sensemaking 0.13.2 → 0.14.0

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.
@@ -47,6 +47,7 @@ _export(exports, {
47
47
  }
48
48
  });
49
49
  var _nodeutil = require("node:util");
50
+ var _columnhintts = require("../column-hint.js");
50
51
  var _indexts = require("../config/index.js");
51
52
  var _indexts1 = require("../db/index.js");
52
53
  var _outputts = require("../output.js");
@@ -466,12 +467,13 @@ function runSql(cfg, sql, params, format, label, preset) {
466
467
  });
467
468
  (0, _outputts.printRowStream)((_statement = statement).iterate.apply(_statement, _to_consumable_array(params)), format, columns);
468
469
  } catch (err) {
470
+ var hinted = (0, _columnhintts.columnHint)(db, err); // pragma needs the db still open
469
471
  db.close();
470
472
  // A saved query or ad-hoc SQL can carry `content MATCH ?` too, so the same FTS5
471
473
  // punctuation trap applies -- the bound parameters are the search terms there. SQL
472
474
  // without MATCH gets its error verbatim; search advice on a plain typo would mislead.
473
- if (/\bMATCH\b/i.test(sql)) throw (0, _searcherrorts.searchError)(err, params.join(' '));
474
- throw err;
475
+ if (/\bMATCH\b/i.test(sql)) throw (0, _searcherrorts.searchError)(hinted, params.join(' '));
476
+ throw hinted;
475
477
  }
476
478
  db.close();
477
479
  }
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/cli/shared.ts"],"sourcesContent":["import type { ParseArgsOptionsConfig } from 'node:util';\nimport { parseArgs } from 'node:util';\nimport type { ResolvedConfig, SearchOverrides } from '../config/index.ts';\nimport { resolvePreset } from '../config/index.ts';\nimport type { OpenResult } from '../db/index.ts';\nimport { open } from '../db/index.ts';\nimport type { Row, RowFormat } from '../output.ts';\nimport { printRowStream } from '../output.ts';\nimport { searchError } from '../search-error.ts';\nimport type { Ctx } from './types.ts';\n\n// Spreadable option fragments -- one flag name keeps one meaning across every command's table.\nexport const FORMAT: ParseArgsOptionsConfig = { format: { type: 'string', default: 'table' } };\nexport const CONFIG: ParseArgsOptionsConfig = { config: { type: 'string' } };\n// The scope vocabulary every scoped command shares: named preset, ad hoc include/exclude\n// globs, a where SQL condition. search adds k on top.\nexport const SCOPE: ParseArgsOptionsConfig = {\n where: { type: 'string' },\n preset: { type: 'string' },\n include: { type: 'string', multiple: true },\n exclude: { type: 'string', multiple: true },\n // Widening, which no other flag can express: --include and --exclude each override their\n // own side of the preset, so neither can drop an exclusion the preset declares.\n 'no-exclude': { type: 'boolean', default: false },\n};\nexport const SEARCH_FLAGS: ParseArgsOptionsConfig = { ...SCOPE, k: { type: 'string' } };\n\ntype Values = Record<string, string | boolean | string[] | undefined>;\n\n// The SCOPE fragment's parsed values as the overrides every scoped command passes down. One\n// place to read them, so a new scope field is added to the fragment and here, not in each\n// command's call.\nexport function scopeOf(values: Values): SearchOverrides {\n return {\n preset: values.preset as string | undefined,\n include: values.include as string[] | undefined,\n exclude: values.exclude as string[] | undefined,\n where: values.where as string | undefined,\n noExclude: values['no-exclude'] === true,\n };\n}\n\n// An unrecognised value used to fall through to `table`, so a typo looked like it worked.\n// parseArgs is strict about flag names; this is the same strictness for the value.\nfunction pickFormat<T extends string>(values: Values, allowed: readonly T[]): T {\n const format = String(values.format);\n if ((allowed as readonly string[]).includes(format)) return format as T;\n console.error(`unknown --format \"${format}\"; expected ${allowed.join(', ')}`);\n process.exit(2);\n}\n\nexport function formatOf(values: Values): 'table' | 'json' {\n return pickFormat(values, ['table', 'json'] as const);\n}\n\n// csv is a row set rendered as rows; map, peek, status, and path render structures instead,\n// and take formatOf.\nexport function rowFormatOf(values: Values): RowFormat {\n return pickFormat(values, ['table', 'json', 'csv'] as const);\n}\n\n// Per-command parseArgs: strict (a foreign flag exits 2), and every table gets --help for free.\nexport function parse(argv: string[], usage: string, options: ParseArgsOptionsConfig): { values: Values; positionals: string[] } {\n let values: Values;\n let positionals: string[];\n try {\n ({ values, positionals } = parseArgs({\n args: argv,\n options: { ...options, help: { type: 'boolean', default: false, short: 'h' } },\n strict: true,\n allowPositionals: true,\n }));\n } catch (err) {\n console.error((err as Error).message);\n console.error(usage);\n process.exit(2);\n }\n if (values.help) {\n console.log(usage);\n process.exit(0);\n }\n return { values, positionals };\n}\n\n// --k must be a positive integer: SQLite reads a bound LIMIT of -1 as \"unlimited\" and 0 as\n// \"nothing\", and parseInt would silently truncate \"5.9\" -- all three are caller mistakes\n// worth a usage error, matching the config-level SavedSearch validation.