sensemaking 0.19.2 → 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.
Files changed (140) hide show
  1. package/dist/cjs/chunk/extract.d.cts +1 -0
  2. package/dist/cjs/chunk/extract.d.ts +1 -0
  3. package/dist/cjs/chunk/extract.js +23 -5
  4. package/dist/cjs/chunk/extract.js.map +1 -1
  5. package/dist/cjs/chunk/index.d.cts +1 -1
  6. package/dist/cjs/chunk/index.d.ts +1 -1
  7. package/dist/cjs/chunk/index.js +3 -0
  8. package/dist/cjs/chunk/index.js.map +1 -1
  9. package/dist/cjs/scan/pool.d.cts +15 -0
  10. package/dist/cjs/scan/pool.d.ts +15 -0
  11. package/dist/cjs/scan/pool.js +263 -0
  12. package/dist/cjs/scan/pool.js.map +1 -0
  13. package/dist/cjs/scan/reparse.d.cts +2 -0
  14. package/dist/cjs/scan/reparse.d.ts +2 -0
  15. package/dist/cjs/scan/reparse.js +15 -76
  16. package/dist/cjs/scan/reparse.js.map +1 -1
  17. package/dist/cjs/store/builder.d.cts +11 -0
  18. package/dist/cjs/store/builder.d.ts +11 -0
  19. package/dist/cjs/store/builder.js +27 -0
  20. package/dist/cjs/store/builder.js.map +1 -0
  21. package/dist/cjs/store/duckdb/lexical.d.cts +1 -0
  22. package/dist/cjs/store/duckdb/lexical.d.ts +1 -0
  23. package/dist/cjs/store/duckdb/lexical.js +21 -5
  24. package/dist/cjs/store/duckdb/lexical.js.map +1 -1
  25. package/dist/cjs/store/duckdb/open.d.cts +9 -8
  26. package/dist/cjs/store/duckdb/open.d.ts +9 -8
  27. package/dist/cjs/store/duckdb/open.js +70 -122
  28. package/dist/cjs/store/duckdb/open.js.map +1 -1
  29. package/dist/cjs/store/duckdb/reconcile.js +4 -0
  30. package/dist/cjs/store/duckdb/reconcile.js.map +1 -1
  31. package/dist/cjs/store/duckdb/store.d.cts +1 -2
  32. package/dist/cjs/store/duckdb/store.d.ts +1 -2
  33. package/dist/cjs/store/duckdb/store.js +1 -26
  34. package/dist/cjs/store/duckdb/store.js.map +1 -1
  35. package/dist/cjs/store/index.d.cts +8 -3
  36. package/dist/cjs/store/index.d.ts +8 -3
  37. package/dist/cjs/store/index.js +36 -7
  38. package/dist/cjs/store/index.js.map +1 -1
  39. package/dist/cjs/store/open.d.cts +15 -0
  40. package/dist/cjs/store/open.d.ts +15 -0
  41. package/dist/cjs/store/open.js +516 -0
  42. package/dist/cjs/store/open.js.map +1 -0
  43. package/dist/cjs/store/reconcile.d.cts +2 -1
  44. package/dist/cjs/store/reconcile.d.ts +2 -1
  45. package/dist/cjs/store/reconcile.js +4 -2
  46. package/dist/cjs/store/reconcile.js.map +1 -1
  47. package/dist/cjs/store/{sqlite/signature.d.cts → signature.d.cts} +0 -3
  48. package/dist/cjs/store/{sqlite/signature.d.ts → signature.d.ts} +0 -3
  49. package/dist/cjs/store/signature.js +162 -0
  50. package/dist/cjs/store/signature.js.map +1 -0
  51. package/dist/cjs/store/sqlite/open.d.cts +9 -9
  52. package/dist/cjs/store/sqlite/open.d.ts +9 -9
  53. package/dist/cjs/store/sqlite/open.js +189 -201
  54. package/dist/cjs/store/sqlite/open.js.map +1 -1
  55. package/dist/cjs/store/sqlite/store.d.cts +1 -1
  56. package/dist/cjs/store/sqlite/store.d.ts +1 -1
  57. package/dist/cjs/store/sqlite/store.js +1 -13
  58. package/dist/cjs/store/sqlite/store.js.map +1 -1
  59. package/dist/cjs/store/turso/open.d.cts +6 -11
  60. package/dist/cjs/store/turso/open.d.ts +6 -11
  61. package/dist/cjs/store/turso/open.js +79 -171
  62. package/dist/cjs/store/turso/open.js.map +1 -1
  63. package/dist/cjs/store/turso/reconcile.d.cts +2 -0
  64. package/dist/cjs/store/turso/reconcile.d.ts +2 -0
  65. package/dist/cjs/store/turso/reconcile.js +16 -5
  66. package/dist/cjs/store/turso/reconcile.js.map +1 -1
  67. package/dist/cjs/store/turso/store.d.cts +1 -2
  68. package/dist/cjs/store/turso/store.d.ts +1 -2
  69. package/dist/cjs/store/turso/store.js +1 -13
  70. package/dist/cjs/store/turso/store.js.map +1 -1
  71. package/dist/cjs/store/types.d.cts +17 -5
  72. package/dist/cjs/store/types.d.ts +17 -5
  73. package/dist/cjs/store/types.js.map +1 -1
  74. package/dist/cjs/text/strip.js +3 -6
  75. package/dist/cjs/text/strip.js.map +1 -1
  76. package/dist/cjs/watch.js +28 -16
  77. package/dist/cjs/watch.js.map +1 -1
  78. package/dist/esm/chunk/extract.d.ts +1 -0
  79. package/dist/esm/chunk/extract.js +10 -2
  80. package/dist/esm/chunk/extract.js.map +1 -1
  81. package/dist/esm/chunk/index.d.ts +1 -1
  82. package/dist/esm/chunk/index.js +1 -1
  83. package/dist/esm/chunk/index.js.map +1 -1
  84. package/dist/esm/scan/pool.d.ts +15 -0
  85. package/dist/esm/scan/pool.js +77 -0
  86. package/dist/esm/scan/pool.js.map +1 -0
  87. package/dist/esm/scan/reparse.d.ts +2 -0
  88. package/dist/esm/scan/reparse.js +11 -58
  89. package/dist/esm/scan/reparse.js.map +1 -1
  90. package/dist/esm/store/builder.d.ts +11 -0
  91. package/dist/esm/store/builder.js +14 -0
  92. package/dist/esm/store/builder.js.map +1 -0
  93. package/dist/esm/store/duckdb/lexical.d.ts +1 -0
  94. package/dist/esm/store/duckdb/lexical.js +22 -9
  95. package/dist/esm/store/duckdb/lexical.js.map +1 -1
  96. package/dist/esm/store/duckdb/open.d.ts +9 -8
  97. package/dist/esm/store/duckdb/open.js +31 -57
  98. package/dist/esm/store/duckdb/open.js.map +1 -1
  99. package/dist/esm/store/duckdb/reconcile.js +4 -0
  100. package/dist/esm/store/duckdb/reconcile.js.map +1 -1
  101. package/dist/esm/store/duckdb/store.d.ts +1 -2
  102. package/dist/esm/store/duckdb/store.js +2 -11
  103. package/dist/esm/store/duckdb/store.js.map +1 -1
  104. package/dist/esm/store/index.d.ts +8 -3
  105. package/dist/esm/store/index.js +22 -7
  106. package/dist/esm/store/index.js.map +1 -1
  107. package/dist/esm/store/open.d.ts +15 -0
  108. package/dist/esm/store/open.js +137 -0
  109. package/dist/esm/store/open.js.map +1 -0
  110. package/dist/esm/store/reconcile.d.ts +2 -1
  111. package/dist/esm/store/reconcile.js +4 -2
  112. package/dist/esm/store/reconcile.js.map +1 -1
  113. package/dist/esm/store/{sqlite/signature.d.ts → signature.d.ts} +0 -3
  114. package/dist/esm/store/{sqlite/signature.js → signature.js} +3 -26
  115. package/dist/esm/store/signature.js.map +1 -0
  116. package/dist/esm/store/sqlite/open.d.ts +9 -9
  117. package/dist/esm/store/sqlite/open.js +81 -113
  118. package/dist/esm/store/sqlite/open.js.map +1 -1
  119. package/dist/esm/store/sqlite/store.d.ts +1 -1
  120. package/dist/esm/store/sqlite/store.js +2 -7
  121. package/dist/esm/store/sqlite/store.js.map +1 -1
  122. package/dist/esm/store/turso/open.d.ts +6 -11
  123. package/dist/esm/store/turso/open.js +30 -78
  124. package/dist/esm/store/turso/open.js.map +1 -1
  125. package/dist/esm/store/turso/reconcile.d.ts +2 -0
  126. package/dist/esm/store/turso/reconcile.js +12 -1
  127. package/dist/esm/store/turso/reconcile.js.map +1 -1
  128. package/dist/esm/store/turso/store.d.ts +1 -2
  129. package/dist/esm/store/turso/store.js +2 -7
  130. package/dist/esm/store/turso/store.js.map +1 -1
  131. package/dist/esm/store/types.d.ts +17 -5
  132. package/dist/esm/store/types.js.map +1 -1
  133. package/dist/esm/text/strip.js +3 -4
  134. package/dist/esm/text/strip.js.map +1 -1
  135. package/dist/esm/watch.js +5 -3
  136. package/dist/esm/watch.js.map +1 -1
  137. package/package.json +2 -2
  138. package/dist/cjs/store/sqlite/signature.js +0 -380
  139. package/dist/cjs/store/sqlite/signature.js.map +0 -1
  140. package/dist/esm/store/sqlite/signature.js.map +0 -1
