{ "schemaVersion": 1, "purpose": "Frozen public controls for a secondary source-local normalized term-frequency density tie-break.", "formula": "For each existing normalized query group g, tf_g counts normalized body-word occurrences whose existing lexicalAliases(normalizedBodyWord) intersect the aliases in g. Count that body-word occurrence at most once per group, regardless of alias-expansion multiplicity. density = sum_g sqrt(tf_g) / sqrt(max(1, normalizedBodyWordCount)).", "ordering": "Lexicographic comparison: preserve the existing primary rank, then compare descending density only for equal primary ranks, then preserve original chunk ordinal. Never add density to or otherwise alter the primary score.", "scope": "Synthetic source snippets and relative file paths. Expected order is a consequence of the frozen lexical formula and existing primary evidence classes; it does not assert semantic correctness.", "cases": [ { "id": "density-breaks-primary-tie", "query": "invoice approval", "chunks": [ { "id": "long-noise", "filePath": "src/long-noise.ts", "content": "invoice approval billing invoice mail report network" }, { "id": "compact", "filePath": "src/compact.ts", "content": "invoice approval" } ], "expectedIds": ["compact", "long-noise"], "expectedPathPatterns": ["src/compact.ts", "src/long-noise.ts"], "expectedProperties": [ "primary evidence ties", "unmatched body vocabulary increases the denominator" ] }, { "id": "declaration-primary-dominates-density", "query": "invoice approval", "chunks": [ { "id": "declared-implementation", "filePath": "src/invoice-approval.ts", "content": "export function apply(invoice) { const approval = true; return applyPolicy(invoice, approval); } billing mail report network logging validation storage adapter response", "metadata": { "functionName": "invoiceApproval" } }, { "id": "dense-body", "filePath": "src/repeated.ts", "content": "invoice approval invoice approval invoice approval" } ], "expectedIds": ["declared-implementation", "dense-body"], "expectedPathPatterns": ["src/invoice-approval.ts", "src/repeated.ts"], "expectedProperties": [ "declaration evidence remains primary", "density cannot promote a weaker primary candidate" ] }, { "id": "exact-anchor-primary-dominates-density", "query": "InvoicePolicy invoice policy", "chunks": [ { "id": "exact-anchor", "filePath": "src/invoice-policy.ts", "content": "invoice policy resolves a rule through a versioned registry and validates the result against the selected tenant before returning it", "metadata": { "symbolName": "InvoicePolicy" } }, { "id": "dense-body", "filePath": "src/repeated-policy.ts", "content": "invoice policy invoice policy invoice policy" } ], "expectedIds": ["exact-anchor", "dense-body"], "expectedPathPatterns": ["src/invoice-policy.ts", "src/repeated-policy.ts"], "expectedProperties": [ "exact indexed anchor remains primary", "body density is only a secondary tie-break" ] }, { "id": "implementation-primary-dominates-test-density", "query": "invoice approval", "chunks": [ { "id": "implementation", "filePath": "src/invoice-approval.ts", "content": "invoice approval validates the request and records the accepted result after checking the account state and policy" }, { "id": "test", "filePath": "src/__tests__/invoice-approval.test.ts", "content": "invoice approval invoice approval invoice approval invoice approval" } ], "expectedIds": ["implementation", "test"], "expectedPathPatterns": ["src/invoice-approval.ts", "src/__tests__/invoice-approval.test.ts"], "expectedProperties": [ "test-file primary penalty remains effective", "repeated test-body terms do not override primary order" ] }, { "id": "own-implementation-outranks-repeated-type-mentions", "query": "cache refresh", "chunks": [ { "id": "own-implementation", "filePath": "src/cache-refresh.ts", "content": "refreshCache updates the cache after validating the new value and then returns the stored entry", "metadata": { "functionName": "refreshCache" } }, { "id": "repeated-type-mentions", "filePath": "src/cache-types.ts", "content": "type CacheRefresh = Cache & Refresh & CacheRefresh & CacheRefresh & CacheRefresh" } ], "expectedIds": ["own-implementation", "repeated-type-mentions"], "expectedPathPatterns": ["src/cache-refresh.ts", "src/cache-types.ts"], "expectedProperties": [ "own declaration evidence remains primary", "repetition in a type body is lexical evidence only" ] }, { "id": "tiny-wrapper-does-not-beat-stronger-declaration", "query": "invoice approval", "chunks": [ { "id": "implementation", "filePath": "src/invoice-approval.ts", "content": "invoice approval validates the