sensemaking 0.24.6 → 0.25.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +20 -11
- package/dist/cjs/cli/build.d.cts +3 -0
- package/dist/cjs/cli/build.d.ts +3 -0
- package/dist/cjs/cli/build.js +195 -0
- package/dist/cjs/cli/build.js.map +1 -0
- package/dist/cjs/cli/index.d.cts +7 -6
- package/dist/cjs/cli/index.d.ts +7 -6
- package/dist/cjs/cli/index.js +12 -6
- package/dist/cjs/cli/index.js.map +1 -1
- package/dist/cjs/cli/init.js +2 -2
- package/dist/cjs/cli/init.js.map +1 -1
- package/dist/cjs/cli/map.js +7 -2
- package/dist/cjs/cli/map.js.map +1 -1
- package/dist/cjs/cli/named.js +63 -21
- package/dist/cjs/cli/named.js.map +1 -1
- package/dist/cjs/cli/path.js +7 -2
- package/dist/cjs/cli/path.js.map +1 -1
- package/dist/cjs/cli/peek.js +7 -2
- package/dist/cjs/cli/peek.js.map +1 -1
- package/dist/cjs/cli/related.js +55 -6
- package/dist/cjs/cli/related.js.map +1 -1
- package/dist/cjs/cli/search.js +54 -15
- package/dist/cjs/cli/search.js.map +1 -1
- package/dist/cjs/cli/shared.d.cts +9 -3
- package/dist/cjs/cli/shared.d.ts +9 -3
- package/dist/cjs/cli/shared.js +109 -74
- package/dist/cjs/cli/shared.js.map +1 -1
- package/dist/cjs/cli/sql.js +8 -3
- package/dist/cjs/cli/sql.js.map +1 -1
- package/dist/cjs/cli/status.js +142 -50
- package/dist/cjs/cli/status.js.map +1 -1
- package/dist/cjs/cli/types.d.cts +4 -2
- package/dist/cjs/cli/types.d.ts +4 -2
- package/dist/cjs/cli.js +10 -4
- package/dist/cjs/cli.js.map +1 -1
- package/dist/cjs/commands/map.js +10 -6
- package/dist/cjs/commands/map.js.map +1 -1
- package/dist/cjs/commands/peek.js +10 -6
- package/dist/cjs/commands/peek.js.map +1 -1
- package/dist/cjs/commands/related.d.cts +1 -0
- package/dist/cjs/commands/related.d.ts +1 -0
- package/dist/cjs/commands/related.js +56 -21
- package/dist/cjs/commands/related.js.map +1 -1
- package/dist/cjs/commands/scope.js +1 -1
- package/dist/cjs/commands/scope.js.map +1 -1
- package/dist/cjs/commands/search.js +102 -84
- package/dist/cjs/commands/search.js.map +1 -1
- package/dist/cjs/config/index.d.cts +1 -1
- package/dist/cjs/config/index.d.ts +1 -1
- package/dist/cjs/config/index.js.map +1 -1
- package/dist/cjs/config/load.d.cts +6 -1
- package/dist/cjs/config/load.d.ts +6 -1
- package/dist/cjs/config/load.js +38 -1
- package/dist/cjs/config/load.js.map +1 -1
- package/dist/cjs/config/types.d.cts +2 -1
- package/dist/cjs/config/types.d.ts +2 -1
- package/dist/cjs/config/types.js +1 -1
- package/dist/cjs/config/types.js.map +1 -1
- package/dist/cjs/config/validate.js +20 -0
- package/dist/cjs/config/validate.js.map +1 -1
- package/dist/cjs/embed/query.d.cts +2 -1
- package/dist/cjs/embed/query.d.ts +2 -1
- package/dist/cjs/embed/query.js +186 -109
- package/dist/cjs/embed/query.js.map +1 -1
- package/dist/cjs/errors.d.cts +1 -1
- package/dist/cjs/errors.d.ts +1 -1
- package/dist/cjs/errors.js.map +1 -1
- package/dist/cjs/features/embed.js.map +1 -1
- package/dist/cjs/graph/traverse.js +11 -6
- package/dist/cjs/graph/traverse.js.map +1 -1
- package/dist/cjs/index.d.cts +3 -3
- package/dist/cjs/index.d.ts +3 -3
- package/dist/cjs/index.js +3 -0
- package/dist/cjs/index.js.map +1 -1
- package/dist/cjs/lib/atomic-write.js +20 -6
- package/dist/cjs/lib/atomic-write.js.map +1 -1
- package/dist/cjs/lib/serial-query.d.cts +1 -0
- package/dist/cjs/lib/serial-query.d.ts +1 -0
- package/dist/cjs/lib/serial-query.js +22 -0
- package/dist/cjs/lib/serial-query.js.map +1 -0
- package/dist/cjs/scan/index.d.cts +5 -0
- package/dist/cjs/scan/index.d.ts +5 -0
- package/dist/cjs/scan/index.js +12 -4
- package/dist/cjs/scan/index.js.map +1 -1
- package/dist/cjs/scan/pool.d.cts +5 -0
- package/dist/cjs/scan/pool.d.ts +5 -0
- package/dist/cjs/scan/pool.js +167 -55
- package/dist/cjs/scan/pool.js.map +1 -1
- package/dist/cjs/store/builder.d.cts +6 -1
- package/dist/cjs/store/builder.d.ts +6 -1
- package/dist/cjs/store/builder.js +135 -5
- package/dist/cjs/store/builder.js.map +1 -1
- package/dist/cjs/store/duckdb/lexical.d.cts +2 -0
- package/dist/cjs/store/duckdb/lexical.d.ts +2 -0
- package/dist/cjs/store/duckdb/lexical.js +139 -7
- package/dist/cjs/store/duckdb/lexical.js.map +1 -1
- package/dist/cjs/store/duckdb/open.d.cts +3 -3
- package/dist/cjs/store/duckdb/open.d.ts +3 -3
- package/dist/cjs/store/duckdb/open.js +61 -42
- package/dist/cjs/store/duckdb/open.js.map +1 -1
- package/dist/cjs/store/duckdb/reconcile.js +2 -2
- package/dist/cjs/store/duckdb/reconcile.js.map +1 -1
- package/dist/cjs/store/duckdb/vectors.d.cts +1 -0
- package/dist/cjs/store/duckdb/vectors.d.ts +1 -0
- package/dist/cjs/store/duckdb/vectors.js +55 -43
- package/dist/cjs/store/duckdb/vectors.js.map +1 -1
- package/dist/cjs/store/index.d.cts +23 -4
- package/dist/cjs/store/index.d.ts +23 -4
- package/dist/cjs/store/index.js +224 -15
- package/dist/cjs/store/index.js.map +1 -1
- package/dist/cjs/store/open.d.cts +9 -1
- package/dist/cjs/store/open.d.ts +9 -1
- package/dist/cjs/store/open.js +479 -128
- package/dist/cjs/store/open.js.map +1 -1
- package/dist/cjs/store/reconcile.js +93 -63
- package/dist/cjs/store/reconcile.js.map +1 -1
- package/dist/cjs/store/sqlite/open.d.cts +3 -3
- package/dist/cjs/store/sqlite/open.d.ts +3 -3
- package/dist/cjs/store/sqlite/open.js +44 -36
- package/dist/cjs/store/sqlite/open.js.map +1 -1
- package/dist/cjs/store/turso/open.d.cts +3 -3
- package/dist/cjs/store/turso/open.d.ts +3 -3
- package/dist/cjs/store/turso/open.js +83 -66
- package/dist/cjs/store/turso/open.js.map +1 -1
- package/dist/cjs/store/turso/store.d.cts +3 -1
- package/dist/cjs/store/turso/store.d.ts +3 -1
- package/dist/cjs/store/turso/store.js +46 -31
- package/dist/cjs/store/turso/store.js.map +1 -1
- package/dist/cjs/store/types.d.cts +12 -3
- package/dist/cjs/store/types.d.ts +12 -3
- package/dist/cjs/store/types.js.map +1 -1
- package/dist/cjs/watch.js +155 -35
- package/dist/cjs/watch.js.map +1 -1
- package/dist/esm/cli/build.d.ts +3 -0
- package/dist/esm/cli/build.js +18 -0
- package/dist/esm/cli/build.js.map +1 -0
- package/dist/esm/cli/index.d.ts +7 -6
- package/dist/esm/cli/index.js +8 -6
- package/dist/esm/cli/index.js.map +1 -1
- package/dist/esm/cli/init.js +2 -2
- package/dist/esm/cli/init.js.map +1 -1
- package/dist/esm/cli/map.js +8 -2
- package/dist/esm/cli/map.js.map +1 -1
- package/dist/esm/cli/named.js +38 -14
- package/dist/esm/cli/named.js.map +1 -1
- package/dist/esm/cli/path.js +8 -2
- package/dist/esm/cli/path.js.map +1 -1
- package/dist/esm/cli/peek.js +8 -2
- package/dist/esm/cli/peek.js.map +1 -1
- package/dist/esm/cli/related.js +21 -3
- package/dist/esm/cli/related.js.map +1 -1
- package/dist/esm/cli/search.js +28 -7
- package/dist/esm/cli/search.js.map +1 -1
- package/dist/esm/cli/shared.d.ts +9 -3
- package/dist/esm/cli/shared.js +44 -24
- package/dist/esm/cli/shared.js.map +1 -1
- package/dist/esm/cli/sql.js +8 -2
- package/dist/esm/cli/sql.js.map +1 -1
- package/dist/esm/cli/status.js +166 -111
- package/dist/esm/cli/status.js.map +1 -1
- package/dist/esm/cli/types.d.ts +4 -2
- package/dist/esm/cli/types.js.map +1 -1
- package/dist/esm/cli.js +9 -4
- package/dist/esm/cli.js.map +1 -1
- package/dist/esm/commands/map.js +5 -1
- package/dist/esm/commands/map.js.map +1 -1
- package/dist/esm/commands/peek.js +5 -1
- package/dist/esm/commands/peek.js.map +1 -1
- package/dist/esm/commands/related.d.ts +1 -0
- package/dist/esm/commands/related.js +19 -10
- package/dist/esm/commands/related.js.map +1 -1
- package/dist/esm/commands/scope.js +1 -1
- package/dist/esm/commands/scope.js.map +1 -1
- package/dist/esm/commands/search.js +11 -10
- package/dist/esm/commands/search.js.map +1 -1
- package/dist/esm/config/index.d.ts +1 -1
- package/dist/esm/config/index.js.map +1 -1
- package/dist/esm/config/load.d.ts +6 -1
- package/dist/esm/config/load.js +32 -3
- package/dist/esm/config/load.js.map +1 -1
- package/dist/esm/config/types.d.ts +2 -1
- package/dist/esm/config/types.js +1 -1
- package/dist/esm/config/types.js.map +1 -1
- package/dist/esm/config/validate.js +20 -0
- package/dist/esm/config/validate.js.map +1 -1
- package/dist/esm/embed/query.d.ts +2 -1
- package/dist/esm/embed/query.js +40 -34
- package/dist/esm/embed/query.js.map +1 -1
- package/dist/esm/errors.d.ts +1 -1
- package/dist/esm/errors.js.map +1 -1
- package/dist/esm/features/embed.js.map +1 -1
- package/dist/esm/graph/traverse.js +5 -1
- package/dist/esm/graph/traverse.js.map +1 -1
- package/dist/esm/index.d.ts +3 -3
- package/dist/esm/index.js +1 -1
- package/dist/esm/index.js.map +1 -1
- package/dist/esm/lib/atomic-write.js +14 -6
- package/dist/esm/lib/atomic-write.js.map +1 -1
- package/dist/esm/lib/serial-query.d.ts +1 -0
- package/dist/esm/lib/serial-query.js +13 -0
- package/dist/esm/lib/serial-query.js.map +1 -0
- package/dist/esm/scan/index.d.ts +5 -0
- package/dist/esm/scan/index.js +7 -3
- package/dist/esm/scan/index.js.map +1 -1
- package/dist/esm/scan/pool.d.ts +5 -0
- package/dist/esm/scan/pool.js +41 -16
- package/dist/esm/scan/pool.js.map +1 -1
- package/dist/esm/store/builder.d.ts +6 -1
- package/dist/esm/store/builder.js +19 -0
- package/dist/esm/store/builder.js.map +1 -1
- package/dist/esm/store/duckdb/lexical.d.ts +2 -0
- package/dist/esm/store/duckdb/lexical.js +32 -4
- package/dist/esm/store/duckdb/lexical.js.map +1 -1
- package/dist/esm/store/duckdb/open.d.ts +3 -3
- package/dist/esm/store/duckdb/open.js +16 -7
- package/dist/esm/store/duckdb/open.js.map +1 -1
- package/dist/esm/store/duckdb/reconcile.js +2 -2
- package/dist/esm/store/duckdb/reconcile.js.map +1 -1
- package/dist/esm/store/duckdb/vectors.d.ts +1 -0
- package/dist/esm/store/duckdb/vectors.js +18 -14
- package/dist/esm/store/duckdb/vectors.js.map +1 -1
- package/dist/esm/store/index.d.ts +23 -4
- package/dist/esm/store/index.js +80 -15
- package/dist/esm/store/index.js.map +1 -1
- package/dist/esm/store/open.d.ts +9 -1
- package/dist/esm/store/open.js +175 -84
- package/dist/esm/store/open.js.map +1 -1
- package/dist/esm/store/reconcile.js +12 -0
- package/dist/esm/store/reconcile.js.map +1 -1
- package/dist/esm/store/sqlite/open.d.ts +3 -3
- package/dist/esm/store/sqlite/open.js +9 -6
- package/dist/esm/store/sqlite/open.js.map +1 -1
- package/dist/esm/store/turso/open.d.ts +3 -3
- package/dist/esm/store/turso/open.js +14 -10
- package/dist/esm/store/turso/open.js.map +1 -1
- package/dist/esm/store/turso/store.d.ts +3 -1
- package/dist/esm/store/turso/store.js +5 -1
- package/dist/esm/store/turso/store.js.map +1 -1
- package/dist/esm/store/types.d.ts +12 -3
- package/dist/esm/store/types.js.map +1 -1
- package/dist/esm/watch.js +57 -19
- package/dist/esm/watch.js.map +1 -1
- package/package.json +1 -1
- package/schema.json +7 -2
- package/skills/sense/SKILL.md +3 -3
- package/skills/sense-bases/SKILL.md +1 -2
- package/skills/sense-setup/SKILL.md +1 -1
- package/skills/sense-setup/references/embeddings.md +1 -1
- package/skills/sense-setup/references/store-benchmarks.md +6 -6