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/scan/reparse.ts"],"sourcesContent":["import { createRequire } from 'node:module';\nimport { availableParallelism } from 'node:os';\nimport { dirname, join } from 'node:path';\n// Type-only: erased at build, but keeps depcheck's usage check satisfied for the tier-2\n// `_require` below (see coding-standards' deferral tiers).\nimport type * as TinypoolNS from 'tinypool';\nimport type { Config } from '../config/index.ts';\nimport type { Feature } from '../features/types.ts';\nimport type { ParseTask, ParseTaskResult, ParseWorkerData } from '../workers/parse.ts';\nimport type { ParsedDoc } from './index.ts';\nimport { parseFile } from './index.ts';\nimport type { FileStat } from './list.ts';\nimport { reviveError } from './worker-error.ts';\n\n// Tinypool is ESM-only; our floor (>=22.20) has native require(esm), so the tier-2 house deferral reaches it.\n// Deferred (not top-level) because most reconciles stay below DEFAULT_THRESHOLD and shouldn't pay for pool machinery they never use.\nconst _require = typeof require === 'undefined' ? createRequire(import.meta.url) : require;\n\n// Store-agnostic per-file parse pass shared by every store's reconcile(). Index-preserving:\n// one pass over `files`, pushed in order.\nexport interface ReparseResult {\n docs: ParsedDoc[];\n warnings: string[];\n // Frontmatter keys not in `knownColumns`, first-seen order -- the order callers ALTER TABLE ADD COLUMN in.\n newColumns: string[];\n}\n\nexport interface ReparseOptions {\n // Overrides DEFAULT_THRESHOLD below, so tests can force either dispatch mode without\n // needing thousands of fixture files. Internal: no caller in src/ passes it.\n threshold?: number;\n // Overrides DEFAULT_MAX_WORKERS below.\n maxWorkers?: number;\n}\n\n// Measured 2026-08-29 on a 14-core Apple M4 Pro: serial and pooled cross between 160-180 files\n// (pool fixed cost ~85-100ms), so 200 keeps margin. 8 workers ties 10 in a 6,566-file sweep; 12+ trends worse.\nconst DEFAULT_THRESHOLD = 200;\nconst DEFAULT_MAX_WORKERS = 8;\n\n// Worker MUST load from dist/cjs/: a worker_threads thread is a fresh realm with no inherited\n// TS loader hook, so a source path cannot work. Resolved on first pooled dispatch, not at import.\nlet workerFile: string | undefined;\nfunction resolveWorkerFile(): string {\n if (workerFile) return workerFile;\n const load = createRequire(import.meta.url);\n for (const rel of ['..', '../..', '../../..']) {\n try {\n if ((load(`${rel}/package.json`) as { name?: string }).name === 'sensemaking') {\n workerFile = join(dirname(load.resolve(`${rel}/package.json`)), 'dist', 'cjs', 'workers', 'parse.js');\n return workerFile;\n }\n } catch {}\n }\n throw new Error('cannot locate the sensemaking package root, so the parse worker cannot be found; run npm run build');\n}\n\ntype FileResult = { doc: ParsedDoc; warnings: string[] };\n\n// Per-file feature filtering, shared by the serial loop (reads `features` from the caller) and\n// the worker path (can only pass `cfg` across the thread boundary, so derives `features` itself).\nexport function featuresForFile(features: Feature[], cfg: Config, file: FileStat): Feature[] {\n return features.filter((feature) => !feature.enabledForFile || feature.enabledForFile(cfg, file));\n}\n\nfunction reparseSerial(files: FileStat[], features: Feature[], cfg: Config, onParsed?: (done: number) => void): FileResult[] {\n const results: FileResult[] = [];\n let done = 0;\n for (const file of files) {\n const fileFeatures = featuresForFile(features, cfg, file);\n results.push(parseFile(file, fileFeatures, cfg));\n onParsed?.(++done);\n }\n return results;\n}\n\n// Never the tree: a worker task carries one FileStat and returns only what parseFile returns --\n// extracted text and feature values. Pool created once here, destroyed once dispatch finishes.\nasync function reparsePooled(files: FileStat[], features: Feature[], cfg: Config, onParsed: ((done: number) => void) | undefined, maxWorkers: number): Promise<FileResult[]> {\n const { Tinypool } = _require('tinypool') as typeof TinypoolNS;\n // cfg and the feature selection are constant for the dispatch, so they cross once per worker\n // as workerData. Features carry closures and cannot cross at all; only their names do.\n const workerData: ParseWorkerData = { cfg, featureNames: features.map((feature) => feature.name) };\n const pool = new Tinypool({ filename: resolveWorkerFile(), minThreads: maxWorkers, maxThreads: maxWorkers, workerData });\n let done = 0;\n try {\n // Promise.all over a mapped array is load-bearing: the resolved array keeps `files` order\n // regardless of task completion order, which first-seen column order depends on downstream.\n const results = await Promise.all(\n files.map(async (file): Promise<FileResult> => {\n const result = (await pool.run(file as ParseTask)) as ParseTaskResult;\n if (!result.ok) throw reviveError(result.error);\n onParsed?.(++done);\n return { doc: result.doc, warnings: result.warnings };\n })\n );\n await pool.destroy();\n return results;\n } catch (err) {\n // Tear the pool down before rethrowing, but a destroy failure must not mask the parse\n // failure that caused it -- not a `finally`, for that reason.\n await pool.destroy().catch(() => {});\n throw err;\n }\n}\n\n// `onParsed` receives the running count (1-based) after each file, mirroring a Progress.tick\n// call; pass one in to keep progress reporting working without this module owning a reporter.\nexport async function reparseFiles(files: FileStat[], features: Feature[], cfg: Config, knownColumns: ReadonlySet<string>, onParsed?: (done: number) => void, options: ReparseOptions = {}): Promise<ReparseResult> {\n const threshold = options.threshold ?? DEFAULT_THRESHOLD;\n const maxWorkers = options.maxWorkers ?? Math.min(DEFAULT_MAX_WORKERS, availableParallelism());\n const results = files.length >= threshold ? await reparsePooled(files, features, cfg, onParsed, maxWorkers) : reparseSerial(files, features, cfg, onParsed);\n\n const docs: ParsedDoc[] = [];\n const warnings: string[] = [];\n const newColumns: string[] = [];\n const seen = new Set(knownColumns);\n for (const { doc, warnings: fileWarnings } of results) {\n warnings.push(...fileWarnings);\n for (const key of Object.keys(doc.data)) {\n if (!seen.has(key)) {\n seen.add(key);\n newColumns.push(key);\n }\n }\n docs.push(doc);\n }\n\n return { docs, warnings, newColumns };\n}\n"],"names":["featuresForFile","reparseFiles","_require","require","createRequire","DEFAULT_THRESHOLD","DEFAULT_MAX_WORKERS","workerFile","resolveWorkerFile","load","rel","name","join","dirname","resolve","Error","features","cfg","file","filter","feature","enabledForFile","reparseSerial","files","onParsed","results","done","fileFeatures","push","parseFile","reparsePooled","maxWorkers","Tinypool","workerData","pool","err","featureNames","map","filename","minThreads","maxThreads","Promise","all","result","run","ok","reviveError","error","doc","warnings","destroy","catch","knownColumns","options","threshold","docs","newColumns","seen","fileWarnings","key","Math","min","availableParallelism","length","Set","Object","keys","data","has","add"],"mappings":";;;;;;;;;;;QA6DgBA;eAAAA;;QA+CMC;eAAAA;;;0BA5GQ;sBACO;wBACP;uBAQJ;6BAEE;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAE5B,8GAA8G;AAC9G,qIAAqI;AACrI,IAAMC,WAAW,OAAOC,YAAY,cAAcC,IAAAA,yBAAa,EAAC,uDAAmBD;AAmBnF,+FAA+F;AAC/F,+GAA+G;AAC/G,IAAME,oBAAoB;AAC1B,IAAMC,sBAAsB;AAE5B,8FAA8F;AAC9F,kGAAkG;AAClG,IAAIC;AACJ,SAASC;IACP,IAAID,YAAY,OAAOA;IACvB,IAAME,OAAOL,IAAAA,yBAAa,EAAC;IAC3B,gBAAkB,QAAA;QAAC;QAAM;QAAS;KAAW,OAA3B,mBAA6B;YAApCM,MAAO;QAChB,IAAI;YACF,IAAI,AAACD,KAAK,AAAC,GAAM,OAAJC,KAAI,kBAAsCC,IAAI,KAAK,eAAe;gBAC7EJ,aAAaK,IAAAA,cAAI,EAACC,IAAAA,iBAAO,EAACJ,KAAKK,OAAO,CAAC,AAAC,GAAM,OAAJJ,KAAI,oBAAkB,QAAQ,OAAO,WAAW;gBAC1F,OAAOH;YACT;QACF,EAAE,eAAM,CAAC;IACX;IACA,MAAM,IAAIQ,MAAM;AAClB;AAMO,SAASf,gBAAgBgB,QAAmB,EAAEC,GAAW,EAAEC,IAAc;IAC9E,OAAOF,SAASG,MAAM,CAAC,SAACC;eAAY,CAACA,QAAQC,cAAc,IAAID,QAAQC,cAAc,CAACJ,KAAKC;;AAC7F;AAEA,SAASI,cAAcC,KAAiB,EAAEP,QAAmB,EAAEC,GAAW,EAAEO,QAAiC;IAC3G,IAAMC