule made every native file its own single-file project and broke go-to-definition, find-references, completion, and diagnostics across every real native workspace. The rule here is the one compilation_closure_files already applies to codegen (decision-log “Native multi-file linking”, 2026-07-23): the discovered module set is the compilation unit, so it is also the editor’s scope. No second discovery mechanism is involved — this partitions the files the db already holds.

The two sets are disjoint, so an INCLUDE in an ink file that names a .brink target (not expressible in native, and meaningless as ink) contributes no edge: the native file is in the native project only.

pub fn is_native(&self, id: FileId) -> bool

Whether id is a native (.brink) module rather than an ink file.

pub (issue #1562 review finding) so per-root callers — brink-lsp, which needs the “does this project’s dialect axis even apply” answer for a project root — can ask it of a FileId without rederiving [crate::queries::file_language] themselves. (The off-db analyze_with_modules pass this originally served retired with option A, 2026-08-24; per-root analysis now runs Self::analysis_for_members.)

pub fn is_all_native(&self) -> bool

Whether every recognized source file (.ink or .brink) this db holds is a native (.brink) module — false for an empty db, one holding even a single ink source file, or one whose tracked files are all non-source documents. A tracked file with neither extension (a project’s own brink.toml, e.g. — issue #2318) does not count either way; see [crate::queries::project_is_all_native]’s doc for the full reasoning and the bug this exemption fixes.

The whole-db view of [crate::queries::project_is_all_native], for a caller that analyzes this db’s entire file set as one unit off-db (IdeSession, whose editor analysis runs [brink_analyzer::analyze_with_modules] over analysis_inputs). That flag is whole-project, so it is only correct when the set is entirely native: a mixed set must analyze as ink, or an ink file would get the native arm of passes that would then mis-judge it.

Distinct from is_native, which answers for one file and is what a per-project caller (brink-lsp’s analysis_loop, which analyzes each project root separately) asks of its root.

pub fn reachable_from(&self, entry: FileId) -> BTreeSet<FileId>

All files reachable from entry via the forward INCLUDE graph, entry included.

A forward DFS over INCLUDE edges (transitive). The result is a BTreeSet, so iteration order is deterministic regardless of graph internals — callers that compare or render the set get stable output.

pub fn analysis_inputs_for( &self, file_ids: &[FileId], ) -> Vec<(FileId, HirFile, SymbolManifest)>

Snapshot analysis inputs for a subset of files.

Like analysis_inputs() but filtered to the given set.

pub fn analysis_inputs(&self) -> Vec<(FileId, HirFile, SymbolManifest)>

Snapshot all analysis inputs for background analysis.

Returns (FileId, HirFile, SymbolManifest) tuples cloned out of the db, so the caller can run brink_analyzer::analyze_with_modules (with module_map, also snapshotted) without holding the lock. Issue #1526: a bare brink_analyzer::analyze() / analyze_with_options over these inputs is module-blind and mints different DefinitionIds than this db’s own queries for native .brink files — see module_map’s doc.

pub fn file_metadata(&self) -> Vec<(FileId, String, String)>

Snapshot file metadata for diagnostic publishing.

Returns (FileId, path, source) tuples for all files in the db.

pub fn symbol_index(&self) -> Arc<SymbolIndex>

The merged project-wide symbol index (layer 2, symbol_index()).

pub fn symbol_index_diagnostics(&self) -> &[Diagnostic]

Indexing diagnostics (duplicate definitions, built-in shadowing) produced alongside symbol_index.

pub fn harvest_index(&self) -> Arc<HarvestIndex>

The project-wide harvest index (layer 2, issue #2114, docs/prose-dialect-spec.md §5): every @NAME cue payload and every inline-markup span kind/attribute name written anywhere in the project, upgraded by the registered host manifest’s markup vocabulary where one is declared. The compiler-side sibling of symbol_index — a completion consumer reads this the same way it reads that index, and gets the same per-file [lowered_query] early cutoff the symbol index has: an edit backdates this memo when it backdates the symbol index’s lowered_query half, but this index also depends on the registered host manifest, so a manifest-only edit backdates this memo without touching the symbol index at all (see harvest_index_query’s own doc for the full dependency set).

pub fn harvest_completion_names(&self) -> Arc<HarvestNames>

The harvest index’s range-free completion projection (issue #2134): every harvested cue and span/attribute name, with every site’s TextRange dropped. This is the query a keystroke-driven completion path should read instead of harvest_index itself — see [harvest_completion_index_query]’s own doc for why the raw index can never Eq-cutoff.