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/duckdb/lexical.ts"],"sourcesContent":["// D1: fts BM25 for ranking, contains() scans for exact substring / phrase verification /\n// unspaced scripts, JS snippets for every doc (no snippet()).\nimport { SenseError } from '../../errors.ts';\nimport { foldForSearch, matchesSearchPhrase, type SearchToken, searchTokens } from '../../text/segment.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport { withTransaction } from '../transaction.ts';\nimport type { Connection, LexicalHit, LexicalQueryOptions } from '../types.ts';\nimport { ensureOrderedBm25, ORDERED_BM25_MACRO } from './ordered-bm25.ts';\n\nconst FIELDS = ['title', 'summary', 'text'] as const;\ntype Field = (typeof FIELDS)[number];\n// Mirrors sqlite's bm25(content, 10.0, 5.0, 1.0, ...) column-weight intent; DuckDB's match_bm25 has no\n// field_weights argument (verified 1.5.5), so each field is scored separately and combined here.\nconst FIELD_WEIGHT: Record<Field, number> = { title: 10, summary: 5, text: 1 };\n\n// FTS5 operator syntax (sqlite.org/fts5.html sec. 3) that words/substrings below would otherwise silently\n// treat as literal terms (PRINCIPLES: no-silent-modes). Quoted contents stay ignored while their boundary remains detectable.\nconst FTS5_OPERATORS: Array<{ label: string; re: RegExp }> = [\n { label: 'prefix query', re: /[\\p{L}\\p{N}_]+\\*(?=\\s|$)/u },\n { label: 'boolean operator', re: /(?:^|\\s)(?:AND|OR|NOT)(?=\\s|$)/ },\n { label: 'NEAR operator', re: /(?:^|\\s)NEAR\\b/ },\n { label: 'initial-token operator', re: /(?:^|\\s)\\^\\S+/ },\n { label: 'column filter', re: /(?:^|\\s)[\\p{L}_]\\w*\\s*:/u },\n { label: 'grouping operator', re: /[()]/ },\n];\n\nfunction unsupportedOperator(terms: string): { label: string; token: string } | null {\n const withoutPhrases = terms.replace(/\"[^\"]*\"/g, 'phrase');\n for (const { label, re } of FTS5_OPERATORS) {\n const m = withoutPhrases.match(re);\n if (m) return { label, token: m[0].trim() };\n }\n return null;\n}\n\ninterface FtsIndexState {\n stale: boolean;\n orderedMacroReady: boolean;\n}\n\n// Durable across connections/processes (meta table, shared.ts); `stale` is the in-memory cache of\n// it for this connection's lifetime, so a watcher's repeat queries don't re-read meta every time.\nconst FTS_STALE_META_KEY = 'fts_stale';\n\n// One state per connection, not per store instance: reconcileContent (reconcile.ts) marks it\n// stale from outside createLexicalIndex's closure, whether a one-shot open or a watcher's builder ran it.\nconst ftsState = new WeakMap<Connection, FtsIndexState>();\nasync function stateFor(conn: Connection): Promise<FtsIndexState> {\n let state = ftsState.get(conn);\n if (!state) {\n // Anything but a persisted '0' (missing key, or '1') means a rebuild is owed: a cache that\n // never built an index, or one a prior process marked stale and never got to clear.\n state = { stale: (await getMeta(conn, FTS_STALE_META_KEY)) !== '0', orderedMacroReady: false };\n ftsState.set(conn, state);\n }\n return state;\n}\n\n// Called whenever this store's reconcileContent changes `content`; must run inside that same\n// transaction (reconcile.ts) so a crash never lands the content write without the stale mark.\nexport async function markContentStale(conn: Connection): Promise<void> {\n ftsState.set(conn, { stale: true, orderedMacroReady: false });\n await setMeta(conn, FTS_STALE_META_KEY, '1');\n}\n\n// A run whose script marks no word boundaries makes match_bm25's whitespace tokenizer index the whole run as one token,\n// so unspaced terms use contains() and quoted phrases use the shared authored-field verifier.\nfunction splitTerms(terms: string): { words: string[]; bareWords: string[]; phraseWords: string[]; substrings: string[]; phrases: SearchToken[][]; emptyPhrase: boolean } {\n const words: string[] = [];\n const bareWords: string[] = [];\n const phraseWords: string[] = [];\n const substrings: string[] = [];\n const phrases: SearchToken[][] = [];\n let emptyPhrase = false;\n const withoutPhrases = terms.replace(/\"([^\"]*)\"/g, (_m, inner: string) => {\n const phrase = inner.trim();\n const phraseTokens = searchTokens(phrase);\n if (phraseTokens.length === 0) {\n emptyPhrase = true;\n return ' ';\n }\n phrases.push(phraseTokens);\n for (const token of phraseTokens) {\n if (token.unspaced) substrings.push(token.text);\n else {\n words.push(token.text);\n phraseWords.push(token.text);\n }\n }\n return ' ';\n });\n for (const tok of withoutPhrases.split(/\\s+/)) {\n if (tok.length === 0) continue;\n for (const token of searchTokens(tok)) {\n if (token.unspaced) substrings.push(token.text);\n else {\n words.push(token.text);\n bareWords.push(token.text);\n }\n }\n }\n return { words, bareWords, phraseWords, substrings, phrases, emptyPhrase };\n}\n\n// No incremental update (verified 1.5.5: PRAGMA create_fts_index is rebuild-only), so this pays\n// the full rebuild whenever `state.stale`, which meta.fts_stale keeps true across processes until cleared below.\nasync function ensureFtsFresh(conn: Connection, state: FtsIndexState): Promise<void> {\n if (!state.stale) return;\n state.orderedMacroReady = false;\n await conn.exec('INSTALL fts; LOAD fts;');\n await withTransaction(conn, async () => {\n // stopwords='none': sqlite's porter/unicode61 tokenizer never removes stopwords either (verified 1.5.5); the fts extension's\n // default 571-word list would otherwise drop common words (e.g. \"and\") from the index but not match_bm25's conjunctive gate, breaking multi-word matches.\n await conn.exec(`PRAGMA create_fts_index('content', 'path', 'title', 'summary', 'text', stopwords='none', overwrite=1)`);\n // Same transaction as the rebuild: a crash before commit leaves the mark set (a redundant\n // rebuild next time), never a cleared mark serving a half-built index as fresh.\n await setMeta(conn, FTS_STALE_META_KEY, '0');\n });\n state.stale = false;\n}\n\n// Field-weighted fragments (title 10 / summary 5 / text 1), with params emitted in the exact left-to-right order the assembled SQL needs for positional `?` binding.\n// score/gate params stay in separate arrays until joined at the end, since the SQL puts every score part before every gate part; sharing one array would shift binds once a query has both words and substrings (PRINCIPLES: no-silent-modes).\nfunction buildScoreAndGate(words: string[], bareWords: string[], phraseWords: string[], substrings: string[]): { scoreSql: string; gateSql: string; params: unknown[] } {\n const scoreParts: string[] = [];\n const gateParts: string[] = [];\n const scoreParams: unknown[] = [];\n const gateParams: unknown[] = [];\n\n if (words.length > 0) {\n const wordQuery = words.map(foldForSearch).join(' ');\n for (const field of FIELDS) {\n scoreParts.push(`${FIELD_WEIGHT[field]}.0 * COALESCE(${ORDERED_BM25_MACRO}(content.\"path\", ?, fields := '${field}', conjunctive := false), 0)`);\n scoreParams.push(wordQuery);\n }\n if (bareWords.length > 0) {\n const nativeGate = `${ORDERED_BM25_MACRO}(content.\"path\", ?, conjunctive := true) IS NOT NULL`;\n gateParts.push(nativeGate);\n gateParams.push(bareWords.map(foldForSearch).join(' '));\n }\n if (phraseWords.length > 0) {\n const nativeGate = `${ORDERED_BM25_MACRO}(content.\"path\", ?, conjunctive := true) IS NOT NULL`;\n const fallback = phraseWords.map((_word) => `(${FIELDS.map((field) => `contains(strip_accents(lower(content.${field})), ?)`).join(' OR ')})`).join(' AND ');\n gateParts.push(`(${nativeGate} OR (${fallback}))`);\n gateParams.push(phraseWords.map(foldForSearch).join(' '));\n gateParams.push(...phraseWords.flatMap((word) => FIELDS.map(() => foldForSearch(word))));\n }\n }\n\n for (const raw of substrings) {\n const needle = raw.toLowerCase();\n for (const field of FIELDS) {\n scoreParts.push(`${FIELD_WEIGHT[field]}.0 * CAST(contains(strip_accents(lower(content.${field})), ?) AS INTEGER)`);\n scoreParams.push(foldForSearch(needle));\n }\n gateParts.push(`(${FIELDS.map((field) => `contains(strip_accents(lower(content.${field})), ?)`).join(' OR ')})`);\n gateParams.push(...FIELDS.map(() => foldForSearch(needle)));\n }\n\n return { scoreSql: scoreParts.join(' + '), gateSql: gateParts.join(' AND ') || 'TRUE', params: [...scoreParams, ...gateParams] };\n}\n\n// Ranked query, scoped by the caller-built SQL fragments (same fragments sqlite's queryLexical takes).\n// Phrase rows carry authored fields through the post-filter; the public result still contains paths only.\nexport async function queryLexical(conn: Connection, terms: string, opts: LexicalQueryOptions, state: FtsIndexState): Promise<LexicalHit[]> {\n const unsupported = unsupportedOperator(terms);\n if (unsupported !== null) {\n throw new SenseError('STORE_CAPABILITY_MISSING', `store \"duckdb\" does not implement FTS5's ${unsupported.label} (\"${unsupported.token}\") in this build; rephrase \"${terms.trim()}\" without it, or set \"store\" to \"sqlite\" in this tree's config to search it as written`);\n }\n const { whereJoin, whereCond, scopeCond, limit } = opts;\n const { words, bareWords, phraseWords, substrings, phrases, emptyPhrase } = splitTerms(terms);\n if (emptyPhrase || (words.length === 0 && substrings.length === 0)) return [];\n\n if (words.length > 0) await ensureFtsFresh(conn, state);\n if (words.length > 0) await ensureOrderedBm25(conn, state);\n\n const { scoreSql, gateSql, params } = buildScoreAndGate(words, bareWords, phraseWords, substrings);\n const selectFields = phrases.length > 0 ? ', content.title, content.summary, content.text' : '';\n const sqlLimit = phrases.length > 0 ? '' : ' LIMIT ?';\n const sql = `SELECT path, score${phrases.length > 0 ? ', title, summary, text' : ''} FROM (\n SELECT content.\"path\" AS path, (${scoreSql}) AS score${selectFields}\n FROM content\n ${whereJoin}\n WHERE ${gateSql}\n ${whereCond} ${scopeCond}\n ) sq ORDER BY score DESC, path${sqlLimit}`;\n const stmt = await conn.prepare(sql);\n const rows = (await stmt.all(...params, ...(phrases.length > 0 ? [] : [limit]))) as Array<LexicalHit & Partial<Record<Field, string>>>;\n const matches = phrases.length === 0 ? rows : rows.filter((row) => phrases.every((phrase) => FIELDS.some((field) => matchesSearchPhrase(row[field] ?? '', phrase))));\n return matches.slice(0, limit).map(({ path }) => ({ path }));\n}\n\n// `stale`'s first read for a connection comes from meta.fts_stale (a fresh connection can't know\n// otherwise), and reconcileContent (reconcile.ts) marks it stale again on every content change.