\nexport function parseK(k: string | undefined, usageError: (message: string) => never): number | undefined {\n if (k === undefined) return undefined;\n const parsed = Number(k);\n if (!Number.isInteger(parsed) || parsed <= 0) usageError(`--k expects a positive integer, got \"${k}\"`);\n return parsed;\n}\n\n// Shared open-query-close envelope for commands that touch the tree.\n\nexport function printWarnings(warnings: string[]): void {\n for (const w of warnings) console.warn(w);\n}\n\nexport async function withDb(ctx: Ctx, configPath: string | undefined, fn: (db: OpenResult['db'], cfg: ResolvedConfig) => void | Promise<void>): Promise<void> {\n const cfg = ctx.resolveConfig(configPath);\n const { db, warnings } = open(cfg);\n printWarnings(warnings);\n try {\n await fn(db, cfg);\n } finally {\n db.close();\n }\n}\n\n// `--preset` on SQL binds the scope rather than applying it: the statement joins `scope`\n// itself. Filtering behind the query's back is not available -- the obvious version, temp views\n// shadowing the base tables, cannot cover `content`, because FTS5 `MATCH` uses the table name\n// as a hidden column and a view has none. A flag that silently scoped three tables of four\n// would look scoped and not be.\nfunction bindScope(db: OpenResult['db'], cfg: ResolvedConfig, preset: string): void {\n const { name } = resolvePreset(cfg, preset); // unknown names throw, listing what is declared\n db.exec('DROP TABLE IF EXISTS temp.scope');\n db.exec('CREATE TEMP TABLE scope (\"path\" TEXT PRIMARY KEY)');\n db.prepare('INSERT INTO temp.scope (\"path\") SELECT \"path\" FROM preset_files WHERE preset = ?').run(name);\n}\n\n// An unbound `?` silently binds NULL, so mismatched param counts fail loudly instead.\nexport function runSql(cfg: ResolvedConfig, sql: string, params: string[], format: RowFormat, label: string, preset?: string): void {\n const placeholderCount = (sql.match(/\\?/g) ?? []).length;\n if (params.length !== placeholderCount) {\n console.error(`${label} expects ${placeholderCount} parameter(s), got ${params.length}`);\n process.exit(2);\n }\n // Naming a preset and never joining it would return the whole index while reading as scoped,\n // which is the one hazard of binding rather than applying. Refuse instead.\n if (preset !== undefined && !/\\bscope\\b/i.test(sql)) {\n console.error(`--preset binds a temporary \"scope\" table of the preset's paths, and ${label} never joins it, so the preset would have no effect`);\n console.error(`add: JOIN scope ON scope.\"path\" = <table>.\"path\"`);\n process.exit(2);\n }\n const { db, warnings } = open(cfg);\n printWarnings(warnings);\n if (preset !== undefined) bindScope(db, cfg, preset);\n // Streamed rather than collected: `sql` is the one caller whose result size is the\n // statement's business, not this package's, so the rows are never held here in bulk.\n // columns() reads the statement's own metadata, so a 0-row csv still has its header.\n // A mid-stream SQLite error (e.g. SQLITE_BUSY outlasting busy_timeout) can still leave\n // partial output; the nonzero exit code is the caller's signal, output completeness is not\n // guaranteed on failure.\n try {\n const statement = db.prepare(sql);\n statement.setReadBigInts(true); // int64 past 2^53 arrives as BigInt instead of throwing at step time\n const columns = statement.columns().map((c) => c.name);\n printRowStream(statement.iterate(...params) as Iterable<Row>, format, columns);\n } catch (err) {\n db.close();\n // A saved query or ad-hoc SQL can carry `content MATCH ?` too, so the same FTS5\n // punctuation trap applies -- the bound parameters are the search terms there. SQL\n // without MATCH gets its error verbatim; search advice on a plain typo would mislead.\n if (/\\bMATCH\\b/i.test(sql)) throw searchError(err as Error, params.join(' '));\n throw err;\n }\n db.close();\n}\n"],"names":["CONFIG","FORMAT","SCOPE","SEARCH_FLAGS","formatOf","parse","parseK","printWarnings","rowFormatOf","runSql","scopeOf","withDb","format","type","default","config","where","preset","include","multiple","exclude","k","values","noExclude","pickFormat","allowed","String","includes","console","error","join","process","exit","argv","usage","options","positionals","parseArgs","args","help","short","strict","allowPositionals","err","message","log","usageError","undefined","parsed","Number","isInteger","warnings","w","warn","ctx","configPath","fn","cfg","open","db","resolveConfig","close","bindScope","name","resolvePreset","exec","prepare","run","sql","params","label","placeholderCount","match","length","test","statement","setReadBigInts","columns","map","c","printRowStream","iterate","searchError"],"mappings":";;;;;;;;;;;QAaaA;eAAAA;;QADAC;eAAAA;;QAIAC;eAAAA;;QASAC;eAAAA;;QA0BGC;eAAAA;;QAWAC;eAAAA;;QAyBAC;eAAAA;;QASAC;eAAAA;;QAvCAC;eAAAA;;QAmEAC;eAAAA;;QA5FAC;eAAAA;;QAoEMC;eAAAA;;;wBAnGI;uBAEI;wBAET;wBAEU;6BACH;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAIrB,IAAMV,SAAiC;IAAEW,QAAQ;QAAEC,MAAM;QAAUC,SAAS;IAAQ;AAAE;AACtF,IAAMd,SAAiC;IAAEe,QAAQ;QAAEF,MAAM;IAAS;AAAE;AAGpE,IAAMX,QAAgC;IAC3Cc,OAAO;QAAEH,MAAM;IAAS;IACxBI,QAAQ;QAAEJ,MAAM;IAAS;IACzBK,SAAS;QAAEL,MAAM;QAAUM,UAAU;IAAK;IAC1CC,SAAS;QAAEP,MAAM;QAAUM,UAAU;IAAK;IAC1C,yFAAyF;IACzF,gFAAgF;IAChF,cAAc;QAAEN,MAAM;QAAWC,SAAS;IAAM;AAClD;AACO,IAAMX,eAAuC,wCAAKD;IAAOmB,GAAG;QAAER,MAAM;IAAS;;AAO7E,SAASH,QAAQY,MAAc;IACpC,OAAO;QACLL,QAAQK,OAAOL,MAAM;QACrBC,SAASI,OAAOJ,OAAO;QACvBE,SAASE,OAAOF,OAAO;QACvBJ,OAAOM,OAAON,KAAK;QACnBO,WAAWD,MAAM,CAAC,aAAa,KAAK;IACtC;AACF;AAEA,0FAA0F;AAC1F,mFAAmF;AACnF,SAASE,WAA6BF,MAAc,EAAEG,OAAqB;IACzE,IAAMb,SAASc,OAAOJ,OAAOV,MAAM;IACnC,IAAI,AAACa,QAA8BE,QAAQ,CAACf,SAAS,OAAOA;IAC5DgB,QAAQC,KAAK,CAAC,AAAC,qBAAyCJ,OAArBb,QAAO,gBAAiC,OAAnBa,QAAQK,IAAI,CAAC;IACrEC,QAAQC,IAAI,CAAC;AACf;AAEO,SAAS5B,SAASkB,MAAc;IACrC,OAAOE,WAAWF,QAAQ;QAAC;QAAS;KAAO;AAC7C;AAIO,SAASd,YAAYc,MAAc;IACxC,OAAOE,WAAWF,QAAQ;QAAC;QAAS;QAAQ;KAAM;AACpD;AAGO,SAASjB,MAAM4B,IAAc,EAAEC,KAAa,EAAEC,OAA+B;IAClF,IAAIb;IACJ,IAAIc;IACJ,IAAI;;cACyBC,IAAAA,mBAAS,EAAC;YACnCC,MAAML;YACNE,SAAS,wCAAKA;gBAASI,MAAM;oBAAE1B,MAAM;oBAAWC,SAAS;oBAAO0B,OAAO;gBAAI;;YAC3EC,QAAQ;YACRC,kBAAkB;QACpB,IALGpB,aAAAA,QAAQc,kBAAAA;IAMb,EAAE,OAAOO,KAAK;QACZf,QAAQC,KAAK,CAAC,AAACc,IAAcC,OAAO;QACpChB,QAAQC,KAAK,CAACK;QACdH,QAAQC,IAAI,CAAC;IACf;IACA,IAAIV,OAAOiB,IAAI,EAAE;QACfX,QAAQiB,GAAG,CAACX;QACZH,QAAQC,IAAI,CAAC;IACf;IACA,OAAO;QAAEV,QAAAA;QAAQc,aAAAA;IAAY;AAC/B;AAKO,SAAS9B,OAAOe,CAAqB,EAAEyB,UAAsC;IAClF,IAAIzB,MAAM0B,WAAW,OAAOA;IAC5B,IAAMC,SAASC,OAAO5B;IACtB,IAAI,CAAC4B,OAAOC,SAAS,CAACF,WAAWA,UAAU,GAAGF,WAAW,AAAC,wCAAyC,OAAFzB,GAAE;IACnG,OAAO2B;AACT;AAIO,SAASzC,cAAc4C,QAAkB;QACzC,kCAAA,2BAAA;;QAAL,QAAK,YAAWA,6BAAX,SAAA,6BAAA,QAAA,yBAAA;YAAA,IAAMC,IAAN;YAAqBxB,QAAQyB,IAAI,CAACD;;;QAAlC;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;AACP;AAEO,SAAezC,OAAO2C,GAAQ,EAAEC,UAA8B,EAAEC,EAAuE;;YACtIC,KACmBC,OAAjBC,IAAIR;;;;oBADNM,MAAMH,IAAIM,aAAa,CAACL;oBACLG,QAAAA,IAAAA,cAAI,EAACD,MAAtBE,KAAiBD,MAAjBC,IAAIR,WAAaO,MAAbP;oBACZ5C,cAAc4C;;;;;;;;;oBAEZ;;wBAAMK,GAAGG,IAAIF;;;oBAAb;;;;;;oBAEAE,GAAGE,KAAK;;;;;;;;;;IAEZ;;AAEA,yFAAyF;AACzF,gGAAgG;AAChG,8FAA8F;AAC9F,2FAA2F;AAC3F,gCAAgC;AAChC,SAASC,UAAUH,EAAoB,EAAEF,GAAmB,EAAExC,MAAc;IAC1E,IAAM,AAAE8C,OAASC,IAAAA,sBAAa,EAACP,KAAKxC,QAA5B8C,MAAqC,gDAAgD;IAC7FJ,GAAGM,IAAI,CAAC;IACRN,GAAGM,IAAI,CAAC;IACRN,GAAGO,OAAO,CAAC,oFAAoFC,GAAG,CAACJ;AACrG;AAGO,SAAStD,OAAOgD,GAAmB,EAAEW,GAAW,EAAEC,MAAgB,EAAEzD,MAAiB,EAAE0D,KAAa,EAAErD,MAAe;QAChGmD;IAA1B,IAAMG,mBAAmB,EAACH,aAAAA,IAAII,KAAK,CAAC,oBAAVJ,wBAAAA,aAAoB,EAAE,EAAEK,MAAM;IACxD,IAAIJ,OAAOI,MAAM,KAAKF,kBAAkB;QACtC3C,QAAQC,KAAK,CAAC,AAAC,GAAmB0C,OAAjBD,OAAM,aAAiDD,OAAtCE,kBAAiB,uBAAmC,OAAdF,OAAOI,MAAM;QACrF1C,QAAQC,IAAI,CAAC;IACf;IACA,6FAA6F;IAC7F,2EAA2E;IAC3E,IAAIf,WAAW8B,aAAa,CAAC,aAAa2B,IAAI,CAACN,MAAM;QACnDxC,QAAQC,KAAK,CAAC,AAAC,wEAA4E,OAANyC,OAAM;QAC3F1C,QAAQC,KAAK,CAAC;QACdE,QAAQC,IAAI,CAAC;IACf;IACA,IAAyB0B,QAAAA,IAAAA,cAAI,EAACD,MAAtBE,KAAiBD,MAAjBC,IAAIR,WAAaO,MAAbP;IACZ5C,cAAc4C;IACd,IAAIlC,WAAW8B,WAAWe,UAAUH,IAAIF,KAAKxC;IAC7C,mFAAmF;IACnF,qFAAqF;IACrF,qFAAqF;IACrF,uFAAuF;IACvF,2FAA2F;IAC3F,yBAAyB;IACzB,IAAI;YAIa0D;QAHf,IAAMA,YAAYhB,GAAGO,OAAO,CAACE;QAC7BO,UAAUC,cAAc,CAAC,OAAO,qEAAqE;QACrG,IAAMC,UAAUF,UAAUE,OAAO,GAAGC,GAAG,CAAC,SAACC;mBAAMA,EAAEhB,IAAI;;QACrDiB,IAAAA,wBAAc,EAACL,CAAAA,aAAAA,WAAUM,OAAO,OAAjBN,YAAkB,qBAAGN,UAA0BzD,QAAQiE;IACxE,EAAE,OAAOlC,KAAK;QACZgB,GAAGE,KAAK;QACR,gFAAgF;QAChF,mFAAmF;QACnF,sFAAsF;QACtF,IAAI,aAAaa,IAAI,CAACN,MAAM,MAAMc,IAAAA,0BAAW,EAACvC,KAAc0B,OAAOvC,IAAI,CAAC;QACxE,MAAMa;IACR;IACAgB,GAAGE,KAAK;AACV"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/cli/shared.ts"],"sourcesContent":["import type { ParseArgsOptionsConfig } from 'node:util';\nimport { parseArgs } from 'node:util';\nimport { columnHint } from '../column-hint.ts';\nimport type { ResolvedConfig, SearchOverrides } from '../config/index.ts';\nimport { resolvePreset } from '../config/index.ts';\nimport type { OpenResult } from '../db/index.ts';\nimport { open } from '../db/index.ts';\nimport type { Row, RowFormat } from '../output.ts';\nimport { printRowStream } from '../output.ts';\nimport { searchError } from '../search-error.ts';\nimport type { Ctx } from './types.ts';\n\n// Spreadable option fragments -- one flag name keeps one meaning across every command's table.\nexport const FORMAT: ParseArgsOptionsConfig = { format: { type: 'string', default: 'table' } };\nexport const CONFIG: ParseArgsOptionsConfig = { config: { type: 'string' } };\n// The scope vocabulary every scoped command shares: named preset, ad hoc include/exclude\n// globs, a where SQL condition. search adds k on top.\nexport const SCOPE: ParseArgsOptionsConfig = {\n where: { type: 'string' },\n preset: { type: 'string' },\n include: { type: 'string', multiple: true },\n exclude: { type: 'string', multiple: true },\n // Widening, which no other flag can express: --include and --exclude each override their\n // own side of the preset, so neither can drop an exclusion the preset declares.\n 'no-exclude': { type: 'boolean', default: false },\n};\nexport const SEARCH_FLAGS: ParseArgsOptionsConfig = { ...SCOPE, k: { type: 'string' } };\n\ntype Values = Record<string, string | boolean | string[] | undefined>;\n\n// The SCOPE fragment's parsed values as the overrides every scoped command passes down. One\n// place to read them, so a new scope field is added to the fragment and here, not in each\n// command's call.