sensemaking 0.19.1 → 0.20.0
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/dist/cjs/chunk/extract.d.cts +1 -0
- package/dist/cjs/chunk/extract.d.ts +1 -0
- package/dist/cjs/chunk/extract.js +23 -5
- package/dist/cjs/chunk/extract.js.map +1 -1
- package/dist/cjs/chunk/index.d.cts +1 -1
- package/dist/cjs/chunk/index.d.ts +1 -1
- package/dist/cjs/chunk/index.js +3 -0
- package/dist/cjs/chunk/index.js.map +1 -1
- package/dist/cjs/index.d.cts +1 -1
- package/dist/cjs/index.d.ts +1 -1
- package/dist/cjs/index.js +3 -0
- package/dist/cjs/index.js.map +1 -1
- package/dist/cjs/scan/pool.d.cts +15 -0
- package/dist/cjs/scan/pool.d.ts +15 -0
- package/dist/cjs/scan/pool.js +263 -0
- package/dist/cjs/scan/pool.js.map +1 -0
- package/dist/cjs/scan/reparse.d.cts +2 -0
- package/dist/cjs/scan/reparse.d.ts +2 -0
- package/dist/cjs/scan/reparse.js +15 -76
- package/dist/cjs/scan/reparse.js.map +1 -1
- package/dist/cjs/store/builder.d.cts +11 -0
- package/dist/cjs/store/builder.d.ts +11 -0
- package/dist/cjs/store/builder.js +27 -0
- package/dist/cjs/store/builder.js.map +1 -0
- package/dist/cjs/store/duckdb/fieldStats.js +87 -10
- package/dist/cjs/store/duckdb/fieldStats.js.map +1 -1
- package/dist/cjs/store/duckdb/lexical.d.cts +1 -0
- package/dist/cjs/store/duckdb/lexical.d.ts +1 -0
- package/dist/cjs/store/duckdb/lexical.js +21 -5
- package/dist/cjs/store/duckdb/lexical.js.map +1 -1
- package/dist/cjs/store/duckdb/open.d.cts +9 -8
- package/dist/cjs/store/duckdb/open.d.ts +9 -8
- package/dist/cjs/store/duckdb/open.js +69 -120
- package/dist/cjs/store/duckdb/open.js.map +1 -1
- package/dist/cjs/store/duckdb/reconcile.d.cts +2 -6
- package/dist/cjs/store/duckdb/reconcile.d.ts +2 -6
- package/dist/cjs/store/duckdb/reconcile.js +45 -581
- package/dist/cjs/store/duckdb/reconcile.js.map +1 -1
- package/dist/cjs/store/duckdb/store.d.cts +1 -2
- package/dist/cjs/store/duckdb/store.d.ts +1 -2
- package/dist/cjs/store/duckdb/store.js +3 -26
- package/dist/cjs/store/duckdb/store.js.map +1 -1
- package/dist/cjs/store/index.d.cts +9 -3
- package/dist/cjs/store/index.d.ts +9 -3
- package/dist/cjs/store/index.js +40 -7
- package/dist/cjs/store/index.js.map +1 -1
- package/dist/cjs/store/open.d.cts +15 -0
- package/dist/cjs/store/open.d.ts +15 -0
- package/dist/cjs/store/open.js +516 -0
- package/dist/cjs/store/open.js.map +1 -0
- package/dist/cjs/store/reconcile.d.cts +7 -0
- package/dist/cjs/store/reconcile.d.ts +7 -0
- package/dist/cjs/store/reconcile.js +752 -0
- package/dist/cjs/store/reconcile.js.map +1 -0
- package/dist/cjs/store/shared.d.cts +1 -0
- package/dist/cjs/store/shared.d.ts +1 -0
- package/dist/cjs/store/shared.js +37 -0
- package/dist/cjs/store/shared.js.map +1 -1
- package/dist/cjs/store/signature.d.cts +3 -0
- package/dist/cjs/store/signature.d.ts +3 -0
- package/dist/cjs/store/signature.js +162 -0
- package/dist/cjs/store/signature.js.map +1 -0
- package/dist/cjs/store/sqlite/open.d.cts +9 -9
- package/dist/cjs/store/sqlite/open.d.ts +9 -9
- package/dist/cjs/store/sqlite/open.js +188 -198
- package/dist/cjs/store/sqlite/open.js.map +1 -1
- package/dist/cjs/store/sqlite/reconcile.d.cts +5 -10
- package/dist/cjs/store/sqlite/reconcile.d.ts +5 -10
- package/dist/cjs/store/sqlite/reconcile.js +48 -850
- package/dist/cjs/store/sqlite/reconcile.js.map +1 -1
- package/dist/cjs/store/sqlite/store.d.cts +1 -1
- package/dist/cjs/store/sqlite/store.d.ts +1 -1
- package/dist/cjs/store/sqlite/store.js +3 -13
- package/dist/cjs/store/sqlite/store.js.map +1 -1
- package/dist/cjs/store/turso/open.d.cts +6 -11
- package/dist/cjs/store/turso/open.d.ts +6 -11
- package/dist/cjs/store/turso/open.js +78 -169
- package/dist/cjs/store/turso/open.js.map +1 -1
- package/dist/cjs/store/turso/reconcile.d.cts +5 -6
- package/dist/cjs/store/turso/reconcile.d.ts +5 -6
- package/dist/cjs/store/turso/reconcile.js +169 -729
- package/dist/cjs/store/turso/reconcile.js.map +1 -1
- package/dist/cjs/store/turso/store.d.cts +1 -2
- package/dist/cjs/store/turso/store.d.ts +1 -2
- package/dist/cjs/store/turso/store.js +1 -12
- package/dist/cjs/store/turso/store.js.map +1 -1
- package/dist/cjs/store/types.d.cts +28 -6
- package/dist/cjs/store/types.d.ts +28 -6
- package/dist/cjs/store/types.js +14 -0
- package/dist/cjs/store/types.js.map +1 -1
- package/dist/cjs/text/strip.js +3 -6
- package/dist/cjs/text/strip.js.map +1 -1
- package/dist/cjs/watch.js +28 -16
- package/dist/cjs/watch.js.map +1 -1
- package/dist/esm/chunk/extract.d.ts +1 -0
- package/dist/esm/chunk/extract.js +10 -2
- package/dist/esm/chunk/extract.js.map +1 -1
- package/dist/esm/chunk/index.d.ts +1 -1
- package/dist/esm/chunk/index.js +1 -1
- package/dist/esm/chunk/index.js.map +1 -1
- package/dist/esm/index.d.ts +1 -1
- package/dist/esm/index.js +1 -1
- package/dist/esm/index.js.map +1 -1
- package/dist/esm/scan/pool.d.ts +15 -0
- package/dist/esm/scan/pool.js +77 -0
- package/dist/esm/scan/pool.js.map +1 -0
- package/dist/esm/scan/reparse.d.ts +2 -0
- package/dist/esm/scan/reparse.js +11 -58
- package/dist/esm/scan/reparse.js.map +1 -1
- package/dist/esm/store/builder.d.ts +11 -0
- package/dist/esm/store/builder.js +14 -0
- package/dist/esm/store/builder.js.map +1 -0
- package/dist/esm/store/duckdb/fieldStats.js +32 -19
- package/dist/esm/store/duckdb/fieldStats.js.map +1 -1
- package/dist/esm/store/duckdb/lexical.d.ts +1 -0
- package/dist/esm/store/duckdb/lexical.js +22 -9
- package/dist/esm/store/duckdb/lexical.js.map +1 -1
- package/dist/esm/store/duckdb/open.d.ts +9 -8
- package/dist/esm/store/duckdb/open.js +32 -57
- package/dist/esm/store/duckdb/open.js.map +1 -1
- package/dist/esm/store/duckdb/reconcile.d.ts +2 -6
- package/dist/esm/store/duckdb/reconcile.js +24 -117
- package/dist/esm/store/duckdb/reconcile.js.map +1 -1
- package/dist/esm/store/duckdb/store.d.ts +1 -2
- package/dist/esm/store/duckdb/store.js +4 -11
- package/dist/esm/store/duckdb/store.js.map +1 -1
- package/dist/esm/store/index.d.ts +9 -3
- package/dist/esm/store/index.js +23 -7
- package/dist/esm/store/index.js.map +1 -1
- package/dist/esm/store/open.d.ts +15 -0
- package/dist/esm/store/open.js +137 -0
- package/dist/esm/store/open.js.map +1 -0
- package/dist/esm/store/reconcile.d.ts +7 -0
- package/dist/esm/store/reconcile.js +132 -0
- package/dist/esm/store/reconcile.js.map +1 -0
- package/dist/esm/store/shared.d.ts +1 -0
- package/dist/esm/store/shared.js +9 -0
- package/dist/esm/store/shared.js.map +1 -1
- package/dist/esm/store/signature.d.ts +3 -0
- package/dist/esm/store/signature.js +35 -0
- package/dist/esm/store/signature.js.map +1 -0
- package/dist/esm/store/sqlite/open.d.ts +9 -9
- package/dist/esm/store/sqlite/open.js +81 -111
- package/dist/esm/store/sqlite/open.js.map +1 -1
- package/dist/esm/store/sqlite/reconcile.d.ts +5 -10
- package/dist/esm/store/sqlite/reconcile.js +31 -201
- package/dist/esm/store/sqlite/reconcile.js.map +1 -1
- package/dist/esm/store/sqlite/store.d.ts +1 -1
- package/dist/esm/store/sqlite/store.js +4 -7
- package/dist/esm/store/sqlite/store.js.map +1 -1
- package/dist/esm/store/turso/open.d.ts +6 -11
- package/dist/esm/store/turso/open.js +30 -77
- package/dist/esm/store/turso/open.js.map +1 -1
- package/dist/esm/store/turso/reconcile.d.ts +5 -6
- package/dist/esm/store/turso/reconcile.js +42 -146
- package/dist/esm/store/turso/reconcile.js.map +1 -1
- package/dist/esm/store/turso/store.d.ts +1 -2
- package/dist/esm/store/turso/store.js +2 -6
- package/dist/esm/store/turso/store.js.map +1 -1
- package/dist/esm/store/types.d.ts +28 -6
- package/dist/esm/store/types.js +12 -1
- package/dist/esm/store/types.js.map +1 -1
- package/dist/esm/text/strip.js +3 -4
- package/dist/esm/text/strip.js.map +1 -1
- package/dist/esm/watch.js +5 -3
- package/dist/esm/watch.js.map +1 -1
- package/package.json +2 -2
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/sqlite/reconcile.ts"],"sourcesContent":["import type { Config } from '../../config/index.ts';\nimport { contentTokenize } from '../../config/index.ts';\nimport { SenseError } from '../../errors.ts';\nimport { activeFeatures } from '../../features/index.ts';\nimport type { ExtractedDoc, ReconcileDelta } from '../../features/types.ts';\nimport { progress } from '../../output/progress.ts';\nimport type { ParsedDoc } from '../../scan/index.ts';\nimport { listFiles, parseFile, RESERVED_COLUMNS } from '../../scan/index.ts';\nimport { reparseFiles } from '../../scan/reparse.ts';\nimport { segmentField } from '../../text/segment.ts';\nimport { getColumns, getMeta, quoteIdent, setMeta } from '../shared.ts';\nimport { BEGIN_WRITE, withTransaction } from '../transaction.ts';\nimport type { Connection } from '../types.ts';\n\n// Feature-owned columns (`_rank`) must stay out of the upsert: a reparse would null the last\n// computed value on every touch, not just the reconciles that recompute it.\nconst CORE_FRONTMATTER_COLUMNS = new Set(['path', '_mtime', '_ctime', '_size', '_parse_error']);\n\n// SQLite's compile-time SQLITE_MAX_COLUMN, default 2000 (https://www.sqlite.org/limits.html).\nconst MAX_FRONTMATTER_COLUMNS = 2000;\n\n// Shared by reconcile() and the tokenize-only rebuild in open(), so both prepare the same\n// literal instead of two copies drifting apart.\nconst INSERT_CONTENT_SQL = `INSERT INTO content (rowid, title, summary, text, \"path\", title_seg, summary_seg, text_seg) VALUES ((SELECT rowid FROM frontmatter WHERE \"path\" = ?), ?, ?, ?, ?, ?, ?, ?)`;\n\n// A single content row's param tuple, matching INSERT_CONTENT_SQL's placeholder order.\n// Assumes the frontmatter row for doc.relPath already exists (rowid subquery).