\nexport function createLexicalIndex(conn: Connection): { query: (terms: string, opts: LexicalQueryOptions) => Promise<LexicalHit[]>; markStale: () => Promise<void> } {\n return {\n query: async (terms, opts) => queryLexical(conn, terms, opts, await stateFor(conn)),\n markStale: () => markContentStale(conn),\n };\n}\n"],"names":["createLexicalIndex","markContentStale","queryLexical","FIELDS","FIELD_WEIGHT","title","summary","text","FTS5_OPERATORS","label","re","unsupportedOperator","terms","withoutPhrases","replace","m","match","token","trim","FTS_STALE_META_KEY","ftsState","WeakMap","stateFor","conn","state","get","getMeta","stale","orderedMacroReady","set","setMeta","splitTerms","words","bareWords","phraseWords","substrings","phrases","emptyPhrase","_m","inner","phrase","phraseTokens","searchTokens","length","push","unspaced","split","tok","ensureFtsFresh","exec","withTransaction","buildScoreAndGate","scoreParts","gateParts","scoreParams","gateParams","wordQuery","map","foldForSearch","join","field","ORDERED_BM25_MACRO","nativeGate","fallback","_word","flatMap","word","raw","needle","toLowerCase","scoreSql","gateSql","params","opts","stmt","unsupported","whereJoin","whereCond","scopeCond","limit","selectFields","sqlLimit","sql","rows","matches","SenseError","ensureOrderedBm25","prepare","all","filter","row","every","some","matchesSearchPhrase","slice","path","query","markStale"],"mappings":"AAAA,yFAAyF;AACzF,8DAA8D;;;;;;;;;;;;QAiM9CA;eAAAA;;QAtIMC;eAAAA;;QAwGAC;eAAAA;;;wBAlKK;yBACwD;wBAClD;6BACD;6BAEsB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAEtD,IAAMC,SAAS;IAAC;IAAS;IAAW;CAAO;AAE3C,uGAAuG;AACvG,iGAAiG;AACjG,IAAMC,eAAsC;IAAEC,OAAO;IAAIC,SAAS;IAAGC,MAAM;AAAE;AAE7E,0GAA0G;AAC1G,8HAA8H;AAC9H,IAAMC,iBAAuD;IAC3D;QAAEC,OAAO;QAAgBC,IAAI;IAA4B;IACzD;QAAED,OAAO;QAAoBC,IAAI;IAAiC;IAClE;QAAED,OAAO;QAAiBC,IAAI;IAAiB;IAC/C;QAAED,OAAO;QAA0BC,IAAI;IAAgB;IACvD;QAAED,OAAO;QAAiBC,IAAI;IAA2B;IACzD;QAAED,OAAO;QAAqBC,IAAI;IAAO;CAC1C;AAED,SAASC,oBAAoBC,KAAa;IACxC,IAAMC,iBAAiBD,MAAME,OAAO,CAAC,YAAY;QAC5C,kCAAA,2BAAA;;QAAL,QAAK,YAAuBN,mCAAvB,SAAA,6BAAA,QAAA,yBAAA,iCAAuC;YAAvC,kBAAA,aAAQC,oBAAAA,OAAOC,iBAAAA;YAClB,IAAMK,IAAIF,eAAeG,KAAK,CAACN;YAC/B,IAAIK,GAAG,OAAO;gBAAEN,OAAAA;gBAAOQ,OAAOF,CAAC,CAAC,EAAE,CAACG,IAAI;YAAG;QAC5C;;QAHK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAIL,OAAO;AACT;AAOA,kGAAkG;AAClG,kGAAkG;AAClG,IAAMC,qBAAqB;AAE3B,6FAA6F;AAC7F,0GAA0G;AAC1G,IAAMC,WAAW,IAAIC;AACrB,SAAeC,SAASC,IAAgB;;YAClCC;;;;oBAAAA,QAAQJ,SAASK,GAAG,CAACF;yBACrB,CAACC,OAAD;;;;;oBAGgB;;wBAAME,IAAAA,iBAAO,EAACH,MAAMJ;;;oBAFtC,2FAA2F;oBAC3F,oFAAoF;oBACpFK,cAAUG,QAAO,AAAC,kBAA6C,UAAKC,oBAAmB;oBACvFR,SAASS,GAAG,CAACN,MAAMC;;;oBAErB;;wBAAOA;;;;IACT;;AAIO,SAAevB,iBAAiBsB,IAAgB;;;;;oBACrDH,SAASS,GAAG,CAACN,MAAM;wBAAEI,OAAO;wBAAMC,mBAAmB;oBAAM;oBAC3D;;wBAAME,IAAAA,iBAAO,EAACP,MAAMJ,oBAAoB;;;oBAAxC;;;;;;IACF;;AAEA,wHAAwH;AACxH,8FAA8F;AAC9F,SAASY,WAAWnB,KAAa;IAC/B,IAAMoB,QAAkB,EAAE;IAC1B,IAAMC,YAAsB,EAAE;IAC9B,IAAMC,cAAwB,EAAE;IAChC,IAAMC,aAAuB,EAAE;IAC/B,IAAMC,UAA2B,EAAE;IACnC,IAAIC,cAAc;IAClB,IAAMxB,iBAAiBD,MAAME,OAAO,CAAC,cAAc,SAACwB,IAAIC;QACtD,IAAMC,SAASD,MAAMrB,IAAI;QACzB,IAAMuB,eAAeC,IAAAA,uBAAY,EAACF;QAClC,IAAIC,aAAaE,MAAM,KAAK,GAAG;YAC7BN,cAAc;YACd,OAAO;QACT;QACAD,QAAQQ,IAAI,CAACH;YACR,kCAAA,2BAAA;;YAAL,QAAK,YAAeA,iCAAf,SAAA,6BAAA,QAAA,yBAAA,iCAA6B;gBAA7B,IAAMxB,QAAN;gBACH,IAAIA,MAAM4B,QAAQ,EAAEV,WAAWS,IAAI,CAAC3B,MAAMV,IAAI;qBACzC;oBACHyB,MAAMY,IAAI,CAAC3B,MAAMV,IAAI;oBACrB2B,YAAYU,IAAI,CAAC3B,MAAMV,IAAI;gBAC7B;YACF;;YANK;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;QAOL,OAAO;IACT;QACK,kCAAA,2BAAA;;QAAL,QAAK,YAAaM,eAAeiC,KAAK,CAAC,2BAAlC,SAAA,6BAAA,QAAA,yBAAA,iCAA0C;YAA1C,IAAMC,MAAN;YACH,IAAIA,IAAIJ,MAAM,KAAK,GAAG;gBACjB,mCAAA,4BAAA;;gBAAL,QAAK,aAAeD,IAAAA,uBAAY,EAACK,yBAA5B,UAAA,8BAAA,SAAA,0BAAA,kCAAkC;oBAAlC,IAAM9B,QAAN;oBACH,IAAIA,MAAM4B,QAAQ,EAAEV,WAAWS,IAAI,CAAC3B,MAAMV,IAAI;yBACzC;wBACHyB,MAAMY,IAAI,CAAC3B,MAAMV,IAAI;wBACrB0B,UAAUW,IAAI,CAAC3B,MAAMV,IAAI;oBAC3B;gBACF;;gBANK;gBAAA;;;yBAAA,8BAAA;wBAAA;;;wBAAA;8BAAA;;;;QAOP;;QATK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAUL,OAAO;QAAEyB,OAAAA;QAAOC,WAAAA;QAAWC,aAAAA;QAAaC,YAAAA;QAAYC,SAAAA;QAASC,aAAAA;IAAY;AAC3E;AAEA,gGAAgG;AAChG,iHAAiH;AACjH,SAAeW,eAAezB,IAAgB,EAAEC,KAAoB;;;;;oBAClE,IAAI,CAACA,MAAMG,KAAK,EAAE;;;oBAClBH,MAAMI,iBAAiB,GAAG;oBAC1B;;wBAAML,KAAK0B,IAAI,CAAC;;;oBAAhB;oBACA;;wBAAMC,IAAAA,8BAAe,EAAC3B,MAAM;;;;;4CAC1B,6HAA6H;4CAC7H,0JAA0J;4CAC1J;;gDAAMA,KAAK0B,IAAI,CAAC;;;4CAAhB;4CACA,0FAA0F;4CAC1F,gFAAgF;4CAChF;;gDAAMnB,IAAAA,iBAAO,EAACP,MAAMJ,oBAAoB;;;4CAAxC;;;;;;4BACF;;;;oBAPA;oBAQAK,MAAMG,KAAK,GAAG;;;;;;IAChB;;AAEA,qKAAqK;AACrK,+OAA+O;AAC/O,SAASwB,kBAAkBnB,KAAe,EAAEC,SAAmB,EAAEC,WAAqB,EAAEC,UAAoB;IAC1G,IAAMiB,aAAuB,EAAE;IAC/B,IAAMC,YAAsB,EAAE;IAC9B,IAAMC,cAAyB,EAAE;IACjC,IAAMC,aAAwB,EAAE;IAEhC,IAAIvB,MAAMW,MAAM,GAAG,GAAG;QACpB,IAAMa,YAAYxB,MAAMyB,GAAG,CAACC,wBAAa,EAAEC,IAAI,CAAC;YAC3C,kCAAA,2BAAA;;YAAL,QAAK,YAAexD,2BAAf,SAAA,6BAAA,QAAA,yBAAA,iCAAuB;gBAAvB,IAAMyD,QAAN;gBACHR,WAAWR,IAAI,CAAC,AAAC,GAAsCiB,OAApCzD,YAAY,CAACwD,MAAM,EAAC,kBAAoEA,OAApDC,iCAAkB,EAAC,oCAAuC,OAAND,OAAM;gBACjHN,YAAYV,IAAI,CAACY;YACnB;;YAHK;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;QAIL,IAAIvB,UAAUU,MAAM,GAAG,GAAG;YACxB,IAAMmB,aAAa,AAAC,GAAqB,OAAnBD,iCAAkB,EAAC;YACzCR,UAAUT,IAAI,CAACkB;YACfP,WAAWX,IAAI,CAACX,UAAUwB,GAAG,CAACC,wBAAa,EAAEC,IAAI,CAAC;QACpD;QACA,IAAIzB,YAAYS,MAAM,GAAG,GAAG;gBAK1BY;YAJA,IAAMO,cAAa,AAAC,GAAqB,OAAnBD,iCAAkB,EAAC;YACzC,IAAME,WAAW7B,YAAYuB,GAAG,CAAC,SAACO;uBAAU,AAAC,IAA6F,OAA1F7D,OAAOsD,GAAG,CAAC,SAACG;2BAAU,AAAC,wCAA6C,OAANA,OAAM;mBAASD,IAAI,CAAC,SAAQ;eAAIA,IAAI,CAAC;YACnJN,UAAUT,IAAI,CAAC,AAAC,IAAqBmB,OAAlBD,aAAW,SAAgB,OAATC,UAAS;YAC9CR,WAAWX,IAAI,CAACV,YAAYuB,GAAG,CAACC,wBAAa,EAAEC,IAAI,CAAC;YACpDJ,CAAAA,cAAAA,YAAWX,IAAI,OAAfW,aAAgB,qBAAGrB,YAAY+B,OAAO,CAAC,SAACC;uBAAS/D,OAAOsD,GAAG,CAAC;2BAAMC,IAAAA,wBAAa,EAACQ;;;QAClF;IACF;QAEK,mCAAA,4BAAA;;;YAAA,IAAMC,MAAN;gBAOHZ;YANA,IAAMa,SAASD,IAAIE,WAAW;gBACzB,kCAAA,2BAAA;;gBAAL,QAAK,YAAelE,2BAAf,SAAA,6BAAA,QAAA,yBAAA,iCAAuB;oBAAvB,IAAMyD,QAAN;oBACHR,WAAWR,IAAI,CAAC,AAAC,GAAuEgB,OAArExD,YAAY,CAACwD,MAAM,EAAC,mDAAuD,OAANA,OAAM;oBAC9FN,YAAYV,IAAI,CAACc,IAAAA,wBAAa,EAACU;gBACjC;;gBAHK;gBAAA;;;yBAAA,6BAAA;wBAAA;;;wBAAA;8BAAA;;;;YAILf,UAAUT,IAAI,CAAC,AAAC,IAA6F,OAA1FzC,OAAOsD,GAAG,CAAC,SAACG;uBAAU,AAAC,wCAA6C,OAANA,OAAM;eAASD,IAAI,CAAC,SAAQ;YAC7GJ,CAAAA,cAAAA,YAAWX,IAAI,OAAfW,aAAgB,qBAAGpD,OAAOsD,GAAG,CAAC;uBAAMC,IAAAA,wBAAa,EAACU;;QACpD;QARA,QAAK,aAAajC,+BAAb,UAAA,8BAAA,SAAA,0BAAA;;QAAA;QAAA;;;iBAAA,8BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAUL,OAAO;QAAEmC,UAAUlB,WAAWO,IAAI,CAAC;QAAQY,SAASlB,UAAUM,IAAI,CAAC,YAAY;QAAQa,QAAQ,AAAC,qBAAGlB,oBAAa,qBAAGC;IAAY;AACjI;AAIO,SAAerD,aAAaqB,IAAgB,EAAEX,KAAa,EAAE6D,IAAyB,EAAEjD,KAAoB;;YAuB7FkD,OAtBdC,aAIEC,WAAWC,WAAWC,WAAWC,OACmChD,aAApEC,OAAOC,WAAWC,aAAaC,YAAYC,SAASC,aAMtBc,oBAA9BmB,UAAUC,SAASC,QACrBQ,cACAC,UACAC,KAOAR,MACAS,MACAC;;;;oBAvBAT,cAAchE,oBAAoBC;oBACxC,IAAI+D,gBAAgB,MAAM;wBACxB,MAAM,IAAIU,oBAAU,CAAC,4BAA4B,AAAC,6CAAkEV,OAAvBA,YAAYlE,KAAK,EAAC,OAAqDG,OAAhD+D,YAAY1D,KAAK,EAAC,gCAA2C,OAAbL,MAAMM,IAAI,IAAG;oBACnL;oBACQ0D,YAA2CH,KAA3CG,WAAWC,YAAgCJ,KAAhCI,WAAWC,YAAqBL,KAArBK,WAAWC,QAAUN,KAAVM;oBACmChD,cAAAA,WAAWnB,QAA/EoB,QAAoED,YAApEC,OAAOC,YAA6DF,YAA7DE,WAAWC,cAAkDH,YAAlDG,aAAaC,aAAqCJ,YAArCI,YAAYC,UAAyBL,YAAzBK,SAASC,cAAgBN,YAAhBM;oBAC5D,IAAIA,eAAgBL,MAAMW,MAAM,KAAK,KAAKR,WAAWQ,MAAM,KAAK,GAAI;;;;yBAEhEX,CAAAA,MAAMW,MAAM,GAAG,CAAA,GAAfX;;;;oBAAkB;;wBAAMgB,eAAezB,MAAMC;;;oBAA3B;;;yBAClBQ,CAAAA,MAAMW,MAAM,GAAG,CAAA,GAAfX;;;;oBAAkB;;wBAAMsD,IAAAA,gCAAiB,EAAC/D,MAAMC;;;oBAA9B;;;oBAEgB2B,qBAAAA,kBAAkBnB,OAAOC,WAAWC,aAAaC,aAA/EmC,WAA8BnB,mBAA9BmB,UAAUC,UAAoBpB,mBAApBoB,SAASC,SAAWrB,mBAAXqB;oBACrBQ,eAAe5C,QAAQO,MAAM,GAAG,IAAI,mDAAmD;oBACvFsC,WAAW7C,QAAQO,MAAM,GAAG,IAAI,KAAK;oBACrCuC,MAAM,AAAC,qBACuBZ,OADHlC,QAAQO,MAAM,GAAG,IAAI,2BAA2B,IAAG,iDAC3BqC,OAArBV,UAAS,cAEzCM,OAFqDI,cAAa,4BAG5DT,OADNK,WAAU,gBAEVC,OADMN,SAAQ,UACDO,OAAbD,WAAU,KACoBI,OADjBH,WAAU,wCACgB,OAATG;oBACrB;;wBAAM1D,KAAKgE,OAAO,CAACL;;;oBAA1BR,OAAO;oBACC;;wBAAMA,CAAAA,QAAAA,MAAKc,GAAG,OAARd,OAAAA,AAAS,qBAAGF,eAAQ,qBAAIpC,QAAQO,MAAM,GAAG;4BAAUoC;;;;oBAAjEI,OAAQ;oBACRC,UAAUhD,QAAQO,MAAM,KAAK,IAAIwC,OAAOA,KAAKM,MAAM,CAAC,SAACC;+BAAQtD,QAAQuD,KAAK,CAAC,SAACnD;mCAAWrC,OAAOyF,IAAI,CAAC,SAAChC;oCAA8B8B;uCAApBG,IAAAA,8BAAmB,GAACH,aAAAA,GAAG,CAAC9B,MAAM,cAAV8B,wBAAAA,aAAc,IAAIlD;;;;oBAC1J;;wBAAO4C,QAAQU,KAAK,CAAC,GAAGf,OAAOtB,GAAG,CAAC;gCAAGsC,aAAAA;mCAAY;gCAAEA,MAAAA;4BAAK;;;;;IAC3D;;AAIO,SAAS/F,mBAAmBuB,IAAgB;IACjD,OAAO;QACLyE,OAAO,SAAPA,MAAcpF,OAAO6D;;;;;;;gCAAsBlD;gCAAMX;gCAAO6D;;4BAAM;;gCAAMnD,SAASC;;;;;gCAA/CrB;oCAAgC;;;;;;;QAC9D+F,WAAW,SAAXA;mBAAiBhG,iBAAiBsB;;IACpC;AACF"}
|
|
1
|
+
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/duckdb/lexical.ts"],"sourcesContent":["// D1: fts BM25 for ranking, contains() scans for exact substring / phrase verification /\n// unspaced scripts, JS snippets for every doc (no snippet()).\nimport { SenseError } from '../../errors.ts';\nimport { foldForSearch, matchesSearchPhrase, type SearchToken, searchTokens } from '../../text/segment.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport { withTransaction } from '../transaction.ts';\nimport type { Connection, LexicalHit, LexicalQueryOptions } from '../types.ts';\nimport { ensureOrderedBm25, ORDERED_BM25_MACRO } from './ordered-bm25.ts';\n\nconst FIELDS = ['title', 'summary', 'text'] as const;\ntype Field = (typeof FIELDS)[number];\n// Mirrors sqlite's bm25(content, 10.0, 5.0, 1.0, ...) column-weight intent; DuckDB's match_bm25 has no\n// field_weights argument (verified 1.5.5), so each field is scored separately and combined here.\nconst FIELD_WEIGHT: Record<Field, number> = { title: 10, summary: 5, text: 1 };\n\n// FTS5 operator syntax (sqlite.org/fts5.html sec. 3) that words/substrings below would otherwise silently\n// treat as literal terms (PRINCIPLES: no-silent-modes). Quoted contents stay ignored while their boundary remains detectable.\nconst FTS5_OPERATORS: Array<{ label: string; re: RegExp }> = [\n { label: 'prefix query', re: /[\\p{L}\\p{N}_]+\\*(?=\\s|$)/u },\n { label: 'boolean operator', re: /(?:^|\\s)(?:AND|OR|NOT)(?=\\s|$)/ },\n { label: 'NEAR operator', re: /(?:^|\\s)NEAR\\b/ },\n { label: 'initial-token operator', re: /(?:^|\\s)\\^\\S+/ },\n { label: 'column filter', re: /(?:^|\\s)[\\p{L}_]\\w*\\s*:/u },\n { label: 'grouping operator', re: /[()]/ },\n];\n\nfunction unsupportedOperator(terms: string): { label: string; token: string } | null {\n const withoutPhrases = terms.replace(/\"[^\"]*\"/g, 'phrase');\n for (const { label, re } of FTS5_OPERATORS) {\n const m = withoutPhrases.match(re);\n if (m) return { label, token: m[0].trim() };\n }\n return null;\n}\n\ninterface FtsIndexState {\n stale: boolean;\n orderedMacroReady: boolean;\n}\n\n// Durable across connections/processes (meta table, shared.ts); `stale` is the in-memory cache of\n// it for this connection's lifetime, so a watcher's repeat queries don't re-read meta every time.\nconst FTS_STALE_META_KEY = 'fts_stale';\n\n// One state per connection, not per store instance: reconcileContent (reconcile.ts) marks it\n// stale from outside createLexicalIndex's closure, whether a one-shot open or a watcher's builder ran it.