,UAAwB,EAAE;IAChC,IAAIC,OAAO;QACN,kCAAA,2BAAA;;QAAL,QAAK,YAAcH,0BAAd,SAAA,6BAAA,QAAA,yBAAA,iCAAqB;YAArB,IAAML,OAAN;YACH,IAAMS,eAAe3B,gBAAgBgB,UAAUC,KAAKC;YACpDO,QAAQG,IAAI,CAACC,IAAAA,kBAAS,EAACX,MAAMS,cAAcV;YAC3CO,qBAAAA,+BAAAA,SAAW,EAAEE;QACf;;QAJK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAKL,OAAOD;AACT;AAEA,gGAAgG;AAChG,+FAA+F;AAC/F,SAAeK,cAAcP,KAAiB,EAAEP,QAAmB,EAAEC,GAAW,EAAEO,QAA8C,EAAEO,UAAkB;;YAC1IC,UAGFC,YACAC,MACFR,MAIID,SAUCU;;;;oBAnBDH,WAAa9B,SAAS,YAAtB8B;oBACR,6FAA6F;oBAC7F,uFAAuF;oBACjFC,aAA8B;wBAAEhB,KAAAA;wBAAKmB,cAAcpB,SAASqB,GAAG,CAAC,SAACjB;mCAAYA,QAAQT,IAAI;;oBAAE;oBAC3FuB,OAAO,IAAIF,SAAS;wBAAEM,UAAU9B;wBAAqB+B,YAAYR;wBAAYS,YAAYT;wBAAYE,YAAAA;oBAAW;oBAClHP,OAAO;;;;;;;;;oBAIO;;wBAAMe,QAAQC,GAAG,CAC/BnB,MAAMc,GAAG,CAAC,SAAOnB;;oCACTyB;;;;4CAAU;;gDAAMT,KAAKU,GAAG,CAAC1B;;;4CAAzByB,SAAU;4CAChB,IAAI,CAACA,OAAOE,EAAE,EAAE,MAAMC,IAAAA,0BAAW,EAACH,OAAOI,KAAK;4CAC9CvB,qBAAAA,+BAAAA,SAAW,EAAEE;4CACb;;gDAAO;oDAAEsB,KAAKL,OAAOK,GAAG;oDAAEC,UAAUN,OAAOM,QAAQ;gDAAC;;;;4BACtD;;;;oBANIxB,UAAU;oBAQhB;;wBAAMS,KAAKgB,OAAO;;;oBAAlB;oBACA;;wBAAOzB;;;oBACAU;oBACP,sFAAsF;oBACtF,8DAA8D;oBAC9D;;wBAAMD,KAAKgB,OAAO,GAAGC,KAAK,CAAC,YAAO;;;oBAAlC;oBACA,MAAMhB;;;;;;;IAEV;;AAIO,SAAelC;wCAAasB,KAAiB,EAAEP,QAAmB,EAAEC,GAAW,EAAEmC,YAAiC,EAAE5B,QAAiC;YAAE6B,SAC1IA,oBACCA,qBADbC,WACAvB,YACAN,eAEA8B,MACAN,UACAO,YACAC,MACD,2BAAA,mBAAA,gBAAA,WAAA,oBAAQT,KAAeU,cAC1BT,WACK,4BAAA,oBAAA,iBAAA,YAAA,QAAMU;;;;;oBAX+IN,UAAAA,oEAA0B,CAAC;oBACjLC,aAAYD,qBAAAA,QAAQC,SAAS,cAAjBD,gCAAAA,qBAAqBhD;oBACjC0B,cAAasB,sBAAAA,QAAQtB,UAAU,cAAlBsB,iCAAAA,sBAAsBO,KAAKC,GAAG,CAACvD,qBAAqBwD,IAAAA,4BAAoB;yBAC3EvC,CAAAA,MAAMwC,MAAM,IAAIT,SAAQ,GAAxB/B;;;;oBAA4B;;wBAAMO,cAAcP,OAAOP,UAAUC,KAAKO,UAAUO;;;2BAApD;;;;;;2BAAkET,cAAcC,OAAOP,UAAUC,KAAKO;;;oBAA5IC;oBAEA8B;oBACAN;oBACAO;oBACAC,OAAO,IAAIO,IAAIZ;oBAChB,kCAAA,2BAAA;;wBAAL,IAAK,YAAyC3B,8BAAzC,6BAAA,QAAA,yBAAA,iCAAkD;0CAAlD,aAAQuB,kBAAAA,KAAeU,2BAAVT;;4BAChBA,CAAAA,YAAAA,UAASrB,IAAI,OAAbqB,WAAc,qBAAGS;4BACZ,mCAAA,4BAAA;;gCAAL,IAAK,aAAaO,OAAOC,IAAI,CAAClB,IAAImB,IAAI,wBAAjC,8BAAA,SAAA,0BAAA,kCAAoC;oCAA9BR,MAAN;oCACH,IAAI,CAACF,KAAKW,GAAG,CAACT,MAAM;wCAClBF,KAAKY,GAAG,CAACV;wCACTH,WAAW5B,IAAI,CAAC+B;oCAClB;gCACF;;gCALK;gCAAA;;;yCAAA,8BAAA;wCAAA;;;wCAAA;8CAAA;;;;4BAMLJ,KAAK3B,IAAI,CAACoB;wBACZ;;wBATK;wBAAA;;;iCAAA,6BAAA;gCAAA;;;gCAAA;sCAAA;;;;oBAWL;;wBAAO;4BAAEO,MAAAA;4BAAMN,UAAAA;4BAAUO,YAAAA;wBAAW;;;;IACtC"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/scan/reparse.ts"],"sourcesContent":["import { availableParallelism } from 'node:os';\nimport type { Config } from '../config/index.ts';\nimport type { Feature } from '../features/types.ts';\nimport type { ParsedDoc } from './index.ts';\nimport { parseFile } from './index.ts';\nimport type { FileStat } from './list.ts';\nimport type { FileResult } from './pool.ts';\nimport { ParsePool } from './pool.ts';\n\n// Store-agnostic per-file parse pass shared by every store's reconcile(). Index-preserving:\n// one pass over `files`, pushed in order.\nexport interface ReparseResult {\n docs: ParsedDoc[];\n warnings: string[];\n // Frontmatter keys not in `knownColumns`, first-seen order -- the order callers ALTER TABLE ADD COLUMN in.\n newColumns: string[];\n}\n\nexport interface ReparseOptions {\n // Overrides DEFAULT_THRESHOLD below, so tests can force either dispatch mode without\n // needing thousands of fixture files. Internal: no caller in src/ passes it.\n threshold?: number;\n // Overrides DEFAULT_MAX_WORKERS below.\n maxWorkers?: number;\n // A builder-owned pool to dispatch through instead of creating and destroying one for this\n // call alone (see store/builder.ts). Absent, reparseFiles manages its own ephemeral pool.\n pool?: ParsePool;\n}\n\n// Measured 2026-08-29 on a 14-core Apple M4 Pro: serial and pooled cross between 160-180 files\n// (pool fixed cost ~85-100ms), so 200 keeps margin. 8 workers ties 10 in a 6,566-file sweep; 12+ trends worse.\nconst DEFAULT_THRESHOLD = 200;\nconst DEFAULT_MAX_WORKERS = 8;\n\n// Per-file feature filtering, shared by the serial loop (reads `features` from the caller) and\n// the worker path (can only pass `cfg` across the thread boundary, so derives `features` itself).\nexport function featuresForFile(features: Feature[], cfg: Config, file: FileStat): Feature[] {\n return features.filter((feature) => !feature.enabledForFile || feature.enabledForFile(cfg, file));\n}\n\nfunction reparseSerial(files: FileStat[], features: Feature[], cfg: Config, onParsed?: (done: number) => void): FileResult[] {\n const results: FileResult[] = [];\n let done = 0;\n for (const file of files) {\n const fileFeatures = featuresForFile(features, cfg, file);\n results.push(parseFile(file, fileFeatures, cfg));\n onParsed?.(++done);\n }\n return results;\n}\n\n// Dispatches through `pool` when the caller owns one (a builder, kept alive across calls); with\n// none given, opens an ephemeral ParsePool for this call alone and always closes it.\nasync function reparsePooled(files: FileStat[], features: Feature[], cfg: Config, onParsed: ((done: number) => void) | undefined, maxWorkers: number, pool: ParsePool | undefined): Promise<FileResult[]> {\n if (pool) return pool.run(files, features, cfg, onParsed, maxWorkers);\n const ephemeral = new ParsePool();\n try {\n const results = await ephemeral.run(files, features, cfg, onParsed, maxWorkers);\n await ephemeral.close();\n return results;\n } catch (err) {\n // Tear the pool down before rethrowing, but a close failure must not mask the parse\n // failure that caused it -- not a `finally`, for that reason.\n await ephemeral.close().catch(() => {});\n throw err;\n }\n}\n\n// `onParsed` receives the running count (1-based) after each file, mirroring a Progress.tick\n// call; pass one in to keep progress reporting working without this module owning a reporter.\nexport async function reparseFiles(files: FileStat[], features: Feature[], cfg: Config, knownColumns: ReadonlySet<string>, onParsed?: (done: number) => void, options: ReparseOptions = {}): Promise<ReparseResult> {\n const threshold = options.threshold ?? DEFAULT_THRESHOLD;\n const maxWorkers = options.maxWorkers ?? Math.min(DEFAULT_MAX_WORKERS, availableParallelism());\n const results = files.length >= threshold ? await reparsePooled(files, features, cfg, onParsed, maxWorkers, options.pool) : reparseSerial(files, features, cfg, onParsed);\n\n const docs: ParsedDoc[] = [];\n const warnings: string[] = [];\n const newColumns: string[] = [];\n const seen = new Set(knownColumns);\n for (const { doc, warnings: fileWarnings } of results) {\n warnings.push(...fileWarnings);\n for (const key of Object.keys(doc.data)) {\n if (!seen.has(key)) {\n seen.add(key);\n newColumns.push(key);\n }\n }\n docs.push(doc);\n }\n\n return { docs, warnings, newColumns };\n}\n"],"names":["featuresForFile","reparseFiles","DEFAULT_THRESHOLD","DEFAULT_MAX_WORKERS","features","cfg","file","filter","feature","enabledForFile","reparseSerial","files","onParsed","results","done","fileFeatures","push","parseFile","reparsePooled","maxWorkers","pool","ephemeral","err","run","ParsePool","close","catch","knownColumns","options","threshold","docs","warnings","newColumns","seen","doc","fileWarnings","key","Math","min","availableParallelism","length","Set","Object","keys","data","has","add"],"mappings":";;;;;;;;;;;QAoCgBA;eAAAA;;QAkCMC;eAAAA;;;sBAtEe;uBAIX;sBAGA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAsB1B,+FAA+F;AAC/F,+GAA+G;AAC/G,IAAMC,oBAAoB;AAC1B,IAAMC,sBAAsB;AAIrB,SAASH,gBAAgBI,QAAmB,EAAEC,GAAW,EAAEC,IAAc;IAC9E,OAAOF,SAASG,MAAM,CAAC,SAACC;eAAY,CAACA,QAAQC,cAAc,IAAID,QAAQC,cAAc,CAACJ,KAAKC;;AAC7F;AAEA,SAASI,cAAcC,KAAiB,EAAEP,QAAmB,EAAEC,GAAW,EAAEO,QAAiC;IAC3G,IAAMC,UAAwB,EAAE;IAChC,IAAIC,OAAO;QACN,kCAAA,2BAAA;;QAAL,QAAK,YAAcH,0BAAd,SAAA,6BAAA,QAAA,yBAAA,iCAAqB;YAArB,IAAML,OAAN;YACH,IAAMS,eAAef,gBAAgBI,UAAUC,KAAKC;YACpDO,QAAQG,IAAI,CAACC,IAAAA,kBAAS,EAACX,MAAMS,cAAcV;YAC3CO,qBAAAA,+BAAAA,SAAW,EAAEE;QACf;;QAJK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAKL,OAAOD;AACT;AAEA,gGAAgG;AAChG,qFAAqF;AACrF,SAAeK,cAAcP,KAAiB,EAAEP,QAAmB,EAAEC,GAAW,EAAEO,QAA8C,EAAEO,UAAkB,EAAEC,IAA2B;;YAEzKC,WAEER,SAGCS;;;;oBANT,IAAIF,MAAM;;wBAAOA,KAAKG,GAAG,CAACZ,OAAOP,UAAUC,KAAKO,UAAUO;;oBACpDE,YAAY,IAAIG,iBAAS;;;;;;;;;oBAEb;;wBAAMH,UAAUE,GAAG,CAACZ,OAAOP,UAAUC,KAAKO,UAAUO;;;oBAA9DN,UAAU;oBAChB;;wBAAMQ,UAAUI,KAAK;;;oBAArB;oBACA;;wBAAOZ;;;oBACAS;oBACP,oFAAoF;oBACpF,8DAA8D;oBAC9D;;wBAAMD,UAAUI,KAAK,GAAGC,KAAK,CAAC,YAAO;;;oBAArC;oBACA,MAAMJ;;;;;;;IAEV;;AAIO,SAAerB;wCAAaU,KAAiB,EAAEP,QAAmB,EAAEC,GAAW,EAAEsB,YAAiC,EAAEf,QAAiC;YAAEgB,SAC1IA,oBACCA,qBADbC,WACAV,YACAN,eAEAiB,MACAC,UACAC,YACAC,MACD,2BAAA,mBAAA,gBAAA,WAAA,oBAAQC,KAAeC,cAC1BJ,WACK,4BAAA,oBAAA,iBAAA,YAAA,QAAMK;;;;;oBAX+IR,UAAAA,oEAA0B,CAAC;oBACjLC,aAAYD,qBAAAA,QAAQC,SAAS,cAAjBD,gCAAAA,qBAAqB1B;oBACjCiB,cAAaS,sBAAAA,QAAQT,UAAU,cAAlBS,iCAAAA,sBAAsBS,KAAKC,GAAG,CAACnC,qBAAqBoC,IAAAA,4BAAoB;yBAC3E5B,CAAAA,MAAM6B,MAAM,IAAIX,SAAQ,GAAxBlB;;;;oBAA4B;;wBAAMO,cAAcP,OAAOP,UAAUC,KAAKO,UAAUO,YAAYS,QAAQR,IAAI;;;2BAA5E;;;;;;2BAAgFV,cAAcC,OAAOP,UAAUC,KAAKO;;;oBAA1JC;oBAEAiB;oBACAC;oBACAC;oBACAC,OAAO,IAAIQ,IAAId;oBAChB,kCAAA,2BAAA;;wBAAL,IAAK,YAAyCd,8BAAzC,6BAAA,QAAA,yBAAA,iCAAkD;0CAAlD,aAAQqB,kBAAAA,KAAeC,2BAAVJ;;4BAChBA,CAAAA,YAAAA,UAASf,IAAI,OAAbe,WAAc,qBAAGI;4BACZ,mCAAA,4BAAA;;gCAAL,IAAK,aAAaO,OAAOC,IAAI,CAACT,IAAIU,IAAI,wBAAjC,8BAAA,SAAA,0BAAA,kCAAoC;oCAA9BR,MAAN;oCACH,IAAI,CAACH,KAAKY,GAAG,CAACT,MAAM;wCAClBH,KAAKa,GAAG,CAACV;wCACTJ,WAAWhB,IAAI,CAACoB;oCAClB;gCACF;;gCALK;gCAAA;;;yCAAA,8BAAA;wCAAA;;;wCAAA;8CAAA;;;;4BAMLN,KAAKd,IAAI,CAACkB;wBACZ;;wBATK;wBAAA;;;iCAAA,6BAAA;gCAAA;;;gCAAA;sCAAA;;;;oBAWL;;wBAAO;4BAAEJ,MAAAA;4BAAMC,UAAAA;4BAAUC,YAAAA;wBAAW;;;;IACtC"}
@@ -0,0 +1,11 @@
1
+ import type { Config } from '../config/index.js';
2
+ import type { Connection, ReconcileDialect } from './types.js';
3
+ export interface Builder {
4
+ build(): Promise<{
5
+ parsed: number;
6
+ warnings: string[];
7
+ }>;
8
+ close(): Promise<void>;
9
+ readonly poolsCreated: number;
10
+ }
11
+ export declare function createBuilder(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect): Builder;
@@ -0,0 +1,11 @@
1
+ import type { Config } from '../config/index.js';
2
+ import type { Connection, ReconcileDialect } from './types.js';
3
+ export interface Builder {
4
+ build(): Promise<{
5
+ parsed: number;
6
+ warnings: string[];
7
+ }>;
8
+ close(): Promise<void>;
9
+ readonly poolsCreated: number;
10
+ }
11
+ export declare function createBuilder(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect): Builder;
@@ -0,0 +1,27 @@
1
+ "use strict";
2
+ Object.defineProperty(exports, "__esModule", {
3
+ value: true
4
+ });
5
+ Object.defineProperty(exports, "createBuilder", {
6
+ enumerable: true,
7
+ get: function() {
8
+ return createBuilder;
9
+ }
10
+ });
11
+ var _poolts = require("../scan/pool.js");
12
+ var _reconcilets = require("./reconcile.js");
13
+ function createBuilder(conn, cfg, baseDir, dialect) {
14
+ var pool = new _poolts.ParsePool();
15
+ return {
16
+ build: function build() {
17
+ return (0, _reconcilets.reconcile)(conn, cfg, baseDir, dialect, pool);
18
+ },
19
+ close: function close() {
20
+ return pool.close();
21
+ },
22
+ get poolsCreated () {
23
+ return pool.poolsCreated;
24
+ }
25
+ };
26
+ }
27
+ /* CJS INTEROP */ if (exports.__esModule && exports.default) { try { Object.defineProperty(exports.default, '__esModule', { value: true }); for (var key in exports) { exports.default[key] = exports[key]; } } catch (_) {}; module.exports = exports.default; }
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/builder.ts"],"sourcesContent":["import type { Config } from '../config/index.ts';\nimport { ParsePool } from '../scan/pool.ts';\nimport { reconcile } from './reconcile.ts';\nimport type { Connection, ReconcileDialect } from './types.ts';\n\n// Owns bringing a store's index current: the write half `Store` (types.ts) deliberately does not\n// carry. A one-shot open calls build() once; a watcher calls it repeatedly on the same instance.\nexport interface Builder {\n build(): Promise<{ parsed: number; warnings: string[] }>;\n // Releases the parse worker pool, if this lifetime ever created one; never the connection.\n close(): Promise<void>;\n // Pools this lifetime has constructed, so reuse is observable rather than inferred.\n readonly poolsCreated: number;\n}\n\n// The pool is created at most once, lazily, on whichever build() call first needs it, and reused\n// by every later build() on this instance.\nexport function createBuilder(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect): Builder {\n const pool = new ParsePool();\n return {\n build: () => reconcile(conn, cfg, baseDir, dialect, pool),\n close: () => pool.close(),\n get poolsCreated() {\n return pool.poolsCreated;\n },\n };\n}\n"],"names":["createBuilder","conn","cfg","baseDir","dialect","pool","ParsePool","build","reconcile","close","poolsCreated"],"mappings":";;;;+BAiBgBA;;;eAAAA;;;sBAhBU;2BACA;AAenB,SAASA,cAAcC,IAAgB,EAAEC,GAAW,EAAEC,OAAe,EAAEC,OAAyB;IACrG,IAAMC,OAAO,IAAIC,iBAAS;IAC1B,OAAO;QACLC,OAAO,SAAPA;mBAAaC,IAAAA,sBAAS,EAACP,MAAMC,KAAKC,SAASC,SAASC;;QACpDI,OAAO,SAAPA;mBAAaJ,KAAKI,KAAK;;QACvB,IAAIC;YACF,OAAOL,KAAKK,YAAY;QAC1B;IACF;AACF"}
@@ -2,6 +2,7 @@ import type { Connection, LexicalHit, LexicalQueryOptions } from '../types.js';
2
2
  interface FtsIndexState {
3
3
  stale: boolean;
4
4
  }
5
+ export declare function markContentStale(conn: Connection): void;
5
6
  export declare function queryLexical(conn: Connection, terms: string, opts: LexicalQueryOptions, state: FtsIndexState): Promise<LexicalHit[]>;
6
7
  export declare function createLexicalIndex(conn: Connection): {
7
8
  query: (terms: string, opts: LexicalQueryOptions) => Promise<LexicalHit[]>;
@@ -2,6 +2,7 @@ import type { Connection, LexicalHit, LexicalQueryOptions } from '../types.js';
2
2
  interface FtsIndexState {
3
3
  stale: boolean;
4
4
  }
5
+ export declare function markContentStale(conn: Connection): void;
5
6
  export declare function queryLexical(conn: Connection, terms: string, opts: LexicalQueryOptions, state: FtsIndexState): Promise<LexicalHit[]>;
6
7
  export declare function createLexicalIndex(conn: Connection): {
7
8
  query: (terms: string, opts: LexicalQueryOptions) => Promise<LexicalHit[]>;
@@ -14,6 +14,9 @@ _export(exports, {
14
14
  get createLexicalIndex () {
15
15
  return createLexicalIndex;
16
16
  },
17
+ get markContentStale () {
18
+ return markContentStale;
19
+ },
17
20
  get queryLexical () {
18
21
  return queryLexical;
19
22
  }
@@ -237,6 +240,22 @@ function unsupportedOperator(terms) {
237
240
  }
238
241
  return null;
239
242
  }
243
+ // One state per connection, not per store instance: reconcileContent (reconcile.ts) marks it
244
+ // stale from outside createLexicalIndex's closure, whether a one-shot open or a watcher's builder ran it.
245
+ var ftsState = new WeakMap();
246
+ function stateFor(conn) {
247
+ var state = ftsState.get(conn);
248
+ if (!state) {
249
+ state = {
250
+ stale: true
251
+ };
252
+ ftsState.set(conn, state);
253
+ }
254
+ return state;
255
+ }
256
+ function markContentStale(conn) {
257
+ stateFor(conn).stale = true;
258
+ }
240
259
  // A run whose script marks no word boundaries makes match_bm25's whitespace tokenizer index the whole run as one token
241
260
  // (same gap FTS5 has without the `_seg` sidecar), so such runs -- and any author-quoted phrase, any script -- go through contains() instead.