pub fn conventions_projection(&self) -> Arc<ConventionsProjection>

The conventions projection (issue #2111, NS-T seam 1/6): every @[convention] handler declared in the project’s one configured conventions module, ascending by order — “THE SOLE EDITOR INTERCHANGE” the design-backport comment on #2111 names (docs/decision-log.md 2026-08-03). Reads the [project] conventions pointer, the project module map, the resolved conventions module’s transitive IMPORT closure (import_closure_query, issue #2111 finding 3), and every file in that closure’s own lowered_query output — see conventions_projection_query’s doc for the exact dependency set, and brink_ir::ConventionsProjection’s doc for the one part of #2111 this still does not deliver: it is not yet serialized into .inkb/StoryData (the attach schema IS now resolved to its fields and types, not merely a struct name — that gap closed in the 2026-08-04 continuation).

pub fn module_map(&self) -> &BTreeMap<FileId, ResolvedModule>

Every file’s resolved module (M-1, docs/modules-spec.md §1/§5) — the map that qualifies DefinitionId identity, built here from file stems, #@module declarations, the INCLUDE graph, and (for native .brink files) the path-derived story::… module.

Exposed (issue #1526) for callers that must run [brink_analyzer::analyze_with_modules] outside the db — the LSP’s background analysis pass and analysis_inputs consumers generally — so their DefinitionIds match the ones this db’s per-def queries (effects, signature, infer_body) are keyed by. Identity is minted here and nowhere else.

The map’s diagnostics half is module_map_diagnostics — an off-db analyze_with_modules pass has to fold it back in itself (issue #1553).

pub fn module_map_diagnostics(&self) -> &[Diagnostic]

Stem-collision diagnostics (E085) produced alongside module_map: a file with no #@module whose stem is some other file’s declared module name.

A db-driven compile picks these up through symbol_index_diagnostics, which folds them in. A caller that instead runs [brink_analyzer::analyze_with_modules] outside the db (the LSP’s background pass, IdeSession’s editor analysis) gets only the analyzer’s own diagnostics, so before issue #1553 the collision was silently dropped on every editor surface. Such callers must snapshot this alongside module_map and extend their result with the entries belonging to their file set.

pub fn resolve( &self, id: FileId, ) -> Option<(Arc<Vec<ResolvedRef>>, &[Diagnostic])>

One file’s resolved references + resolution diagnostics (layer 2, resolve(FileId)).

pub fn signature(&self, def: DefinitionId) -> Option<Arc<Sig>>

Per-declaration signature stub (layer 2, signature(def)). None for an unknown definition id.

pub fn local_signature(&self, id: FileId, def: DefinitionId) -> Option<Arc<Sig>>

Signature stub for a local (Param/Temp) def, declared in id (issue #530): the per-file locals path signature itself can’t take — see local_signature_query’s doc for why a local’s DefinitionId needs a caller-supplied file. None for an unknown file id or a def not declared as a local in that file (including a declaration id — those stay signature’s job).

pub fn analysis_for_members(&self, members: &[FileId]) -> &AnalysisResult

Full cross-file analysis over all files, honoring the registered AnalysisOptions. Memoized; module-aware — identical to brink_analyzer::analyze_with_modules over analysis_inputs and module_map by construction. For native .brink files this is not identical to analyze_with_options (module-blind), which mints different DefinitionIds — see module_map’s doc (issue #1526). Subset analysis for one project root’s member files (option A total, 2026-08-24) — the retired analyze_with_modules monolith’s composition, relocated into the member-set-keyed subset_analysis_query (see its doc). Members are canonicalized (sorted, deduped) before interning, so caller ordering never mints a distinct memo. For the whole file set, prefer analysis — the FG-decomposed incremental chain.

pub fn analysis(&self) -> &AnalysisResult

pub fn resolutions_index(&self) -> Arc<ResolvedProject>

Index + resolutions, no diagnostics (issue #632 / FG-3 — the RESOLUTIONS/INDEX half of analysis, split off from the diagnostics half so a diagnostics-only AnalysisOptions edit leaves this Arc’s pointer identity untouched).

pub fn per_file_diagnostics(&self, id: FileId) -> Option<Arc<Vec<Diagnostic>>>

One file’s per-file diagnostic contributors — structural validation, the dialect gate, and (brink dialect only) annotation-content checks (issue #632 / FG-3). None for an unknown file id. A body edit in a different file leaves this Arc’s pointer identity untouched.

pub fn file_value_meta( &self, id: FileId, ) -> Option<Arc<BTreeMap<DefinitionId, SymbolMeta>>>

One file’s VAR/CONST/LIST initializer/doc enrichment (issue #750 / FG-3 completion) — purely presentational symbol_meta entries, no diagnostics. None for an unknown file id. A body edit in a different file leaves this Arc’s pointer identity untouched.

pub fn file_call_site_diagnostics( &self, id: FileId, ) -> Option<Arc<Vec<Diagnostic>>>