\nconst ftsState = new WeakMap<Connection, FtsIndexState>();\nasync function stateFor(conn: Connection): Promise<FtsIndexState> {\n let state = ftsState.get(conn);\n if (!state) {\n // Anything but a persisted '0' (missing key, or '1') means a rebuild is owed: a cache that\n // never built an index, or one a prior process marked stale and never got to clear.\n state = { stale: (await getMeta(conn, FTS_STALE_META_KEY)) !== '0', orderedMacroReady: false };\n ftsState.set(conn, state);\n }\n return state;\n}\n\nasync function hasFtsArtifact(conn: Connection): Promise<boolean> {\n const stmt = await conn.prepare(`\n SELECT 1 AS ready\n FROM duckdb_functions()\n WHERE schema_name = 'fts_main_content'\n AND function_name = 'match_bm25'\n AND function_type = 'macro'\n LIMIT 1\n `);\n return (await stmt.get()) !== undefined;\n}\n\n// Called whenever this store's reconcileContent changes `content`; must run inside that same\n// transaction (reconcile.ts) so a crash never lands the content write without the stale mark.\nexport async function markContentStale(conn: Connection): Promise<void> {\n ftsState.set(conn, { stale: true, orderedMacroReady: false });\n await setMeta(conn, FTS_STALE_META_KEY, '1');\n}\n\n// A run whose script marks no word boundaries makes match_bm25's whitespace tokenizer index the whole run as one token,\n// so unspaced terms use contains() and quoted phrases use the shared authored-field verifier.\nfunction splitTerms(terms: string): { words: string[]; bareWords: string[]; phraseWords: string[]; substrings: string[]; phrases: SearchToken[][]; emptyPhrase: boolean } {\n const words: string[] = [];\n const bareWords: string[] = [];\n const phraseWords: string[] = [];\n const substrings: string[] = [];\n const phrases: SearchToken[][] = [];\n let emptyPhrase = false;\n const withoutPhrases = terms.replace(/\"([^\"]*)\"/g, (_m, inner: string) => {\n const phrase = inner.trim();\n const phraseTokens = searchTokens(phrase);\n if (phraseTokens.length === 0) {\n emptyPhrase = true;\n return ' ';\n }\n phrases.push(phraseTokens);\n for (const token of phraseTokens) {\n if (token.unspaced) substrings.push(token.text);\n else {\n words.push(token.text);\n phraseWords.push(token.text);\n }\n }\n return ' ';\n });\n for (const tok of withoutPhrases.split(/\\s+/)) {\n if (tok.length === 0) continue;\n for (const token of searchTokens(tok)) {\n if (token.unspaced) substrings.push(token.text);\n else {\n words.push(token.text);\n bareWords.push(token.text);\n }\n }\n }\n return { words, bareWords, phraseWords, substrings, phrases, emptyPhrase };\n}\n\n// No incremental update (verified 1.5.5: PRAGMA create_fts_index is rebuild-only), so this pays\n// the full rebuild whenever `state.stale`, which meta.fts_stale keeps true across processes until cleared below.\nasync function rebuildFts(conn: Connection, state: FtsIndexState): Promise<void> {\n await conn.exec('INSTALL fts; LOAD fts;');\n if (!state.stale && (await hasFtsArtifact(conn))) return;\n state.stale = true;\n state.orderedMacroReady = false;\n await withTransaction(conn, async () => {\n // stopwords='none': sqlite's porter/unicode61 tokenizer never removes stopwords either (verified 1.5.5); the fts extension's\n // default 571-word list would otherwise drop common words (e.g. \"and\") from the index but not match_bm25's conjunctive gate, breaking multi-word matches.\n await conn.exec(`PRAGMA create_fts_index('content', 'path', 'title', 'summary', 'text', stopwords='none', overwrite=1)`);\n // Same transaction as the rebuild: a crash before commit leaves the mark set (a redundant\n // rebuild next time), never a cleared mark serving a half-built index as fresh.\n await setMeta(conn, FTS_STALE_META_KEY, '0');\n });\n state.stale = false;\n}\n\nexport async function prepareFts(conn: Connection): Promise<void> {\n await rebuildFts(conn, await stateFor(conn));\n}\n\nexport async function assertFtsReady(conn: Connection): Promise<void> {\n const state = await stateFor(conn);\n if (state.stale) throw new SenseError('INDEX_NOT_READY', 'the DuckDB lexical index is missing or stale; run `sense build` (or library build(config)) before lexical querying with build disabled');\n try {\n await conn.exec('LOAD fts;');\n if (!(await hasFtsArtifact(conn))) {\n throw new SenseError('INDEX_NOT_READY', 'the DuckDB lexical index artifact is missing; run `sense build` (or library build(config)) before lexical querying with build disabled');\n }\n } catch (err) {\n if (err instanceof SenseError) throw err;\n throw new SenseError('INDEX_NOT_READY', `the DuckDB lexical extension is not prepared (${(err as Error).message}); run \\`sense build\\` (or library build(config))`);\n }\n}\n\n// Field-weighted fragments (title 10 / summary 5 / text 1), with params emitted in the exact left-to-right order the assembled SQL needs for positional `?` binding.\n// score/gate params stay in separate arrays until joined at the end, since the SQL puts every score part before every gate part; sharing one array would shift binds once a query has both words and substrings (PRINCIPLES: no-silent-modes).\nfunction buildScoreAndGate(words: string[], bareWords: string[], phraseWords: string[], substrings: string[]): { scoreSql: string; gateSql: string; params: unknown[] } {\n const scoreParts: string[] = [];\n const gateParts: string[] = [];\n const scoreParams: unknown[] = [];\n const gateParams: unknown[] = [];\n\n if (words.length > 0) {\n const wordQuery = words.map(foldForSearch).join(' ');\n for (const field of FIELDS) {\n scoreParts.push(`${FIELD_WEIGHT[field]}.0 * COALESCE(${ORDERED_BM25_MACRO}(content.\"path\", ?, fields := '${field}', conjunctive := false), 0)`);\n scoreParams.push(wordQuery);\n }\n if (bareWords.length > 0) {\n const nativeGate = `${ORDERED_BM25_MACRO}(content.\"path\", ?, conjunctive := true) IS NOT NULL`;\n gateParts.push(nativeGate);\n gateParams.push(bareWords.map(foldForSearch).join(' '));\n }\n if (phraseWords.length > 0) {\n const nativeGate = `${ORDERED_BM25_MACRO}(content.\"path\", ?, conjunctive := true) IS NOT NULL`;\n const fallback = phraseWords.map((_word) => `(${FIELDS.map((field) => `contains(strip_accents(lower(content.${field})), ?)`).join(' OR ')})`).join(' AND ');\n gateParts.push(`(${nativeGate} OR (${fallback}))`);\n gateParams.push(phraseWords.map(foldForSearch).join(' '));\n gateParams.push(...phraseWords.flatMap((word) => FIELDS.map(() => foldForSearch(word))));\n }\n }\n\n for (const raw of substrings) {\n const needle = raw.toLowerCase();\n for (const field of FIELDS) {\n scoreParts.push(`${FIELD_WEIGHT[field]}.0 * CAST(contains(strip_accents(lower(content.${field})), ?) AS INTEGER)`);\n scoreParams.push(foldForSearch(needle));\n }\n gateParts.push(`(${FIELDS.map((field) => `contains(strip_accents(lower(content.${field})), ?)`).join(' OR ')})`);\n gateParams.push(...FIELDS.map(() => foldForSearch(needle)));\n }\n\n return { scoreSql: scoreParts.join(' + '), gateSql: gateParts.join(' AND ') || 'TRUE', params: [...scoreParams, ...gateParams] };\n}\n\n// Ranked query, scoped by the caller-built SQL fragments (same fragments sqlite's queryLexical takes).\n// Phrase rows carry authored fields through the post-filter; the public result still contains paths only.\nexport async function queryLexical(conn: Connection, terms: string, opts: LexicalQueryOptions, state: FtsIndexState): Promise<LexicalHit[]> {\n const unsupported = unsupportedOperator(terms);\n if (unsupported !== null) {\n throw new SenseError('STORE_CAPABILITY_MISSING', `store \"duckdb\" does not implement FTS5's ${unsupported.label} (\"${unsupported.token}\") in this build; rephrase \"${terms.trim()}\" without it, or set \"store\" to \"sqlite\" in this tree's config to search it as written`);\n }\n const { whereJoin, whereCond, scopeCond, limit } = opts;\n const { words, bareWords, phraseWords, substrings, phrases, emptyPhrase } = splitTerms(terms);\n if (emptyPhrase || (words.length === 0 && substrings.length === 0)) return [];\n\n if (words.length > 0) await assertFtsReady(conn);\n if (words.length > 0) await ensureOrderedBm25(conn, state);\n\n const { scoreSql, gateSql, params } = buildScoreAndGate(words, bareWords, phraseWords, substrings);\n const selectFields = phrases.length > 0 ? ', content.title, content.summary, content.text' : '';\n const sqlLimit = phrases.length > 0 ? '' : ' LIMIT ?';\n const sql = `SELECT path, score${phrases.length > 0 ? ', title, summary, text' : ''} FROM (\n SELECT content.\"path\" AS path, (${scoreSql}) AS score${selectFields}\n FROM content\n ${whereJoin}\n WHERE ${gateSql}\n ${whereCond} ${scopeCond}\n ) sq ORDER BY score DESC, path${sqlLimit}`;\n const stmt = await conn.prepare(sql);\n const rows = (await stmt.all(...params, ...(phrases.length > 0 ? [] : [limit]))) as Array<LexicalHit & Partial<Record<Field, string>>>;\n const matches = phrases.length === 0 ? rows : rows.filter((row) => phrases.every((phrase) => FIELDS.some((field) => matchesSearchPhrase(row[field] ?? '', phrase))));\n return matches.slice(0, limit).map(({ path }) => ({ path }));\n}\n\n// `stale`'s first read for a connection comes from meta.fts_stale (a fresh connection can't know\n// otherwise), and reconcileContent (reconcile.ts) marks it stale again on every content change.\nexport function createLexicalIndex(conn: Connection): { query: (terms: string, opts: LexicalQueryOptions) => Promise<LexicalHit[]>; markStale: () => Promise<void> } {\n return {\n query: async (terms, opts) => queryLexical(conn, terms, opts, await stateFor(conn)),\n markStale: () => markContentStale(conn),\n };\n}\n"],"names":["assertFtsReady","createLexicalIndex","markContentStale","prepareFts","queryLexical","FIELDS","FIELD_WEIGHT","title","summary","text","FTS5_OPERATORS","label","re","unsupportedOperator","terms","withoutPhrases","replace","m","match","token","trim","FTS_STALE_META_KEY","ftsState","WeakMap","stateFor","conn","state","get","getMeta","stale","orderedMacroReady","set","hasFtsArtifact","stmt","prepare","undefined","setMeta","splitTerms","words","bareWords","phraseWords","substrings","phrases","emptyPhrase","_m","inner","phrase","phraseTokens","searchTokens","length","push","unspaced","split","tok","rebuildFts","exec","withTransaction","err","SenseError","message","buildScoreAndGate","scoreParts","gateParts","scoreParams","gateParams","wordQuery","map","foldForSearch","join","field","ORDERED_BM25_MACRO","nativeGate","fallback","_word","flatMap","word","raw","needle","toLowerCase","scoreSql","gateSql","params","opts","unsupported","whereJoin","whereCond","scopeCond","limit","selectFields","sqlLimit","sql","rows","matches","ensureOrderedBm25","all","filter","row","every","some","matchesSearchPhrase","slice","path","query","markStale"],"mappings":"AAAA,yFAAyF;AACzF,8DAA8D;;;;;;;;;;;;QAyIxCA;eAAAA;;QAuFNC;eAAAA;;QAzJMC;eAAAA;;QA8DAC;eAAAA;;QA6DAC;eAAAA;;;wBAjMK;yBACwD;wBAClD;6BACD;6BAEsB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAEtD,IAAMC,SAAS;IAAC;IAAS;IAAW;CAAO;AAE3C,uGAAuG;AACvG,iGAAiG;AACjG,IAAMC,eAAsC;IAAEC,OAAO;IAAIC,SAAS;IAAGC,MAAM;AAAE;AAE7E,0GAA0G;AAC1G,8HAA8H;AAC9H,IAAMC,iBAAuD;IAC3D;QAAEC,OAAO;QAAgBC,IAAI;IAA4B;IACzD;QAAED,OAAO;QAAoBC,IAAI;IAAiC;IAClE;QAAED,OAAO;QAAiBC,IAAI;IAAiB;IAC/C;QAAED,OAAO;QAA0BC,IAAI;IAAgB;IACvD;QAAED,OAAO;QAAiBC,IAAI;IAA2B;IACzD;QAAED,OAAO;QAAqBC,IAAI;IAAO;CAC1C;AAED,SAASC,oBAAoBC,KAAa;IACxC,IAAMC,iBAAiBD,MAAME,OAAO,CAAC,YAAY;QAC5C,kCAAA,2BAAA;;QAAL,QAAK,YAAuBN,mCAAvB,SAAA,6BAAA,QAAA,yBAAA,iCAAuC;YAAvC,kBAAA,aAAQC,oBAAAA,OAAOC,iBAAAA;YAClB,IAAMK,IAAIF,eAAeG,KAAK,CAACN;YAC/B,IAAIK,GAAG,OAAO;gBAAEN,OAAAA;gBAAOQ,OAAOF,CAAC,CAAC,EAAE,CAACG,IAAI;YAAG;QAC5C;;QAHK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAIL,OAAO;AACT;AAOA,kGAAkG;AAClG,kGAAkG;AAClG,IAAMC,qBAAqB;AAE3B,6FAA6F;AAC7F,0GAA0G;AAC1G,IAAMC,WAAW,IAAIC;AACrB,SAAeC,SAASC,IAAgB;;YAClCC;;;;oBAAAA,QAAQJ,SAASK,GAAG,CAACF;yBACrB,CAACC,OAAD;;;;;oBAGgB;;wBAAME,IAAAA,iBAAO,EAACH,MAAMJ;;;oBAFtC,2FAA2F;oBAC3F,oFAAoF;oBACpFK,cAAUG,QAAO,AAAC,kBAA6C,UAAKC,oBAAmB;oBACvFR,SAASS,GAAG,CAACN,MAAMC;;;oBAErB;;wBAAOA;;;;IACT;;AAEA,SAAeM,eAAeP,IAAgB;;YACtCQ;;;;oBAAO;;wBAAMR,KAAKS,OAAO,CAAC;;;oBAA1BD,OAAO;oBAQL;;wBAAMA,KAAKN,GAAG;;;oBAAtB;;wBAAQ,kBAAsBQ;;;;IAChC;;AAIO,SAAejC,iBAAiBuB,IAAgB;;;;;oBACrDH,SAASS,GAAG,CAACN,MAAM;wBAAEI,OAAO;wBAAMC,mBAAmB;oBAAM;oBAC3D;;wBAAMM,IAAAA,iBAAO,EAACX,MAAMJ,oBAAoB;;;oBAAxC;;;;;;IACF;;AAEA,wHAAwH;AACxH,8FAA8F;AAC9F,SAASgB,WAAWvB,KAAa;IAC/B,IAAMwB,QAAkB,EAAE;IAC1B,IAAMC,YAAsB,EAAE;IAC9B,IAAMC,cAAwB,EAAE;IAChC,IAAMC,aAAuB,EAAE;IAC/B,IAAMC,UAA2B,EAAE;IACnC,IAAIC,cAAc;IAClB,IAAM5B,iBAAiBD,MAAME,OAAO,CAAC,cAAc,SAAC4B,IAAIC;QACtD,IAAMC,SAASD,MAAMzB,IAAI;QACzB,IAAM2B,eAAeC,IAAAA,uBAAY,EAACF;QAClC,IAAIC,aAAaE,MAAM,KAAK,GAAG;YAC7BN,cAAc;YACd,OAAO;QACT;QACAD,QAAQQ,IAAI,CAACH;YACR,kCAAA,2BAAA;;YAAL,QAAK,YAAeA,iCAAf,SAAA,6BAAA,QAAA,yBAAA,iCAA6B;gBAA7B,IAAM5B,QAAN;gBACH,IAAIA,MAAMgC,QAAQ,EAAEV,WAAWS,IAAI,CAAC/B,MAAMV,IAAI;qBACzC;oBACH6B,MAAMY,IAAI,CAAC/B,MAAMV,IAAI;oBACrB+B,YAAYU,IAAI,CAAC/B,MAAMV,IAAI;gBAC7B;YACF;;YANK;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;QAOL,OAAO;IACT;QACK,kCAAA,2BAAA;;QAAL,QAAK,YAAaM,eAAeqC,KAAK,CAAC,2BAAlC,SAAA,6BAAA,QAAA,yBAAA,iCAA0C;YAA1C,IAAMC,MAAN;YACH,IAAIA,IAAIJ,MAAM,KAAK,GAAG;gBACjB,mCAAA,4BAAA;;gBAAL,QAAK,aAAeD,IAAAA,uBAAY,EAACK,yBAA5B,UAAA,8BAAA,SAAA,0BAAA,kCAAkC;oBAAlC,IAAMlC,QAAN;oBACH,IAAIA,MAAMgC,QAAQ,EAAEV,WAAWS,IAAI,CAAC/B,MAAMV,IAAI;yBACzC;wBACH6B,MAAMY,IAAI,CAAC/B,MAAMV,IAAI;wBACrB8B,UAAUW,IAAI,CAAC/B,MAAMV,IAAI;oBAC3B;gBACF;;gBANK;gBAAA;;;yBAAA,8BAAA;wBAAA;;;wBAAA;8BAAA;;;;QAOP;;QATK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAUL,OAAO;QAAE6B,OAAAA;QAAOC,WAAAA;QAAWC,aAAAA;QAAaC,YAAAA;QAAYC,SAAAA;QAASC,aAAAA;IAAY;AAC3E;AAEA,gGAAgG;AAChG,iHAAiH;AACjH,SAAeW,WAAW7B,IAAgB,EAAEC,KAAoB;;;;;;oBAC9D;;wBAAMD,KAAK8B,IAAI,CAAC;;;oBAAhB;2BACI,CAAC7B,MAAMG,KAAK;+BAAZ;;;;oBAAiB;;wBAAMG,eAAeP;;;2BAArB;;;oBAArB,UAAkD;;;oBAClDC,MAAMG,KAAK,GAAG;oBACdH,MAAMI,iBAAiB,GAAG;oBAC1B;;wBAAM0B,IAAAA,8BAAe,EAAC/B,MAAM;;;;;4CAC1B,6HAA6H;4CAC7H,0JAA0J;4CAC1J;;gDAAMA,KAAK8B,IAAI,CAAC;;;4CAAhB;4CACA,0FAA0F;4CAC1F,gFAAgF;4CAChF;;gDAAMnB,IAAAA,iBAAO,EAACX,MAAMJ,oBAAoB;;;4CAAxC;;;;;;4BACF;;;;oBAPA;oBAQAK,MAAMG,KAAK,GAAG;;;;;;IAChB;;AAEO,SAAe1B,WAAWsB,IAAgB;;;;;;;wBAC9BA;;oBAAM;;wBAAMD,SAASC;;;oBAAtC;;wBAAM6B;4BAAiB;;;;oBAAvB;;;;;;IACF;;AAEO,SAAetD,eAAeyB,IAAgB;;YAC7CC,OAOG+B;;;;oBAPK;;wBAAMjC,SAASC;;;oBAAvBC,QAAQ;oBACd,IAAIA,MAAMG,KAAK,EAAE,MAAM,IAAI6B,oBAAU,CAAC,mBAAmB;;;;;;;;;oBAEvD;;wBAAMjC,KAAK8B,IAAI,CAAC;;;oBAAhB;oBACM;;wBAAMvB,eAAeP;;;oBAA3B,IAAI,CAAE,eAA6B;wBACjC,MAAM,IAAIiC,oBAAU,CAAC,mBAAmB;oBAC1C;;;;;;oBACOD;oBACP,IAAIA,AAAG,YAAHA,KAAeC,oBAAU,GAAE,MAAMD;oBACrC,MAAM,IAAIC,oBAAU,CAAC,mBAAmB,AAAC,iDAAuE,OAAvB,AAACD,IAAcE,OAAO,EAAC;;;;;;;IAEpH;;AAEA,qKAAqK;AACrK,+OAA+O;AAC/O,SAASC,kBAAkBtB,KAAe,EAAEC,SAAmB,EAAEC,WAAqB,EAAEC,UAAoB;IAC1G,IAAMoB,aAAuB,EAAE;IAC/B,IAAMC,YAAsB,EAAE;IAC9B,IAAMC,cAAyB,EAAE;IACjC,IAAMC,aAAwB,EAAE;IAEhC,IAAI1B,MAAMW,MAAM,GAAG,GAAG;QACpB,IAAMgB,YAAY3B,MAAM4B,GAAG,CAACC,wBAAa,EAAEC,IAAI,CAAC;YAC3C,kCAAA,2BAAA;;YAAL,QAAK,YAAe/D,2BAAf,SAAA,6BAAA,QAAA,yBAAA,iCAAuB;gBAAvB,IAAMgE,QAAN;gBACHR,WAAWX,IAAI,CAAC,AAAC,GAAsCoB,OAApChE,YAAY,CAAC+D,MAAM,EAAC,kBAAoEA,OAApDC,iCAAkB,EAAC,oCAAuC,OAAND,OAAM;gBACjHN,YAAYb,IAAI,CAACe;YACnB;;YAHK;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;QAIL,IAAI1B,UAAUU,MAAM,GAAG,GAAG;YACxB,IAAMsB,aAAa,AAAC,GAAqB,OAAnBD,iCAAkB,EAAC;YACzCR,UAAUZ,IAAI,CAACqB;YACfP,WAAWd,IAAI,CAACX,UAAU2B,GAAG,CAACC,wBAAa,EAAEC,IAAI,CAAC;QACpD;QACA,IAAI5B,YAAYS,MAAM,GAAG,GAAG;gBAK1Be;YAJA,IAAMO,cAAa,AAAC,GAAqB,OAAnBD,iCAAkB,EAAC;YACzC,IAAME,WAAWhC,YAAY0B,GAAG,CAAC,SAACO;uBAAU,AAAC,IAA6F,OAA1FpE,OAAO6D,GAAG,CAAC,SAACG;2BAAU,AAAC,wCAA6C,OAANA,OAAM;mBAASD,IAAI,CAAC,SAAQ;eAAIA,IAAI,CAAC;YACnJN,UAAUZ,IAAI,CAAC,AAAC,IAAqBsB,OAAlBD,aAAW,SAAgB,OAATC,UAAS;YAC9CR,WAAWd,IAAI,CAACV,YAAY0B,GAAG,CAACC,wBAAa,EAAEC,IAAI,CAAC;YACpDJ,CAAAA,cAAAA,YAAWd,IAAI,OAAfc,aAAgB,qBAAGxB,YAAYkC,OAAO,CAAC,SAACC;uBAAStE,OAAO6D,GAAG,CAAC;2BAAMC,IAAAA,wBAAa,EAACQ;;;QAClF;IACF;QAEK,mCAAA,4BAAA;;;YAAA,IAAMC,MAAN;gBAOHZ;YANA,IAAMa,SAASD,IAAIE,WAAW;gBACzB,kCAAA,2BAAA;;gBAAL,QAAK,YAAezE,2BAAf,SAAA,6BAAA,QAAA,yBAAA,iCAAuB;oBAAvB,IAAMgE,QAAN;oBACHR,WAAWX,IAAI,CAAC,AAAC,GAAuEmB,OAArE/D,YAAY,CAAC+D,MAAM,EAAC,mDAAuD,OAANA,OAAM;oBAC9FN,YAAYb,IAAI,CAACiB,IAAAA,wBAAa,EAACU;gBACjC;;gBAHK;gBAAA;;;yBAAA,6BAAA;wBAAA;;;wBAAA;8BAAA;;;;YAILf,UAAUZ,IAAI,CAAC,AAAC,IAA6F,OAA1F7C,OAAO6D,GAAG,CAAC,SAACG;uBAAU,AAAC,wCAA6C,OAANA,OAAM;eAASD,IAAI,CAAC,SAAQ;YAC7GJ,CAAAA,cAAAA,YAAWd,IAAI,OAAfc,aAAgB,qBAAG3D,OAAO6D,GAAG,CAAC;uBAAMC,IAAAA,wBAAa,EAACU;;QACpD;QARA,QAAK,aAAapC,+BAAb,UAAA,8BAAA,SAAA,0BAAA;;QAAA;QAAA;;;iBAAA,8BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAUL,OAAO;QAAEsC,UAAUlB,WAAWO,IAAI,CAAC;QAAQY,SAASlB,UAAUM,IAAI,CAAC,YAAY;QAAQa,QAAQ,AAAC,qBAAGlB,oBAAa,qBAAGC;IAAY;AACjI;AAIO,SAAe5D,aAAaqB,IAAgB,EAAEX,KAAa,EAAEoE,IAAyB,EAAExD,KAAoB;;YAuB7FO,OAtBdkD,aAIEC,WAAWC,WAAWC,WAAWC,OACmClD,aAApEC,OAAOC,WAAWC,aAAaC,YAAYC,SAASC,aAMtBiB,oBAA9BmB,UAAUC,SAASC,QACrBO,cACAC,UACAC,KAOAzD,MACA0D,MACAC;;;;oBAvBAT,cAActE,oBAAoBC;oBACxC,IAAIqE,gBAAgB,MAAM;wBACxB,MAAM,IAAIzB,oBAAU,CAAC,4BAA4B,AAAC,6CAAkEyB,OAAvBA,YAAYxE,KAAK,EAAC,OAAqDG,OAAhDqE,YAAYhE,KAAK,EAAC,gCAA2C,OAAbL,MAAMM,IAAI,IAAG;oBACnL;oBACQgE,YAA2CF,KAA3CE,WAAWC,YAAgCH,KAAhCG,WAAWC,YAAqBJ,KAArBI,WAAWC,QAAUL,KAAVK;oBACmClD,cAAAA,WAAWvB,QAA/EwB,QAAoED,YAApEC,OAAOC,YAA6DF,YAA7DE,WAAWC,cAAkDH,YAAlDG,aAAaC,aAAqCJ,YAArCI,YAAYC,UAAyBL,YAAzBK,SAASC,cAAgBN,YAAhBM;oBAC5D,IAAIA,eAAgBL,MAAMW,MAAM,KAAK,KAAKR,WAAWQ,MAAM,KAAK,GAAI;;;;yBAEhEX,CAAAA,MAAMW,MAAM,GAAG,CAAA,GAAfX;;;;oBAAkB;;wBAAMtC,eAAeyB;;;oBAArB;;;yBAClBa,CAAAA,MAAMW,MAAM,GAAG,CAAA,GAAfX;;;;oBAAkB;;wBAAMuD,IAAAA,gCAAiB,EAACpE,MAAMC;;;oBAA9B;;;oBAEgBkC,qBAAAA,kBAAkBtB,OAAOC,WAAWC,aAAaC,aAA/EsC,WAA8BnB,mBAA9BmB,UAAUC,UAAoBpB,mBAApBoB,SAASC,SAAWrB,mBAAXqB;oBACrBO,eAAe9C,QAAQO,MAAM,GAAG,IAAI,mDAAmD;oBACvFwC,WAAW/C,QAAQO,MAAM,GAAG,IAAI,KAAK;oBACrCyC,MAAM,AAAC,qBACuBX,OADHrC,QAAQO,MAAM,GAAG,IAAI,2BAA2B,IAAG,iDAC3BuC,OAArBT,UAAS,cAEzCK,OAFqDI,cAAa,4BAG5DR,OADNI,WAAU,gBAEVC,OADML,SAAQ,UACDM,OAAbD,WAAU,KACoBI,OADjBH,WAAU,wCACgB,OAATG;oBACrB;;wBAAMhE,KAAKS,OAAO,CAACwD;;;oBAA1BzD,OAAO;oBACC;;wBAAMA,CAAAA,QAAAA,MAAK6D,GAAG,OAAR7D,OAAAA,AAAS,qBAAGgD,eAAQ,qBAAIvC,QAAQO,MAAM,GAAG;4BAAUsC;;;;oBAAjEI,OAAQ;oBACRC,UAAUlD,QAAQO,MAAM,KAAK,IAAI0C,OAAOA,KAAKI,MAAM,CAAC,SAACC;+BAAQtD,QAAQuD,KAAK,CAAC,SAACnD;mCAAWzC,OAAO6F,IAAI,CAAC,SAAC7B;oCAA8B2B;uCAApBG,IAAAA,8BAAmB,GAACH,aAAAA,GAAG,CAAC3B,MAAM,cAAV2B,wBAAAA,aAAc,IAAIlD;;;;oBAC1J;;wBAAO8C,QAAQQ,KAAK,CAAC,GAAGb,OAAOrB,GAAG,CAAC;gCAAGmC,aAAAA;mCAAY;gCAAEA,MAAAA;4BAAK;;;;;IAC3D;;AAIO,SAASpG,mBAAmBwB,IAAgB;IACjD,OAAO;QACL6E,OAAO,SAAPA,MAAcxF,OAAOoE;;;;;;;gCAAsBzD;gCAAMX;gCAAOoE;;4BAAM;;gCAAM1D,SAASC;;;;;gCAA/CrB;oCAAgC;;;;;;;QAC9DmG,WAAW,SAAXA;mBAAiBrG,iBAAiBuB;;IACpC;AACF"}
|
|
@@ -1,13 +1,13 @@
|
|
|
1
1
|
import type { DuckDBConnection, DuckDBInstance } from '@duckdb/node-api';
|
|
2
2
|
import type { ResolvedConfig } from '../../config/index.js';
|
|
3
|
-
import type { OpenResult } from '../open.js';
|
|
3
|
+
import type { InternalOpenOptions, OpenResult } from '../open.js';
|
|
4
4
|
import type { OpenDialect } from '../types.js';
|
|
5
5
|
export declare const DB_FILENAME = "cache.duckdb";
|
|
6
|
-
export declare const SCHEMA_VERSION = "
|
|
6
|
+
export declare const SCHEMA_VERSION = "5";
|
|
7
7
|
export type { OpenResult };
|
|
8
8
|
interface DuckdbHandle {
|
|
9
9
|
instance: DuckDBInstance;
|
|
10
10
|
duckdb: DuckDBConnection;
|
|
11
11
|
}
|
|
12
12
|
export declare const duckdbOpenDialect: OpenDialect<DuckdbHandle>;
|
|
13
|
-
export declare function openDuckdb(cfg: ResolvedConfig): Promise<OpenResult>;
|
|
13
|
+
export declare function openDuckdb(cfg: ResolvedConfig, options?: InternalOpenOptions): Promise<OpenResult>;
|
|
@@ -1,13 +1,13 @@
|
|
|
1
1
|
import type { DuckDBConnection, DuckDBInstance } from '@duckdb/node-api';
|
|
2
2
|
import type { ResolvedConfig } from '../../config/index.js';
|
|
3
|
-
import type { OpenResult } from '../open.js';
|
|
3
|
+
import type { InternalOpenOptions, OpenResult } from '../open.js';
|
|
4
4
|
import type { OpenDialect } from '../types.js';
|
|
5
5
|
export declare const DB_FILENAME = "cache.duckdb";
|
|
6
|
-
export declare const SCHEMA_VERSION = "
|
|
6
|
+
export declare const SCHEMA_VERSION = "5";
|
|
7
7
|
export type { OpenResult };
|
|
8
8
|
interface DuckdbHandle {
|
|
9
9
|
instance: DuckDBInstance;
|
|
10
10
|
duckdb: DuckDBConnection;
|
|
11
11
|
}
|
|
12
12
|
export declare const duckdbOpenDialect: OpenDialect<DuckdbHandle>;
|
|
13
|
-
export declare function openDuckdb(cfg: ResolvedConfig): Promise<OpenResult>;
|
|
13
|
+
export declare function openDuckdb(cfg: ResolvedConfig, options?: InternalOpenOptions): Promise<OpenResult>;
|
|
@@ -29,10 +29,12 @@ var _indexts1 = require("../../features/index.js");
|
|
|
29
29
|
var _opents = require("../open.js");
|
|
30
30
|
var _sharedts = require("../shared.js");
|
|
31
31
|
var _connectionts = require("./connection.js");
|
|
32
|
+
var _lexicalts = require("./lexical.js");
|
|
32
33
|
var _nativets = require("./native.js");
|
|
33
34
|
var _reconcilets = require("./reconcile.js");
|
|
34
35
|
var _sqlfunctionsts = require("./sql-functions.js");
|
|
35
36
|
var _storets = require("./store.js");
|
|
37
|
+
var _vectorsts = require("./vectors.js");
|
|
36
38
|
function asyncGeneratorStep(gen, resolve, reject, _next, _throw, key, arg) {
|
|
37
39
|
try {
|
|
38
40
|
var info = gen[key](arg);
|
|
@@ -165,10 +167,11 @@ function _ts_generator(thisArg, body) {
|
|
|
165
167
|
}
|
|
166
168
|
}
|
|
167
169
|
var DB_FILENAME = 'cache.duckdb';
|
|
168
|
-
var SCHEMA_VERSION = '
|
|
169
|
-
// `content` is a plain table (not FTS-virtual): the
|
|
170
|
+
var SCHEMA_VERSION = '5';
|
|
171
|
+
// `content` is a plain table (not FTS-virtual): build prepares the persistent FTS index over it,
|
|
172
|
+
// while contains() verification reads the table directly.
|
|
170
173
|
// No tokenizer resolution: this store always uses the fts extension's default (porter) stemmer.