242
261
  function splitTerms(terms) {
@@ -480,15 +499,12 @@ function queryLexical(conn, terms, opts, state) {
480
499
  })();
481
500
  }
482
501
  function createLexicalIndex(conn) {
483
- var state = {
484
- stale: true
485
- };
486
502
  return {
487
503
  query: function query(terms, opts) {
488
- return queryLexical(conn, terms, opts, state);
504
+ return queryLexical(conn, terms, opts, stateFor(conn));
489
505
  },
490
506
  markStale: function markStale() {
491
- state.stale = true;
507
+ return markContentStale(conn);
492
508
  }
493
509
  };
494
510
  }
@@ -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 excerpts for every doc (no snippet()).\nimport { SenseError } from '../../errors.ts';\nimport { hasUnspacedRun } from '../../text/segment.ts';\nimport { withTransaction } from '../transaction.ts';\nimport type { Connection, LexicalHit, LexicalQueryOptions } from '../types.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). Checked with quoted spans blanked; those go through contains() and are supported.\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];\n\nfunction unsupportedOperator(terms: string): { label: string; token: string } | null {\n const withoutPhrases = terms.replace(/\"[^\"]*\"/g, ' ');\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}\n\n// A run whose script marks no word boundaries makes match_bm25's whitespace tokenizer index the whole run as one token\n// (same gap FTS5 has without the `_seg` sidecar), so such runs -- and any author-quoted phrase, any script -- go through contains() instead.\nfunction splitTerms(terms: string): { words: string[]; substrings: string[] } {\n const words: string[] = [];\n const substrings: string[] = [];\n const withoutPhrases = terms.replace(/\"([^\"]*)\"/g, (_m, inner: string) => {\n const phrase = inner.trim();\n if (phrase.length === 0) return ' ';\n substrings.push(phrase);\n if (!hasUnspacedRun(phrase)) for (const w of phrase.split(/\\s+/)) if (w.length > 0) words.push(w);\n return ' ';\n });\n for (const tok of withoutPhrases.split(/\\s+/)) {\n if (tok.length === 0) continue;\n if (hasUnspacedRun(tok)) substrings.push(tok);\n else words.push(tok);\n }\n return { words, substrings };\n}\n\n// No incremental update (verified 1.5.5: PRAGMA create_fts_index is rebuild-only), so this pays the full rebuild -- but only\n// once per store instance's first lexical query, and only when content actually changed since (FtsIndexState / store.ts's markStale()).\nasync function ensureFtsFresh(conn: Connection, state: FtsIndexState): Promise<void> {\n if (!state.stale) return;\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 });\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[], 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.join(' ');\n for (const field of FIELDS) {\n scoreParts.push(`${FIELD_WEIGHT[field]}.0 * COALESCE(fts_main_content.match_bm25(content.\"path\", ?, fields := '${field}', conjunctive := false), 0)`);\n scoreParams.push(wordQuery);\n }\n gateParts.push(`fts_main_content.match_bm25(content.\"path\", ?, conjunctive := true) IS NOT NULL`);\n gateParams.push(wordQuery);\n }\n\n for (const raw of substrings) {\n const needle = raw.toLowerCase();\n for (const field of FIELDS) {\n scoreParts.push(`${FIELD_WEIGHT[field]}.0 * CAST(contains(lower(content.${field}), ?) AS INTEGER)`);\n scoreParams.push(needle);\n }\n gateParts.push(`contains(lower(content.title) || ' ' || lower(content.summary) || ' ' || lower(content.text), ?)`);\n gateParams.push(needle);\n }\n\n return { scoreSql: scoreParts.join(' + '), gateSql: gateParts.join(' AND '), params: [...scoreParams, ...gateParams] };\n}\n\n// Ranked word-match query, scoped by the caller-built SQL fragments (same fragments sqlite's queryLexical takes).\n// `hit` is always NULL: no snippet() equivalent exists, so every row goes through the caller's JS excerpt fallback (commands/search.ts).\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 \"${terms.trim()}\" without it, or set \"store\" to \"sqlite\" in this tree's config to search it as written`);\n }\n const { whereJoin, whereCond, scopeCond, limit } = opts;\n const { words, substrings } = splitTerms(terms);\n if (words.length === 0 && substrings.length === 0) return [];\n\n if (words.length > 0) await ensureFtsFresh(conn, state);\n\n const { scoreSql, gateSql, params } = buildScoreAndGate(words, substrings);\n const sql = `SELECT path, NULL AS hit FROM (\n SELECT content.\"path\" AS path, (${scoreSql}) AS score\n FROM content\n ${whereJoin}\n WHERE ${gateSql}\n ${whereCond} ${scopeCond}\n ) sq ORDER BY score DESC, path LIMIT ?`;\n const stmt = await conn.prepare(sql);\n return (await stmt.all(...params, limit)) as unknown as LexicalHit[];\n}\n\n// One instance per store: `stale` starts true (a fresh connection can't know whether an on-disk fts schema matches `content`),\n// and store.ts's reconcile wrapper flips it back to true whenever content changes, so the next query rebuilds first.\nexport function createLexicalIndex(conn: Connection): { query: (terms: string, opts: LexicalQueryOptions) => Promise<LexicalHit[]>; markStale: () => void } {\n const state: FtsIndexState = { stale: true };\n return {\n query: (terms, opts) => queryLexical(conn, terms, opts, state),\n markStale: () => {\n state.stale = true;\n },\n };\n}\n"],"names":["createLexicalIndex","queryLexical","FIELDS","FIELD_WEIGHT","title","summary","text","FTS5_OPERATORS","label","re","unsupportedOperator","terms","withoutPhrases","replace","m","match","token","trim","splitTerms","words","substrings","_m","inner","phrase","length","push","hasUnspacedRun","split","w","tok","ensureFtsFresh","conn","state","stale","exec","withTransaction","buildScoreAndGate","scoreParts","gateParts","scoreParams","gateParams","wordQuery","join","field","raw","needle","toLowerCase","scoreSql","gateSql","params","opts","stmt","unsupported","whereJoin","whereCond","scopeCond","limit","sql","SenseError","prepare","all","query","markStale"],"mappings":"AAAA,yFAAyF;AACzF,8DAA8D;;;;;;;;;;;;QA8H9CA;eAAAA;;QAzBMC;eAAAA;;;wBApGK;yBACI;6BACC;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAGhC,IAAMC,SAAS;IAAC;IAAS;IAAW;CAAO;AAE3C,uGAAuG;AACvG,iGAAiG;AACjG,IAAMC,eAAsC;IAAEC,OAAO;IAAIC,SAAS;IAAGC,MAAM;AAAE;AAE7E,0GAA0G;AAC1G,0IAA0I;AAC1I,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;CAC1D;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;AAMA,uHAAuH;AACvH,6IAA6I;AAC7I,SAASC,WAAWP,KAAa;IAC/B,IAAMQ,QAAkB,EAAE;IAC1B,IAAMC,aAAuB,EAAE;IAC/B,IAAMR,iBAAiBD,MAAME,OAAO,CAAC,cAAc,SAACQ,IAAIC;QACtD,IAAMC,SAASD,MAAML,IAAI;QACzB,IAAIM,OAAOC,MAAM,KAAK,GAAG,OAAO;QAChCJ,WAAWK,IAAI,CAACF;QAChB,IAAI,CAACG,IAAAA,yBAAc,EAACH,SAAS;gBAAK,kCAAA,2BAAA;;gBAAL,QAAK,YAAWA,OAAOI,KAAK,CAAC,2BAAxB,SAAA,6BAAA,QAAA,yBAAA;oBAAA,IAAMC,IAAN;oBAAgC,IAAIA,EAAEJ,MAAM,GAAG,GAAGL,MAAMM,IAAI,CAACG;;;gBAA7D;gBAAA;;;yBAAA,6BAAA;wBAAA;;;wBAAA;8BAAA;;;;QAA+D;QACjG,OAAO;IACT;QACK,kCAAA,2BAAA;;QAAL,QAAK,YAAahB,eAAee,KAAK,CAAC,2BAAlC,SAAA,6BAAA,QAAA,yBAAA,iCAA0C;YAA1C,IAAME,MAAN;YACH,IAAIA,IAAIL,MAAM,KAAK,GAAG;YACtB,IAAIE,IAAAA,yBAAc,EAACG,MAAMT,WAAWK,IAAI,CAACI;iBACpCV,MAAMM,IAAI,CAACI;QAClB;;QAJK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAKL,OAAO;QAAEV,OAAAA;QAAOC,YAAAA;IAAW;AAC7B;AAEA,6HAA6H;AAC7H,wIAAwI;AACxI,SAAeU,eAAeC,IAAgB,EAAEC,KAAoB;;;;;oBAClE,IAAI,CAACA,MAAMC,KAAK,EAAE;;;oBAClB;;wBAAMF,KAAKG,IAAI,CAAC;;;oBAAhB;oBACA;;wBAAMC,IAAAA,8BAAe,EAACJ,MAAM;;;;;4CAC1B,6HAA6H;4CAC7H,0JAA0J;4CAC1J;;gDAAMA,KAAKG,IAAI,CAAC;;;4CAAhB;;;;;;4BACF;;;;oBAJA;oBAKAF,MAAMC,KAAK,GAAG;;;;;;IAChB;;AAEA,qKAAqK;AACrK,+OAA+O;AAC/O,SAASG,kBAAkBjB,KAAe,EAAEC,UAAoB;IAC9D,IAAMiB,aAAuB,EAAE;IAC/B,IAAMC,YAAsB,EAAE;IAC9B,IAAMC,cAAyB,EAAE;IACjC,IAAMC,aAAwB,EAAE;IAEhC,IAAIrB,MAAMK,MAAM,GAAG,GAAG;QACpB,IAAMiB,YAAYtB,MAAMuB,IAAI,CAAC;YACxB,kCAAA,2BAAA;;YAAL,QAAK,YAAexC,2BAAf,SAAA,6BAAA,QAAA,yBAAA,iCAAuB;gBAAvB,IAAMyC,QAAN;gBACHN,WAAWZ,IAAI,CAAC,AAAC,GAAgGkB,OAA9FxC,YAAY,CAACwC,MAAM,EAAC,6EAAgF,OAANA,OAAM;gBACvHJ,YAAYd,IAAI,CAACgB;YACnB;;YAHK;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;QAILH,UAAUb,IAAI,CAAC;QACfe,WAAWf,IAAI,CAACgB;IAClB;QAEK,mCAAA,4BAAA;;QAAL,QAAK,aAAarB,+BAAb,UAAA,8BAAA,SAAA,0BAAA,kCAAyB;YAAzB,IAAMwB,MAAN;YACH,IAAMC,SAASD,IAAIE,WAAW;gBACzB,mCAAA,4BAAA;;gBAAL,QAAK,aAAe5C,2BAAf,UAAA,8BAAA,SAAA,0BAAA,kCAAuB;oBAAvB,IAAMyC,SAAN;oBACHN,WAAWZ,IAAI,CAAC,AAAC,GAAyDkB,OAAvDxC,YAAY,CAACwC,OAAM,EAAC,qCAAyC,OAANA,QAAM;oBAChFJ,YAAYd,IAAI,CAACoB;gBACnB;;gBAHK;gBAAA;;;yBAAA,8BAAA;wBAAA;;;wBAAA;8BAAA;;;;YAILP,UAAUb,IAAI,CAAC;YACfe,WAAWf,IAAI,CAACoB;QAClB;;QARK;QAAA;;;iBAAA,8BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAUL,OAAO;QAAEE,UAAUV,WAAWK,IAAI,CAAC;QAAQM,SAASV,UAAUI,IAAI,CAAC;QAAUO,QAAQ,AAAC,qBAAGV,oBAAa,qBAAGC;IAAY;AACvH;AAIO,SAAevC,aAAa8B,IAAgB,EAAEpB,KAAa,EAAEuC,IAAyB,EAAElB,KAAoB;;YAoBnGmB,OAnBRC,aAIEC,WAAWC,WAAWC,WAAWC,OACXtC,aAAtBC,OAAOC,YAKuBgB,oBAA9BW,UAAUC,SAASC,QACrBQ,KAOAN;;;;oBAlBAC,cAAc1C,oBAAoBC;oBACxC,IAAIyC,gBAAgB,MAAM;wBACxB,MAAM,IAAIM,oBAAU,CAAC,4BAA4B,AAAC,6CAAkEN,OAAvBA,YAAY5C,KAAK,EAAC,OAAqDG,OAAhDyC,YAAYpC,KAAK,EAAC,gCAA2C,OAAbL,MAAMM,IAAI,IAAG;oBACnL;oBACQoC,YAA2CH,KAA3CG,WAAWC,YAAgCJ,KAAhCI,WAAWC,YAAqBL,KAArBK,WAAWC,QAAUN,KAAVM;oBACXtC,cAAAA,WAAWP,QAAjCQ,QAAsBD,YAAtBC,OAAOC,aAAeF,YAAfE;oBACf,IAAID,MAAMK,MAAM,KAAK,KAAKJ,WAAWI,MAAM,KAAK,GAAG;;;;yBAE/CL,CAAAA,MAAMK,MAAM,GAAG,CAAA,GAAfL;;;;oBAAkB;;wBAAMW,eAAeC,MAAMC;;;oBAA3B;;;oBAEgBI,qBAAAA,kBAAkBjB,OAAOC,aAAvD2B,WAA8BX,mBAA9BW,UAAUC,UAAoBZ,mBAApBY,SAASC,SAAWb,mBAAXa;oBACrBQ,MAAM,AAAC,wEAGTJ,OAFgCN,UAAS,sCAGnCC,OADNK,WAAU,gBAEVC,OADMN,SAAQ,UACDO,OAAbD,WAAU,KAAa,OAAVC,WAAU;oBAEd;;wBAAMxB,KAAK4B,OAAO,CAACF;;;oBAA1BN,OAAO;oBACL;;wBAAMA,CAAAA,QAAAA,MAAKS,GAAG,OAART,OAAAA,AAAS,qBAAGF;4BAAQO;;;;oBAAlC;;wBAAQ;;;;IACV;;AAIO,SAASxD,mBAAmB+B,IAAgB;IACjD,IAAMC,QAAuB;QAAEC,OAAO;IAAK;IAC3C,OAAO;QACL4B,OAAO,SAAPA,MAAQlD,OAAOuC;mBAASjD,aAAa8B,MAAMpB,OAAOuC,MAAMlB;;QACxD8B,WAAW,SAAXA;YACE9B,MAAMC,KAAK,GAAG;QAChB;IACF;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 excerpts for every doc (no snippet()).\nimport { SenseError } from '../../errors.ts';\nimport { hasUnspacedRun } from '../../text/segment.ts';\nimport { withTransaction } from '../transaction.ts';\nimport type { Connection, LexicalHit, LexicalQueryOptions } from '../types.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). Checked with quoted spans blanked; those go through contains() and are supported.\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];\n\nfunction unsupportedOperator(terms: string): { label: string; token: string } | null {\n const withoutPhrases = terms.replace(/\"[^\"]*\"/g, ' ');\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}\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>();\nfunction stateFor(conn: Connection): FtsIndexState {\n let state = ftsState.get(conn);\n if (!state) {\n state = { stale: true };\n ftsState.set(conn, state);\n }\n return state;\n}\n\n// Called whenever this store's reconcileContent changes `content`, so the next lexical query\n// rebuilds instead of searching stale content.\nexport function markContentStale(conn: Connection): void {\n stateFor(conn).stale = true;\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// (same gap FTS5 has without the `_seg` sidecar), so such runs -- and any author-quoted phrase, any script -- go through contains() instead.\nfunction splitTerms(terms: string): { words: string[]; substrings: string[] } {\n const words: string[] = [];\n const substrings: string[] = [];\n const withoutPhrases = terms.replace(/\"([^\"]*)\"/g, (_m, inner: string) => {\n const phrase = inner.trim();\n if (phrase.length === 0) return ' ';\n substrings.push(phrase);\n if (!hasUnspacedRun(phrase)) for (const w of phrase.split(/\\s+/)) if (w.length > 0) words.push(w);\n return ' ';\n });\n for (const tok of withoutPhrases.split(/\\s+/)) {\n if (tok.length === 0) continue;\n if (hasUnspacedRun(tok)) substrings.push(tok);\n else words.push(tok);\n }\n return { words, substrings };\n}\n\n// No incremental update (verified 1.5.5: PRAGMA create_fts_index is rebuild-only), so this pays the full rebuild -- but only\n// once per store instance's first lexical query, and only when content actually changed since (FtsIndexState / store.ts's markStale()).\nasync function ensureFtsFresh(conn: Connection, state: FtsIndexState): Promise<void> {\n if (!state.stale) return;\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 });\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[], 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.join(' ');\n for (const field of FIELDS) {\n scoreParts.push(`${FIELD_WEIGHT[field]}.0 * COALESCE(fts_main_content.match_bm25(content.\"path\", ?, fields := '${field}', conjunctive := false), 0)`);\n scoreParams.push(wordQuery);\n }\n gateParts.push(`fts_main_content.match_bm25(content.\"path\", ?, conjunctive := true) IS NOT NULL`);\n gateParams.push(wordQuery);\n }\n\n for (const raw of substrings) {\n const needle = raw.toLowerCase();\n for (const field of FIELDS) {\n scoreParts.push(`${FIELD_WEIGHT[field]}.0 * CAST(contains(lower(content.${field}), ?) AS INTEGER)`);\n scoreParams.push(needle);\n }\n gateParts.push(`contains(lower(content.title) || ' ' || lower(content.summary) || ' ' || lower(content.text), ?)`);\n gateParams.push(needle);\n }\n\n return { scoreSql: scoreParts.join(' + '), gateSql: gateParts.join(' AND '), params: [...scoreParams, ...gateParams] };\n}\n\n// Ranked word-match query, scoped by the caller-built SQL fragments (same fragments sqlite's queryLexical takes).\n// `hit` is always NULL: no snippet() equivalent exists, so every row goes through the caller's JS excerpt fallback (commands/search.ts).\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 \"${terms.trim()}\" without it, or set \"store\" to \"sqlite\" in this tree's config to search it as written`);\n }\n const { whereJoin, whereCond, scopeCond, limit } = opts;\n const { words, substrings } = splitTerms(terms);\n if (words.length === 0 && substrings.length === 0) return [];\n\n if (words.length > 0) await ensureFtsFresh(conn, state);\n\n const { scoreSql, gateSql, params } = buildScoreAndGate(words, substrings);\n const sql = `SELECT path, NULL AS hit FROM (\n SELECT content.\"path\" AS path, (${scoreSql}) AS score\n FROM content\n ${whereJoin}\n WHERE ${gateSql}\n ${whereCond} ${scopeCond}\n ) sq ORDER BY score DESC, path LIMIT ?