\nfunction contentRow(doc: ParsedDoc, segmenting: boolean): unknown[] {\n return [doc.relPath, doc.search.title, doc.search.summary, doc.search.text, doc.relPath, segmenting ? segmentField(doc.search.title) : '', segmenting ? segmentField(doc.search.summary) : '', segmenting ? segmentField(doc.search.text) : ''];\n}\n\nexport async function reconcile(conn: Connection, cfg: Config, baseDir: string): Promise<{ parsed: number; warnings: string[] }> {\n const files = listFiles(cfg, baseDir);\n const currentSet = new Set(files.map((f) => f.relPath));\n\n const existingStmt = await conn.prepare('SELECT \"path\", \"_mtime\", \"_size\" FROM frontmatter');\n const existingRows = (await existingStmt.all()) as Array<{ path: string; _mtime: number; _size: number }>;\n const existing = new Map(existingRows.map((r) => [r.path, r]));\n const vanished = existingRows.filter((r) => !currentSet.has(r.path)).map((r) => r.path);\n\n const toReparse = files.filter((f) => {\n const row = existing.get(f.relPath);\n return !row || row._mtime !== f.mtimeMs || row._size !== f.size;\n });\n\n if (vanished.length === 0 && toReparse.length === 0) return { parsed: 0, warnings: [] };\n\n const features = activeFeatures(cfg);\n const seenColumns = await getColumns(conn);\n\n // Bulk reparses (a sync, a cold build) are the long silences a query can hit; short\n // reconciles stay silent (progress() has a threshold).\n const report = progress('reparsing files', toReparse.length);\n const { docs: parsedDocs, warnings, newColumns } = await reparseFiles(toReparse, features, cfg, seenColumns, report.tick);\n report.finish();\n for (const col of newColumns) seenColumns.add(col);\n\n const allColumns = [...seenColumns];\n // Fence before ALTERing: SQLite's own failure past this point is a raw\n // \"too many columns on sqlite_altertab_frontmatter\" with no indication of the boundary or the levers.\n if (allColumns.length > MAX_FRONTMATTER_COLUMNS) {\n throw new SenseError(\n 'COLUMN_LIMIT',\n `frontmatter would need ${allColumns.length} columns, crossing SQLite's compile-time SQLITE_MAX_COLUMN limit (default ${MAX_FRONTMATTER_COLUMNS}; see https://www.sqlite.org/limits.html). Narrow the presets' include globs so fewer/other files are indexed, or fix whatever is generating unbounded frontmatter keys.`\n );\n }\n // Columns the frontmatter upsert actually writes: core + parsed frontmatter keys, never a\n // feature-owned reserved column (see CORE_FRONTMATTER_COLUMNS above).\n const writableColumns = allColumns.filter((c) => CORE_FRONTMATTER_COLUMNS.has(c) || !RESERVED_COLUMNS.has(c));\n // ON CONFLICT UPDATE (not OR REPLACE) keeps the row's rowid stable across reparses --\n // content rows are coupled to that rowid below.\n const insertSql = `INSERT INTO frontmatter (${writableColumns.map(quoteIdent).join(', ')}) VALUES (${writableColumns.map(() => '?').join(', ')}) ON CONFLICT(\"path\") DO UPDATE SET ${writableColumns\n .filter((c) => c !== 'path')\n .map((c) => `${quoteIdent(c)} = excluded.${quoteIdent(c)}`)\n .join(', ')}`;\n\n const added = toReparse.filter((f) => !existing.has(f.relPath)).map((f) => f.relPath);\n const delta: ReconcileDelta = { files, reparsed: parsedDocs.map((d) => d.relPath), added, vanished };\n const addedSet = new Set(added);\n const reparsedExisting = parsedDocs.map((d) => d.relPath).filter((p) => !addedSet.has(p));\n\n const txStart = Date.now();\n await withTransaction(\n conn,\n async () => {\n // Re-read inside the write transaction: newColumns came from a read taken before it opened, so a\n // concurrent reconcile may have added some of them since. ALTER has no IF NOT EXISTS.\n const present = await getColumns(conn);\n for (const col of newColumns) if (!present.has(col)) await conn.exec(`ALTER TABLE frontmatter ADD COLUMN ${quoteIdent(col)}`);\n\n // content's rowid lookup depends on its coupled frontmatter row, so the content\n // deletes/inserts run while every path still has one -- the vanished frontmatter delete comes last.\n if (vanished.length > 0) {\n await conn.runBatch(\n 'DELETE FROM content WHERE rowid = (SELECT rowid FROM frontmatter WHERE \"path\" = ?)',\n vanished.map((p) => [p])\n );\n }\n if (reparsedExisting.length > 0) {\n // FTS5 has no upsert, so delete-before-insert, coupled to the frontmatter rowid (indexed via\n // its PRIMARY KEY) rather than the UNINDEXED `path` column, which a per-row DELETE would scan to find.\n await conn.runBatch(\n 'DELETE FROM content WHERE rowid = (SELECT rowid FROM frontmatter WHERE \"path\" = ?)',\n reparsedExisting.map((p) => [p])\n );\n }\n\n if (parsedDocs.length > 0) {\n const rows = parsedDocs.map((doc) =>\n writableColumns.map((col) => {\n if (col === 'path') return doc.relPath;\n if (col === '_mtime') return doc.mtimeMs;\n if (col === '_ctime') return doc.ctimeMs;\n if (col === '_size') return doc.size;\n // Written per parse, unlike _rank, which a feature pass owns and the upsert skips.\n if (col === '_parse_error') return doc.parseError;\n return doc.data[col] ?? null;\n })\n );\n await conn.runBatch(insertSql, rows);\n\n // A non-default tokenizer means the tree has chosen its own scheme; a phrase query over\n // grapheme runs would be nonsense against trigram, so the sidecars stay empty.\n const segmenting = contentTokenize(cfg) === undefined;\n await conn.runBatch(\n INSERT_CONTENT_SQL,\n parsedDocs.map((doc) => contentRow(doc, segmenting))\n );\n }\n\n if (vanished.length > 0)\n await conn.runBatch(\n 'DELETE FROM frontmatter WHERE \"path\" = ?',\n vanished.map((p) => [p])\n );\n\n // A preset edit forces a full rebuild, so an unchanged doc's coverage is already correct;\n // new docs have nothing to clear, which keeps cold builds linear.\n const presetTouched = [...vanished, ...reparsedExisting];\n if (presetTouched.length > 0)\n await conn.runBatch(\n 'DELETE FROM preset_files WHERE \"path\" = ?',\n presetTouched.map((p) => [p])\n );\n const presetRows: unknown[][] = [];\n for (const doc of parsedDocs) for (const presetName of doc.presets) presetRows.push([doc.relPath, presetName]);\n if (presetRows.length > 0) await conn.runBatch('INSERT INTO preset_files (\"path\", preset) VALUES (?, ?)', presetRows);\n\n const removedPaths = [...vanished, ...reparsedExisting];\n if (removedPaths.length > 0) for (const feature of features) await feature.remove?.(conn, removedPaths, delta);\n for (const feature of features) {\n const docsForFeature: ExtractedDoc[] = parsedDocs.map((doc) => ({ path: doc.relPath, extracted: doc.extracted[feature.name] }));\n await feature.store?.(conn, docsForFeature, delta);\n }\n for (const feature of features) await feature.afterReconcile?.(conn, delta);\n },\n BEGIN_WRITE\n );\n\n // Reconcile's own write-transaction duration, for open()'s derived busy_timeout (F):\n // keep the observed max so a big watcher reconcile's lock hold is what the next open bounds its wait against.\n const durationMs = Date.now() - txStart;\n const prevRaw = await getMeta(conn, 'reconcile_max_ms');\n // -1, not 0, so a genuinely 0ms first reconcile (sub-millisecond, common on a tiny tree)\n // still gets recorded instead of losing to the \"nothing recorded yet\" default.\n const prevMax = prevRaw === null ? -1 : Number(prevRaw);\n if (durationMs > prevMax) await setMeta(conn, 'reconcile_max_ms', String(durationMs));\n\n return { parsed: parsedDocs.length, warnings };\n}\n\n// Segment keys that moved between two feature signatures (see config.featureSignature's\n// format: global features, embed provider, tokenize, then one segment per preset).\nexport function changedSignatureKeys(before: string, after: string): Set<string> {\n const keyOf = (part: string) => (part.startsWith('preset:') ? part.split(':').slice(0, 2).join(':') : part.split(':')[0]);\n const parse = (sig: string) => new Map(sig.split('|').map((part) => [keyOf(part), part]));\n const a = parse(before);\n const b = parse(after);\n const changed = new Set<string>();\n for (const [key, val] of b) if (a.get(key) !== val) changed.add(key);\n for (const key of a.keys()) if (!b.has(key)) changed.add(key);\n return changed;\n}\n\n// Whether the embed segment only gained its resolved weight identity: same provider and\n// model, no identity recorded before, one now -- adopted into meta without a rebuild.\nexport function embedIdentityAdopted(before: string, after: string): boolean {\n const embedPart = (sig: string) => sig.split('|').find((p) => p.startsWith('embed:'));\n const b = embedPart(before);\n const a = embedPart(after);\n if (b === undefined || a === undefined) return false;\n const at = a.indexOf('@');\n return b.indexOf('@') === -1 && at !== -1 && a.slice(0, at) === b;\n}\n\n// Names what moved, for the rebuild notice.\nexport function signatureDiff(before: string, after: string): string {\n const changed = changedSignatureKeys(before, after);\n const label = (key: string) => (key === 'embed' ? 'embed settings' : key === 'tokenize' ? 'content tokenizer' : key.startsWith('preset:') ? `preset \"${key.slice(7)}\"` : 'features');\n return changed.size === 0 ? 'features' : [...changed].map(label).join(', ');\n}\n\n// Tokenize-only rebuild: only `content` is repopulated (the other tables are tokenizer-independent)\n// and no feature extractors run. Returns per-file warnings, since mtimes stay untouched and reconcile would never reparse them.