\nexport function scopeOf(values: Values): SearchOverrides {\n return {\n preset: values.preset as string | undefined,\n include: values.include as string[] | undefined,\n exclude: values.exclude as string[] | undefined,\n where: values.where as string | undefined,\n noExclude: values['no-exclude'] === true,\n };\n}\n\n// An unrecognised value used to fall through to `table`, so a typo looked like it worked.\n// parseArgs is strict about flag names; this is the same strictness for the value.\nfunction pickFormat<T extends string>(values: Values, allowed: readonly T[]): T {\n const format = String(values.format);\n if ((allowed as readonly string[]).includes(format)) return format as T;\n console.error(`unknown --format \"${format}\"; expected ${allowed.join(', ')}`);\n process.exit(2);\n}\n\nexport function formatOf(values: Values): 'table' | 'json' {\n return pickFormat(values, ['table', 'json'] as const);\n}\n\n// csv is a row set rendered as rows; map, peek, status, and path render structures instead,\n// and take formatOf.\nexport function rowFormatOf(values: Values): RowFormat {\n return pickFormat(values, ['table', 'json', 'csv'] as const);\n}\n\n// Per-command parseArgs: strict (a foreign flag exits 2), and every table gets --help for free.\nexport function parse(argv: string[], usage: string, options: ParseArgsOptionsConfig): { values: Values; positionals: string[] } {\n let values: Values;\n let positionals: string[];\n try {\n ({ values, positionals } = parseArgs({\n args: argv,\n options: { ...options, help: { type: 'boolean', default: false, short: 'h' } },\n strict: true,\n allowPositionals: true,\n }));\n } catch (err) {\n console.error((err as Error).message);\n console.error(usage);\n process.exit(2);\n }\n if (values.help) {\n console.log(usage);\n process.exit(0);\n }\n return { values, positionals };\n}\n\n// --k must be a positive integer: SQLite reads a bound LIMIT of -1 as \"unlimited\" and 0 as\n// \"nothing\", and parseInt would silently truncate \"5.9\" -- all three are caller mistakes\n// worth a usage error, matching the config-level SavedSearch validation.\nexport function parseK(k: string | undefined, usageError: (message: string) => never): number | undefined {\n if (k === undefined) return undefined;\n const parsed = Number(k);\n if (!Number.isInteger(parsed) || parsed <= 0) usageError(`--k expects a positive integer, got \"${k}\"`);\n return parsed;\n}\n\n// Shared open-query-close envelope for commands that touch the tree.\n\nexport function printWarnings(warnings: string[]): void {\n for (const w of warnings) console.warn(w);\n}\n\nexport async function withDb(ctx: Ctx, configPath: string | undefined, fn: (db: OpenResult['db'], cfg: ResolvedConfig) => void | Promise<void>): Promise<void> {\n const cfg = ctx.resolveConfig(configPath);\n const { db, warnings } = open(cfg);\n printWarnings(warnings);\n try {\n await fn(db, cfg);\n } finally {\n db.close();\n }\n}\n\n// `--preset` on SQL binds the scope rather than applying it: the statement joins `scope`\n// itself. Filtering behind the query's back is not available -- the obvious version, temp views\n// shadowing the base tables, cannot cover `content`, because FTS5 `MATCH` uses the table name\n// as a hidden column and a view has none. A flag that silently scoped three tables of four\n// would look scoped and not be.\nfunction bindScope(db: OpenResult['db'], cfg: ResolvedConfig, preset: string): void {\n const { name } = resolvePreset(cfg, preset); // unknown names throw, listing what is declared\n db.exec('DROP TABLE IF EXISTS temp.scope');\n db.exec('CREATE TEMP TABLE scope (\"path\" TEXT PRIMARY KEY)');\n db.prepare('INSERT INTO temp.scope (\"path\") SELECT \"path\" FROM preset_files WHERE preset = ?').run(name);\n}\n\n// An unbound `?` silently binds NULL, so mismatched param counts fail loudly instead.\nexport function runSql(cfg: ResolvedConfig, sql: string, params: string[], format: RowFormat, label: string, preset?: string): void {\n const placeholderCount = (sql.match(/\\?/g) ?? []).length;\n if (params.length !== placeholderCount) {\n console.error(`${label} expects ${placeholderCount} parameter(s), got ${params.length}`);\n process.exit(2);\n }\n // Naming a preset and never joining it would return the whole index while reading as scoped,\n // which is the one hazard of binding rather than applying. Refuse instead.\n if (preset !== undefined && !/\\bscope\\b/i.test(sql)) {\n console.error(`--preset binds a temporary \"scope\" table of the preset's paths, and ${label} never joins it, so the preset would have no effect`);\n console.error(`add: JOIN scope ON scope.\"path\" = <table>.\"path\"`);\n process.exit(2);\n }\n const { db, warnings } = open(cfg);\n printWarnings(warnings);\n if (preset !== undefined) bindScope(db, cfg, preset);\n // Streamed rather than collected: `sql` is the one caller whose result size is the\n // statement's business, not this package's, so the rows are never held here in bulk.