`;\n const stmt = await conn.prepare(sql);\n return (await stmt.all(...params, limit)) as unknown as LexicalHit[];\n}\n\n// `stale` starts true (a fresh connection can't know whether an on-disk fts schema matches\n// `content`), 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: () => void } {\n return {\n query: (terms, opts) => queryLexical(conn, terms, opts, 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","ftsState","WeakMap","stateFor","conn","state","get","stale","set","splitTerms","words","substrings","_m","inner","phrase","length","push","hasUnspacedRun","split","w","tok","ensureFtsFresh","exec","withTransaction","buildScoreAndGate","scoreParts","gateParts","scoreParams","gateParams","wordQuery","join","field","raw","needle","toLowerCase","scoreSql","gateSql","params","opts","stmt","unsupported","whereJoin","whereCond","scopeCond","limit","sql","SenseError","prepare","all","query","markStale"],"mappings":"AAAA,yFAAyF;AACzF,8DAA8D;;;;;;;;;;;;QAgJ9CA;eAAAA;;QA/FAC;eAAAA;;QAsEMC;eAAAA;;;wBAtHK;yBACI;6BACC;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAGhC,IAAMC,SAAS;IAAC;IAAS;IAAW;CAAO;AAE3C,uGAAuG;AACvG,iGAAiG;AACjG,IAAMC,eAAsC;IAAEC,OAAO;IAAIC,SAAS;IAAGC,MAAM;AAAE;AAE7E,0GAA0G;AAC1G,0IAA0I;AAC1I,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;CAC1D;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;AAMA,6FAA6F;AAC7F,0GAA0G;AAC1G,IAAMC,WAAW,IAAIC;AACrB,SAASC,SAASC,IAAgB;IAChC,IAAIC,QAAQJ,SAASK,GAAG,CAACF;IACzB,IAAI,CAACC,OAAO;QACVA,QAAQ;YAAEE,OAAO;QAAK;QACtBN,SAASO,GAAG,CAACJ,MAAMC;IACrB;IACA,OAAOA;AACT;AAIO,SAAStB,iBAAiBqB,IAAgB;IAC/CD,SAASC,MAAMG,KAAK,GAAG;AACzB;AAEA,uHAAuH;AACvH,6IAA6I;AAC7I,SAASE,WAAWf,KAAa;IAC/B,IAAMgB,QAAkB,EAAE;IAC1B,IAAMC,aAAuB,EAAE;IAC/B,IAAMhB,iBAAiBD,MAAME,OAAO,CAAC,cAAc,SAACgB,IAAIC;QACtD,IAAMC,SAASD,MAAMb,IAAI;QACzB,IAAIc,OAAOC,MAAM,KAAK,GAAG,OAAO;QAChCJ,WAAWK,IAAI,CAACF;QAChB,IAAI,CAACG,IAAAA,yBAAc,EAACH,SAAS;gBAAK,kCAAA,2BAAA;;gBAAL,QAAK,YAAWA,OAAOI,KAAK,CAAC,2BAAxB,SAAA,6BAAA,QAAA,yBAAA;oBAAA,IAAMC,IAAN;oBAAgC,IAAIA,EAAEJ,MAAM,GAAG,GAAGL,MAAMM,IAAI,CAACG;;;gBAA7D;gBAAA;;;yBAAA,6BAAA;wBAAA;;;wBAAA;8BAAA;;;;QAA+D;QACjG,OAAO;IACT;QACK,kCAAA,2BAAA;;QAAL,QAAK,YAAaxB,eAAeuB,KAAK,CAAC,2BAAlC,SAAA,6BAAA,QAAA,yBAAA,iCAA0C;YAA1C,IAAME,MAAN;YACH,IAAIA,IAAIL,MAAM,KAAK,GAAG;YACtB,IAAIE,IAAAA,yBAAc,EAACG,MAAMT,WAAWK,IAAI,CAACI;iBACpCV,MAAMM,IAAI,CAACI;QAClB;;QAJK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAKL,OAAO;QAAEV,OAAAA;QAAOC,YAAAA;IAAW;AAC7B;AAEA,6HAA6H;AAC7H,wIAAwI;AACxI,SAAeU,eAAejB,IAAgB,EAAEC,KAAoB;;;;;oBAClE,IAAI,CAACA,MAAME,KAAK,EAAE;;;oBAClB;;wBAAMH,KAAKkB,IAAI,CAAC;;;oBAAhB;oBACA;;wBAAMC,IAAAA,8BAAe,EAACnB,MAAM;;;;;4CAC1B,6HAA6H;4CAC7H,0JAA0J;4CAC1J;;gDAAMA,KAAKkB,IAAI,CAAC;;;4CAAhB;;;;;;4BACF;;;;oBAJA;oBAKAjB,MAAME,KAAK,GAAG;;;;;;IAChB;;AAEA,qKAAqK;AACrK,+OAA+O;AAC/O,SAASiB,kBAAkBd,KAAe,EAAEC,UAAoB;IAC9D,IAAMc,aAAuB,EAAE;IAC/B,IAAMC,YAAsB,EAAE;IAC9B,IAAMC,cAAyB,EAAE;IACjC,IAAMC,aAAwB,EAAE;IAEhC,IAAIlB,MAAMK,MAAM,GAAG,GAAG;QACpB,IAAMc,YAAYnB,MAAMoB,IAAI,CAAC;YACxB,kCAAA,2BAAA;;YAAL,QAAK,YAAe7C,2BAAf,SAAA,6BAAA,QAAA,yBAAA,iCAAuB;gBAAvB,IAAM8C,QAAN;gBACHN,WAAWT,IAAI,CAAC,AAAC,GAAgGe,OAA9F7C,YAAY,CAAC6C,MAAM,EAAC,6EAAgF,OAANA,OAAM;gBACvHJ,YAAYX,IAAI,CAACa;YACnB;;YAHK;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;QAILH,UAAUV,IAAI,CAAC;QACfY,WAAWZ,IAAI,CAACa;IAClB;QAEK,mCAAA,4BAAA;;QAAL,QAAK,aAAalB,+BAAb,UAAA,8BAAA,SAAA,0BAAA,kCAAyB;YAAzB,IAAMqB,MAAN;YACH,IAAMC,SAASD,IAAIE,WAAW;gBACzB,mCAAA,4BAAA;;gBAAL,QAAK,aAAejD,2BAAf,UAAA,8BAAA,SAAA,0BAAA,kCAAuB;oBAAvB,IAAM8C,SAAN;oBACHN,WAAWT,IAAI,CAAC,AAAC,GAAyDe,OAAvD7C,YAAY,CAAC6C,OAAM,EAAC,qCAAyC,OAANA,QAAM;oBAChFJ,YAAYX,IAAI,CAACiB;gBACnB;;gBAHK;gBAAA;;;yBAAA,8BAAA;wBAAA;;;wBAAA;8BAAA;;;;YAILP,UAAUV,IAAI,CAAC;YACfY,WAAWZ,IAAI,CAACiB;QAClB;;QARK;QAAA;;;iBAAA,8BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAUL,OAAO;QAAEE,UAAUV,WAAWK,IAAI,CAAC;QAAQM,SAASV,UAAUI,IAAI,CAAC;QAAUO,QAAQ,AAAC,qBAAGV,oBAAa,qBAAGC;IAAY;AACvH;AAIO,SAAe5C,aAAaoB,IAAgB,EAAEV,KAAa,EAAE4C,IAAyB,EAAEjC,KAAoB;;YAoBnGkC,OAnBRC,aAIEC,WAAWC,WAAWC,WAAWC,OACXnC,aAAtBC,OAAOC,YAKuBa,oBAA9BW,UAAUC,SAASC,QACrBQ,KAOAN;;;;oBAlBAC,cAAc/C,oBAAoBC;oBACxC,IAAI8C,gBAAgB,MAAM;wBACxB,MAAM,IAAIM,oBAAU,CAAC,4BAA4B,AAAC,6CAAkEN,OAAvBA,YAAYjD,KAAK,EAAC,OAAqDG,OAAhD8C,YAAYzC,KAAK,EAAC,gCAA2C,OAAbL,MAAMM,IAAI,IAAG;oBACnL;oBACQyC,YAA2CH,KAA3CG,WAAWC,YAAgCJ,KAAhCI,WAAWC,YAAqBL,KAArBK,WAAWC,QAAUN,KAAVM;oBACXnC,cAAAA,WAAWf,QAAjCgB,QAAsBD,YAAtBC,OAAOC,aAAeF,YAAfE;oBACf,IAAID,MAAMK,MAAM,KAAK,KAAKJ,WAAWI,MAAM,KAAK,GAAG;;;;yBAE/CL,CAAAA,MAAMK,MAAM,GAAG,CAAA,GAAfL;;;;oBAAkB;;wBAAMW,eAAejB,MAAMC;;;oBAA3B;;;oBAEgBmB,qBAAAA,kBAAkBd,OAAOC,aAAvDwB,WAA8BX,mBAA9BW,UAAUC,UAAoBZ,mBAApBY,SAASC,SAAWb,mBAAXa;oBACrBQ,MAAM,AAAC,wEAGTJ,OAFgCN,UAAS,sCAGnCC,OADNK,WAAU,gBAEVC,OADMN,SAAQ,UACDO,OAAbD,WAAU,KAAa,OAAVC,WAAU;oBAEd;;wBAAMvC,KAAK2C,OAAO,CAACF;;;oBAA1BN,OAAO;oBACL;;wBAAMA,CAAAA,QAAAA,MAAKS,GAAG,OAART,OAAAA,AAAS,qBAAGF;4BAAQO;;;;oBAAlC;;wBAAQ;;;;IACV;;AAIO,SAAS9D,mBAAmBsB,IAAgB;IACjD,OAAO;QACL6C,OAAO,SAAPA,MAAQvD,OAAO4C;mBAAStD,aAAaoB,MAAMV,OAAO4C,MAAMnC,SAASC;;QACjE8C,WAAW,SAAXA;mBAAiBnE,iBAAiBqB;;IACpC;AACF"}
@@ -1,12 +1,13 @@
1
+ import type { DuckDBConnection, DuckDBInstance } from '@duckdb/node-api';
1
2
  import type { ResolvedConfig } from '../../config/index.js';
2
- import type { Store } from '../types.js';
3
+ import type { OpenResult } from '../open.js';
4
+ import type { OpenDialect } from '../types.js';
3
5
  export declare const DB_FILENAME = "cache.duckdb";
4
- export declare const SCHEMA_VERSION = "2";
5
- export interface OpenResult {
6
- store: Store;
7
- cfg: ResolvedConfig;
8
- dbPath: string;
9
- parsed: number;
10
- warnings: string[];
6
+ export declare const SCHEMA_VERSION = "3";
7
+ export type { OpenResult };
8
+ interface DuckdbHandle {
9
+ instance: DuckDBInstance;
10
+ duckdb: DuckDBConnection;
11
11
  }
12
+ export declare const duckdbOpenDialect: OpenDialect<DuckdbHandle>;
12
13
  export declare function openDuckdb(cfg: ResolvedConfig): Promise<OpenResult>;
@@ -1,12 +1,13 @@
1
+ import type { DuckDBConnection, DuckDBInstance } from '@duckdb/node-api';
1
2
  import type { ResolvedConfig } from '../../config/index.js';
2
- import type { Store } from '../types.js';
3
+ import type { OpenResult } from '../open.js';
4
+ import type { OpenDialect } from '../types.js';
3
5
  export declare const DB_FILENAME = "cache.duckdb";
4
- export declare const SCHEMA_VERSION = "2";
5
- export interface OpenResult {
6
- store: Store;
7
- cfg: ResolvedConfig;
8
- dbPath: string;
9
- parsed: number;
10
- warnings: string[];
6
+ export declare const SCHEMA_VERSION = "3";
7
+ export type { OpenResult };
8
+ interface DuckdbHandle {
9
+ instance: DuckDBInstance;
10
+ duckdb: DuckDBConnection;
11
11
  }
12
+ export declare const duckdbOpenDialect: OpenDialect<DuckdbHandle>;
12
13
  export declare function openDuckdb(cfg: ResolvedConfig): Promise<OpenResult>;
@@ -15,23 +15,22 @@ _export(exports, {
15
15
  get SCHEMA_VERSION () {
16
16
  return SCHEMA_VERSION;
17
17
  },
18
+ get duckdbOpenDialect () {
19
+ return duckdbOpenDialect;
20
+ },
18
21
  get openDuckdb () {
19
22
  return openDuckdb;
20
23
  }
21
24
  });
22
- var _nodefs = require("node:fs");
23
- var _nodepath = require("node:path");
24
25
  var _indexts = require("../../config/index.js");
25
- var _handoffts = require("../../embed/handoff.js");
26
26
  var _typests = require("../../embed/types.js");
27
27
  var _errorsts = require("../../errors.js");
28
28
  var _indexts1 = require("../../features/index.js");
29
- var _cachets = require("../cache.js");
30
- var _reconcilets = require("../reconcile.js");
29
+ var _opents = require("../open.js");
31
30
  var _sharedts = require("../shared.js");
32
31
  var _connectionts = require("./connection.js");
33
32
  var _nativets = require("./native.js");
34
- var _reconcilets1 = require("./reconcile.js");
33
+ var _reconcilets = require("./reconcile.js");
35
34
  var _sqlfunctionsts = require("./sql-functions.js");
36
35
  var _storets = require("./store.js");
37
36
  function asyncGeneratorStep(gen, resolve, reject, _next, _throw, key, arg) {
@@ -166,10 +165,10 @@ function _ts_generator(thisArg, body) {
166
165
  }
167
166
  }
168
167
  var DB_FILENAME = 'cache.duckdb';
169
- var SCHEMA_VERSION = '2';
168
+ var SCHEMA_VERSION = '3';
170
169
  // `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.