\nexport async function rebuildContentTable(conn: Connection, cfg: Config, baseDir: string): Promise<string[]> {\n const stmt = await conn.prepare('SELECT \"path\" FROM frontmatter');\n const known = new Set(((await stmt.all()) as Array<{ path: string }>).map((r) => r.path));\n const files = listFiles(cfg, baseDir).filter((f) => known.has(f.relPath));\n const segmenting = contentTokenize(cfg) === undefined;\n const warnings: string[] = [];\n const rows: unknown[][] = [];\n for (const file of files) {\n const { doc, warnings: fileWarnings } = parseFile(file);\n warnings.push(...fileWarnings);\n rows.push(contentRow(doc, segmenting));\n }\n await withTransaction(\n conn,\n async () => {\n if (rows.length > 0) await conn.runBatch(INSERT_CONTENT_SQL, rows);\n },\n BEGIN_WRITE\n );\n return warnings;\n}\n"],"names":["contentTokenize","SenseError","activeFeatures","progress","listFiles","parseFile","RESERVED_COLUMNS","reparseFiles","segmentField","getColumns","getMeta","quoteIdent","setMeta","BEGIN_WRITE","withTransaction","CORE_FRONTMATTER_COLUMNS","Set","MAX_FRONTMATTER_COLUMNS","INSERT_CONTENT_SQL","contentRow","doc","segmenting","relPath","search","title","summary","text","reconcile","conn","cfg","baseDir","files","currentSet","map","f","existingStmt","prepare","existingRows","all","existing","Map","r","path","vanished","filter","has","toReparse","row","get","_mtime","mtimeMs","_size","size","length","parsed","warnings","features","seenColumns","report","docs","parsedDocs","newColumns","tick","finish","col","add","allColumns","writableColumns","c","insertSql","join","added","delta","reparsed","d","addedSet","reparsedExisting","p","txStart","Date","now","feature","present","exec","runBatch","rows","ctimeMs","parseError","data","undefined","presetTouched","presetRows","presetName","presets","push","removedPaths","remove","docsForFeature","extracted","name","store","afterReconcile","durationMs","prevRaw","prevMax","Number","String","changedSignatureKeys","before","after","keyOf","part","startsWith","split","slice","parse","sig","a","b","changed","key","val","keys","embedIdentityAdopted","embedPart","find","at","indexOf","signatureDiff","label","rebuildContentTable","stmt","known","file","fileWarnings"],"mappings":"AACA,SAASA,eAAe,QAAQ,wBAAwB;AACxD,SAASC,UAAU,QAAQ,kBAAkB;AAC7C,SAASC,cAAc,QAAQ,0BAA0B;AAEzD,SAASC,QAAQ,QAAQ,2BAA2B;AAEpD,SAASC,SAAS,EAAEC,SAAS,EAAEC,gBAAgB,QAAQ,sBAAsB;AAC7E,SAASC,YAAY,QAAQ,wBAAwB;AACrD,SAASC,YAAY,QAAQ,wBAAwB;AACrD,SAASC,UAAU,EAAEC,OAAO,EAAEC,UAAU,EAAEC,OAAO,QAAQ,eAAe;AACxE,SAASC,WAAW,EAAEC,eAAe,QAAQ,oBAAoB;AAGjE,6FAA6F;AAC7F,4EAA4E;AAC5E,MAAMC,2BAA2B,IAAIC,IAAI;IAAC;IAAQ;IAAU;IAAU;IAAS;CAAe;AAE9F,8FAA8F;AAC9F,MAAMC,0BAA0B;AAEhC,0FAA0F;AAC1F,gDAAgD;AAChD,MAAMC,qBAAqB,CAAC,0KAA0K,CAAC;AAEvM,uFAAuF;AACvF,+EAA+E;AAC/E,SAASC,WAAWC,GAAc,EAAEC,UAAmB;IACrD,OAAO;QAACD,IAAIE,OAAO;QAAEF,IAAIG,MAAM,CAACC,KAAK;QAAEJ,IAAIG,MAAM,CAACE,OAAO;QAAEL,IAAIG,MAAM,CAACG,IAAI;QAAEN,IAAIE,OAAO;QAAED,aAAab,aAAaY,IAAIG,MAAM,CAACC,KAAK,IAAI;QAAIH,aAAab,aAAaY,IAAIG,MAAM,CAACE,OAAO,IAAI;QAAIJ,aAAab,aAAaY,IAAIG,MAAM,CAACG,IAAI,IAAI;KAAG;AACjP;AAEA,OAAO,eAAeC,UAAUC,IAAgB,EAAEC,GAAW,EAAEC,OAAe;IAC5E,MAAMC,QAAQ3B,UAAUyB,KAAKC;IAC7B,MAAME,aAAa,IAAIhB,IAAIe,MAAME,GAAG,CAAC,CAACC,IAAMA,EAAEZ,OAAO;IAErD,MAAMa,eAAe,MAAMP,KAAKQ,OAAO,CAAC;IACxC,MAAMC,eAAgB,MAAMF,aAAaG,GAAG;IAC5C,MAAMC,WAAW,IAAIC,IAAIH,aAAaJ,GAAG,CAAC,CAACQ,IAAM;YAACA,EAAEC,IAAI;YAAED;SAAE;IAC5D,MAAME,WAAWN,aAAaO,MAAM,CAAC,CAACH,IAAM,CAACT,WAAWa,GAAG,CAACJ,EAAEC,IAAI,GAAGT,GAAG,CAAC,CAACQ,IAAMA,EAAEC,IAAI;IAEtF,MAAMI,YAAYf,MAAMa,MAAM,CAAC,CAACV;QAC9B,MAAMa,MAAMR,SAASS,GAAG,CAACd,EAAEZ,OAAO;QAClC,OAAO,CAACyB,OAAOA,IAAIE,MAAM,KAAKf,EAAEgB,OAAO,IAAIH,IAAII,KAAK,KAAKjB,EAAEkB,IAAI;IACjE;IAEA,IAAIT,SAASU,MAAM,KAAK,KAAKP,UAAUO,MAAM,KAAK,GAAG,OAAO;QAAEC,QAAQ;QAAGC,UAAU,EAAE;IAAC;IAEtF,MAAMC,WAAWtD,eAAe2B;IAChC,MAAM4B,cAAc,MAAMhD,WAAWmB;IAErC,oFAAoF;IACpF,uDAAuD;IACvD,MAAM8B,SAASvD,SAAS,mBAAmB2C,UAAUO,MAAM;IAC3D,MAAM,EAAEM,MAAMC,UAAU,EAAEL,QAAQ,EAAEM,UAAU,EAAE,GAAG,MAAMtD,aAAauC,WAAWU,UAAU3B,KAAK4B,aAAaC,OAAOI,IAAI;IACxHJ,OAAOK,MAAM;IACb,KAAK,MAAMC,OAAOH,WAAYJ,YAAYQ,GAAG,CAACD;IAE9C,MAAME,aAAa;WAAIT;KAAY;IACnC,uEAAuE;IACvE,sGAAsG;IACtG,IAAIS,WAAWb,MAAM,GAAGpC,yBAAyB;QAC/C,MAAM,IAAIhB,WACR,gBACA,CAAC,uBAAuB,EAAEiE,WAAWb,MAAM,CAAC,0EAA0E,EAAEpC,wBAAwB,wKAAwK,CAAC;IAE7T;IACA,0FAA0F;IAC1F,sEAAsE;IACtE,MAAMkD,kBAAkBD,WAAWtB,MAAM,CAAC,CAACwB,IAAMrD,yBAAyB8B,GAAG,CAACuB,MAAM,CAAC9D,iBAAiBuC,GAAG,CAACuB;IAC1G,sFAAsF;IACtF,gDAAgD;IAChD,MAAMC,YAAY,CAAC,yBAAyB,EAAEF,gBAAgBlC,GAAG,CAACtB,YAAY2D,IAAI,CAAC,MAAM,UAAU,EAAEH,gBAAgBlC,GAAG,CAAC,IAAM,KAAKqC,IAAI,CAAC,MAAM,oCAAoC,EAAEH,gBAClLvB,MAAM,CAAC,CAACwB,IAAMA,MAAM,QACpBnC,GAAG,CAAC,CAACmC,IAAM,GAAGzD,WAAWyD,GAAG,YAAY,EAAEzD,WAAWyD,IAAI,EACzDE,IAAI,CAAC,OAAO;IAEf,MAAMC,QAAQzB,UAAUF,MAAM,CAAC,CAACV,IAAM,CAACK,SAASM,GAAG,CAACX,EAAEZ,OAAO,GAAGW,GAAG,CAAC,CAACC,IAAMA,EAAEZ,OAAO;IACpF,MAAMkD,QAAwB;QAAEzC;QAAO0C,UAAUb,WAAW3B,GAAG,CAAC,CAACyC,IAAMA,EAAEpD,OAAO;QAAGiD;QAAO5B;IAAS;IACnG,MAAMgC,WAAW,IAAI3D,IAAIuD;IACzB,MAAMK,mBAAmBhB,WAAW3B,GAAG,CAAC,CAACyC,IAAMA,EAAEpD,OAAO,EAAEsB,MAAM,CAAC,CAACiC,IAAM,CAACF,SAAS9B,GAAG,CAACgC;IAEtF,MAAMC,UAAUC,KAAKC,GAAG;IACxB,MAAMlE,gBACJc,MACA;YAiEqEqD,iBAK7BA;QArEtC,iGAAiG;QACjG,sFAAsF;QACtF,MAAMC,UAAU,MAAMzE,WAAWmB;QACjC,KAAK,MAAMoC,OAAOH,WAAY,IAAI,CAACqB,QAAQrC,GAAG,CAACmB,MAAM,MAAMpC,KAAKuD,IAAI,CAAC,CAAC,mCAAmC,EAAExE,WAAWqD,MAAM;QAE5H,gFAAgF;QAChF,oGAAoG;QACpG,IAAIrB,SAASU,MAAM,GAAG,GAAG;YACvB,MAAMzB,KAAKwD,QAAQ,CACjB,sFACAzC,SAASV,GAAG,CAAC,CAAC4C,IAAM;oBAACA;iBAAE;QAE3B;QACA,IAAID,iBAAiBvB,MAAM,GAAG,GAAG;YAC/B,6FAA6F;YAC7F,uGAAuG;YACvG,MAAMzB,KAAKwD,QAAQ,CACjB,sFACAR,iBAAiB3C,GAAG,CAAC,CAAC4C,IAAM;oBAACA;iBAAE;QAEnC;QAEA,IAAIjB,WAAWP,MAAM,GAAG,GAAG;YACzB,MAAMgC,OAAOzB,WAAW3B,GAAG,CAAC,CAACb,MAC3B+C,gBAAgBlC,GAAG,CAAC,CAAC+B;wBAOZ5C;oBANP,IAAI4C,QAAQ,QAAQ,OAAO5C,IAAIE,OAAO;oBACtC,IAAI0C,QAAQ,UAAU,OAAO5C,IAAI8B,OAAO;oBACxC,IAAIc,QAAQ,UAAU,OAAO5C,IAAIkE,OAAO;oBACxC,IAAItB,QAAQ,SAAS,OAAO5C,IAAIgC,IAAI;oBACpC,mFAAmF;oBACnF,IAAIY,QAAQ,gBAAgB,OAAO5C,IAAImE,UAAU;oBACjD,QAAOnE,gBAAAA,IAAIoE,IAAI,CAACxB,IAAI,cAAb5C,2BAAAA,gBAAiB;gBAC1B;YAEF,MAAMQ,KAAKwD,QAAQ,CAACf,WAAWgB;YAE/B,wFAAwF;YACxF,+EAA+E;YAC/E,MAAMhE,aAAarB,gBAAgB6B,SAAS4D;YAC5C,MAAM7D,KAAKwD,QAAQ,CACjBlE,oBACA0C,WAAW3B,GAAG,CAAC,CAACb,MAAQD,WAAWC,KAAKC;QAE5C;QAEA,IAAIsB,SAASU,MAAM,GAAG,GACpB,MAAMzB,KAAKwD,QAAQ,CACjB,4CACAzC,SAASV,GAAG,CAAC,CAAC4C,IAAM;gBAACA;aAAE;QAG3B,0FAA0F;QAC1F,kEAAkE;QAClE,MAAMa,gBAAgB;eAAI/C;eAAaiC;SAAiB;QACxD,IAAIc,cAAcrC,MAAM,GAAG,GACzB,MAAMzB,KAAKwD,QAAQ,CACjB,6CACAM,cAAczD,GAAG,CAAC,CAAC4C,IAAM;gBAACA;aAAE;QAEhC,MAAMc,aAA0B,EAAE;QAClC,KAAK,MAAMvE,OAAOwC,WAAY,KAAK,MAAMgC,cAAcxE,IAAIyE,OAAO,CAAEF,WAAWG,IAAI,CAAC;YAAC1E,IAAIE,OAAO;YAAEsE;SAAW;QAC7G,IAAID,WAAWtC,MAAM,GAAG,GAAG,MAAMzB,KAAKwD,QAAQ,CAAC,2DAA2DO;QAE1G,MAAMI,eAAe;eAAIpD;eAAaiC;SAAiB;QACvD,IAAImB,aAAa1C,MAAM,GAAG,GAAG,KAAK,MAAM4B,WAAWzB,SAAU,QAAMyB,kBAAAA,QAAQe,MAAM,cAAdf,sCAAAA,qBAAAA,SAAiBrD,MAAMmE,cAAcvB;QACxG,KAAK,MAAMS,WAAWzB,SAAU;gBAExByB;YADN,MAAMgB,iBAAiCrC,WAAW3B,GAAG,CAAC,CAACb,MAAS,CAAA;oBAAEsB,MAAMtB,IAAIE,OAAO;oBAAE4E,WAAW9E,IAAI8E,SAAS,CAACjB,QAAQkB,IAAI,CAAC;gBAAC,CAAA;YAC5H,QAAMlB,iBAAAA,QAAQmB,KAAK,cAAbnB,qCAAAA,oBAAAA,SAAgBrD,MAAMqE,gBAAgBzB;QAC9C;QACA,KAAK,MAAMS,WAAWzB,SAAU,QAAMyB,0BAAAA,QAAQoB,cAAc,cAAtBpB,8CAAAA,6BAAAA,SAAyBrD,MAAM4C;IACvE,GACA3D;IAGF,qFAAqF;IACrF,8GAA8G;IAC9G,MAAMyF,aAAavB,KAAKC,GAAG,KAAKF;IAChC,MAAMyB,UAAU,MAAM7F,QAAQkB,MAAM;IACpC,yFAAyF;IACzF,+EAA+E;IAC/E,MAAM4E,UAAUD,YAAY,OAAO,CAAC,IAAIE,OAAOF;IAC/C,IAAID,aAAaE,SAAS,MAAM5F,QAAQgB,MAAM,oBAAoB8E,OAAOJ;IAEzE,OAAO;QAAEhD,QAAQM,WAAWP,MAAM;QAAEE;IAAS;AAC/C;AAEA,wFAAwF;AACxF,mFAAmF;AACnF,OAAO,SAASoD,qBAAqBC,MAAc,EAAEC,KAAa;IAChE,MAAMC,QAAQ,CAACC,OAAkBA,KAAKC,UAAU,CAAC,aAAaD,KAAKE,KAAK,CAAC,KAAKC,KAAK,CAAC,GAAG,GAAG5C,IAAI,CAAC,OAAOyC,KAAKE,KAAK,CAAC,IAAI,CAAC,EAAE;IACxH,MAAME,QAAQ,CAACC,MAAgB,IAAI5E,IAAI4E,IAAIH,KAAK,CAAC,KAAKhF,GAAG,CAAC,CAAC8E,OAAS;gBAACD,MAAMC;gBAAOA;aAAK;IACvF,MAAMM,IAAIF,MAAMP;IAChB,MAAMU,IAAIH,MAAMN;IAChB,MAAMU,UAAU,IAAIvG;IACpB,KAAK,MAAM,CAACwG,KAAKC,IAAI,IAAIH,EAAG,IAAID,EAAErE,GAAG,CAACwE,SAASC,KAAKF,QAAQtD,GAAG,CAACuD;IAChE,KAAK,MAAMA,OAAOH,EAAEK,IAAI,GAAI,IAAI,CAACJ,EAAEzE,GAAG,CAAC2E,MAAMD,QAAQtD,GAAG,CAACuD;IACzD,OAAOD;AACT;AAEA,wFAAwF;AACxF,sFAAsF;AACtF,OAAO,SAASI,qBAAqBf,MAAc,EAAEC,KAAa;IAChE,MAAMe,YAAY,CAACR,MAAgBA,IAAIH,KAAK,CAAC,KAAKY,IAAI,CAAC,CAAChD,IAAMA,EAAEmC,UAAU,CAAC;IAC3E,MAAMM,IAAIM,UAAUhB;IACpB,MAAMS,IAAIO,UAAUf;IACpB,IAAIS,MAAM7B,aAAa4B,MAAM5B,WAAW,OAAO;IAC/C,MAAMqC,KAAKT,EAAEU,OAAO,CAAC;IACrB,OAAOT,EAAES,OAAO,CAAC,SAAS,CAAC,KAAKD,OAAO,CAAC,KAAKT,EAAEH,KAAK,CAAC,GAAGY,QAAQR;AAClE;AAEA,4CAA4C;AAC5C,OAAO,SAASU,cAAcpB,MAAc,EAAEC,KAAa;IACzD,MAAMU,UAAUZ,qBAAqBC,QAAQC;IAC7C,MAAMoB,QAAQ,CAACT,MAAiBA,QAAQ,UAAU,mBAAmBA,QAAQ,aAAa,sBAAsBA,IAAIR,UAAU,CAAC,aAAa,CAAC,QAAQ,EAAEQ,IAAIN,KAAK,CAAC,GAAG,CAAC,CAAC,GAAG;IACzK,OAAOK,QAAQnE,IAAI,KAAK,IAAI,aAAa;WAAImE;KAAQ,CAACtF,GAAG,CAACgG,OAAO3D,IAAI,CAAC;AACxE;AAEA,oGAAoG;AACpG,gIAAgI;AAChI,OAAO,eAAe4D,oBAAoBtG,IAAgB,EAAEC,GAAW,EAAEC,OAAe;IACtF,MAAMqG,OAAO,MAAMvG,KAAKQ,OAAO,CAAC;IAChC,MAAMgG,QAAQ,IAAIpH,IAAI,AAAE,CAAA,MAAMmH,KAAK7F,GAAG,EAAC,EAA+BL,GAAG,CAAC,CAACQ,IAAMA,EAAEC,IAAI;IACvF,MAAMX,QAAQ3B,UAAUyB,KAAKC,SAASc,MAAM,CAAC,CAACV,IAAMkG,MAAMvF,GAAG,CAACX,EAAEZ,OAAO;IACvE,MAAMD,aAAarB,gBAAgB6B,SAAS4D;IAC5C,MAAMlC,WAAqB,EAAE;IAC7B,MAAM8B,OAAoB,EAAE;IAC5B,KAAK,MAAMgD,QAAQtG,MAAO;QACxB,MAAM,EAAEX,GAAG,EAAEmC,UAAU+E,YAAY,EAAE,GAAGjI,UAAUgI;QAClD9E,SAASuC,IAAI,IAAIwC;QACjBjD,KAAKS,IAAI,CAAC3E,WAAWC,KAAKC;IAC5B;IACA,MAAMP,gBACJc,MACA;QACE,IAAIyD,KAAKhC,MAAM,GAAG,GAAG,MAAMzB,KAAKwD,QAAQ,CAAClE,oBAAoBmE;IAC/D,GACAxE;IAEF,OAAO0C;AACT"}
|
|
1
|
+