\n // columns() reads the statement's own metadata, so a 0-row csv still has its header.\n // A mid-stream SQLite error (e.g. SQLITE_BUSY outlasting busy_timeout) can still leave\n // partial output; the nonzero exit code is the caller's signal, output completeness is not\n // guaranteed on failure.\n try {\n const statement = db.prepare(sql);\n statement.setReadBigInts(true); // int64 past 2^53 arrives as BigInt instead of throwing at step time\n const columns = statement.columns().map((c) => c.name);\n printRowStream(statement.iterate(...params) as Iterable<Row>, format, columns);\n } catch (err) {\n const hinted = columnHint(db, err as Error); // pragma needs the db still open\n db.close();\n // A saved query or ad-hoc SQL can carry `content MATCH ?` too, so the same FTS5\n // punctuation trap applies -- the bound parameters are the search terms there. SQL\n // without MATCH gets its error verbatim; search advice on a plain typo would mislead.\n if (/\\bMATCH\\b/i.test(sql)) throw searchError(hinted, params.join(' '));\n throw hinted;\n }\n db.close();\n}\n"],"names":["CONFIG","FORMAT","SCOPE","SEARCH_FLAGS","formatOf","parse","parseK","printWarnings","rowFormatOf","runSql","scopeOf","withDb","format","type","default","config","where","preset","include","multiple","exclude","k","values","noExclude","pickFormat","allowed","String","includes","console","error","join","process","exit","argv","usage","options","positionals","parseArgs","args","help","short","strict","allowPositionals","err","message","log","usageError","undefined","parsed","Number","isInteger","warnings","w","warn","ctx","configPath","fn","cfg","open","db","resolveConfig","close","bindScope","name","resolvePreset","exec","prepare","run","sql","params","label","placeholderCount","match","length","test","statement","setReadBigInts","columns","map","c","printRowStream","iterate","hinted","columnHint","searchError"],"mappings":";;;;;;;;;;;QAcaA;eAAAA;;QADAC;eAAAA;;QAIAC;eAAAA;;QASAC;eAAAA;;QA0BGC;eAAAA;;QAWAC;eAAAA;;QAyBAC;eAAAA;;QASAC;eAAAA;;QAvCAC;eAAAA;;QAmEAC;eAAAA;;QA5FAC;eAAAA;;QAoEMC;eAAAA;;;wBApGI;4BACC;uBAEG;wBAET;wBAEU;6BACH;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAIrB,IAAMV,SAAiC;IAAEW,QAAQ;QAAEC,MAAM;QAAUC,SAAS;IAAQ;AAAE;AACtF,IAAMd,SAAiC;IAAEe,QAAQ;QAAEF,MAAM;IAAS;AAAE;AAGpE,IAAMX,QAAgC;IAC3Cc,OAAO;QAAEH,MAAM;IAAS;IACxBI,QAAQ;QAAEJ,MAAM;IAAS;IACzBK,SAAS;QAAEL,MAAM;QAAUM,UAAU;IAAK;IAC1CC,SAAS;QAAEP,MAAM;QAAUM,UAAU;IAAK;IAC1C,yFAAyF;IACzF,gFAAgF;IAChF,cAAc;QAAEN,MAAM;QAAWC,SAAS;IAAM;AAClD;AACO,IAAMX,eAAuC,wCAAKD;IAAOmB,GAAG;QAAER,MAAM;IAAS;;AAO7E,SAASH,QAAQY,MAAc;IACpC,OAAO;QACLL,QAAQK,OAAOL,MAAM;QACrBC,SAASI,OAAOJ,OAAO;QACvBE,SAASE,OAAOF,OAAO;QACvBJ,OAAOM,OAAON,KAAK;QACnBO,WAAWD,MAAM,CAAC,aAAa,KAAK;IACtC;AACF;AAEA,0FAA0F;AAC1F,mFAAmF;AACnF,SAASE,WAA6BF,MAAc,EAAEG,OAAqB;IACzE,IAAMb,SAASc,OAAOJ,OAAOV,MAAM;IACnC,IAAI,AAACa,QAA8BE,QAAQ,CAACf,SAAS,OAAOA;IAC5DgB,QAAQC,KAAK,CAAC,AAAC,qBAAyCJ,OAArBb,QAAO,gBAAiC,OAAnBa,QAAQK,IAAI,CAAC;IACrEC,QAAQC,IAAI,CAAC;AACf;AAEO,SAAS5B,SAASkB,MAAc;IACrC,OAAOE,WAAWF,QAAQ;QAAC;QAAS;KAAO;AAC7C;AAIO,SAASd,YAAYc,MAAc;IACxC,OAAOE,WAAWF,QAAQ;QAAC;QAAS;QAAQ;KAAM;AACpD;AAGO,SAASjB,MAAM4B,IAAc,EAAEC,KAAa,EAAEC,OAA+B;IAClF,IAAIb;IACJ,IAAIc;IACJ,IAAI;;cACyBC,IAAAA,mBAAS,EAAC;YACnCC,MAAML;YACNE,SAAS,wCAAKA;gBAASI,MAAM;oBAAE1B,MAAM;oBAAWC,SAAS;oBAAO0B,OAAO;gBAAI;;YAC3EC,QAAQ;YACRC,kBAAkB;QACpB,IALGpB,aAAAA,QAAQc,kBAAAA;IAMb,EAAE,OAAOO,KAAK;QACZf,QAAQC,KAAK,CAAC,AAACc,IAAcC,OAAO;QACpChB,QAAQC,KAAK,CAACK;QACdH,QAAQC,IAAI,CAAC;IACf;IACA,IAAIV,OAAOiB,IAAI,EAAE;QACfX,QAAQiB,GAAG,CAACX;QACZH,QAAQC,IAAI,CAAC;IACf;IACA,OAAO;QAAEV,QAAAA;QAAQc,aAAAA;IAAY;AAC/B;AAKO,SAAS9B,OAAOe,CAAqB,EAAEyB,UAAsC;IAClF,IAAIzB,MAAM0B,WAAW,OAAOA;IAC5B,IAAMC,SAASC,OAAO5B;IACtB,IAAI,CAAC4B,OAAOC,SAAS,CAACF,WAAWA,UAAU,GAAGF,WAAW,AAAC,wCAAyC,OAAFzB,GAAE;IACnG,OAAO2B;AACT;AAIO,SAASzC,cAAc4C,QAAkB;QACzC,kCAAA,2BAAA;;QAAL,QAAK,YAAWA,6BAAX,SAAA,6BAAA,QAAA,yBAAA;YAAA,IAAMC,IAAN;YAAqBxB,QAAQyB,IAAI,CAACD;;;QAAlC;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;AACP;AAEO,SAAezC,OAAO2C,GAAQ,EAAEC,UAA8B,EAAEC,EAAuE;;YACtIC,KACmBC,OAAjBC,IAAIR;;;;oBADNM,MAAMH,IAAIM,aAAa,CAACL;oBACLG,QAAAA,IAAAA,cAAI,EAACD,MAAtBE,KAAiBD,MAAjBC,IAAIR,WAAaO,MAAbP;oBACZ5C,cAAc4C;;;;;;;;;oBAEZ;;wBAAMK,GAAGG,IAAIF;;;oBAAb;;;;;;oBAEAE,GAAGE,KAAK;;;;;;;;;;IAEZ;;AAEA,yFAAyF;AACzF,gGAAgG;AAChG,8FAA8F;AAC9F,2FAA2F;AAC3F,gCAAgC;AAChC,SAASC,UAAUH,EAAoB,EAAEF,GAAmB,EAAExC,MAAc;IAC1E,IAAM,AAAE8C,OAASC,IAAAA,sBAAa,EAACP,KAAKxC,QAA5B8C,MAAqC,gDAAgD;IAC7FJ,GAAGM,IAAI,CAAC;IACRN,GAAGM,IAAI,CAAC;IACRN,GAAGO,OAAO,CAAC,oFAAoFC,GAAG,CAACJ;AACrG;AAGO,SAAStD,OAAOgD,GAAmB,EAAEW,GAAW,EAAEC,MAAgB,EAAEzD,MAAiB,EAAE0D,KAAa,EAAErD,MAAe;QAChGmD;IAA1B,IAAMG,mBAAmB,EAACH,aAAAA,IAAII,KAAK,CAAC,oBAAVJ,wBAAAA,aAAoB,EAAE,EAAEK,MAAM;IACxD,IAAIJ,OAAOI,MAAM,KAAKF,kBAAkB;QACtC3C,QAAQC,KAAK,CAAC,AAAC,GAAmB0C,OAAjBD,OAAM,aAAiDD,OAAtCE,kBAAiB,uBAAmC,OAAdF,OAAOI,MAAM;QACrF1C,QAAQC,IAAI,CAAC;IACf;IACA,6FAA6F;IAC7F,2EAA2E;IAC3E,IAAIf,WAAW8B,aAAa,CAAC,aAAa2B,IAAI,CAACN,MAAM;QACnDxC,QAAQC,KAAK,CAAC,AAAC,wEAA4E,OAANyC,OAAM;QAC3F1C,QAAQC,KAAK,CAAC;QACdE,QAAQC,IAAI,CAAC;IACf;IACA,IAAyB0B,QAAAA,IAAAA,cAAI,EAACD,MAAtBE,KAAiBD,MAAjBC,IAAIR,WAAaO,MAAbP;IACZ5C,cAAc4C;IACd,IAAIlC,WAAW8B,WAAWe,UAAUH,IAAIF,KAAKxC;IAC7C,mFAAmF;IACnF,qFAAqF;IACrF,qFAAqF;IACrF,uFAAuF;IACvF,2FAA2F;IAC3F,yBAAyB;IACzB,IAAI;YAIa0D;QAHf,IAAMA,YAAYhB,GAAGO,OAAO,CAACE;QAC7BO,UAAUC,cAAc,CAAC,OAAO,qEAAqE;QACrG,IAAMC,UAAUF,UAAUE,OAAO,GAAGC,GAAG,CAAC,SAACC;mBAAMA,EAAEhB,IAAI;;QACrDiB,IAAAA,wBAAc,EAACL,CAAAA,aAAAA,WAAUM,OAAO,OAAjBN,YAAkB,qBAAGN,UAA0BzD,QAAQiE;IACxE,EAAE,OAAOlC,KAAK;QACZ,IAAMuC,SAASC,IAAAA,wBAAU,EAACxB,IAAIhB,MAAe,iCAAiC;QAC9EgB,GAAGE,KAAK;QACR,gFAAgF;QAChF,mFAAmF;QACnF,sFAAsF;QACtF,IAAI,aAAaa,IAAI,CAACN,MAAM,MAAMgB,IAAAA,0BAAW,EAACF,QAAQb,OAAOvC,IAAI,CAAC;QAClE,MAAMoD;IACR;IACAvB,GAAGE,KAAK;AACV"}