One file’s external call-site literal checks (E041/E042, issue #750 / FG-3 completion). None for an unknown file id; empty when the external_check severity is Off. A body edit in a different file leaves this Arc’s pointer identity untouched.

pub fn call_site_metas(&self) -> Arc<BTreeMap<String, SymbolMeta>>

The range-free, name-keyed external metas feeding the per-file call-site checks (issue #750 / FG-3 completion) — the cutoff seam between the (often re-executed, full-ranged-index-reading) enrichment pass and every file’s call-site memo. Exposed for the dependency-edge tests; pointer identity across an edit proves the seam backdated.

pub fn diagnostics(&self, id: FileId) -> Option<&[Diagnostic]>

Per-file diagnostics including this file’s share of analysis diagnostics (layer 3, diagnostics(FileId)). Raw — no suppression filtering.

pub fn type_inference(&self) -> &InferenceResult

Whole-project type inference (TM-1, typed-mode-spec §2/§9 step 1). Advisory-only substrate: infer_body/type_diagnostics are thin per-def/per-file views over this. Lazy — nothing in story_data, lir_product, or diagnostics reads it, so calling this (directly or via infer_body/type_diagnostics) is the only thing that triggers the underlying computation.

pub fn infer_body(&self, def: DefinitionId) -> Option<Arc<BodyTypes>>

Per-def inferred body types (infer_body(def)). None for a def with no inferable body (not a knot/stitch, or an unknown id).

pub fn inferred_signature(&self, def: DefinitionId) -> Option<Arc<InferredSig>>

Per-def inferred signature (inferred_signature(def), FG-2 issue #631) — the firewall-facing per-def view: params + return type only, no locals, no ranges. This is the boundary TM-2’s annotation-override consumer reads. None for a def with no inferable body (not a knot/stitch, or an unknown id) — same None contract as signature/infer_body.

pub fn effects(&self, def: DefinitionId) -> Option<Arc<EffectRow>>

Per-def effect row (effects(def), T2-1, docs/effects-spec.md §2/§4, issue #860) — the advisory {reads, writes, calls} summary of the atomic effects def (and everything it transitively calls) may perform, sited beside inferred_signature. Conservative-total (spec §3): the row over-reports, never under-reports; a call through a function value or an unknown callee makes it pessimal (EffectRow::opaque). None for a def with no inferable body (not a knot/stitch, or an unknown id) — same contract as inferred_signature.

Advisory-only: nothing in story_data/lir_product/diagnostics reads this, so the row is additive metadata that leaves compiled output byte-identical. Lazy — calling this is the only thing that triggers the underlying atom harvest + per-SCC fixpoint.

pub fn ufcs_verdict( &self, file: FileId, range: TextRange, ) -> Option<&UfcsVerdict>

The B3a UFCS resolution verdict for the call site at range in file (issue #1507) — reads the same memoized ufcs_resolution_query (#1506) LIR lowering already shares, rather than re-running the analyzer’s ufcs pass a second time for IDE hover/go-to-def. None when the pass recorded no verdict at this exact range: not a UFCS-shaped call site, an unresolved one (already diagnosed E140–E143 elsewhere), or the project has no dotted-callee call anywhere (ufcs_resolution_query’s own laziness gate).

pub fn ufcs_call_sites_for_target( &self, target: DefinitionId, ) -> Vec<(FileId, TextRange)>

Every UFCS call site (recv.verb(args)) whose verdict desugars to a free function targeting target, project-wide (issue #1539) — reads the same memoized ufcs_resolution_query table ufcs_verdict does. The find_references/ rename counterpart to that single-site lookup: renaming or listing references to a free function must also reach every UFCS call site that resolves to it, not just its plain ResolutionMap references.

pub fn type_diagnostics(&self, id: FileId) -> Option<&[Diagnostic]>

Per-file type diagnostics (type_diagnostics(FileId)). Advisory-only in this slice — always empty (see type_diagnostics_query’s docs).

pub fn lir_product(&self) -> Option<&LirProduct>

Whole-project LIR lowering (layer 3). None until an entry point is set via set_entry.

pub fn has_errors(&self) -> bool

Whether the project has at least one Error-severity diagnostic after suppression filtering (issue #791 / FG-4a) — the narrow boolean projection lir_product’s gate reads instead of the full diagnostics vector. false (never None) when no entry point is set, matching partition_diagnostics’s empty-errors default in that case. Exposed for the dependency-edge tests; a diagnostics-content edit that leaves this boolean unchanged proves the cutoff seam backdated.

pub fn story_data(&self) -> Option<&CompileProduct>

Whole-project compile to brink_format::StoryData (layer 3, story_data()). None until an entry point is set via set_entry.

Trait Implementations

impl Default for ProjectDb

fn default() -> ProjectDb

Returns the “default value” for a type.