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/sqlite/reconcile.ts"],"sourcesContent":["import type { Config } from '../../config/index.ts';\nimport { contentTokenize } from '../../config/index.ts';\nimport { SenseError } from '../../errors.ts';\nimport type { ReconcileDelta } from '../../features/types.ts';\nimport type { ParsedDoc } from '../../scan/index.ts';\nimport { segmentField } from '../../text/segment.ts';\nimport { recordReconcileDuration } from '../shared.ts';\nimport { BEGIN_WRITE } from '../transaction.ts';\nimport type { Connection, ReconcileDialect } from '../types.ts';\n\n// This store's dialect (types.ts's ReconcileDialect) for the shared orchestration in\n// store/reconcile.ts. content is FTS5-virtual with \"_seg\" sidecars for lexical.ts.\n\n// SQLite's compile-time SQLITE_MAX_COLUMN, default 2000 (https://www.sqlite.org/limits.html).\nconst MAX_FRONTMATTER_COLUMNS = 2000;\n\n// Shared by reconcileContent() and the tokenize-only rebuild in open.ts, so both prepare the\n// same literal instead of two copies drifting apart.\nexport const INSERT_CONTENT_SQL = `INSERT INTO content (rowid, title, summary, text, \"path\", title_seg, summary_seg, text_seg) VALUES ((SELECT rowid FROM frontmatter WHERE \"path\" = ?), ?, ?, ?, ?, ?, ?, ?)`;\n\n// A single content row's param tuple, matching INSERT_CONTENT_SQL's placeholder order.\n// Assumes the frontmatter row for doc.relPath already exists (rowid subquery).\nexport function contentRow(doc: ParsedDoc, segmenting: boolean): unknown[] {\n return [doc.relPath, doc.search.title, doc.search.summary, doc.search.text, doc.relPath, segmenting ? segmentField(doc.search.title) : '', segmenting ? segmentField(doc.search.summary) : '', segmenting ? segmentField(doc.search.text) : ''];\n}\n\nasync function reconcileContent(conn: Connection, touched: string[], docs: ParsedDoc[], _delta: ReconcileDelta, cfg: Config): Promise<void> {\n // FTS5 has no upsert, so delete-before-insert, coupled to the frontmatter rowid (indexed via\n // its PRIMARY KEY) rather than the UNINDEXED `path` column, which a per-row DELETE would scan to find.\n if (touched.length > 0)\n await conn.runBatch(\n 'DELETE FROM content WHERE rowid = (SELECT rowid FROM frontmatter WHERE \"path\" = ?)',\n touched.map((p) => [p])\n );\n\n if (docs.length > 0) {\n // A non-default tokenizer means the tree has chosen its own scheme; a phrase query over\n // grapheme runs would be nonsense against trigram, so the sidecars stay empty.\n const segmenting = contentTokenize(cfg) === undefined;\n await conn.runBatch(\n INSERT_CONTENT_SQL,\n docs.map((doc) => contentRow(doc, segmenting))\n );\n }\n}\n\nexport const sqliteDialect: ReconcileDialect = {\n beginMode: () => BEGIN_WRITE,\n checkColumnLimit(count) {\n // Fence before ALTERing: SQLite's own failure past this point is a raw\n // \"too many columns on sqlite_altertab_frontmatter\" with no indication of the boundary or the levers.\n if (count > MAX_FRONTMATTER_COLUMNS) {\n throw new SenseError(\n 'COLUMN_LIMIT',\n `frontmatter would need ${count} columns, crossing SQLite's compile-time SQLITE_MAX_COLUMN limit (default ${MAX_FRONTMATTER_COLUMNS}; see https://www.sqlite.org/limits.html). Narrow the presets' include globs so fewer/other files are indexed, or fix whatever is generating unbounded frontmatter keys.`\n );\n }\n },\n columnTypeSuffix: () => '',\n reconcileContent,\n recordDuration: recordReconcileDuration,\n};\n"],"names":["contentTokenize","SenseError","segmentField","recordReconcileDuration","BEGIN_WRITE","MAX_FRONTMATTER_COLUMNS","INSERT_CONTENT_SQL","contentRow","doc","segmenting","relPath","search","title","summary","text","reconcileContent","conn","touched","docs","_delta","cfg","length","runBatch","map","p","undefined","sqliteDialect","beginMode","checkColumnLimit","count","columnTypeSuffix","recordDuration"],"mappings":"AACA,SAASA,eAAe,QAAQ,wBAAwB;AACxD,SAASC,UAAU,QAAQ,kBAAkB;AAG7C,SAASC,YAAY,QAAQ,wBAAwB;AACrD,SAASC,uBAAuB,QAAQ,eAAe;AACvD,SAASC,WAAW,QAAQ,oBAAoB;AAGhD,qFAAqF;AACrF,mFAAmF;AAEnF,8FAA8F;AAC9F,MAAMC,0BAA0B;AAEhC,6FAA6F;AAC7F,qDAAqD;AACrD,OAAO,MAAMC,qBAAqB,CAAC,0KAA0K,CAAC,CAAC;AAE/M,uFAAuF;AACvF,+EAA+E;AAC/E,OAAO,SAASC,WAAWC,GAAc,EAAEC,UAAmB;IAC5D,OAAO;QAACD,IAAIE,OAAO;QAAEF,IAAIG,MAAM,CAACC,KAAK;QAAEJ,IAAIG,MAAM,CAACE,OAAO;QAAEL,IAAIG,MAAM,CAACG,IAAI;QAAEN,IAAIE,OAAO;QAAED,aAAaP,aAAaM,IAAIG,MAAM,CAACC,KAAK,IAAI;QAAIH,aAAaP,aAAaM,IAAIG,MAAM,CAACE,OAAO,IAAI;QAAIJ,aAAaP,aAAaM,IAAIG,MAAM,CAACG,IAAI,IAAI;KAAG;AACjP;AAEA,eAAeC,iBAAiBC,IAAgB,EAAEC,OAAiB,EAAEC,IAAiB,EAAEC,MAAsB,EAAEC,GAAW;IACzH,6FAA6F;IAC7F,uGAAuG;IACvG,IAAIH,QAAQI,MAAM,GAAG,GACnB,MAAML,KAAKM,QAAQ,CACjB,sFACAL,QAAQM,GAAG,CAAC,CAACC,IAAM;YAACA;SAAE;IAG1B,IAAIN,KAAKG,MAAM,GAAG,GAAG;QACnB,wFAAwF;QACxF,+EAA+E;QAC/E,MAAMZ,aAAaT,gBAAgBoB,SAASK;QAC5C,MAAMT,KAAKM,QAAQ,CACjBhB,oBACAY,KAAKK,GAAG,CAAC,CAACf,MAAQD,WAAWC,KAAKC;IAEtC;AACF;AAEA,OAAO,MAAMiB,gBAAkC;IAC7CC,WAAW,IAAMvB;IACjBwB,kBAAiBC,KAAK;QACpB,uEAAuE;QACvE,sGAAsG;QACtG,IAAIA,QAAQxB,yBAAyB;YACnC,MAAM,IAAIJ,WACR,gBACA,CAAC,uBAAuB,EAAE4B,MAAM,0EAA0E,EAAExB,wBAAwB,wKAAwK,CAAC;QAEjT;IACF;IACAyB,kBAAkB,IAAM;IACxBf;IACAgB,gBAAgB5B;AAClB,EAAE"}
|
|
@@ -2,4 +2,4 @@ import type { DatabaseSync } from 'node:sqlite';
|
|
|
2
2
|
import type { Config } from '../../config/index.js';
|
|
3
3
|
import type { Capability, Connection, Store } from '../types.js';
|
|
4
4
|
export declare const CAPABILITIES: ReadonlySet<Capability>;
|
|
5
|
-
export declare function createStore(db: DatabaseSync, conn: Connection, cfg: Config
|
|
5
|
+
export declare function createStore(db: DatabaseSync, conn: Connection, cfg: Config): Store;
|
|
@@ -4,18 +4,18 @@ import { withTransaction } from '../transaction.js';
|
|
|
4
4
|
import { hasVectorRow, pendingRows } from '../vectors.js';
|
|
5
5
|
import { fieldStats } from './fieldStats.js';
|
|
6
6
|
import { queryLexical } from './lexical.js';
|
|
7
|
-
import { reconcile } from './reconcile.js';
|
|
8
7
|
import { scanCandidates, scanSimilar, writeVectorBatch } from './vectors.js';
|
|
9
8
|
export const CAPABILITIES = new Set([
|
|
10
9
|
'phrases',
|
|
11
10
|
'snippets',
|
|
12
11
|
'watch-concurrency',
|
|
13
12
|
'lexical',
|
|
14
|
-
'vectors'
|
|
13
|
+
'vectors',
|
|
14
|
+
'sql-functions'
|
|
15
15
|
]);
|
|
16
16
|
// Wraps the synchronous DatabaseSync connection in the async Store interface, sharing one
|
|
17
|
-
// Connection instance (conn) with
|
|
18
|
-
export function createStore(db, conn, cfg
|
|
17
|
+
// Connection instance (conn) with the builder's own reconcile call so transaction depth is tracked against the same object everywhere.
|
|
18
|
+
export function createStore(db, conn, cfg) {
|
|
19
19
|
return {
|
|
20
20
|
name: 'sqlite',
|
|
21
21
|
capabilities: CAPABILITIES,
|
|
@@ -31,9 +31,6 @@ export function createStore(db, conn, cfg, baseDir) {
|
|
|
31
31
|
async transaction (fn) {
|
|
32
32
|
return withTransaction(conn, fn);
|
|
33
33
|
},
|
|
34
|
-
async reconcile () {
|
|
35
|
-
return reconcile(conn, cfg, baseDir);
|
|
36
|
-
},
|
|
37
34
|
docs: {
|
|
38
35
|
async columns () {
|
|
39
36
|
return [
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/sqlite/store.ts"],"sourcesContent":["import type { DatabaseSync } from 'node:sqlite';\nimport type { Config } from '../../config/index.ts';\nimport { contentTokenize } from '../../config/index.ts';\nimport { getColumns } from '../shared.ts';\nimport { withTransaction } from '../transaction.ts';\nimport type { Capability, Connection, Statement, Store, VectorWriteRow } from '../types.ts';\nimport { hasVectorRow, pendingRows } from '../vectors.ts';\nimport { fieldStats } from './fieldStats.ts';\nimport { queryLexical } from './lexical.ts';\nimport {
|
|
1
|
+
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/sqlite/store.ts"],"sourcesContent":["import type { DatabaseSync } from 'node:sqlite';\nimport type { Config } from '../../config/index.ts';\nimport { contentTokenize } from '../../config/index.ts';\nimport { getColumns } from '../shared.ts';\nimport { withTransaction } from '../transaction.ts';\nimport type { Capability, Connection, Statement, Store, VectorWriteRow } from '../types.ts';\nimport { hasVectorRow, pendingRows } from '../vectors.ts';\nimport { fieldStats } from './fieldStats.ts';\nimport { queryLexical } from './lexical.ts';\nimport { scanCandidates, scanSimilar, writeVectorBatch } from './vectors.ts';\n\nexport const CAPABILITIES: ReadonlySet<Capability> = new Set(['phrases', 'snippets', 'watch-concurrency', 'lexical', 'vectors', 'sql-functions']);\n\n// Wraps the synchronous DatabaseSync connection in the async Store interface, sharing one\n// Connection instance (conn) with the builder's own reconcile call so transaction depth is tracked against the same object everywhere.\nexport function createStore(db: DatabaseSync, conn: Connection, cfg: Config): Store {\n return {\n name: 'sqlite',\n capabilities: CAPABILITIES,\n async exec(sql: string): Promise<void> {\n await conn.exec(sql);\n },\n async prepare(sql: string): Promise<Statement> {\n return conn.prepare(sql);\n },\n async runBatch(sql: string, paramRows: unknown[][]): Promise<void> {\n await conn.runBatch(sql, paramRows);\n },\n async transaction<T>(fn: () => Promise<T>): Promise<T> {\n return withTransaction(conn, fn);\n },\n docs: {\n async columns() {\n return [...(await getColumns(conn))];\n },\n async fieldStats(columns, scopeWhere) {\n return fieldStats(conn, columns, scopeWhere);\n },\n },\n lexical: {\n async query(terms, opts) {\n return queryLexical(conn, terms, opts, contentTokenize(cfg) === undefined);\n },\n },\n vectors: {\n async pending() {\n return pendingRows(conn);\n },\n async writeVectors(rows: VectorWriteRow[]) {\n await writeVectorBatch(conn, rows);\n },\n async candidates(qv, storeDims, fetch, allowed) {\n return scanCandidates(conn, qv, storeDims, fetch, allowed);\n },\n async similar(path, opts) {\n return scanSimilar(conn, path, opts);\n },\n async hasVector(path) {\n return hasVectorRow(conn, path);\n },\n },\n async engineStatus() {\n // Read back rather than recomputed: this is what open() actually set (3x the largest\n // recorded reconcile, floored at 30s, capped at 10min), not a value re-derived here.\n const row = db.prepare('PRAGMA busy_timeout').get() as { timeout: number };\n return { busy_timeout: `${row.timeout}ms (derived: 3x the largest reconcile this cache has recorded, floored at 30000ms)` };\n },\n raw: {\n async prepare(sql: string) {\n const stmt = db.prepare(sql);\n stmt.setReadBigInts(true); // int64 past 2^53 arrives as BigInt instead of throwing at step time\n return {\n columns: () => stmt.columns(),\n // sense sql streams through an async iterator wrapping the sync iterate().\n iterate: async function* (...params: unknown[]) {\n yield* stmt.iterate(...