@@ -0,0 +1,2 @@
1
+ import type { DatabaseSync } from 'node:sqlite';
2
+ export declare function columnHint(db: DatabaseSync, err: Error): Error;
@@ -0,0 +1,2 @@
1
+ import type { DatabaseSync } from 'node:sqlite';
2
+ export declare function columnHint(db: DatabaseSync, err: Error): Error;
@@ -0,0 +1,104 @@
1
+ "use strict";
2
+ Object.defineProperty(exports, "__esModule", {
3
+ value: true
4
+ });
5
+ Object.defineProperty(exports, "columnHint", {
6
+ enumerable: true,
7
+ get: function() {
8
+ return columnHint;
9
+ }
10
+ });
11
+ function _array_like_to_array(arr, len) {
12
+ if (len == null || len > arr.length) len = arr.length;
13
+ for(var i = 0, arr2 = new Array(len); i < len; i++)arr2[i] = arr[i];
14
+ return arr2;
15
+ }
16
+ function _array_without_holes(arr) {
17
+ if (Array.isArray(arr)) return _array_like_to_array(arr);
18
+ }
19
+ function _iterable_to_array(iter) {
20
+ if (typeof Symbol !== "undefined" && iter[Symbol.iterator] != null || iter["@@iterator"] != null) return Array.from(iter);
21
+ }
22
+ function _non_iterable_spread() {
23
+ throw new TypeError("Invalid attempt to spread non-iterable instance.\\nIn order to be iterable, non-array objects must have a [Symbol.iterator]() method.");
24
+ }
25
+ function _to_consumable_array(arr) {
26
+ return _array_without_holes(arr) || _iterable_to_array(arr) || _unsupported_iterable_to_array(arr) || _non_iterable_spread();
27
+ }
28
+ function _unsupported_iterable_to_array(o, minLen) {
29
+ if (!o) return;
30
+ if (typeof o === "string") return _array_like_to_array(o, minLen);
31
+ var n = Object.prototype.toString.call(o).slice(8, -1);
32
+ if (n === "Object" && o.constructor) n = o.constructor.name;
33
+ if (n === "Map" || n === "Set") return Array.from(n);
34
+ if (n === "Arguments" || /^(?:Ui|I)nt(?:8|16|32)(?:Clamped)?Array$/.test(n)) return _array_like_to_array(o, minLen);
35
+ }
36
+ // A frontmatter key with punctuation (`plugin-id`, `a.b`) is a real column, but naming it
37
+ // unquoted in SQL parses as an expression: `plugin-id` is `plugin - id`, and SQLite's error
38
+ // names a fragment (`plugin`) the user never wrote. This maps that fragment back to the
39
+ // columns it could have come from, so the error names the actual fix instead.
40
+ function startsWithBoundary(name, prefix) {
41
+ return name.length > prefix.length && name.startsWith(prefix) && /\W/.test(name[prefix.length]);
42
+ }
43
+ function columnHint(db, err) {
44
+ var match = /no such column: (\S+)/.exec(err.message);
45
+ if (!match) return err;
46
+ var token = match[1];
47
+ // `f.plugin` -> `plugin`; left as-is when there's no leading alias segment, which also
48
+ // covers a column literally named `a.b` -- the full-token check below matches that case
49
+ // directly, so stripping here never needs to special-case it.
50
+ var aliasStripped = token.replace(/^[A-Za-z_]\w*\./, '');
51
+ var forms = new Set([
52
+ token,
53
+ aliasStripped
54
+ ]);
55
+ var columns = db.prepare('PRAGMA table_info(frontmatter)').all().map(function(c) {
56
+ return c.name;
57
+ });
58
+ var candidates = new Set();
59
+ var _iteratorNormalCompletion = true, _didIteratorError = false, _iteratorError = undefined;
60
+ try {
61
+ for(var _iterator = columns[Symbol.iterator](), _step; !(_iteratorNormalCompletion = (_step = _iterator.next()).done); _iteratorNormalCompletion = true){
62
+ var name = _step.value;
63
+ var _iteratorNormalCompletion1 = true, _didIteratorError1 = false, _iteratorError1 = undefined;
64
+ try {
65
+ for(var _iterator1 = forms[Symbol.iterator](), _step1; !(_iteratorNormalCompletion1 = (_step1 = _iterator1.next()).done); _iteratorNormalCompletion1 = true){
66
+ var form = _step1.value;
67
+ if (name === form || startsWithBoundary(name, form)) candidates.add(name);
68
+ }
69
+ } catch (err) {
70
+ _didIteratorError1 = true;
71
+ _iteratorError1 = err;
72
+ } finally{
73
+ try {
74
+ if (!_iteratorNormalCompletion1 && _iterator1.return != null) {
75
+ _iterator1.return();
76
+ }
77
+ } finally{
78
+ if (_didIteratorError1) {
79
+ throw _iteratorError1;
80
+ }
81
+ }
82
+ }
83
+ }
84
+ } catch (err) {
85
+ _didIteratorError = true;
86
+ _iteratorError = err;
87
+ } finally{
88
+ try {
89
+ if (!_iteratorNormalCompletion && _iterator.return != null) {
90
+ _iterator.return();
91
+ }
92
+ } finally{
93
+ if (_didIteratorError) {
94
+ throw _iteratorError;
95
+ }
96
+ }
97
+ }
98
+ if (candidates.size === 0) return err;
99
+ var quoted = _to_consumable_array(candidates).map(function(name) {
100
+ return '"'.concat(name.split('"').join('""'), '"');
101
+ }).join(', ');