account, checks the pending state, records the decision, and returns the persisted result", "metadata": { "functionName": "invoiceApproval" } }, { "id": "tiny-wrapper", "filePath": "src/wrapper.ts", "content": "invoice approval" } ], "expectedIds": ["implementation", "tiny-wrapper"], "expectedPathPatterns": ["src/invoice-approval.ts", "src/wrapper.ts"], "expectedProperties": [ "declaration primary rank dominates the wrapper's higher density", "shortness alone does not imply a better result" ] }, { "id": "metadata-repetition-is-not-body-frequency", "query": "invoice approval", "chunks": [ { "id": "repeated-metadata-first", "filePath": "src/a.ts", "content": "invoice approval validates a request", "metadata": { "componentName": "InvoiceApproval InvoiceApproval InvoiceApproval" } }, { "id": "single-metadata-second", "filePath": "src/b.ts", "content": "invoice approval validates a request", "metadata": { "componentName": "InvoiceApproval" } } ], "expectedIds": ["repeated-metadata-first", "single-metadata-second"], "expectedPathPatterns": ["src/a.ts", "src/b.ts"], "expectedProperties": [ "body density is identical", "metadata repetition does not increase term frequency", "original ordinal breaks the tie" ] }, { "id": "body-inflection-alias-matching-preserves-ordinal", "query": "archive policy", "chunks": [ { "id": "inflected-first", "filePath": "src/a.ts", "content": "archiving policy" }, { "id": "canonical-second", "filePath": "src/b.ts", "content": "archive policy" } ], "expectedIds": ["inflected-first", "canonical-second"], "expectedPathPatterns": ["src/a.ts", "src/b.ts"], "expectedProperties": [ "existing body lexical aliases are intersected with the query-group aliases", "archiving and archive each contribute one occurrence to the archive group", "both bodies have equal density and original ordinal is preserved" ] }, { "id": "overlapping-aliases-count-one-occurrence-per-group", "query": "archive archiving policy", "chunks": [ { "id": "archiving-first", "filePath": "src/a.ts", "content": "archiving policy" }, { "id": "archive-second", "filePath": "src/b.ts", "content": "archive policy" } ], "expectedIds": ["archiving-first", "archive-second"], "expectedPathPatterns": ["src/a.ts", "src/b.ts"], "expectedProperties": [ "archive and archiving belong to one existing alias group", "one body occurrence contributes once to that group", "equal density preserves ordinal" ] }, { "id": "repeat-one-group-does-not-win-on-raw-count", "query": "archive policy", "chunks": [ { "id": "balanced", "filePath": "src/balanced.ts", "content": "archive policy" }, { "id": "repeated-archive", "filePath": "src/repeated.ts", "content": "archive archive policy" } ], "expectedIds": ["balanced", "repeated-archive"], "expectedPathPatterns": ["src/balanced.ts", "src/repeated.ts"], "expectedProperties": [ "balanced density is 2/sqrt(2)", "repeated density is (sqrt(2)+1)/sqrt(3)", "square-root TF and body length are both applied" ] }, { "id": "unrelated-billing-mail-report-network-words-only-add-length", "query": "archive policy", "chunks": [ { "id": "compact", "filePath": "src/compact.ts", "content": "archive policy" }, { "id": "unrelated-vocabulary", "filePath": "src/unrelated.ts", "content": "archive policy billing invoice mail report network" } ], "expectedIds": ["compact", "unrelated-vocabulary"], "expectedPathPatterns": ["src/compact.ts", "src/unrelated.ts"], "expectedProperties": [ "billing, mail, report, and network words are not query clues", "unmatched body words increase the denominator only" ] }, { "id": "overlapping-chunks-are-scored-separately", "query": "invoice approval", "chunks": [ { "id": "long-overlap", "filePath": "src/ledger.ts", "startLine": 1, "content": "invoice approval\ninvoice approval\nreturn approved\nledger entries retain stable boundaries" }, { "id": "short-overlap", "filePath": "src/ledger.ts", "startLine": 2, "content": "invoice approval\nreturn approved" } ], "expectedIds": ["short-overlap"], "expectedPathPatterns": ["src/ledger.ts"], "expectedProperties": [ "the second chunk exactly matches overlapping source lines", "rescue returns one representative per source file", "density ranks body evidence per chunk before file deduplication", "query term counts are not aggregated across chunks" ] }, { "id": "identical-full-bodies-preserve-original-ordinal", "query": "invoice approval", "chunks": [ { "id": "z-first", "filePath": "src/z.ts", "content": "invoice approval validates the request" }, { "id": "a-second", "filePath": "src/a.ts", "content": "invoice approval validates the request" } ], "expectedIds": ["z-first", "a-second"], "expectedPathPatterns": ["src/z.ts", "src/a.ts"], "expectedProperties": [ "identical normalized bodies have equal density", "lexicographic path or id does not replace ordinal tie-breaking" ] }, { "id": "body-evidence-beats-metadata-only-density-zero", "query": "invoice approval", "chunks": [ { "id": "metadata-only", "filePath": "src/metadata.ts", "content": " \n ", "metadata": { "componentName": "InvoiceApproval" } }, { "id": "body-present", "filePath": "src/body.ts", "content": "invoice approval", "metadata": { "componentName": "InvoiceApproval" } } ], "expectedIds": ["body-present", "metadata-only"], "expectedPathPatterns": ["src/body.ts", "src/metadata.ts"], "expectedDensityById": { "metadata-only": 0 }, "expectedProperties": [ "primary evidence ties", "empty body uses max(1, body length)", "metadata does not enter the density numerator or denominator" ] }, { "id": "empty-metadata-only-bodies-have-finite-zero-density", "query": "invoice approval", "chunks": [ { "id": "empty-first", "filePath": "src/empty-a.ts", "content": "", "metadata": { "componentName": "InvoiceApproval" } }, { "id": "empty-second", "filePath": "src/empty-b.ts", "content": " \n ", "metadata": { "componentName": "InvoiceApproval" } } ], "expectedIds": ["empty-first", "empty-second"], "expectedPathPatterns": ["src/empty-a.ts", "src/empty-b.ts"], "expectedDensityById": { "empty-first": 0, "empty-second": 0 }, "expectedProperties": [ "density is finite and exactly zero", "equal primary evidence and density preserve ordinal" ] }, { "id": "equal-primary-tiny-wrapper-wins-density-with-explicit-limitation", "query": "archive policy", "chunks": [ { "id": "tiny-wrapper", "filePath": "src/wrapper.ts", "content": "archive policy" }, { "id": "repeated-long-body", "filePath": "src/implementation.ts", "content": "archive archive policy policy filler filler filler filler" } ], "expectedIds": ["tiny-wrapper", "repeated-long-body"], "expectedPathPatterns": ["src/wrapper.ts", "src/implementation.ts"], "limitation": "Primary lexical evidence ties, so the fixed formula ranks the short wrapper first (sqrt(1)+sqrt(1) over sqrt(2), versus sqrt(2)+sqrt(2) over sqrt(8)). This order is not evidence that the wrapper is semantically more useful; preserving this counterexample prevents an unsupported semantic claim.", "expectedProperties": [ "equal primary ranks invoke density as a separate secondary comparison", "the higher-density result may be a wrapper", "density does not establish semantic usefulness" ] }, { "id": "ambiguous-primary-density-order-is-not-semantic-proof", "query": "invoice approval", "chunks": [ { "id": "comment-repetition", "filePath": "src/comment.ts", "content": "// invoice approval invoice approval invoice approval" }, { "id": "implementation", "filePath": "src/implementation.ts", "content": "export function approve(invoice) { return policyAllowsApproval(invoice); }" } ], "expectedIds": ["comment-repetition", "implementation"], "expectedPathPatterns": ["src/comment.ts", "src/implementation.ts"], "limitation": "Primary lexical evidence ties. The fixed formula places the repeated comment first; the available evidence cannot establish whether the comment or implementation is semantically more useful. This expected order is algorithmic, not a semantic winner claim.", "expectedProperties": [ "ambiguous exact-primary tie is retained explicitly", "density does not prove action meaning or semantic relevance" ] } ], "sourceValidationCases": [ { "id": "stale-sources-within-nomination-pool-are-skipped", "query": "invoice approval", "staleChunkGenerator": { "count": 30, "idPattern": "stale-{index}", "filePathPattern": "src/stale-{index}.ts", "content": "invoice approval" }, "currentChunks": [ { "id": "current", "filePath": "src/current.ts", "content": "invoice approval validates current source" } ], "mutateFilePathPattern": "src/stale-{index}.ts", "replacementSource": "export const unrelated = true;", "expectedIds": ["current"], "expectedProperties": [ "31 candidates fit inside the existing nomination pool", "stale sources are excluded before results are returned", "density cannot bypass source validation", "no beyond-window refill is required" ] }, { "id": "changed-only-source-cannot-supply-density-evidence", "query": "invoice approval", "chunks": [ { "id": "changed", "filePath": "src/changed.ts", "content": "invoice approval validates the indexed source" } ], "mutateFilePath": "src/changed.ts", "replacementSource": "export const unrelated = true;", "expectedIds": [], "expectedProperties": [ "changed source is invalidated by the existing harness", "density cannot bypass source integrity checks" ] } ] }