(params as Parameters<typeof stmt.iterate>));\n },\n };\n },\n },\n async close() {\n db.close();\n },\n };\n}\n"],"names":["contentTokenize","getColumns","withTransaction","hasVectorRow","pendingRows","fieldStats","queryLexical","scanCandidates","scanSimilar","writeVectorBatch","CAPABILITIES","Set","createStore","db","conn","cfg","name","capabilities","exec","sql","prepare","runBatch","paramRows","transaction","fn","docs","columns","scopeWhere","lexical","query","terms","opts","undefined","vectors","pending","writeVectors","rows","candidates","qv","storeDims","fetch","allowed","similar","path","hasVector","engineStatus","row","get","busy_timeout","timeout","raw","stmt","setReadBigInts","iterate","params","close"],"mappings":"AAEA,SAASA,eAAe,QAAQ,wBAAwB;AACxD,SAASC,UAAU,QAAQ,eAAe;AAC1C,SAASC,eAAe,QAAQ,oBAAoB;AAEpD,SAASC,YAAY,EAAEC,WAAW,QAAQ,gBAAgB;AAC1D,SAASC,UAAU,QAAQ,kBAAkB;AAC7C,SAASC,YAAY,QAAQ,eAAe;AAC5C,SAASC,cAAc,EAAEC,WAAW,EAAEC,gBAAgB,QAAQ,eAAe;AAE7E,OAAO,MAAMC,eAAwC,IAAIC,IAAI;IAAC;IAAW;IAAY;IAAqB;IAAW;IAAW;CAAgB,EAAE;AAElJ,0FAA0F;AAC1F,uIAAuI;AACvI,OAAO,SAASC,YAAYC,EAAgB,EAAEC,IAAgB,EAAEC,GAAW;IACzE,OAAO;QACLC,MAAM;QACNC,cAAcP;QACd,MAAMQ,MAAKC,GAAW;YACpB,MAAML,KAAKI,IAAI,CAACC;QAClB;QACA,MAAMC,SAAQD,GAAW;YACvB,OAAOL,KAAKM,OAAO,CAACD;QACtB;QACA,MAAME,UAASF,GAAW,EAAEG,SAAsB;YAChD,MAAMR,KAAKO,QAAQ,CAACF,KAAKG;QAC3B;QACA,MAAMC,aAAeC,EAAoB;YACvC,OAAOtB,gBAAgBY,MAAMU;QAC/B;QACAC,MAAM;YACJ,MAAMC;gBACJ,OAAO;uBAAK,MAAMzB,WAAWa;iBAAO;YACtC;YACA,MAAMT,YAAWqB,OAAO,EAAEC,UAAU;gBAClC,OAAOtB,WAAWS,MAAMY,SAASC;YACnC;QACF;QACAC,SAAS;YACP,MAAMC,OAAMC,KAAK,EAAEC,IAAI;gBACrB,OAAOzB,aAAaQ,MAAMgB,OAAOC,MAAM/B,gBAAgBe,SAASiB;YAClE;QACF;QACAC,SAAS;YACP,MAAMC;gBACJ,OAAO9B,YAAYU;YACrB;YACA,MAAMqB,cAAaC,IAAsB;gBACvC,MAAM3B,iBAAiBK,MAAMsB;YAC/B;YACA,MAAMC,YAAWC,EAAE,EAAEC,SAAS,EAAEC,KAAK,EAAEC,OAAO;gBAC5C,OAAOlC,eAAeO,MAAMwB,IAAIC,WAAWC,OAAOC;YACpD;YACA,MAAMC,SAAQC,IAAI,EAAEZ,IAAI;gBACtB,OAAOvB,YAAYM,MAAM6B,MAAMZ;YACjC;YACA,MAAMa,WAAUD,IAAI;gBAClB,OAAOxC,aAAaW,MAAM6B;YAC5B;QACF;QACA,MAAME;YACJ,qFAAqF;YACrF,qFAAqF;YACrF,MAAMC,MAAMjC,GAAGO,OAAO,CAAC,uBAAuB2B,GAAG;YACjD,OAAO;gBAAEC,cAAc,GAAGF,IAAIG,OAAO,CAAC,kFAAkF,CAAC;YAAC;QAC5H;QACAC,KAAK;YACH,MAAM9B,SAAQD,GAAW;gBACvB,MAAMgC,OAAOtC,GAAGO,OAAO,CAACD;gBACxBgC,KAAKC,cAAc,CAAC,OAAO,qEAAqE;gBAChG,OAAO;oBACL1B,SAAS,IAAMyB,KAAKzB,OAAO;oBAC3B,2EAA2E;oBAC3E2B,SAAS,gBAAiB,GAAGC,MAAiB;wBAC5C,OAAOH,KAAKE,OAAO,IAAKC;oBAC1B;gBACF;YACF;QACF;QACA,MAAMC;YACJ1C,GAAG0C,KAAK;QACV;IACF;AACF"}
|
|
@@ -1,14 +1,9 @@
|
|
|
1
|
+
import type { Database } from '@tursodatabase/database';
|
|
1
2
|
import type { ResolvedConfig } from '../../config/index.js';
|
|
2
|
-
import type {
|
|
3
|
+
import type { OpenResult } from '../open.js';
|
|
4
|
+
import type { OpenDialect } from '../types.js';
|
|
3
5
|
export declare const DB_FILENAME = "cache.turso.db";
|
|
4
|
-
export declare const SCHEMA_VERSION = "
|
|
5
|
-
export
|
|
6
|
-
|
|
7
|
-
cfg: ResolvedConfig;
|
|
8
|
-
dbPath: string;
|
|
9
|
-
parsed: number;
|
|
10
|
-
warnings: string[];
|
|
11
|
-
}
|
|
12
|
-
export declare const CONTENT_FTS_DDL: readonly ["CREATE INDEX IF NOT EXISTS content_fts ON content USING fts (title, summary, text) WITH (weights = 'title=10.0,summary=5.0,text=1.0')", "CREATE INDEX IF NOT EXISTS content_fts_ngram ON content USING fts (title_ngram, summary_ngram, text_ngram) WITH (tokenizer='ngram', weights='title_ngram=10.0,summary_ngram=5.0,text_ngram=1.0')"];
|
|
13
|
-
export declare const CONTENT_FTS_NAMES: readonly ["content_fts", "content_fts_ngram"];
|
|
6
|
+
export declare const SCHEMA_VERSION = "4";
|
|
7
|
+
export type { OpenResult };
|
|
8
|
+
export declare const tursoOpenDialect: OpenDialect<Database>;
|
|
14
9
|
export declare function openTurso(cfg: ResolvedConfig): Promise<OpenResult>;
|
|
@@ -1,33 +1,20 @@
|
|
|
1
|
-
import {
|
|
2
|
-
import { join } from 'node:path';
|
|
3
|
-
import { featureSignature, STATE_DIR } from '../../config/index.js';
|
|
4
|
-
import { rekeyChunkText } from '../../embed/handoff.js';
|
|
1
|
+
import { featureSignature } from '../../config/index.js';
|
|
5
2
|
import { STORE_DIMS } from '../../embed/types.js';
|
|
6
3
|
import { SenseError } from '../../errors.js';
|
|
7
4
|
import { activeFeatures, FEATURES } from '../../features/index.js';
|
|
8
|
-
import {
|
|
5
|
+
import { openWithDialect } from '../open.js';
|
|
9
6
|
import { getMeta, setMeta } from '../shared.js';
|
|
10
7
|
import { createConnection } from './connection.js';
|
|
11
8
|
import { TURSO_PACKAGE, tursoApi } from './native.js';
|
|
12
|
-
import {
|
|
9
|
+
import { CONTENT_FTS_DDL, tursoDialect } from './reconcile.js';
|
|
13
10
|
import { createStore } from './store.js';
|
|
14
11
|
export const DB_FILENAME = 'cache.turso.db';
|
|
15
12
|
// Independent of sqlite's and duckdb's SCHEMA_VERSION: each store's cache shape evolves
|
|
16
13
|
// separately. Covers the FTS indexes, the "_ngram" sidecar columns, and embeddings.vector's width.
|
|
17
|
-
export const SCHEMA_VERSION = '
|
|
18
|
-
//
|
|
19
|
-
//
|
|
20
|
-
|
|
21
|
-
// Tantivy maintains them per inserted row, which is quadratic in what is already indexed.
|
|
22
|
-
export const CONTENT_FTS_DDL = [
|
|
23
|
-
`CREATE INDEX IF NOT EXISTS content_fts ON content USING fts (title, summary, text) WITH (weights = 'title=10.0,summary=5.0,text=1.0')`,
|
|
24
|
-
`CREATE INDEX IF NOT EXISTS content_fts_ngram ON content USING fts (title_ngram, summary_ngram, text_ngram) WITH (tokenizer='ngram', weights='title_ngram=10.0,summary_ngram=5.0,text_ngram=1.0')`
|
|
25
|
-
];
|
|
26
|
-
export const CONTENT_FTS_NAMES = [
|
|
27
|
-
'content_fts',
|
|
28
|
-
'content_fts_ngram'
|
|
29
|
-
];
|
|
30
|
-
async function ensureSchema(conn, cfg) {
|
|
14
|
+
export const SCHEMA_VERSION = '4';
|
|
15
|
+
// This store's Handle (types.ts's OpenDialect<Handle>) is the connected Database itself: no
|
|
16
|
+
// extra native state to thread, unlike duckdb's separate instance/connection pair.
|
|
17
|
+
async function ensureSchema(_handle, conn, cfg) {
|
|
31
18
|
await conn.exec(`CREATE TABLE IF NOT EXISTS frontmatter ("path" TEXT PRIMARY KEY, "_mtime" REAL, "_ctime" REAL, "_size" INTEGER, "_parse_error" TEXT)`);
|
|
32
19
|
await conn.exec(`CREATE TABLE IF NOT EXISTS content ("path" TEXT PRIMARY KEY, title TEXT, summary TEXT, text TEXT, title_ngram TEXT, summary_ngram TEXT, text_ngram TEXT)`);
|
|
33
20
|
for (const ddl of CONTENT_FTS_DDL)await conn.exec(ddl);
|
|
@@ -45,12 +32,13 @@ async function ensureSchema(conn, cfg) {
|
|
|
45
32
|
if (await getMeta(conn, 'schema_version') === null) await setMeta(conn, 'schema_version', SCHEMA_VERSION);
|
|
46
33
|
if (await getMeta(conn, 'features') === null) await setMeta(conn, 'features', featureSignature(cfg, FEATURES));
|
|
47
34
|
}
|
|
48
|
-
async function
|
|
49
|
-
|
|
50
|
-
|
|
51
|
-
|
|
52
|
-
});
|
|
53
|
-
|
|
35
|
+
async function close(handle) {
|
|
36
|
+
await handle.close();
|
|
37
|
+
}
|
|
38
|
+
async function setDerivedBusyTimeout(_handle, conn, ms) {
|
|
39
|
+
await conn.exec(`PRAGMA busy_timeout = ${ms}`);
|
|
40
|
+
}
|
|
41
|
+
async function connect(dbPath, _cfg) {
|
|
54
42
|
// Dynamic, not a top-level import: a sqlite or duckdb tree must never attempt to resolve this
|
|
55
43
|
// optional dependency until a turso tree is actually opened. Installed on first use if missing.
|
|
56
44
|
let turso;
|
|
@@ -73,57 +61,22 @@ async function connect(cfg) {
|
|
|
73
61
|
} catch (err) {
|
|
74
62
|
throw new SenseError('STORE_DEPENDENCY_MISSING', `store "turso" failed to open ${dbPath}: ${err.message}`);
|
|
75
63
|
}
|
|
76
|
-
// A throw below must release this handle, or the leaked WAL makes the db file undeletable on
|
|
77
|
-
// Windows. closed guards the catch against double-closing.
|
|
78
|
-
let closed = false;
|
|
79
|
-
try {
|
|
80
|
-
var _ref;
|
|
81
|
-
const conn = createConnection(db);
|
|
82
|
-
await conn.exec('CREATE TABLE IF NOT EXISTS meta (key TEXT PRIMARY KEY, value TEXT)');
|
|
83
|
-
const version = await getMeta(conn, 'schema_version');
|
|
84
|
-
const features = await getMeta(conn, 'features');
|
|
85
|
-
const wantFeatures = featureSignature(cfg, FEATURES);
|
|
86
|
-
if (version !== null && version !== SCHEMA_VERSION || features !== null && features !== wantFeatures) {
|
|
87
|
-
// Reconcile only reparses changed files, so a stale cache is rebuilt rather than patched
|
|
88
|
-
// incrementally. No tokenize-only or embed-identity-adopted partial path: any mismatch rebuilds.
|
|
89
|
-
const reason = version !== null && version !== SCHEMA_VERSION ? 'cache format changed (new sensemaking version)' : 'config change (features, tokenizer, or presets)';
|
|
90
|
-
console.error(`sense: ${reason}; rebuilding the index`);
|
|
91
|
-
closed = true;
|
|
92
|
-
await db.close();
|
|
93
|
-
clearCache(cfg);
|
|
94
|
-
return connect(cfg);
|
|
95
|
-
}
|
|
96
|
-
await ensureSchema(conn, cfg);
|
|
97
|
-
// 3x the largest reconcile this cache has recorded, floored at 30s and capped at 10min.
|
|
98
|
-
// Installed before reconcile() -- that call is the one that races a watcher's transaction.
|
|
99
|
-
const recordedMaxMs = Number((_ref = await getMeta(conn, 'reconcile_max_ms')) !== null && _ref !== void 0 ? _ref : '0');
|
|
100
|
-
await conn.exec(`PRAGMA busy_timeout = ${Math.min(Math.max(30000, 3 * recordedMaxMs), 600000)}`);
|
|
101
|
-
const { parsed, warnings } = await reconcile(conn, cfg, cfg.baseDir);
|
|
102
|
-
return {
|
|
103
|
-
db,
|
|
104
|
-
conn,
|
|
105
|
-
cfg,
|
|
106
|
-
dbPath,
|
|
107
|
-
parsed,
|
|
108
|
-
warnings
|
|
109
|
-
};
|
|
110
|
-
} catch (err) {
|
|
111
|
-
if (!closed) await db.close();
|
|
112
|
-
throw err;
|
|
113
|
-
}
|
|
114
|
-
}
|
|
115
|
-
// The turso store's open: connects (fully async, see connect() above), then wraps the resulting
|
|
116
|
-
// connection in the Store interface.
|
|
117
|
-
export async function openTurso(cfg) {
|
|
118
|
-
const { db, conn, cfg: resolvedCfg, dbPath, parsed, warnings } = await connect(cfg);
|
|
119
|
-
const store = createStore(db, conn, resolvedCfg, resolvedCfg.baseDir);
|
|
120
|
-
// reconcile ran before this object existed, so its chunk text is keyed by the connection.