102
+ return new Error("".concat(err.message, " -- ").concat(quoted, " needs double quotes in SQL: unquoted, SQLite parses the punctuation as an operator instead of column syntax."));
103
+ }
104
+ /* 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/column-hint.ts"],"sourcesContent":["import type { DatabaseSync } from 'node:sqlite';\n\n// A frontmatter key with punctuation (`plugin-id`, `a.b`) is a real column, but naming it\n// unquoted in SQL parses as an expression: `plugin-id` is `plugin - id`, and SQLite's error\n// names a fragment (`plugin`) the user never wrote. This maps that fragment back to the\n// columns it could have come from, so the error names the actual fix instead.\n\nfunction startsWithBoundary(name: string, prefix: string): boolean {\n return name.length > prefix.length && name.startsWith(prefix) && /\\W/.test(name[prefix.length]);\n}\n\nexport function columnHint(db: DatabaseSync, err: Error): Error {\n const match = /no such column: (\\S+)/.exec(err.message);\n if (!match) return err;\n const token = match[1];\n // `f.plugin` -> `plugin`; left as-is when there's no leading alias segment, which also\n // covers a column literally named `a.b` -- the full-token check below matches that case\n // directly, so stripping here never needs to special-case it.\n const aliasStripped = token.replace(/^[A-Za-z_]\\w*\\./, '');\n const forms = new Set([token, aliasStripped]);\n const columns = (db.prepare('PRAGMA table_info(frontmatter)').all() as Array<{ name: string }>).map((c) => c.name);\n const candidates = new Set<string>();\n for (const name of columns) {\n for (const form of forms) {\n if (name === form || startsWithBoundary(name, form)) candidates.add(name);\n }\n }\n if (candidates.size === 0) return err;\n const quoted = [...candidates].map((name) => `\"${name.split('\"').join('\"\"')}\"`).join(', ');\n return new Error(`${err.message} -- ${quoted} needs double quotes in SQL: unquoted, SQLite parses the punctuation as an operator instead of column syntax.`);\n}\n"],"names":["columnHint","startsWithBoundary","name","prefix","length","startsWith","test","db","err","match","exec","message","token","aliasStripped","replace","forms","Set","columns","prepare","all","map","c","candidates","form","add","size","quoted","split","join","Error"],"mappings":";;;;+BAWgBA;;;eAAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAThB,0FAA0F;AAC1F,4FAA4F;AAC5F,wFAAwF;AACxF,8EAA8E;AAE9E,SAASC,mBAAmBC,IAAY,EAAEC,MAAc;IACtD,OAAOD,KAAKE,MAAM,GAAGD,OAAOC,MAAM,IAAIF,KAAKG,UAAU,CAACF,WAAW,KAAKG,IAAI,CAACJ,IAAI,CAACC,OAAOC,MAAM,CAAC;AAChG;AAEO,SAASJ,WAAWO,EAAgB,EAAEC,GAAU;IACrD,IAAMC,QAAQ,wBAAwBC,IAAI,CAACF,IAAIG,OAAO;IACtD,IAAI,CAACF,OAAO,OAAOD;IACnB,IAAMI,QAAQH,KAAK,CAAC,EAAE;IACtB,uFAAuF;IACvF,wFAAwF;IACxF,8DAA8D;IAC9D,IAAMI,gBAAgBD,MAAME,OAAO,CAAC,mBAAmB;IACvD,IAAMC,QAAQ,IAAIC,IAAI;QAACJ;QAAOC;KAAc;IAC5C,IAAMI,UAAU,AAACV,GAAGW,OAAO,CAAC,kCAAkCC,GAAG,GAA+BC,GAAG,CAAC,SAACC;eAAMA,EAAEnB,IAAI;;IACjH,IAAMoB,aAAa,IAAIN;QAClB,kCAAA,2BAAA;;QAAL,QAAK,YAAcC,4BAAd,SAAA,6BAAA,QAAA,yBAAA,iCAAuB;YAAvB,IAAMf,OAAN;gBACE,mCAAA,4BAAA;;gBAAL,QAAK,aAAca,0BAAd,UAAA,8BAAA,SAAA,0BAAA,kCAAqB;oBAArB,IAAMQ,OAAN;oBACH,IAAIrB,SAASqB,QAAQtB,mBAAmBC,MAAMqB,OAAOD,WAAWE,GAAG,CAACtB;gBACtE;;gBAFK;gBAAA;;;yBAAA,8BAAA;wBAAA;;;wBAAA;8BAAA;;;;QAGP;;QAJK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAKL,IAAIoB,WAAWG,IAAI,KAAK,GAAG,OAAOjB;IAClC,IAAMkB,SAAS,AAAC,qBAAGJ,YAAYF,GAAG,CAAC,SAAClB;eAAS,AAAC,IAA8B,OAA3BA,KAAKyB,KAAK,CAAC,KAAKC,IAAI,CAAC,OAAM;OAAIA,IAAI,CAAC;IACrF,OAAO,IAAIC,MAAM,AAAC,GAAoBH,OAAlBlB,IAAIG,OAAO,EAAC,QAAa,OAAPe,QAAO;AAC/C"}
@@ -35,6 +35,7 @@ _export(exports, {
35
35
  }
36
36
  });
37
37
  var _nodepath = require("node:path");
38
+ var _columnhintts = require("../column-hint.js");
38
39
  var _indexts = require("../config/index.js");
39
40
  function _array_like_to_array(arr, len) {
40
41
  if (len == null || len > arr.length) len = arr.length;
@@ -101,7 +102,12 @@ function rawScope(db, cfg, overrides, allPaths) {
101
102
  }
102
103
  function narrowByWhere(db, paths, where) {
103
104
  if (!where) return paths;
104
- var whereRows = db.prepare('SELECT "path" FROM frontmatter f WHERE ('.concat(where, ")")).all();
105
+ var whereRows;
106
+ try {
107
+ whereRows = db.prepare('SELECT "path" FROM frontmatter f WHERE ('.concat(where, ")")).all();
108
+ } catch (err) {
109
+ throw (0, _columnhintts.columnHint)(db, err);
110
+ }
105
111
  var wherePaths = new Set(whereRows.map(function(r) {
106
112
  return r.path;
107
113
  }));