|
|
171
|
-
function ensureSchema(
|
|
174
|
+
function ensureSchema(handle, conn, cfg) {
|
|
172
175
|
return _async_to_generator(function() {
|
|
173
176
|
var _iteratorNormalCompletion, _didIteratorError, _iteratorError, _iterator, _step, feature, err;
|
|
174
177
|
return _ts_generator(this, function(_state) {
|
|
@@ -188,75 +191,87 @@ function ensureSchema(_handle, conn, cfg) {
|
|
|
188
191
|
_state.sent();
|
|
189
192
|
return [
|
|
190
193
|
4,
|
|
191
|
-
conn.exec('CREATE TABLE IF NOT EXISTS
|
|
194
|
+
conn.exec('CREATE TABLE IF NOT EXISTS indexed_sources ("path" TEXT PRIMARY KEY, text TEXT NOT NULL)')
|
|
192
195
|
];
|
|
193
196
|
case 3:
|
|
194
197
|
_state.sent();
|
|
195
198
|
return [
|
|
196
199
|
4,
|
|
197
|
-
conn.exec('CREATE
|
|
200
|
+
conn.exec('CREATE TABLE IF NOT EXISTS preset_files ("path" TEXT, preset TEXT, PRIMARY KEY ("path", preset))')
|
|
198
201
|
];
|
|
199
202
|
case 4:
|
|
200
203
|
_state.sent();
|
|
201
|
-
|
|
202
|
-
|
|
204
|
+
return [
|
|
205
|
+
4,
|
|
206
|
+
conn.exec('CREATE INDEX IF NOT EXISTS preset_files_preset ON preset_files(preset)')
|
|
207
|
+
];
|
|
203
208
|
case 5:
|
|
209
|
+
_state.sent();
|
|
210
|
+
_iteratorNormalCompletion = true, _didIteratorError = false, _iteratorError = undefined;
|
|
211
|
+
_state.label = 6;
|
|
212
|
+
case 6:
|
|
204
213
|
_state.trys.push([
|
|
205
|
-
|
|
206
|
-
|
|
207
|
-
|
|
208
|
-
|
|
214
|
+
6,
|
|
215
|
+
14,
|
|
216
|
+
15,
|
|
217
|
+
16
|
|
209
218
|
]);
|
|
210
219
|
_iterator = (0, _indexts1.activeFeatures)(cfg)[Symbol.iterator]();
|
|
211
|
-
_state.label =
|
|
212
|
-
case
|
|
220
|
+
_state.label = 7;
|
|
221
|
+
case 7:
|
|
213
222
|
if (!!(_iteratorNormalCompletion = (_step = _iterator.next()).done)) return [
|
|
214
223
|
3,
|
|
215
|
-
|
|
224
|
+
13
|
|
216
225
|
];
|
|
217
226
|
feature = _step.value;
|
|
218
227
|
if (!(feature.name === 'embed')) return [
|
|
219
228
|
3,
|
|
220
|
-
|
|
229
|
+
10
|
|
221
230
|
];
|
|
222
231
|
return [
|
|
223
232
|
4,
|
|
224
233
|
conn.exec('CREATE TABLE IF NOT EXISTS embeddings ("path" TEXT, chunk INTEGER, start_line INTEGER, end_line INTEGER, scale REAL, vector FLOAT['.concat(_typests.STORE_DIMS, '], PRIMARY KEY ("path", chunk))'))
|
|
225
234
|
];
|
|
226
|
-
case
|
|
235
|
+
case 8:
|
|
236
|
+
_state.sent();
|
|
237
|
+
return [
|
|
238
|
+
4,
|
|
239
|
+
(0, _vectorsts.createVectorWriteStage)(handle.duckdb, _typests.STORE_DIMS)
|
|
240
|
+
];
|
|
241
|
+
case 9:
|
|
227
242
|
_state.sent();
|
|
228
243
|
return [
|
|
229
244
|
3,
|
|
230
|
-
|
|
245
|
+
12
|
|
231
246
|
];
|
|
232
|
-
case
|
|
247
|
+
case 10:
|
|
233
248
|
return [
|
|
234
249
|
4,
|
|
235
250
|
feature.schema(conn)
|
|
236
251
|
];
|
|
237
|
-
case
|
|
252
|
+
case 11:
|
|
238
253
|
_state.sent();
|
|
239
|
-
_state.label =
|
|
240
|
-
case
|
|
254
|
+
_state.label = 12;
|
|
255
|
+
case 12:
|
|
241
256
|
_iteratorNormalCompletion = true;
|
|
242
257
|
return [
|
|
243
258
|
3,
|
|
244
|
-
|
|
259
|
+
7
|
|
245
260
|
];
|
|
246
|
-
case
|
|
261
|
+
case 13:
|
|
247
262
|
return [
|
|
248
263
|
3,
|
|
249
|
-
|
|
264
|
+
16
|
|
250
265
|
];
|
|
251
|
-
case
|
|
266
|
+
case 14:
|
|
252
267
|
err = _state.sent();
|
|
253
268
|
_didIteratorError = true;
|
|
254
269
|
_iteratorError = err;
|
|
255
270
|
return [
|
|
256
271
|
3,
|
|
257
|
-
|
|
272
|
+
16
|
|
258
273
|
];
|
|
259
|
-
case
|
|
274
|
+
case 15:
|
|
260
275
|
try {
|
|
261
276
|
if (!_iteratorNormalCompletion && _iterator.return != null) {
|
|
262
277
|
_iterator.return();
|
|
@@ -269,41 +284,41 @@ function ensureSchema(_handle, conn, cfg) {
|
|
|
269
284
|
return [
|
|
270
285
|
7
|
|
271
286
|
];
|
|
272
|
-
case
|
|
287
|
+
case 16:
|
|
273
288
|
return [
|
|
274
289
|
4,
|
|
275
290
|
(0, _sharedts.getMeta)(conn, 'schema_version')
|
|
276
291
|
];
|
|
277
|
-
case
|
|
292
|
+
case 17:
|
|
278
293
|
if (!(_state.sent() === null)) return [
|
|
279
294
|
3,
|
|
280
|
-
|
|
295
|
+
19
|
|
281
296
|
];
|
|
282
297
|
return [
|
|
283
298
|
4,
|
|
284
299
|
(0, _sharedts.setMeta)(conn, 'schema_version', SCHEMA_VERSION)
|
|
285
300
|
];
|
|
286
|
-
case
|
|
301
|
+
case 18:
|
|
287
302
|
_state.sent();
|
|
288
|
-
_state.label =
|
|
289
|
-
case
|
|
303
|
+
_state.label = 19;
|
|
304
|
+
case 19:
|
|
290
305
|
return [
|
|
291
306
|
4,
|
|
292
307
|
(0, _sharedts.getMeta)(conn, 'features')
|
|
293
308
|
];
|
|
294
|
-
case
|
|
309
|
+
case 20:
|
|
295
310
|
if (!(_state.sent() === null)) return [
|
|
296
311
|
3,
|
|
297
|
-
|
|
312
|
+
22
|
|
298
313
|
];
|
|
299
314
|
return [
|
|
300
315
|
4,
|
|
301
316
|
(0, _sharedts.setMeta)(conn, 'features', (0, _indexts.featureSignature)(cfg, _indexts1.FEATURES))
|
|
302
317
|
];
|
|
303
|
-
case
|
|
318
|
+
case 21:
|
|
304
319
|
_state.sent();
|
|
305
|
-
_state.label =
|
|
306
|
-
case
|
|
320
|
+
_state.label = 22;
|
|
321
|
+
case 22:
|
|
307
322
|
return [
|
|
308
323
|
2
|
|
309
324
|
];
|
|
@@ -323,7 +338,7 @@ function close(handle) {
|
|
|
323
338
|
});
|
|
324
339
|
})();
|
|
325
340
|
}
|
|
326
|
-
function connect(dbPath, _cfg) {
|
|
341
|
+
function connect(dbPath, _cfg, options) {
|
|
327
342
|
return _async_to_generator(function() {
|
|
328
343
|
var DuckDBInstance, err, duckdb, instance, err1, conn, err2;
|
|
329
344
|
return _ts_generator(this, function(_state) {
|
|
@@ -358,7 +373,9 @@ function connect(dbPath, _cfg) {
|
|
|
358
373
|
]);
|
|
359
374
|
return [
|
|
360
375
|
4,
|
|
361
|
-
DuckDBInstance.create(dbPath, {
|
|
376
|
+
DuckDBInstance.create(dbPath, options.observational ? {
|
|
377
|
+
access_mode: 'READ_ONLY'
|
|
378
|
+
} : {
|
|
362
379
|
storage_compatibility_version: 'v1.5.0'
|
|
363
380
|
})
|
|
364
381
|
];
|
|
@@ -438,16 +455,18 @@ var duckdbOpenDialect = {
|
|
|
438
455
|
return /Could not set lock on file|being used by another process/.test(err.message);
|
|
439
456
|
},
|
|
440
457
|
ensureSchema: ensureSchema,
|
|
458
|
+
prepareLexical: _lexicalts.prepareFts,
|
|
459
|
+
assertLexicalReady: _lexicalts.assertFtsReady,
|
|
441
460
|
createStore: function createStore(handle, conn) {
|
|
442
461
|
return (0, _storets.createStore)(handle.instance, handle.duckdb, conn);
|
|
443
462
|
}
|
|
444
463
|
};
|
|
445
|
-
function openDuckdb(cfg) {
|
|
464
|
+
function openDuckdb(cfg, options) {
|
|
446
465
|
return _async_to_generator(function() {
|
|
447
466
|
return _ts_generator(this, function(_state) {
|
|
448
467
|
return [
|
|
449
468
|
2,
|
|
450
|
-
(0, _opents.openWithDialect)(cfg, duckdbOpenDialect)
|
|
469
|
+
(0, _opents.openWithDialect)(cfg, duckdbOpenDialect, options)
|
|
451
470
|
];
|
|
452
471
|
});
|
|
453
472
|
})();
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/duckdb/open.ts"],"sourcesContent":["import type { DuckDBConnection, DuckDBInstance } from '@duckdb/node-api';\nimport type { Config, ResolvedConfig } from '../../config/index.ts';\nimport { featureSignature } from '../../config/index.ts';\nimport { STORE_DIMS } from '../../embed/types.ts';\nimport { SenseError } from '../../errors.ts';\nimport { activeFeatures, FEATURES } from '../../features/index.ts';\nimport type { OpenResult } from '../open.ts';\nimport { openWithDialect } from '../open.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport type { Connection, OpenDialect } from '../types.ts';\nimport { createConnection } from './connection.ts';\nimport { DUCKDB_PACKAGE, duckdbApi } from './native.ts';\nimport { duckdbDialect } from './reconcile.ts';\nimport { registerFunctions } from './sql-functions.ts';\nimport { createStore } from './store.ts';\n\nexport const DB_FILENAME = 'cache.duckdb';\n// Independent of sqlite's SCHEMA_VERSION -- the two stores' cache shapes evolve separately (VARIANT frontmatter columns vs untyped).\n// The store name already joins the feature signature, so switching a config's `store` key rebuilds rather than reusing the other engine's cache.\nexport const SCHEMA_VERSION = '4';\n\nexport type { OpenResult };\n\n// The connection plus this store's own native handles (types.ts's OpenDialect<Handle>): the\n// instance owns the WAL and must outlive the connection borrowed from it.\ninterface DuckdbHandle {\n instance: DuckDBInstance;\n duckdb: DuckDBConnection;\n}\n\n// `content` is a plain table (not FTS-virtual): the fts index is built over it lazily, only when a lexical query runs (lexical.ts), and read directly for contains() verification either way.\n// No tokenizer resolution: this store always uses the fts extension's default (porter) stemmer.\nasync function ensureSchema(_handle: DuckdbHandle, conn: Connection, cfg: Config): Promise<void> {\n await conn.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" DOUBLE, \"_ctime\" DOUBLE, \"_size\" INTEGER, \"_parse_error\" TEXT)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS content (\"path\" TEXT PRIMARY KEY, title TEXT, summary TEXT, text TEXT)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS preset_files (\"path\" TEXT, preset TEXT, PRIMARY KEY (\"path\", preset))`);\n await conn.exec('CREATE INDEX IF NOT EXISTS preset_files_preset ON preset_files(preset)');\n for (const feature of activeFeatures(cfg)) {\n // Native FLOAT[STORE_DIMS] instead of the embed feature's engine-neutral BLOB+scale DDL (vectors.ts). `scale` is kept, unused,\n // so the feature's shared reconcile-time INSERT/DELETE (features/embed.ts) names a column that exists on both stores.\n if (feature.name === 'embed') {\n await conn.exec(`CREATE TABLE IF NOT EXISTS embeddings (\"path\" TEXT, chunk INTEGER, start_line INTEGER, end_line INTEGER, scale REAL, vector FLOAT[${STORE_DIMS}], PRIMARY KEY (\"path\", chunk))`);\n continue;\n }\n await feature.schema(conn);\n }\n if ((await getMeta(conn, 'schema_version')) === null) await setMeta(conn, 'schema_version', SCHEMA_VERSION);\n if ((await getMeta(conn, 'features')) === null) await setMeta(conn, 'features', featureSignature(cfg, FEATURES));\n}\n\n// Order matters: the connection must be gone before the instance closes the WAL.\nasync function close(handle: DuckdbHandle): Promise<void> {\n handle.duckdb.disconnectSync();\n handle.instance.closeSync();\n}\n\nasync function connect(dbPath: string, _cfg: ResolvedConfig): Promise<{ handle: DuckdbHandle; conn: Connection }> {\n // Dynamic, not a top-level import: sqlite trees must never attempt to resolve this optional peer dependency, so nothing imports\n // it as a value until a duckdb tree opens (types-only imports are erased). Installed on first use if missing, shared with sql-functions.ts and vectors.ts via native.ts's duckdbApi.\n let DuckDBInstance: typeof import('@duckdb/node-api').DuckDBInstance;\n try {\n ({ DuckDBInstance } = await duckdbApi());\n } catch (err) {\n if (err instanceof SenseError) throw err;\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"duckdb\" needs the ${DUCKDB_PACKAGE} package (${(err as Error).message})`);\n }\n\n let duckdb: DuckDBConnection;\n let instance: DuckDBInstance | undefined;\n try {\n // On-disk files default to an older storage format for cross-version compatibility, which rejects VARIANT columns (\"VARIANT\n // columns are not supported in storage versions prior to v1.5.0\"); this store's dynamic frontmatter columns need VARIANT (reconcile.ts), so the floor is pinned explicitly.\n instance = await DuckDBInstance.create(dbPath, { storage_compatibility_version: 'v1.5.0' });\n duckdb = await instance.connect();\n } catch (err) {\n // create() may have succeeded before connect() failed: close it, or its WAL stays open.\n instance?.closeSync();\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"duckdb\" failed to open ${dbPath}: ${(err as Error).message}`);\n }\n // A throw below would leak the open instance, whose WAL then locks the .duckdb file undeletable on Windows.\n try {\n await registerFunctions(duckdb);\n const conn = createConnection(duckdb);\n return { handle: { instance, duckdb }, conn };\n } catch (err) {\n await close({ instance, duckdb });\n throw err;\n }\n}\n\n// This store's dialect (types.ts's OpenDialect) for the shared orchestration in store/open.ts.\nexport const duckdbOpenDialect: OpenDialect<DuckdbHandle> = {\n filename: DB_FILENAME,\n schemaVersion: SCHEMA_VERSION,\n reconcileDialect: duckdbDialect,\n connect,\n close,\n // duckdb's file lock spans the connection's life, and each platform words the refusal its own\n // way: posix \"Could not set lock on file ... Conflicting lock is held in <exe> (PID n)\", Windows\n // \"Cannot open file ... being used by another process\". Both are the same condition.