|
|
121
|
-
rekeyChunkText(conn, store);
|
|
122
64
|
return {
|
|
123
|
-
|
|
124
|
-
|
|
125
|
-
dbPath,
|
|
126
|
-
parsed,
|
|
127
|
-
warnings
|
|
65
|
+
handle: db,
|
|
66
|
+
conn: createConnection(db)
|
|
128
67
|
};
|
|
129
68
|
}
|
|
69
|
+
// This store's dialect (types.ts's OpenDialect) for the shared orchestration in store/open.ts.
|
|
70
|
+
export const tursoOpenDialect = {
|
|
71
|
+
filename: DB_FILENAME,
|
|
72
|
+
schemaVersion: SCHEMA_VERSION,
|
|
73
|
+
reconcileDialect: tursoDialect,
|
|
74
|
+
connect,
|
|
75
|
+
close,
|
|
76
|
+
ensureSchema,
|
|
77
|
+
setDerivedBusyTimeout,
|
|
78
|
+
createStore: (handle, conn)=>createStore(handle, conn)
|
|
79
|
+
};
|
|
80
|
+
export async function openTurso(cfg) {
|
|
81
|
+
return openWithDialect(cfg, tursoOpenDialect);
|
|
82
|
+
}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/turso/open.ts"],"sourcesContent":["import { mkdirSync } from 'node:fs';\nimport { join } from 'node:path';\nimport type { Database } from '@tursodatabase/database';\nimport type { Config, ResolvedConfig } from '../../config/index.ts';\nimport { featureSignature, STATE_DIR } from '../../config/index.ts';\nimport { rekeyChunkText } from '../../embed/handoff.ts';\nimport { STORE_DIMS } from '../../embed/types.ts';\nimport { SenseError } from '../../errors.ts';\nimport { activeFeatures, FEATURES } from '../../features/index.ts';\nimport { clearCache } from '../cache.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport type { Connection, Store } from '../types.ts';\nimport { createConnection } from './connection.ts';\nimport { TURSO_PACKAGE, tursoApi } from './native.ts';\nimport { reconcile } from './reconcile.ts';\nimport { createStore } from './store.ts';\n\nexport const DB_FILENAME = 'cache.turso.db';\n// Independent of sqlite's and duckdb's SCHEMA_VERSION: each store's cache shape evolves\n// separately. Covers the FTS indexes, the \"_ngram\" sidecar columns, and embeddings.vector's width.\nexport const SCHEMA_VERSION = '3';\n\nexport interface OpenResult {\n store: Store;\n cfg: ResolvedConfig;\n dbPath: string;\n parsed: number;\n warnings: string[];\n}\n\n// The ngram index is scoped to disjoint \"_ngram\" sidecar columns: a second index over the same\n// columns makes a bare substring match a whole word, defeating the prefix-query rejection.\n// The two FTS indexes, named here because reconcile drops and rebuilds them around a bulk load:\n// Tantivy maintains them per inserted row, which is quadratic in what is already indexed.\nexport const CONTENT_FTS_DDL = [\n `CREATE INDEX IF NOT EXISTS content_fts ON content USING fts (title, summary, text) WITH (weights = 'title=10.0,summary=5.0,text=1.0')`,\n `CREATE INDEX IF NOT EXISTS content_fts_ngram ON content USING fts (title_ngram, summary_ngram, text_ngram) WITH (tokenizer='ngram', weights='title_ngram=10.0,summary_ngram=5.0,text_ngram=1.0')`,\n] as const;\nexport const CONTENT_FTS_NAMES = ['content_fts', 'content_fts_ngram'] as const;\n\nasync function ensureSchema(conn: Connection, cfg: Config): Promise<void> {\n await conn.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" REAL, \"_ctime\" REAL, \"_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, title_ngram TEXT, summary_ngram TEXT, text_ngram TEXT)`);\n for (const ddl of CONTENT_FTS_DDL) await conn.exec(ddl);\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 F32_BLOB(STORE_DIMS) instead of the embed feature's engine-neutral BLOB DDL. `scale`\n // is kept unused, so the shared reconcile-time INSERT/DELETE names a column both stores have.\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 F32_BLOB(${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\ninterface ConnectResult {\n db: Database;\n conn: Connection;\n cfg: ResolvedConfig;\n dbPath: string;\n parsed: number;\n warnings: string[];\n}\n\nasync function connect(cfg: ResolvedConfig): Promise<ConnectResult> {\n const stateDir = join(cfg.baseDir, STATE_DIR);\n mkdirSync(stateDir, { recursive: true });\n const dbPath = join(stateDir, DB_FILENAME);\n\n // Dynamic, not a top-level import: a sqlite or duckdb tree must never attempt to resolve this\n // optional dependency until a turso tree is actually opened. Installed on first use if missing.\n let turso: Awaited<ReturnType<typeof tursoApi>>;\n try {\n turso = await tursoApi();\n } catch (err) {\n if (err instanceof SenseError) throw err;\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"turso\" needs the ${TURSO_PACKAGE} package (${(err as Error).message})`);\n }\n\n let db: Database;\n try {\n // Floored at the same 30s sqlite opens with. `timeout` is connect-time only in this client;\n // the derived value is set via runtime PRAGMA below. `index_method` is required for ensureSchema()'s FTS indexes (T1).\n db = await turso.connect(dbPath, { timeout: 30_000, experimental: ['index_method'] });\n } catch (err) {\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"turso\" failed to open ${dbPath}: ${(err as Error).message}`);\n }\n // A throw below must release this handle, or the leaked WAL makes the db file undeletable on\n // Windows. closed guards the catch against double-closing.\n let closed = false;\n try {\n const conn = createConnection(db);\n\n await conn.exec('CREATE TABLE IF NOT EXISTS meta (key TEXT PRIMARY KEY, value TEXT)');\n\n const version = await getMeta(conn, 'schema_version');\n const features = await getMeta(conn, 'features');\n const wantFeatures = featureSignature(cfg, FEATURES);\n if ((version !== null && version !== SCHEMA_VERSION) || (features !== null && features !== wantFeatures)) {\n // Reconcile only reparses changed files, so a stale cache is rebuilt rather than patched\n // incrementally. No tokenize-only or embed-identity-adopted partial path: any mismatch rebuilds.\n const reason = version !== null && version !== SCHEMA_VERSION ? 'cache format changed (new sensemaking version)' : 'config change (features, tokenizer, or presets)';\n console.error(`sense: ${reason}; rebuilding the index`);\n closed = true;\n await db.close();\n clearCache(cfg);\n return connect(cfg);\n }\n\n await ensureSchema(conn, cfg);\n\n // 3x the largest reconcile this cache has recorded, floored at 30s and capped at 10min.\n // Installed before reconcile() -- that call is the one that races a watcher's transaction.\n const recordedMaxMs = Number((await getMeta(conn, 'reconcile_max_ms')) ?? '0');\n await conn.exec(`PRAGMA busy_timeout = ${Math.min(Math.max(30000, 3 * recordedMaxMs), 600_000)}`);\n\n const { parsed, warnings } = await reconcile(conn, cfg, cfg.baseDir);\n\n return { db, conn, cfg, dbPath, parsed, warnings };\n } catch (err) {\n if (!closed) await db.close();\n throw err;\n }\n}\n\n// The turso store's open: connects (fully async, see connect() above), then wraps the resulting\n// connection in the Store interface.\nexport async function openTurso(cfg: ResolvedConfig): Promise<OpenResult> {\n const { db, conn, cfg: resolvedCfg, dbPath, parsed, warnings } = await connect(cfg);\n const store = createStore(db, conn, resolvedCfg, resolvedCfg.baseDir);\n // reconcile ran before this object existed, so its chunk text is keyed by the connection.\n rekeyChunkText(conn, store);\n return { store: store, cfg: resolvedCfg, dbPath, parsed, warnings };\n}\n"],"names":["mkdirSync","join","featureSignature","STATE_DIR","rekeyChunkText","STORE_DIMS","SenseError","activeFeatures","FEATURES","clearCache","getMeta","setMeta","createConnection","TURSO_PACKAGE","tursoApi","reconcile","createStore","DB_FILENAME","SCHEMA_VERSION","CONTENT_FTS_DDL","CONTENT_FTS_NAMES","ensureSchema","conn","cfg","exec","ddl","feature","name","schema","connect","stateDir","baseDir","recursive","dbPath","turso","err","message","db","timeout","experimental","closed","version","features","wantFeatures","reason","console","error","close","recordedMaxMs","Number","Math","min","max","parsed","warnings","openTurso","resolvedCfg","store"],"mappings":"AAAA,SAASA,SAAS,QAAQ,UAAU;AACpC,SAASC,IAAI,QAAQ,YAAY;AAGjC,SAASC,gBAAgB,EAAEC,SAAS,QAAQ,wBAAwB;AACpE,SAASC,cAAc,QAAQ,yBAAyB;AACxD,SAASC,UAAU,QAAQ,uBAAuB;AAClD,SAASC,UAAU,QAAQ,kBAAkB;AAC7C,SAASC,cAAc,EAAEC,QAAQ,QAAQ,0BAA0B;AACnE,SAASC,UAAU,QAAQ,cAAc;AACzC,SAASC,OAAO,EAAEC,OAAO,QAAQ,eAAe;AAEhD,SAASC,gBAAgB,QAAQ,kBAAkB;AACnD,SAASC,aAAa,EAAEC,QAAQ,QAAQ,cAAc;AACtD,SAASC,SAAS,QAAQ,iBAAiB;AAC3C,SAASC,WAAW,QAAQ,aAAa;AAEzC,OAAO,MAAMC,cAAc,iBAAiB;AAC5C,wFAAwF;AACxF,mGAAmG;AACnG,OAAO,MAAMC,iBAAiB,IAAI;AAUlC,+FAA+F;AAC/F,2FAA2F;AAC3F,gGAAgG;AAChG,0FAA0F;AAC1F,OAAO,MAAMC,kBAAkB;IAC7B,CAAC,qIAAqI,CAAC;IACvI,CAAC,gMAAgM,CAAC;CACnM,CAAU;AACX,OAAO,MAAMC,oBAAoB;IAAC;IAAe;CAAoB,CAAU;AAE/E,eAAeC,aAAaC,IAAgB,EAAEC,GAAW;IACvD,MAAMD,KAAKE,IAAI,CAAC,CAAC,oIAAoI,CAAC;IACtJ,MAAMF,KAAKE,IAAI,CAAC,CAAC,wJAAwJ,CAAC;IAC1K,KAAK,MAAMC,OAAON,gBAAiB,MAAMG,KAAKE,IAAI,CAACC;IACnD,MAAMH,KAAKE,IAAI,CAAC,CAAC,gGAAgG,CAAC;IAClH,MAAMF,KAAKE,IAAI,CAAC;IAChB,KAAK,MAAME,WAAWnB,eAAegB,KAAM;QACzC,8FAA8F;QAC9F,8FAA8F;QAC9F,IAAIG,QAAQC,IAAI,KAAK,SAAS;YAC5B,MAAML,KAAKE,IAAI,CAAC,CAAC,qIAAqI,EAAEnB,WAAW,+BAA+B,CAAC;YACnM;QACF;QACA,MAAMqB,QAAQE,MAAM,CAACN;IACvB;IACA,IAAI,AAAC,MAAMZ,QAAQY,MAAM,sBAAuB,MAAM,MAAMX,QAAQW,MAAM,kBAAkBJ;IAC5F,IAAI,AAAC,MAAMR,QAAQY,MAAM,gBAAiB,MAAM,MAAMX,QAAQW,MAAM,YAAYpB,iBAAiBqB,KAAKf;AACxG;AAWA,eAAeqB,QAAQN,GAAmB;IACxC,MAAMO,WAAW7B,KAAKsB,IAAIQ,OAAO,EAAE5B;IACnCH,UAAU8B,UAAU;QAAEE,WAAW;IAAK;IACtC,MAAMC,SAAShC,KAAK6B,UAAUb;IAE9B,8FAA8F;IAC9F,gGAAgG;IAChG,IAAIiB;IACJ,IAAI;QACFA,QAAQ,MAAMpB;IAChB,EAAE,OAAOqB,KAAK;QACZ,IAAIA,eAAe7B,YAAY,MAAM6B;QACrC,MAAM,IAAI7B,WAAW,4BAA4B,CAAC,wBAAwB,EAAEO,cAAc,UAAU,EAAE,AAACsB,IAAcC,OAAO,CAAC,CAAC,CAAC;IACjI;IAEA,IAAIC;IACJ,IAAI;QACF,4FAA4F;QAC5F,uHAAuH;QACvHA,KAAK,MAAMH,MAAML,OAAO,CAACI,QAAQ;YAAEK,SAAS;YAAQC,cAAc;gBAAC;aAAe;QAAC;IACrF,EAAE,OAAOJ,KAAK;QACZ,MAAM,IAAI7B,WAAW,4BAA4B,CAAC,6BAA6B,EAAE2B,OAAO,EAAE,EAAE,AAACE,IAAcC,OAAO,EAAE;IACtH;IACA,6FAA6F;IAC7F,2DAA2D;IAC3D,IAAII,SAAS;IACb,IAAI;YAuB4B;QAtB9B,MAAMlB,OAAOV,iBAAiByB;QAE9B,MAAMf,KAAKE,IAAI,CAAC;QAEhB,MAAMiB,UAAU,MAAM/B,QAAQY,MAAM;QACpC,MAAMoB,WAAW,MAAMhC,QAAQY,MAAM;QACrC,MAAMqB,eAAezC,iBAAiBqB,KAAKf;QAC3C,IAAI,AAACiC,YAAY,QAAQA,YAAYvB,kBAAoBwB,aAAa,QAAQA,aAAaC,cAAe;YACxG,yFAAyF;YACzF,iGAAiG;YACjG,MAAMC,SAASH,YAAY,QAAQA,YAAYvB,iBAAiB,mDAAmD;YACnH2B,QAAQC,KAAK,CAAC,CAAC,OAAO,EAAEF,OAAO,sBAAsB,CAAC;YACtDJ,SAAS;YACT,MAAMH,GAAGU,KAAK;YACdtC,WAAWc;YACX,OAAOM,QAAQN;QACjB;QAEA,MAAMF,aAAaC,MAAMC;QAEzB,wFAAwF;QACxF,2FAA2F;QAC3F,MAAMyB,gBAAgBC,QAAQ,OAAA,MAAMvC,QAAQY,MAAM,iCAApB,kBAAA,OAA4C;QAC1E,MAAMA,KAAKE,IAAI,CAAC,CAAC,sBAAsB,EAAE0B,KAAKC,GAAG,CAACD,KAAKE,GAAG,CAAC,OAAO,IAAIJ,gBAAgB,SAAU;QAEhG,MAAM,EAAEK,MAAM,EAAEC,QAAQ,EAAE,GAAG,MAAMvC,UAAUO,MAAMC,KAAKA,IAAIQ,OAAO;QAEnE,OAAO;YAAEM;YAAIf;YAAMC;YAAKU;YAAQoB;YAAQC;QAAS;IACnD,EAAE,OAAOnB,KAAK;QACZ,IAAI,CAACK,QAAQ,MAAMH,GAAGU,KAAK;QAC3B,MAAMZ;IACR;AACF;AAEA,gGAAgG;AAChG,qCAAqC;AACrC,OAAO,eAAeoB,UAAUhC,GAAmB;IACjD,MAAM,EAAEc,EAAE,EAAEf,IAAI,EAAEC,KAAKiC,WAAW,EAAEvB,MAAM,EAAEoB,MAAM,EAAEC,QAAQ,EAAE,GAAG,MAAMzB,QAAQN;IAC/E,MAAMkC,QAAQzC,YAAYqB,IAAIf,MAAMkC,aAAaA,YAAYzB,OAAO;IACpE,0FAA0F;IAC1F3B,eAAekB,MAAMmC;IACrB,OAAO;QAAEA,OAAOA;QAAOlC,KAAKiC;QAAavB;QAAQoB;QAAQC;IAAS;AACpE"}