171
170
  // No tokenizer resolution: this store always uses the fts extension's default (porter) stemmer; sqlite's configurable `content.tokenize` is not read here.
172
- function ensureSchema(conn, cfg) {
171
+ function ensureSchema(_handle, conn, cfg) {
173
172
  return _async_to_generator(function() {
174
173
  var _iteratorNormalCompletion, _didIteratorError, _iteratorError, _iterator, _step, feature, err;
175
174
  return _ts_generator(this, function(_state) {
@@ -312,45 +311,50 @@ function ensureSchema(conn, cfg) {
312
311
  });
313
312
  })();
314
313
  }
315
- function connect(cfg) {
314
+ // Order matters: the connection must be gone before the instance closes the WAL.
315
+ function close(handle) {
316
+ return _async_to_generator(function() {
317
+ return _ts_generator(this, function(_state) {
318
+ handle.duckdb.disconnectSync();
319
+ handle.instance.closeSync();
320
+ return [
321
+ 2
322
+ ];
323
+ });
324
+ })();
325
+ }
326
+ function connect(dbPath, _cfg) {
316
327
  return _async_to_generator(function() {
317
- var stateDir, dbPath, DuckDBInstance, err, duckdb, instance, err1, closed, conn, version, features, wantFeatures, reason, _ref, parsed, warnings, err2;
328
+ var DuckDBInstance, err, duckdb, instance, err1, conn, err2;
318
329
  return _ts_generator(this, function(_state) {
319
330
  switch(_state.label){
320
331
  case 0:
321
- stateDir = (0, _nodepath.join)(cfg.baseDir, _indexts.STATE_DIR);
322
- (0, _nodefs.mkdirSync)(stateDir, {
323
- recursive: true
324
- });
325
- dbPath = (0, _nodepath.join)(stateDir, DB_FILENAME);
326
- _state.label = 1;
327
- case 1:
328
332
  _state.trys.push([
329
- 1,
330
- 3,
333
+ 0,
334
+ 2,
331
335
  ,
332
- 4
336
+ 3
333
337
  ]);
334
338
  return [
335
339
  4,
336
340
  (0, _nativets.duckdbApi)()
337
341
  ];
338
- case 2:
342
+ case 1:
339
343
  DuckDBInstance = _state.sent().DuckDBInstance;
340
344
  return [
341
345
  3,
342
- 4
346
+ 3
343
347
  ];
344
- case 3:
348
+ case 2:
345
349
  err = _state.sent();
346
350
  if (_instanceof(err, _errorsts.SenseError)) throw err;
347
351
  throw new _errorsts.SenseError('STORE_DEPENDENCY_MISSING', 'store "duckdb" needs the '.concat(_nativets.DUCKDB_PACKAGE, " package (").concat(err.message, ")"));
348
- case 4:
352
+ case 3:
349
353
  _state.trys.push([
350
- 4,
351
- 7,
354
+ 3,
355
+ 6,
352
356
  ,
353
- 8
357
+ 7
354
358
  ]);
355
359
  return [
356
360
  4,
@@ -358,7 +362,7 @@ function connect(cfg) {
358
362
  storage_compatibility_version: 'v1.5.0'
359
363
  })
360
364
  ];
361
- case 5:
365
+ case 4:
362
366
  // On-disk files default to an older storage format for cross-version compatibility, which rejects VARIANT columns ("VARIANT
363
367
  // 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.
364
368
  instance = _state.sent();
@@ -366,102 +370,54 @@ function connect(cfg) {
366
370
  4,
367
371
  instance.connect()
368
372
  ];
369
- case 6:
373
+ case 5:
370
374
  duckdb = _state.sent();
371
375
  return [
372
376
  3,
373
- 8
377
+ 7
374
378
  ];
375
- case 7:
379
+ case 6:
376
380
  err1 = _state.sent();
377
381
  // create() may have succeeded before connect() failed: close it, or its WAL stays open.
378
382
  instance === null || instance === void 0 ? void 0 : instance.closeSync();
379
383
  throw new _errorsts.SenseError('STORE_DEPENDENCY_MISSING', 'store "duckdb" failed to open '.concat(dbPath, ": ").concat(err1.message));
380
- case 8:
381
- // A throw below (schema, reconcile's COLUMN_LIMIT) would leak the open instance, whose WAL
382
- // then locks the .duckdb file undeletable on Windows. closed marks the rebuild branch.
383
- closed = false;
384
- _state.label = 9;
385
- case 9:
384
+ case 7:
386
385
  _state.trys.push([
386
+ 7,
387
387
  9,
388
- 16,
389
388
  ,
390
- 17
389
+ 11
391
390
  ]);
392
391
  return [
393
392
  4,
394
393
  (0, _sqlfunctionsts.registerFunctions)(duckdb)
395
394
  ];
396
- case 10:
395
+ case 8:
397
396
  _state.sent();
398
397
  conn = (0, _connectionts.createConnection)(duckdb);
399
- return [
400
- 4,
401
- conn.exec('CREATE TABLE IF NOT EXISTS meta (key TEXT PRIMARY KEY, value TEXT)')
402
- ];
403
- case 11:
404
- _state.sent();
405
- return [
406
- 4,
407
- (0, _sharedts.getMeta)(conn, 'schema_version')
408
- ];
409
- case 12:
410
- version = _state.sent();
411
- return [
412
- 4,
413
- (0, _sharedts.getMeta)(conn, 'features')
414
- ];
415
- case 13:
416
- features = _state.sent();
417
- wantFeatures = (0, _indexts.featureSignature)(cfg, _indexts1.FEATURES);
418
- if (version !== null && version !== SCHEMA_VERSION || features !== null && features !== wantFeatures) {
419
- // Reconcile only reparses changed files, so a stale cache can't be patched incrementally --
420
- // rebuild instead, same as sqlite's open() (see its comment for why this is announced).
421
- reason = version !== null && version !== SCHEMA_VERSION ? 'cache format changed (new sensemaking version)' : 'config change (features, tokenizer, or presets)';
422
- console.error("sense: ".concat(reason, "; rebuilding the index"));
423
- // Close before clearCache deletes the files underneath the still-open instance.
424
- closed = true;
425
- duckdb.disconnectSync();
426
- instance.closeSync();
427
- (0, _cachets.clearCache)(cfg);
428
- return [
429
- 2,
430
- connect(cfg)
431
- ];
432
- }
433
- return [
434
- 4,
435
- ensureSchema(conn, cfg)
436
- ];
437
- case 14:
438
- _state.sent();
439
- return [
440
- 4,
441
- (0, _reconcilets.reconcile)(conn, cfg, cfg.baseDir, _reconcilets1.duckdbDialect)
442
- ];
443
- case 15:
444
- _ref = _state.sent(), parsed = _ref.parsed, warnings = _ref.warnings;
445
398
  return [
446
399
  2,
447
400
  {
448
- instance: instance,
449
- duckdb: duckdb,
450
- conn: conn,
451
- cfg: cfg,
452
- dbPath: dbPath,
453
- parsed: parsed,
454
- warnings: warnings
401
+ handle: {
402
+ instance: instance,
403
+ duckdb: duckdb
404
+ },
405
+ conn: conn
455
406
  }
456
407
  ];
457
- case 16:
408
+ case 9:
458
409
  err2 = _state.sent();
459
- if (!closed) {
460
- duckdb.disconnectSync();
461
- instance === null || instance === void 0 ? void 0 : instance.closeSync();
462
- }
410
+ return [
411
+ 4,
412
+ close({
413
+ instance: instance,
414
+ duckdb: duckdb
415
+ })
416
+ ];
417
+ case 10:
418
+ _state.sent();
463
419
  throw err2;
464
- case 17:
420
+ case 11:
465
421
  return [
466
422
  2
467
423
  ];
@@ -469,32 +425,24 @@ function connect(cfg) {
469
425
  });
470
426
  })();
471
427
  }
428
+ var duckdbOpenDialect = {
429
+ filename: DB_FILENAME,
430
+ schemaVersion: SCHEMA_VERSION,
431
+ reconcileDialect: _reconcilets.duckdbDialect,
432
+ connect: connect,
433
+ close: close,
434
+ ensureSchema: ensureSchema,
435
+ createStore: function createStore(handle, conn) {
436
+ return (0, _storets.createStore)(handle.instance, handle.duckdb, conn);
437
+ }
438
+ };
472
439
  function openDuckdb(cfg) {
473
440
  return _async_to_generator(function() {
474
- var _ref, instance, duckdb, conn, resolvedCfg, dbPath, parsed, warnings, store;
475
441
  return _ts_generator(this, function(_state) {
476
- switch(_state.label){
477
- case 0:
478
- return [
479
- 4,
480
- connect(cfg)
481
- ];
482
- case 1:
483
- _ref = _state.sent(), instance = _ref.instance, duckdb = _ref.duckdb, conn = _ref.conn, resolvedCfg = _ref.cfg, dbPath = _ref.dbPath, parsed = _ref.parsed, warnings = _ref.warnings;
484
- store = (0, _storets.createStore)(instance, duckdb, conn, resolvedCfg, resolvedCfg.baseDir);
485
- // reconcile ran before this object existed, so its chunk text is keyed by the connection.
486
- (0, _handoffts.rekeyChunkText)(conn, store);
487
- return [
488
- 2,
489
- {
490
- store: store,
491
- cfg: resolvedCfg,
492
- dbPath: dbPath,
493
- parsed: parsed,
494
- warnings: warnings
495
- }
496
- ];
497
- }
442
+ return [
443
+ 2,
444
+ (0, _opents.openWithDialect)(cfg, duckdbOpenDialect)
445
+ ];
498
446
  });
499
447
  })();
500
448
  }