\n isLocked: (err) => /Could not set lock on file|being used by another process/.test(err.message),\n ensureSchema,\n createStore: (handle, conn) => createStore(handle.instance, handle.duckdb, conn),\n};\n\nexport async function openDuckdb(cfg: ResolvedConfig): Promise<OpenResult> {\n return openWithDialect(cfg, duckdbOpenDialect);\n}\n"],"names":["DB_FILENAME","SCHEMA_VERSION","duckdbOpenDialect","openDuckdb","ensureSchema","_handle","conn","cfg","feature","exec","activeFeatures","name","STORE_DIMS","schema","getMeta","setMeta","featureSignature","FEATURES","close","handle","duckdb","disconnectSync","instance","closeSync","connect","dbPath","_cfg","DuckDBInstance","err","duckdbApi","SenseError","DUCKDB_PACKAGE","message","create","storage_compatibility_version","registerFunctions","createConnection","filename","schemaVersion","reconcileDialect","duckdbDialect","isLocked","test","createStore","openWithDialect"],"mappings":";;;;;;;;;;;QAgBaA;eAAAA;;QAGAC;eAAAA;;QAwEAC;eAAAA;;QAcSC;eAAAA;;;uBAvGW;uBACN;wBACA;wBACc;sBAET;wBACC;4BAEA;wBACS;2BACZ;8BACI;uBACN;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAErB,IAAMH,cAAc;AAGpB,IAAMC,iBAAiB;AAW9B,8LAA8L;AAC9L,gGAAgG;AAChG,SAAeG,aAAaC,OAAqB,EAAEC,IAAgB,EAAEC,GAAW;;YAKzE,2BAAA,mBAAA,gBAAA,WAAA,OAAMC;;;;oBAJX;;wBAAMF,KAAKG,IAAI,CAAC;;;oBAAhB;oBACA;;wBAAMH,KAAKG,IAAI,CAAC;;;oBAAhB;oBACA;;wBAAMH,KAAKG,IAAI,CAAC;;;oBAAhB;oBACA;;wBAAMH,KAAKG,IAAI,CAAC;;;oBAAhB;oBACK,kCAAA,2BAAA;;;;;;;;;oBAAA,YAAiBC,IAAAA,wBAAc,EAACH;;;2BAAhC,6BAAA,QAAA;;;;oBAAMC,UAAN;yBAGCA,CAAAA,QAAQG,IAAI,KAAK,OAAM,GAAvBH;;;;oBACF;;wBAAMF,KAAKG,IAAI,CAAC,AAAC,qIAA+I,OAAXG,mBAAU,EAAC;;;oBAAhK;oBACA;;;;;oBAEF;;wBAAMJ,QAAQK,MAAM,CAACP;;;oBAArB;;;oBAPG;;;;;;;;;;;;oBAAA;oBAAA;;;;;;;6BAAA,6BAAA;4BAAA;;;4BAAA;kCAAA;;;;;;;oBASA;;wBAAMQ,IAAAA,iBAAO,EAACR,MAAM;;;yBAArB,CAAA,AAAC,kBAA2C,IAAG,GAA/C;;;;oBAAkD;;wBAAMS,IAAAA,iBAAO,EAACT,MAAM,kBAAkBL;;;oBAAtC;;;oBACjD;;wBAAMa,IAAAA,iBAAO,EAACR,MAAM;;;yBAArB,CAAA,AAAC,kBAAqC,IAAG,GAAzC;;;;oBAA4C;;wBAAMS,IAAAA,iBAAO,EAACT,MAAM,YAAYU,IAAAA,yBAAgB,EAACT,KAAKU,kBAAQ;;;oBAA9D;;;;;;;;IAClD;;AAEA,iFAAiF;AACjF,SAAeC,MAAMC,MAAoB;;;YACvCA,OAAOC,MAAM,CAACC,cAAc;YAC5BF,OAAOG,QAAQ,CAACC,SAAS;;;;;IAC3B;;AAEA,SAAeC,QAAQC,MAAc,EAAEC,IAAoB;;YAGrDC,gBAGKC,KAKLR,QACAE,UAMKM,MAQDtB,MAECsB;;;;;;;;;;oBAvBe;;wBAAMC,IAAAA,mBAAS;;;oBAAlCF,iBAAmB,cAAnBA;;;;;;oBACIC;oBACP,IAAIA,AAAG,YAAHA,KAAeE,oBAAU,GAAE,MAAMF;oBACrC,MAAM,IAAIE,oBAAU,CAAC,4BAA4B,AAAC,4BAAsD,OAA3BC,wBAAc,EAAC,cAAmC,OAAvB,AAACH,IAAcI,OAAO,EAAC;;;;;;;;oBAQpH;;wBAAML,eAAeM,MAAM,CAACR,QAAQ;4BAAES,+BAA+B;wBAAS;;;oBAFzF,4HAA4H;oBAC5H,4KAA4K;oBAC5KZ,WAAW;oBACF;;wBAAMA,SAASE,OAAO;;;oBAA/BJ,SAAS;;;;;;oBACFQ;oBACP,wFAAwF;oBACxFN,qBAAAA,+BAAAA,SAAUC,SAAS;oBACnB,MAAM,IAAIO,oBAAU,CAAC,4BAA4B,AAAC,iCAA2C,OAAXL,QAAO,MAA2B,OAAvB,AAACG,KAAcI,OAAO;;;;;;;;oBAInH;;wBAAMG,IAAAA,iCAAiB,EAACf;;;oBAAxB;oBACMd,OAAO8B,IAAAA,8BAAgB,EAAChB;oBAC9B;;wBAAO;4BAAED,QAAQ;gCAAEG,UAAAA;gCAAUF,QAAAA;4BAAO;4BAAGd,MAAAA;wBAAK;;;oBACrCsB;oBACP;;wBAAMV,MAAM;4BAAEI,UAAAA;4BAAUF,QAAAA;wBAAO;;;oBAA/B;oBACA,MAAMQ;;;;;;;IAEV;;AAGO,IAAM1B,oBAA+C;IAC1DmC,UAAUrC;IACVsC,eAAerC;IACfsC,kBAAkBC,0BAAa;IAC/BhB,SAAAA;IACAN,OAAAA;IACA,8FAA8F;IAC9F,iGAAiG;IACjG,qFAAqF;IACrFuB,UAAU,SAAVA,SAAWb;eAAQ,2DAA2Dc,IAAI,CAACd,IAAII,OAAO;;IAC9F5B,cAAAA;IACAuC,aAAa,SAAbA,YAAcxB,QAAQb;eAASqC,IAAAA,oBAAW,EAACxB,OAAOG,QAAQ,EAAEH,OAAOC,MAAM,EAAEd;;AAC7E;AAEO,SAAeH,WAAWI,GAAmB;;;YAClD;;gBAAOqC,IAAAA,uBAAe,EAACrC,KAAKL;;;IAC9B"}
|
|
1
|
+
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/duckdb/open.ts"],"sourcesContent":["import type { DuckDBConnection, DuckDBInstance } from '@duckdb/node-api';\nimport type { Config, ResolvedConfig } from '../../config/index.ts';\nimport { featureSignature } from '../../config/index.ts';\nimport { STORE_DIMS } from '../../embed/types.ts';\nimport { SenseError } from '../../errors.ts';\nimport { activeFeatures, FEATURES } from '../../features/index.ts';\nimport type { InternalOpenOptions, OpenResult } from '../open.ts';\nimport { openWithDialect } from '../open.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport type { Connection, OpenDialect } from '../types.ts';\nimport { createConnection } from './connection.ts';\nimport { assertFtsReady, prepareFts } from './lexical.ts';\nimport { DUCKDB_PACKAGE, duckdbApi } from './native.ts';\nimport { duckdbDialect } from './reconcile.ts';\nimport { registerFunctions } from './sql-functions.ts';\nimport { createStore } from './store.ts';\nimport { createVectorWriteStage } from './vectors.ts';\n\nexport const DB_FILENAME = 'cache.duckdb';\n// Independent of sqlite's SCHEMA_VERSION -- the two stores' cache shapes evolve separately (VARIANT frontmatter columns vs untyped).\n// The store name already joins the feature signature, so switching a config's `store` key rebuilds rather than reusing the other engine's cache.\nexport const SCHEMA_VERSION = '5';\n\nexport type { OpenResult };\n\n// The connection plus this store's own native handles (types.ts's OpenDialect<Handle>): the\n// instance owns the WAL and must outlive the connection borrowed from it.\ninterface DuckdbHandle {\n instance: DuckDBInstance;\n duckdb: DuckDBConnection;\n}\n\n// `content` is a plain table (not FTS-virtual): build prepares the persistent FTS index over it,\n// while contains() verification reads the table directly.\n// No tokenizer resolution: this store always uses the fts extension's default (porter) stemmer.\nasync function ensureSchema(handle: DuckdbHandle, conn: Connection, cfg: Config): Promise<void> {\n await conn.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" DOUBLE, \"_ctime\" DOUBLE, \"_size\" INTEGER, \"_parse_error\" TEXT)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS content (\"path\" TEXT PRIMARY KEY, title TEXT, summary TEXT, text TEXT)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS indexed_sources (\"path\" TEXT PRIMARY KEY, text TEXT NOT NULL)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS preset_files (\"path\" TEXT, preset TEXT, PRIMARY KEY (\"path\", preset))`);\n await conn.exec('CREATE INDEX IF NOT EXISTS preset_files_preset ON preset_files(preset)');\n for (const feature of activeFeatures(cfg)) {\n // Native FLOAT[STORE_DIMS] instead of the embed feature's engine-neutral BLOB+scale DDL (vectors.ts). `scale` is kept, unused,\n // so the feature's shared reconcile-time INSERT/DELETE (features/embed.ts) names a column that exists on both stores.\n if (feature.name === 'embed') {\n await conn.exec(`CREATE TABLE IF NOT EXISTS embeddings (\"path\" TEXT, chunk INTEGER, start_line INTEGER, end_line INTEGER, scale REAL, vector FLOAT[${STORE_DIMS}], PRIMARY KEY (\"path\", chunk))`);\n await createVectorWriteStage(handle.duckdb, STORE_DIMS);\n continue;\n }\n await feature.schema(conn);\n }\n if ((await getMeta(conn, 'schema_version')) === null) await setMeta(conn, 'schema_version', SCHEMA_VERSION);\n if ((await getMeta(conn, 'features')) === null) await setMeta(conn, 'features', featureSignature(cfg, FEATURES));\n}\n\n// Order matters: the connection must be gone before the instance closes the WAL.\nasync function close(handle: DuckdbHandle): Promise<void> {\n handle.duckdb.disconnectSync();\n handle.instance.closeSync();\n}\n\nasync function connect(dbPath: string, _cfg: ResolvedConfig, options: { existingOnly: boolean; observational: boolean }): Promise<{ handle: DuckdbHandle; conn: Connection }> {\n // Dynamic, not a top-level import: sqlite trees must never attempt to resolve this optional peer dependency, so nothing imports\n // it as a value until a duckdb tree opens (types-only imports are erased). Installed on first use if missing, shared with sql-functions.ts and vectors.ts via native.ts's duckdbApi.\n let DuckDBInstance: typeof import('@duckdb/node-api').DuckDBInstance;\n try {\n ({ DuckDBInstance } = await duckdbApi());\n } catch (err) {\n if (err instanceof SenseError) throw err;\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"duckdb\" needs the ${DUCKDB_PACKAGE} package (${(err as Error).message})`);\n }\n\n let duckdb: DuckDBConnection;\n let instance: DuckDBInstance | undefined;\n try {\n // On-disk files default to an older storage format for cross-version compatibility, which rejects VARIANT columns (\"VARIANT\n // columns are not supported in storage versions prior to v1.5.0\"); this store's dynamic frontmatter columns need VARIANT (reconcile.ts), so the floor is pinned explicitly.\n instance = await DuckDBInstance.create(dbPath, options.observational ? { access_mode: 'READ_ONLY' } : { storage_compatibility_version: 'v1.5.0' });\n duckdb = await instance.connect();\n } catch (err) {\n // create() may have succeeded before connect() failed: close it, or its WAL stays open.\n instance?.closeSync();\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"duckdb\" failed to open ${dbPath}: ${(err as Error).message}`);\n }\n // A throw below would leak the open instance, whose WAL then locks the .duckdb file undeletable on Windows.\n try {\n await registerFunctions(duckdb);\n const conn = createConnection(duckdb);\n return { handle: { instance, duckdb }, conn };\n } catch (err) {\n await close({ instance, duckdb });\n throw err;\n }\n}\n\n// This store's dialect (types.ts's OpenDialect) for the shared orchestration in store/open.ts.\nexport const duckdbOpenDialect: OpenDialect<DuckdbHandle> = {\n filename: DB_FILENAME,\n schemaVersion: SCHEMA_VERSION,\n reconcileDialect: duckdbDialect,\n connect,\n close,\n // duckdb's file lock spans the connection's life, and each platform words the refusal its own\n // way: posix \"Could not set lock on file ... Conflicting lock is held in <exe> (PID n)\", Windows\n // \"Cannot open file ... being used by another process\". Both are the same condition.\n isLocked: (err) => /Could not set lock on file|being used by another process/.test(err.message),\n ensureSchema,\n prepareLexical: prepareFts,\n assertLexicalReady: assertFtsReady,\n createStore: (handle, conn) => createStore(handle.instance, handle.duckdb, conn),\n};\n\nexport async function openDuckdb(cfg: ResolvedConfig, options?: InternalOpenOptions): Promise<OpenResult> {\n return openWithDialect(cfg, duckdbOpenDialect, options);\n}\n"],"names":["DB_FILENAME","SCHEMA_VERSION","duckdbOpenDialect","openDuckdb","ensureSchema","handle","conn","cfg","feature","exec","activeFeatures","name","STORE_DIMS","createVectorWriteStage","duckdb","schema","getMeta","setMeta","featureSignature","FEATURES","close","disconnectSync","instance","closeSync","connect","dbPath","_cfg","options","DuckDBInstance","err","duckdbApi","SenseError","DUCKDB_PACKAGE","message","create","observational","access_mode","storage_compatibility_version","registerFunctions","createConnection","filename","schemaVersion","reconcileDialect","duckdbDialect","isLocked","test","prepareLexical","prepareFts","assertLexicalReady","assertFtsReady","createStore","openWithDialect"],"mappings":";;;;;;;;;;;QAkBaA;eAAAA;;QAGAC;eAAAA;;QA2EAC;eAAAA;;QAgBSC;eAAAA;;;uBA9GW;uBACN;wBACA;wBACc;sBAET;wBACC;4BAEA;yBACU;wBACD;2BACZ;8BACI;uBACN;yBACW;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAEhC,IAAMH,cAAc;AAGpB,IAAMC,iBAAiB;AAW9B,iGAAiG;AACjG,0DAA0D;AAC1D,gGAAgG;AAChG,SAAeG,aAAaC,MAAoB,EAAEC,IAAgB,EAAEC,GAAW;;YAMxE,2BAAA,mBAAA,gBAAA,WAAA,OAAMC;;;;oBALX;;wBAAMF,KAAKG,IAAI,CAAC;;;oBAAhB;oBACA;;wBAAMH,KAAKG,IAAI,CAAC;;;oBAAhB;oBACA;;wBAAMH,KAAKG,IAAI,CAAC;;;oBAAhB;oBACA;;wBAAMH,KAAKG,IAAI,CAAC;;;oBAAhB;oBACA;;wBAAMH,KAAKG,IAAI,CAAC;;;oBAAhB;oBACK,kCAAA,2BAAA;;;;;;;;;oBAAA,YAAiBC,IAAAA,wBAAc,EAACH;;;2BAAhC,6BAAA,QAAA;;;;oBAAMC,UAAN;yBAGCA,CAAAA,QAAQG,IAAI,KAAK,OAAM,GAAvBH;;;;oBACF;;wBAAMF,KAAKG,IAAI,CAAC,AAAC,qIAA+I,OAAXG,mBAAU,EAAC;;;oBAAhK;oBACA;;wBAAMC,IAAAA,iCAAsB,EAACR,OAAOS,MAAM,EAAEF,mBAAU;;;oBAAtD;oBACA;;;;;oBAEF;;wBAAMJ,QAAQO,MAAM,CAACT;;;oBAArB;;;oBARG;;;;;;;;;;;;oBAAA;oBAAA;;;;;;;6BAAA,6BAAA;4BAAA;;;4BAAA;kCAAA;;;;;;;oBAUA;;wBAAMU,IAAAA,iBAAO,EAACV,MAAM;;;yBAArB,CAAA,AAAC,kBAA2C,IAAG,GAA/C;;;;oBAAkD;;wBAAMW,IAAAA,iBAAO,EAACX,MAAM,kBAAkBL;;;oBAAtC;;;oBACjD;;wBAAMe,IAAAA,iBAAO,EAACV,MAAM;;;yBAArB,CAAA,AAAC,kBAAqC,IAAG,GAAzC;;;;oBAA4C;;wBAAMW,IAAAA,iBAAO,EAACX,MAAM,YAAYY,IAAAA,yBAAgB,EAACX,KAAKY,kBAAQ;;;oBAA9D;;;;;;;;IAClD;;AAEA,iFAAiF;AACjF,SAAeC,MAAMf,MAAoB;;;YACvCA,OAAOS,MAAM,CAACO,cAAc;YAC5BhB,OAAOiB,QAAQ,CAACC,SAAS;;;;;IAC3B;;AAEA,SAAeC,QAAQC,MAAc,EAAEC,IAAoB,EAAEC,OAA0D;;YAGjHC,gBAGKC,KAKLf,QACAQ,UAMKO,MAQDvB,MAECuB;;;;;;;;;;oBAvBe;;wBAAMC,IAAAA,mBAAS;;;oBAAlCF,iBAAmB,cAAnBA;;;;;;oBACIC;oBACP,IAAIA,AAAG,YAAHA,KAAeE,oBAAU,GAAE,MAAMF;oBACrC,MAAM,IAAIE,oBAAU,CAAC,4BAA4B,AAAC,4BAAsD,OAA3BC,wBAAc,EAAC,cAAmC,OAAvB,AAACH,IAAcI,OAAO,EAAC;;;;;;;;oBAQpH;;wBAAML,eAAeM,MAAM,CAACT,QAAQE,QAAQQ,aAAa,GAAG;4BAAEC,aAAa;wBAAY,IAAI;4BAAEC,+BAA+B;wBAAS;;;oBAFhJ,4HAA4H;oBAC5H,4KAA4K;oBAC5Kf,WAAW;oBACF;;wBAAMA,SAASE,OAAO;;;oBAA/BV,SAAS;;;;;;oBACFe;oBACP,wFAAwF;oBACxFP,qBAAAA,+BAAAA,SAAUC,SAAS;oBACnB,MAAM,IAAIQ,oBAAU,CAAC,4BAA4B,AAAC,iCAA2C,OAAXN,QAAO,MAA2B,OAAvB,AAACI,KAAcI,OAAO;;;;;;;;oBAInH;;wBAAMK,IAAAA,iCAAiB,EAACxB;;;oBAAxB;oBACMR,OAAOiC,IAAAA,8BAAgB,EAACzB;oBAC9B;;wBAAO;4BAAET,QAAQ;gCAAEiB,UAAAA;gCAAUR,QAAAA;4BAAO;4BAAGR,MAAAA;wBAAK;;;oBACrCuB;oBACP;;wBAAMT,MAAM;4BAAEE,UAAAA;4BAAUR,QAAAA;wBAAO;;;oBAA/B;oBACA,MAAMe;;;;;;;IAEV;;AAGO,IAAM3B,oBAA+C;IAC1DsC,UAAUxC;IACVyC,eAAexC;IACfyC,kBAAkBC,0BAAa;IAC/BnB,SAAAA;IACAJ,OAAAA;IACA,8FAA8F;IAC9F,iGAAiG;IACjG,qFAAqF;IACrFwB,UAAU,SAAVA,SAAWf;eAAQ,2DAA2DgB,IAAI,CAAChB,IAAII,OAAO;;IAC9F7B,cAAAA;IACA0C,gBAAgBC,qBAAU;IAC1BC,oBAAoBC,yBAAc;IAClCC,aAAa,SAAbA,YAAc7C,QAAQC;eAAS4C,IAAAA,oBAAW,EAAC7C,OAAOiB,QAAQ,EAAEjB,OAAOS,MAAM,EAAER;;AAC7E;AAEO,SAAeH,WAAWI,GAAmB,EAAEoB,OAA6B;;;YACjF;;gBAAOwB,IAAAA,uBAAe,EAAC5C,KAAKL,mBAAmByB;;;IACjD"}
|
|
@@ -189,8 +189,8 @@ function reconcileContent(conn, touched, docs, _delta) {
|
|
|
189
189
|
];
|
|
190
190
|
case 3:
|
|
191
191
|
_state.sent();
|
|
192
|
-
//
|
|
193
|
-
//
|
|
192
|
+
// Build prepares the FTS index after reconcile. This persisted mark commits with the content
|
|
193
|
+
// change, so an interrupted build cannot expose stale lexical results.