|
|
1
|
+
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/turso/open.ts"],"sourcesContent":["import type { Database } from '@tursodatabase/database';\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 { TURSO_PACKAGE, tursoApi } from './native.ts';\nimport { CONTENT_FTS_DDL, tursoDialect } from './reconcile.ts';\nimport { createStore } from './store.ts';\n\nexport const DB_FILENAME = 'cache.turso.db';\n// Independent of sqlite's and duckdb's SCHEMA_VERSION: each store's cache shape evolves\n// separately. Covers the FTS indexes, the \"_ngram\" sidecar columns, and embeddings.vector's width.\nexport const SCHEMA_VERSION = '4';\n\nexport type { OpenResult };\n\n// This store's Handle (types.ts's OpenDialect<Handle>) is the connected Database itself: no\n// extra native state to thread, unlike duckdb's separate instance/connection pair.\nasync function ensureSchema(_handle: Database, conn: Connection, cfg: Config): Promise<void> {\n await conn.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" REAL, \"_ctime\" REAL, \"_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, title_ngram TEXT, summary_ngram TEXT, text_ngram TEXT)`);\n for (const ddl of CONTENT_FTS_DDL) await conn.exec(ddl);\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 F32_BLOB(STORE_DIMS) instead of the embed feature's engine-neutral BLOB DDL. `scale`\n // is kept unused, so the shared reconcile-time INSERT/DELETE names a column both stores have.\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 F32_BLOB(${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\nasync function close(handle: Database): Promise<void> {\n await handle.close();\n}\n\nasync function setDerivedBusyTimeout(_handle: Database, conn: Connection, ms: number): Promise<void> {\n await conn.exec(`PRAGMA busy_timeout = ${ms}`);\n}\n\nasync function connect(dbPath: string, _cfg: ResolvedConfig): Promise<{ handle: Database; conn: Connection }> {\n // Dynamic, not a top-level import: a sqlite or duckdb tree must never attempt to resolve this\n // optional dependency until a turso tree is actually opened. Installed on first use if missing.\n let turso: Awaited<ReturnType<typeof tursoApi>>;\n try {\n turso = await tursoApi();\n } catch (err) {\n if (err instanceof SenseError) throw err;\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"turso\" needs the ${TURSO_PACKAGE} package (${(err as Error).message})`);\n }\n\n let db: Database;\n try {\n // Floored at the same 30s sqlite opens with. `timeout` is connect-time only in this client;\n // the derived value is set via runtime PRAGMA below. `index_method` is required for ensureSchema()'s FTS indexes (T1).\n db = await turso.connect(dbPath, { timeout: 30_000, experimental: ['index_method'] });\n } catch (err) {\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"turso\" failed to open ${dbPath}: ${(err as Error).message}`);\n }\n return { handle: db, conn: createConnection(db) };\n}\n\n// This store's dialect (types.ts's OpenDialect) for the shared orchestration in store/open.ts.\nexport const tursoOpenDialect: OpenDialect<Database> = {\n filename: DB_FILENAME,\n schemaVersion: SCHEMA_VERSION,\n reconcileDialect: tursoDialect,\n connect,\n close,\n ensureSchema,\n setDerivedBusyTimeout,\n createStore: (handle, conn) => createStore(handle, conn),\n};\n\nexport async function openTurso(cfg: ResolvedConfig): Promise<OpenResult> {\n return openWithDialect(cfg, tursoOpenDialect);\n}\n"],"names":["featureSignature","STORE_DIMS","SenseError","activeFeatures","FEATURES","openWithDialect","getMeta","setMeta","createConnection","TURSO_PACKAGE","tursoApi","CONTENT_FTS_DDL","tursoDialect","createStore","DB_FILENAME","SCHEMA_VERSION","ensureSchema","_handle","conn","cfg","exec","ddl","feature","name","schema","close","handle","setDerivedBusyTimeout","ms","connect","dbPath","_cfg","turso","err","message","db","timeout","experimental","tursoOpenDialect","filename","schemaVersion","reconcileDialect","openTurso"],"mappings":"AAEA,SAASA,gBAAgB,QAAQ,wBAAwB;AACzD,SAASC,UAAU,QAAQ,uBAAuB;AAClD,SAASC,UAAU,QAAQ,kBAAkB;AAC7C,SAASC,cAAc,EAAEC,QAAQ,QAAQ,0BAA0B;AAEnE,SAASC,eAAe,QAAQ,aAAa;AAC7C,SAASC,OAAO,EAAEC,OAAO,QAAQ,eAAe;AAEhD,SAASC,gBAAgB,QAAQ,kBAAkB;AACnD,SAASC,aAAa,EAAEC,QAAQ,QAAQ,cAAc;AACtD,SAASC,eAAe,EAAEC,YAAY,QAAQ,iBAAiB;AAC/D,SAASC,WAAW,QAAQ,aAAa;AAEzC,OAAO,MAAMC,cAAc,iBAAiB;AAC5C,wFAAwF;AACxF,mGAAmG;AACnG,OAAO,MAAMC,iBAAiB,IAAI;AAIlC,4FAA4F;AAC5F,mFAAmF;AACnF,eAAeC,aAAaC,OAAiB,EAAEC,IAAgB,EAAEC,GAAW;IAC1E,MAAMD,KAAKE,IAAI,CAAC,CAAC,oIAAoI,CAAC;IACtJ,MAAMF,KAAKE,IAAI,CAAC,CAAC,wJAAwJ,CAAC;IAC1K,KAAK,MAAMC,OAAOV,gBAAiB,MAAMO,KAAKE,IAAI,CAACC;IACnD,MAAMH,KAAKE,IAAI,CAAC,CAAC,gGAAgG,CAAC;IAClH,MAAMF,KAAKE,IAAI,CAAC;IAChB,KAAK,MAAME,WAAWnB,eAAegB,KAAM;QACzC,8FAA8F;QAC9F,8FAA8F;QAC9F,IAAIG,QAAQC,IAAI,KAAK,SAAS;YAC5B,MAAML,KAAKE,IAAI,CAAC,CAAC,qIAAqI,EAAEnB,WAAW,+BAA+B,CAAC;YACnM;QACF;QACA,MAAMqB,QAAQE,MAAM,CAACN;IACvB;IACA,IAAI,AAAC,MAAMZ,QAAQY,MAAM,sBAAuB,MAAM,MAAMX,QAAQW,MAAM,kBAAkBH;IAC5F,IAAI,AAAC,MAAMT,QAAQY,MAAM,gBAAiB,MAAM,MAAMX,QAAQW,MAAM,YAAYlB,iBAAiBmB,KAAKf;AACxG;AAEA,eAAeqB,MAAMC,MAAgB;IACnC,MAAMA,OAAOD,KAAK;AACpB;AAEA,eAAeE,sBAAsBV,OAAiB,EAAEC,IAAgB,EAAEU,EAAU;IAClF,MAAMV,KAAKE,IAAI,CAAC,CAAC,sBAAsB,EAAEQ,IAAI;AAC/C;AAEA,eAAeC,QAAQC,MAAc,EAAEC,IAAoB;IACzD,8FAA8F;IAC9F,gGAAgG;IAChG,IAAIC;IACJ,IAAI;QACFA,QAAQ,MAAMtB;IAChB,EAAE,OAAOuB,KAAK;QACZ,IAAIA,eAAe/B,YAAY,MAAM+B;QACrC,MAAM,IAAI/B,WAAW,4BAA4B,CAAC,wBAAwB,EAAEO,cAAc,UAAU,EAAE,AAACwB,IAAcC,OAAO,CAAC,CAAC,CAAC;IACjI;IAEA,IAAIC;IACJ,IAAI;QACF,4FAA4F;QAC5F,uHAAuH;QACvHA,KAAK,MAAMH,MAAMH,OAAO,CAACC,QAAQ;YAAEM,SAAS;YAAQC,cAAc;gBAAC;aAAe;QAAC;IACrF,EAAE,OAAOJ,KAAK;QACZ,MAAM,IAAI/B,WAAW,4BAA4B,CAAC,6BAA6B,EAAE4B,OAAO,EAAE,EAAE,AAACG,IAAcC,OAAO,EAAE;IACtH;IACA,OAAO;QAAER,QAAQS;QAAIjB,MAAMV,iBAAiB2B;IAAI;AAClD;AAEA,+FAA+F;AAC/F,OAAO,MAAMG,mBAA0C;IACrDC,UAAUzB;IACV0B,eAAezB;IACf0B,kBAAkB7B;IAClBiB;IACAJ;IACAT;IACAW;IACAd,aAAa,CAACa,QAAQR,OAASL,YAAYa,QAAQR;AACrD,EAAE;AAEF,OAAO,eAAewB,UAAUvB,GAAmB;IACjD,OAAOd,gBAAgBc,KAAKmB;AAC9B"}
|
|
@@ -1,6 +1,5 @@
|
|
|
1
|
-
import type {
|
|
2
|
-
|
|
3
|
-
export declare
|
|
4
|
-
|
|
5
|
-
|
|
6
|
-
}>;
|
|
1
|
+
import type { ReconcileDialect } from '../types.js';
|
|
2
|
+
export declare const CONTENT_FTS_DDL: readonly ["CREATE INDEX IF NOT EXISTS content_fts ON content USING fts (title, summary, text) WITH (weights = 'title=10.0,summary=5.0,text=1.0')", "CREATE INDEX IF NOT EXISTS content_fts_ngram ON content USING fts (title_ngram, summary_ngram, text_ngram) WITH (tokenizer='ngram', weights='title_ngram=10.0,summary_ngram=5.0,text_ngram=1.0')"];
|
|
3
|
+
export declare const CONTENT_FTS_NAMES: readonly ["content_fts", "content_fts_ngram"];
|
|
4
|
+
export declare const FTS_REBUILD_THRESHOLD = 250;
|
|
5
|
+
export declare const tursoDialect: ReconcileDialect;
|
|
@@ -1,21 +1,21 @@
|
|
|
1
1
|
import { SenseError } from '../../errors.js';
|
|
2
|
-
import { activeFeatures } from '../../features/index.js';
|
|
3
|
-
import { progress } from '../../output/progress.js';
|
|
4
|
-
import { listFiles, RESERVED_COLUMNS } from '../../scan/index.js';
|
|
5
|
-
import { reparseFiles } from '../../scan/reparse.js';
|
|
6
2
|
import { hasUnspacedRun } from '../../text/segment.js';
|
|
7
|
-
import {
|
|
8
|
-
import { BEGIN_WRITE
|
|
9
|
-
|
|
10
|
-
//
|
|
11
|
-
//
|
|
12
|
-
|
|
13
|
-
|
|
14
|
-
|
|
15
|
-
|
|
16
|
-
'
|
|
17
|
-
'
|
|
18
|
-
]
|
|
3
|
+
import { recordReconcileDuration } from '../shared.js';
|
|
4
|
+
import { BEGIN_WRITE } from '../transaction.js';
|
|
5
|
+
// This store's dialect (types.ts's ReconcileDialect) for the shared orchestration in
|
|
6
|
+
// store/reconcile.ts. content is a plain table with "_ngram" sidecars for lexical.ts.
|
|
7
|
+
// The ngram index is scoped to disjoint "_ngram" sidecar columns: a second index over the same
|
|
8
|
+
// columns makes a bare substring match a whole word, defeating the prefix-query rejection.
|
|
9
|
+
// The two FTS indexes, named here (not open.ts) because reconcile drops and rebuilds them around
|
|
10
|
+
// a bulk load: Tantivy maintains them per inserted row, which is quadratic in what is already indexed.
|
|
11
|
+
export const CONTENT_FTS_DDL = [
|
|
12
|
+
`CREATE INDEX IF NOT EXISTS content_fts ON content USING fts (title, summary, text) WITH (weights = 'title=10.0,summary=5.0,text=1.0')`,
|
|
13
|
+
`CREATE INDEX IF NOT EXISTS content_fts_ngram ON content USING fts (title_ngram, summary_ngram, text_ngram) WITH (tokenizer='ngram', weights='title_ngram=10.0,summary_ngram=5.0,text_ngram=1.0')`
|
|
14
|
+
];
|
|
15
|
+
export const CONTENT_FTS_NAMES = [
|
|
16
|
+
'content_fts',
|
|
17
|
+
'content_fts_ngram'
|
|
18
|
+
];
|
|
19
19
|
// '' when the field has no unspaced-script run (the common case), so the ngram index carries
|
|
20
20
|
// nothing for it -- same "pay for nothing when absent" shape as sqlite's segmentField sidecars.