|
|
194
194
|
return [
|
|
195
195
|
4,
|
|
196
196
|
(0, _lexicalts.markContentStale)(conn)
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/duckdb/reconcile.ts"],"sourcesContent":["import { SenseError } from '../../errors.ts';\nimport type { ReconcileDelta } from '../../features/types.ts';\nimport type { ParsedDoc } from '../../scan/index.ts';\nimport { appendRows, quoteIdent } from '../shared.ts';\nimport type { Connection, ReconcileDialect } from '../types.ts';\nimport { markContentStale } from './lexical.ts';\n\n// This store's dialect (types.ts's ReconcileDialect) for the shared orchestration in\n// store/reconcile.ts. `content` is a plain table, not FTS-virtual, so rows are maintained here\n// unconditionally; DuckDB's ALTER TABLE needs a declared type, so every dynamic frontmatter\n// column is VARIANT (holds mapValue()'s mixed JS types for one key across files).\n\n// No rowid coupling needed (unlike sqlite's content, which links to frontmatter's rowid):\n// `path` is content's own primary key, so this is a plain per-doc row.\nconst CONTENT_COLUMNS = ['path', 'title', 'summary', 'text'];\nconst INSERT_CONTENT_SQL = `INSERT INTO content (${CONTENT_COLUMNS.map(quoteIdent).join(', ')}) VALUES (?, ?, ?, ?)`;\n\nfunction contentRow(doc: ParsedDoc): unknown[] {\n return [doc.relPath, doc.search.title, doc.search.summary, doc.search.text];\n}\n\n// No compile-time column cap in DuckDB (unlike SQLite's SQLITE_MAX_COLUMN); kept as a sanity fence anyway\n// so a runaway frontmatter generator fails with a clear message instead of an unbounded ALTER TABLE loop.\nconst MAX_FRONTMATTER_COLUMNS = 10_000;\n\nasync function reconcileContent(conn: Connection, touched: string[], docs: ParsedDoc[], _delta: ReconcileDelta): Promise<void> {\n if (touched.length > 0)\n await conn.runBatch(\n 'DELETE FROM content WHERE \"path\" = ?',\n touched.map((p) => [p])\n );\n // The delete above cleared every touched path, and duckdb holds the cache file for its whole\n // connection, so no second writer can have landed one of these rows: nothing can conflict.\n await appendRows(conn, 'content', CONTENT_COLUMNS, INSERT_CONTENT_SQL, docs.map(contentRow));\n //
|
|
1
|
+
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/duckdb/reconcile.ts"],"sourcesContent":["import { SenseError } from '../../errors.ts';\nimport type { ReconcileDelta } from '../../features/types.ts';\nimport type { ParsedDoc } from '../../scan/index.ts';\nimport { appendRows, quoteIdent } from '../shared.ts';\nimport type { Connection, ReconcileDialect } from '../types.ts';\nimport { markContentStale } from './lexical.ts';\n\n// This store's dialect (types.ts's ReconcileDialect) for the shared orchestration in\n// store/reconcile.ts. `content` is a plain table, not FTS-virtual, so rows are maintained here\n// unconditionally; DuckDB's ALTER TABLE needs a declared type, so every dynamic frontmatter\n// column is VARIANT (holds mapValue()'s mixed JS types for one key across files).\n\n// No rowid coupling needed (unlike sqlite's content, which links to frontmatter's rowid):\n// `path` is content's own primary key, so this is a plain per-doc row.\nconst CONTENT_COLUMNS = ['path', 'title', 'summary', 'text'];\nconst INSERT_CONTENT_SQL = `INSERT INTO content (${CONTENT_COLUMNS.map(quoteIdent).join(', ')}) VALUES (?, ?, ?, ?)`;\n\nfunction contentRow(doc: ParsedDoc): unknown[] {\n return [doc.relPath, doc.search.title, doc.search.summary, doc.search.text];\n}\n\n// No compile-time column cap in DuckDB (unlike SQLite's SQLITE_MAX_COLUMN); kept as a sanity fence anyway\n// so a runaway frontmatter generator fails with a clear message instead of an unbounded ALTER TABLE loop.\nconst MAX_FRONTMATTER_COLUMNS = 10_000;\n\nasync function reconcileContent(conn: Connection, touched: string[], docs: ParsedDoc[], _delta: ReconcileDelta): Promise<void> {\n if (touched.length > 0)\n await conn.runBatch(\n 'DELETE FROM content WHERE \"path\" = ?',\n touched.map((p) => [p])\n );\n // The delete above cleared every touched path, and duckdb holds the cache file for its whole\n // connection, so no second writer can have landed one of these rows: nothing can conflict.\n await appendRows(conn, 'content', CONTENT_COLUMNS, INSERT_CONTENT_SQL, docs.map(contentRow));\n // Build prepares the FTS index after reconcile. This persisted mark commits with the content\n // change, so an interrupted build cannot expose stale lexical results.\n await markContentStale(conn);\n}\n\n// DuckDB rejects more than one ALTER command per statement (\"Parser Error: Only one ALTER\n// command per statement is supported\", measured), so every name's clause joins into one string\n// and runs as a single exec() -- one column-add \"leg\" through the driver instead of `names.length`.\n// VARIANT is the only type that can hold the mixed bigint/number/string/null shapes mapValue()\n// produces for one key across files.\nasync function addColumns(conn: Connection, names: string[]): Promise<void> {\n if (names.length === 0) return;\n await conn.exec(names.map((name) => `ALTER TABLE frontmatter ADD COLUMN ${quoteIdent(name)} VARIANT`).join('; '));\n}\n\nexport const duckdbDialect: ReconcileDialect = {\n beginMode: () => 'BEGIN',\n checkColumnLimit(count) {\n if (count > MAX_FRONTMATTER_COLUMNS) {\n throw new SenseError('COLUMN_LIMIT', `frontmatter would need ${count} columns, crossing this store's sanity limit (${MAX_FRONTMATTER_COLUMNS}). Narrow the presets' include globs so fewer/other files are indexed, or fix whatever is generating unbounded frontmatter keys.`);\n }\n },\n addColumns,\n reconcileContent,\n};\n"],"names":["duckdbDialect","CONTENT_COLUMNS","INSERT_CONTENT_SQL","map","quoteIdent","join","contentRow","doc","relPath","search","title","summary","text","MAX_FRONTMATTER_COLUMNS","reconcileContent","conn","touched","docs","_delta","length","runBatch","p","appendRows","markContentStale","addColumns","names","exec","name","beginMode","checkColumnLimit","count","SenseError"],"mappings":";;;;+BAiDaA;;;eAAAA;;;wBAjDc;wBAGY;yBAEN;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAEjC,qFAAqF;AACrF,+FAA+F;AAC/F,4FAA4F;AAC5F,kFAAkF;AAElF,0FAA0F;AAC1F,uEAAuE;AACvE,IAAMC,kBAAkB;IAAC;IAAQ;IAAS;IAAW;CAAO;AAC5D,IAAMC,qBAAqB,AAAC,wBAAkE,OAA3CD,gBAAgBE,GAAG,CAACC,oBAAU,EAAEC,IAAI,CAAC,OAAM;AAE9F,SAASC,WAAWC,GAAc;IAChC,OAAO;QAACA,IAAIC,OAAO;QAAED,IAAIE,MAAM,CAACC,KAAK;QAAEH,IAAIE,MAAM,CAACE,OAAO;QAAEJ,IAAIE,MAAM,CAACG,IAAI;KAAC;AAC7E;AAEA,0GAA0G;AAC1G,0GAA0G;AAC1G,IAAMC,0BAA0B;AAEhC,SAAeC,iBAAiBC,IAAgB,EAAEC,OAAiB,EAAEC,IAAiB,EAAEC,MAAsB;;;;;yBACxGF,CAAAA,QAAQG,MAAM,GAAG,CAAA,GAAjBH;;;;oBACF;;wBAAMD,KAAKK,QAAQ,CACjB,wCACAJ,QAAQb,GAAG,CAAC,SAACkB;mCAAM;gCAACA;6BAAE;;;;oBAFxB;;;oBAIF,6FAA6F;oBAC7F,2FAA2F;oBAC3F;;wBAAMC,IAAAA,oBAAU,EAACP,MAAM,WAAWd,iBAAiBC,oBAAoBe,KAAKd,GAAG,CAACG;;;oBAAhF;oBACA,6FAA6F;oBAC7F,uEAAuE;oBACvE;;wBAAMiB,IAAAA,2BAAgB,EAACR;;;oBAAvB;;;;;;IACF;;AAEA,0FAA0F;AAC1F,+FAA+F;AAC/F,oGAAoG;AACpG,+FAA+F;AAC/F,qCAAqC;AACrC,SAAeS,WAAWT,IAAgB,EAAEU,KAAe;;;;;oBACzD,IAAIA,MAAMN,MAAM,KAAK,GAAG;;;oBACxB;;wBAAMJ,KAAKW,IAAI,CAACD,MAAMtB,GAAG,CAAC,SAACwB;mCAAS,AAAC,sCAAsD,OAAjBvB,IAAAA,oBAAU,EAACuB,OAAM;2BAAWtB,IAAI,CAAC;;;oBAA3G;;;;;;IACF;;AAEO,IAAML,gBAAkC;IAC7C4B,WAAW,SAAXA;eAAiB;;IACjBC,kBAAAA,SAAAA,iBAAiBC,KAAK;QACpB,IAAIA,QAAQjB,yBAAyB;YACnC,MAAM,IAAIkB,oBAAU,CAAC,gBAAgB,AAAC,0BAA+ElB,OAAtDiB,OAAM,kDAAwE,OAAxBjB,yBAAwB;QAC/I;IACF;IACAW,YAAAA;IACAV,kBAAAA;AACF"}
|
|
@@ -1,5 +1,6 @@
|
|
|
1
1
|
import type { DuckDBConnection } from '@duckdb/node-api';
|
|
2
2
|
import type { Connection, VectorCandidate, VectorSimilar, VectorWriteRow } from '../types.js';
|
|
3
|
+
export declare function createVectorWriteStage(duckdb: DuckDBConnection, dims: number): Promise<void>;
|
|
3
4
|
export declare function writeVectorBatch(duckdb: DuckDBConnection, conn: Connection, dims: number, rows: VectorWriteRow[]): Promise<void>;
|
|
4
5
|
export declare function scanCandidates(duckdb: DuckDBConnection, qv: Float32Array, dims: number, fetch: number, allowed?: Set<string>): Promise<VectorCandidate[]>;
|
|
5
6
|
export declare function scanSimilar(duckdb: DuckDBConnection, conn: Connection, dims: number, path: string, opts: {
|
|
@@ -1,5 +1,6 @@
|
|
|
1
1
|
import type { DuckDBConnection } from '@duckdb/node-api';
|
|
2
2
|
import type { Connection, VectorCandidate, VectorSimilar, VectorWriteRow } from '../types.js';
|
|
3
|
+
export declare function createVectorWriteStage(duckdb: DuckDBConnection, dims: number): Promise<void>;
|
|
3
4
|
export declare function writeVectorBatch(duckdb: DuckDBConnection, conn: Connection, dims: number, rows: VectorWriteRow[]): Promise<void>;
|
|
4
5
|
export declare function scanCandidates(duckdb: DuckDBConnection, qv: Float32Array, dims: number, fetch: number, allowed?: Set<string>): Promise<VectorCandidate[]>;
|
|
5
6
|
export declare function scanSimilar(duckdb: DuckDBConnection, conn: Connection, dims: number, path: string, opts: {
|