|
|
21
21
|
function ngramSidecar(text) {
|
|
@@ -24,8 +24,9 @@ function ngramSidecar(text) {
|
|
|
24
24
|
// Not a compile-time cap on ALTER TABLE ADD COLUMN (spike-measured: turso accepts 10,000 with
|
|
25
25
|
// no error); the real fence is a SELECT projecting more than this many result columns, which fails to prepare.
|
|
26
26
|
const MAX_FRONTMATTER_COLUMNS = 2000;
|
|
27
|
-
// Changed files above which rebuilding the FTS index beats maintaining it per row
|
|
28
|
-
|
|
27
|
+
// Changed files above which rebuilding the FTS index beats maintaining it per row. Measured
|
|
28
|
+
// 2026-08-30; the derivation and its bounds are pinned in this file's spec.
|
|
29
|
+
export const FTS_REBUILD_THRESHOLD = 250;
|
|
29
30
|
// No rowid coupling (unlike sqlite's FTS5 content): `path` is content's own primary key.
|
|
30
31
|
const INSERT_CONTENT_SQL = `INSERT INTO content ("path", title, summary, text, title_ngram, summary_ngram, text_ngram) VALUES (?, ?, ?, ?, ?, ?, ?)`;
|
|
31
32
|
function contentRow(doc) {
|
|
@@ -40,133 +41,28 @@ function contentRow(doc) {
|
|
|
40
41
|
ngramSidecar(text)
|
|
41
42
|
];
|
|
42
43
|
}
|
|
43
|
-
|
|
44
|
-
const
|
|
45
|
-
|
|
46
|
-
|
|
47
|
-
const
|
|
48
|
-
const
|
|
49
|
-
|
|
50
|
-
|
|
44
|
+
async function reconcileContent(conn, touched, docs, delta) {
|
|
45
|
+
const churn = delta.reparsed.length + delta.vanished.length;
|
|
46
|
+
// Tantivy indexes per inserted row at a cost that grows with the batch, so a large insert is
|
|
47
|
+
// superlinear and a rebuild wins past the threshold.
|
|
48
|
+
const bulk = delta.files.length === 0 || churn > FTS_REBUILD_THRESHOLD;
|
|
49
|
+
if (bulk) for (const name of CONTENT_FTS_NAMES)await conn.exec(`DROP INDEX IF EXISTS ${name}`);
|
|
50
|
+
// content is a plain table keyed by its own path (no rowid subquery, unlike sqlite's FTS5
|
|
51
|
+
// content), so vanished and reparsed docs delete in one pass.
|
|
52
|
+
if (touched.length > 0) await conn.runBatch('DELETE FROM content WHERE "path" = ?', touched.map((p)=>[
|
|
53
|
+
p
|
|
51
54
|
]));
|
|
52
|
-
|
|
53
|
-
const
|
|
54
|
-
const row = existing.get(f.relPath);
|
|
55
|
-
return !row || row._mtime !== f.mtimeMs || row._size !== f.size;
|
|
56
|
-
});
|
|
57
|
-
if (vanished.length === 0 && toReparse.length === 0) return {
|
|
58
|
-
parsed: 0,
|
|
59
|
-
warnings: []
|
|
60
|
-
};
|
|
61
|
-
const features = activeFeatures(cfg);
|
|
62
|
-
const seenColumns = await getColumns(conn);
|
|
63
|
-
// Bulk reparses (a sync, a cold build) are the long silences a query can hit; short
|
|
64
|
-
// reconciles stay silent (progress() has a threshold).
|
|
65
|
-
const report = progress('reparsing files', toReparse.length);
|
|
66
|
-
const { docs: parsedDocs, warnings, newColumns } = await reparseFiles(toReparse, features, cfg, seenColumns, report.tick);
|
|
67
|
-
report.finish();
|
|
68
|
-
for (const col of newColumns)seenColumns.add(col);
|
|
69
|
-
const allColumns = [
|
|
70
|
-
...seenColumns
|
|
71
|
-
];
|
|
72
|
-
if (allColumns.length > MAX_FRONTMATTER_COLUMNS) {
|
|
73
|
-
throw new SenseError('COLUMN_LIMIT', `frontmatter would need ${allColumns.length} columns, crossing turso's SELECT result-set column limit (${MAX_FRONTMATTER_COLUMNS}; ALTER TABLE ADD COLUMN itself accepts far more, but a query projecting past this many columns fails to prepare). Narrow the presets' include globs so fewer/other files are indexed, or fix whatever is generating unbounded frontmatter keys.`);
|
|
74
|
-
}
|
|
75
|
-
// Columns the frontmatter upsert actually writes: core + parsed frontmatter keys, never a
|
|
76
|
-
// feature-owned reserved column (see CORE_FRONTMATTER_COLUMNS above).
|
|
77
|
-
const writableColumns = allColumns.filter((c)=>CORE_FRONTMATTER_COLUMNS.has(c) || !RESERVED_COLUMNS.has(c));
|
|
78
|
-
// ON CONFLICT UPDATE, rather than the delete-and-insert `OR` clause, keeps the row's identity stable across reparses.
|
|
79
|
-
const insertSql = `INSERT INTO frontmatter (${writableColumns.map(quoteIdent).join(', ')}) VALUES (${writableColumns.map(()=>'?').join(', ')}) ON CONFLICT("path") DO UPDATE SET ${writableColumns.filter((c)=>c !== 'path').map((c)=>`${quoteIdent(c)} = excluded.${quoteIdent(c)}`).join(', ')}`;
|
|
80
|
-
const added = toReparse.filter((f)=>!existing.has(f.relPath)).map((f)=>f.relPath);
|
|
81
|
-
const delta = {
|
|
82
|
-
files,
|
|
83
|
-
reparsed: parsedDocs.map((d)=>d.relPath),
|
|
84
|
-
added,
|
|
85
|
-
vanished
|
|
86
|
-
};
|
|
87
|
-
const addedSet = new Set(added);
|
|
88
|
-
const reparsedExisting = parsedDocs.map((d)=>d.relPath).filter((p)=>!addedSet.has(p));
|
|
89
|
-
const txStart = Date.now();
|
|
90
|
-
await withTransaction(conn, async ()=>{
|
|
91
|
-
var _feature_remove, _feature_afterReconcile;
|
|
92
|
-
for (const col of newColumns)await conn.exec(`ALTER TABLE frontmatter ADD COLUMN ${quoteIdent(col)}`);
|
|
93
|
-
if (parsedDocs.length > 0) {
|
|
94
|
-
const rows = parsedDocs.map((doc)=>writableColumns.map((col)=>{
|
|
95
|
-
var _doc_data_col;
|
|
96
|
-
if (col === 'path') return doc.relPath;
|
|
97
|
-
if (col === '_mtime') return doc.mtimeMs;
|
|
98
|
-
if (col === '_ctime') return doc.ctimeMs;
|
|
99
|
-
if (col === '_size') return doc.size;
|
|
100
|
-
// Written per parse, unlike _rank, which a feature pass owns and the upsert skips.
|
|
101
|
-
if (col === '_parse_error') return doc.parseError;
|
|
102
|
-
return (_doc_data_col = doc.data[col]) !== null && _doc_data_col !== void 0 ? _doc_data_col : null;
|
|
103
|
-
}));
|
|
104
|
-
await conn.runBatch(insertSql, rows);
|
|
105
|
-
}
|
|
106
|
-
// Tantivy maintains the FTS index per inserted row and the cost grows with the batch, so a
|
|
107
|
-
// large insert is superlinear: 6,566 notes took 945s with the index live, 3.3s built after.
|
|
108
|
-
// Measured crossover against a full rebuild, in changed files: ~250 on a 6,566-note tree
|
|
109
|
-
// (100 -> 895ms, 300 -> 4.1s, 1,300 -> 45.5s, rebuild 3.35s), ~295 on 13,132 (250 -> 5.2s,
|
|
110
|
-
// 400 -> 10.0s, rebuild 6.7s). Doubling the corpus moved it 250 to 295, so a constant fits
|
|
111
|
-
// where a ratio does not: 20% would allow 1,313 files here, costing 45s to save 3.3s.
|
|
112
|
-
// Erring low is the cheap direction, since a needless rebuild is linear and predictable.
|
|
113
|
-
const churn = delta.reparsed.length + delta.vanished.length;
|
|
114
|
-
const bulk = delta.files.length === 0 || churn > FTS_REBUILD_THRESHOLD;
|
|
115
|
-
if (bulk) for (const name of CONTENT_FTS_NAMES)await conn.exec(`DROP INDEX IF EXISTS ${name}`);
|
|
116
|
-
// content is a plain table keyed by its own path (no rowid subquery, unlike sqlite's FTS5
|
|
117
|
-
// content), so vanished and reparsed docs delete in one pass.
|
|
118
|
-
const contentTouched = [
|
|
119
|
-
...vanished,
|
|
120
|
-
...reparsedExisting
|
|
121
|
-
];
|
|
122
|
-
if (contentTouched.length > 0) await conn.runBatch('DELETE FROM content WHERE "path" = ?', contentTouched.map((p)=>[
|
|
123
|
-
p
|
|
124
|
-
]));
|
|
125
|
-
if (parsedDocs.length > 0) await conn.runBatch(INSERT_CONTENT_SQL, parsedDocs.map(contentRow));
|
|
126
|
-
if (bulk) for (const ddl of CONTENT_FTS_DDL)await conn.exec(ddl);
|
|
127
|
-
if (vanished.length > 0) await conn.runBatch('DELETE FROM frontmatter WHERE "path" = ?', vanished.map((p)=>[
|
|
128
|
-
p
|
|
129
|
-
]));
|
|
130
|
-
// A preset edit forces a full rebuild, so an unchanged doc's coverage is already correct;
|
|
131
|
-
// new docs have nothing to clear, which keeps cold builds linear.
|
|
132
|
-
const presetTouched = [
|
|
133
|
-
...vanished,
|
|
134
|
-
...reparsedExisting
|
|
135
|
-
];
|
|
136
|
-
if (presetTouched.length > 0) await conn.runBatch('DELETE FROM preset_files WHERE "path" = ?', presetTouched.map((p)=>[
|
|
137
|
-
p
|
|
138
|
-
]));
|
|
139
|
-
const presetRows = [];
|
|
140
|
-
for (const doc of parsedDocs)for (const presetName of doc.presets)presetRows.push([
|
|
141
|
-
doc.relPath,
|
|
142
|
-
presetName
|
|
143
|
-
]);
|
|
144
|
-
if (presetRows.length > 0) await conn.runBatch('INSERT INTO preset_files ("path", preset) VALUES (?, ?)', presetRows);
|
|
145
|
-
const removedPaths = [
|
|
146
|
-
...vanished,
|
|
147
|
-
...reparsedExisting
|
|
148
|
-
];
|
|
149
|
-
if (removedPaths.length > 0) for (const feature of features)await ((_feature_remove = feature.remove) === null || _feature_remove === void 0 ? void 0 : _feature_remove.call(feature, conn, removedPaths, delta));
|
|
150
|
-
for (const feature of features){
|
|
151
|
-
var _feature_store;
|
|
152
|
-
const docsForFeature = parsedDocs.map((doc)=>({
|
|
153
|
-
path: doc.relPath,
|
|
154
|
-
extracted: doc.extracted[feature.name]
|
|
155
|
-
}));
|
|
156
|
-
await ((_feature_store = feature.store) === null || _feature_store === void 0 ? void 0 : _feature_store.call(feature, conn, docsForFeature, delta));
|
|
157
|
-
}
|
|
158
|
-
for (const feature of features)await ((_feature_afterReconcile = feature.afterReconcile) === null || _feature_afterReconcile === void 0 ? void 0 : _feature_afterReconcile.call(feature, conn, delta));
|
|
159
|
-
}, BEGIN_WRITE);
|
|
160
|
-
// Reconcile's own write-transaction duration, for open()'s derived busy_timeout: the observed
|
|
161
|
-
// max is what the next open bounds its wait against.
|
|
162
|
-
const durationMs = Date.now() - txStart;
|
|
163
|
-
const prevRaw = await getMeta(conn, 'reconcile_max_ms');
|
|
164
|
-
// -1, not 0, so a genuinely 0ms first reconcile (sub-millisecond, common on a tiny tree)
|
|
165
|
-
// gets recorded rather than losing to the "nothing recorded yet" default.
|
|
166
|
-
const prevMax = prevRaw === null ? -1 : Number(prevRaw);
|
|
167
|
-
if (durationMs > prevMax) await setMeta(conn, 'reconcile_max_ms', String(durationMs));
|
|
168
|
-
return {
|
|
169
|
-
parsed: parsedDocs.length,
|
|
170
|
-
warnings
|
|
171
|
-
};
|
|
55
|
+
if (docs.length > 0) await conn.runBatch(INSERT_CONTENT_SQL, docs.map(contentRow));
|
|
56
|
+
if (bulk) for (const ddl of CONTENT_FTS_DDL)await conn.exec(ddl);
|
|
172
57
|
}
|
|
58
|
+
export const tursoDialect = {
|
|
59
|
+
beginMode: ()=>BEGIN_WRITE,
|
|
60
|
+
checkColumnLimit (count) {
|
|
61
|
+
if (count > MAX_FRONTMATTER_COLUMNS) {
|
|
62
|
+
throw new SenseError('COLUMN_LIMIT', `frontmatter would need ${count} columns, crossing turso's SELECT result-set column limit (${MAX_FRONTMATTER_COLUMNS}; ALTER TABLE ADD COLUMN itself accepts far more, but a query projecting past this many columns fails to prepare). Narrow the presets' include globs so fewer/other files are indexed, or fix whatever is generating unbounded frontmatter keys.`);
|
|
63
|
+
}
|
|
64
|
+
},
|
|
65
|
+
columnTypeSuffix: ()=>'',
|
|
66
|
+
reconcileContent,
|
|
67
|
+
recordDuration: recordReconcileDuration
|
|
68
|
+
};
|