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/commands/scope.ts"],"sourcesContent":["import { matchesGlob } from 'node:path';\nimport type { DatabaseSync } from 'node:sqlite';\nimport type { ResolvedConfig, SearchOverrides } from '../config/index.ts';\nimport { anyPresetEmbeds, resolveSearch } from '../config/index.ts';\n\nexport const INTERNAL_COLUMNS = new Set(['path', '_mtime', '_ctime', '_size', '_rank', '_parse_error']);\n\n// node:path's matchesGlob is experimental (stable behind an unstable-API flag) as of the\n// engines floor (Node >=22.20); scope filtering only ever needs single-pattern matching, so\n// it's used here in JS rather than running a directory walk in the query path.\nexport function inScope(path: string, include: string[], exclude?: string[]): boolean {\n if (!include.some((g) => matchesGlob(path, g))) return false;\n if (exclude?.some((g) => matchesGlob(path, g))) return false;\n return true;\n}\n\nexport function scopeHasEmbeddings(db: DatabaseSync, cfg: ResolvedConfig, scopedPaths: Set<string>): boolean {\n if (!anyPresetEmbeds(cfg)) return false; // the embeddings table doesn't exist at all in this case\n const rows = db.prepare('SELECT DISTINCT \"path\" FROM embeddings').all() as Array<{ path: string }>;\n return rows.some((r) => scopedPaths.has(r.path));\n}\n\n// Scope only (no --where): preset_files for a named preset, JS glob matching for an ad hoc\n// include/exclude override. Shared by scopedPaths() and search(), which also needs the\n// pre-where set to size the candidate-pool filter.\nexport function rawScope(db: DatabaseSync, cfg: ResolvedConfig, overrides: SearchOverrides, allPaths?: string[]): Set<string> {\n const effective = resolveSearch(cfg, overrides);\n const { include, exclude } = effective;\n const adHocScope = overrides.include !== undefined || overrides.exclude !== undefined || overrides.noExclude === true;\n if (!adHocScope) return new Set((db.prepare('SELECT \"path\" FROM preset_files WHERE preset = ?').all(effective.presetName) as Array<{ path: string }>).map((r) => r.path));\n const paths = allPaths ?? (db.prepare('SELECT \"path\" FROM frontmatter').all() as Array<{ path: string }>).map((r) => r.path);\n return new Set(paths.filter((p) => inScope(p, include, exclude)));\n}\n\nexport function narrowByWhere(db: DatabaseSync, paths: Set<string>, where: string | undefined): Set<string> {\n if (!where) return paths;\n const whereRows = db.prepare(`SELECT \"path\" FROM frontmatter f WHERE (${where})`).all() as Array<{ path: string }>;\n const wherePaths = new Set(whereRows.map((r) => r.path));\n return new Set([...paths].filter((p) => wherePaths.has(p)));\n}\n\n// The scope resolver for non-search commands (path, peek, map): same coverage rule search()\n// applies, then narrowed by the resolved `where`.\nexport function scopedPaths(db: DatabaseSync, cfg: ResolvedConfig, overrides: SearchOverrides): Set<string> {\n const effective = resolveSearch(cfg, overrides);\n return narrowByWhere(db, rawScope(db, cfg, overrides), effective.where);\n}\n\n// Materializes a path set into a named temp table (same shape as traverse.ts's allowed_nodes)\n// so a query can join/filter against it cheaply instead of binding a parameter per path.\nexport function materializeScope(db: DatabaseSync, table: string, paths: Set<string>): void {\n db.exec(`DROP TABLE IF EXISTS ${table}`);\n db.exec(`CREATE TEMP TABLE ${table} (\"path\" TEXT PRIMARY KEY)`);\n db.prepare(`INSERT INTO ${table} SELECT DISTINCT value FROM json_each(?1)`).run(JSON.stringify([...paths]));\n}\n\nexport function setupMapScope(db: DatabaseSync, paths: Set<string>): void {\n materializeScope(db, '_map_scope', paths);\n}\n"],"names":["INTERNAL_COLUMNS","inScope","materializeScope","narrowByWhere","rawScope","scopeHasEmbeddings","scopedPaths","setupMapScope","Set","path","include","exclude","some","g","matchesGlob","db","cfg","anyPresetEmbeds","rows","prepare","all","r","has","overrides","allPaths","effective","resolveSearch","adHocScope","undefined","noExclude","presetName","map","paths","filter","p","where","whereRows","wherePaths","table","exec","run","JSON","stringify"],"mappings":";;;;;;;;;;;QAKaA;eAAAA;;QAKGC;eAAAA;;QAwCAC;eAAAA;;QAhBAC;eAAAA;;QATAC;eAAAA;;QATAC;eAAAA;;QA2BAC;eAAAA;;QAaAC;eAAAA;;;wBAxDY;uBAGmB;;;;;;;;;;;;;;;;;;;;;;;;;;AAExC,IAAMP,mBAAmB,IAAIQ,IAAI;IAAC;IAAQ;IAAU;IAAU;IAAS;IAAS;CAAe;AAK/F,SAASP,QAAQQ,IAAY,EAAEC,OAAiB,EAAEC,OAAkB;IACzE,IAAI,CAACD,QAAQE,IAAI,CAAC,SAACC;eAAMC,IAAAA,qBAAW,EAACL,MAAMI;QAAK,OAAO;IACvD,IAAIF,oBAAAA,8BAAAA,QAASC,IAAI,CAAC,SAACC;eAAMC,IAAAA,qBAAW,EAACL,MAAMI;QAAK,OAAO;IACvD,OAAO;AACT;AAEO,SAASR,mBAAmBU,EAAgB,EAAEC,GAAmB,EAAEV,WAAwB;IAChG,IAAI,CAACW,IAAAA,wBAAe,EAACD,MAAM,OAAO,OAAO,yDAAyD;IAClG,IAAME,OAAOH,GAAGI,OAAO,CAAC,0CAA0CC,GAAG;IACrE,OAAOF,KAAKN,IAAI,CAAC,SAACS;eAAMf,YAAYgB,GAAG,CAACD,EAAEZ,IAAI;;AAChD;AAKO,SAASL,SAASW,EAAgB,EAAEC,GAAmB,EAAEO,SAA0B,EAAEC,QAAmB;IAC7G,IAAMC,YAAYC,IAAAA,sBAAa,EAACV,KAAKO;IACrC,IAAQb,UAAqBe,UAArBf,SAASC,UAAYc,UAAZd;IACjB,IAAMgB,aAAaJ,UAAUb,OAAO,KAAKkB,aAAaL,UAAUZ,OAAO,KAAKiB,aAAaL,UAAUM,SAAS,KAAK;IACjH,IAAI,CAACF,YAAY,OAAO,IAAInB,IAAI,AAACO,GAAGI,OAAO,CAAC,oDAAoDC,GAAG,CAACK,UAAUK,UAAU,EAA8BC,GAAG,CAAC,SAACV;eAAMA,EAAEZ,IAAI;;IACvK,IAAMuB,QAAQR,qBAAAA,sBAAAA,WAAY,AAACT,GAAGI,OAAO,CAAC,kCAAkCC,GAAG,GAA+BW,GAAG,CAAC,SAACV;eAAMA,EAAEZ,IAAI;;IAC3H,OAAO,IAAID,IAAIwB,MAAMC,MAAM,CAAC,SAACC;eAAMjC,QAAQiC,GAAGxB,SAASC;;AACzD;AAEO,SAASR,cAAcY,EAAgB,EAAEiB,KAAkB,EAAEG,KAAyB;IAC3F,IAAI,CAACA,OAAO,OAAOH;IACnB,IAAMI,YAAYrB,GAAGI,OAAO,CAAC,AAAC,2CAAgD,OAANgB,OAAM,MAAIf,GAAG;IACrF,IAAMiB,aAAa,IAAI7B,IAAI4B,UAAUL,GAAG,CAAC,SAACV;eAAMA,EAAEZ,IAAI;;IACtD,OAAO,IAAID,IAAI,AAAC,qBAAGwB,OAAOC,MAAM,CAAC,SAACC;eAAMG,WAAWf,GAAG,CAACY;;AACzD;AAIO,SAAS5B,YAAYS,EAAgB,EAAEC,GAAmB,EAAEO,SAA0B;IAC3F,IAAME,YAAYC,IAAAA,sBAAa,EAACV,KAAKO;IACrC,OAAOpB,cAAcY,IAAIX,SAASW,IAAIC,KAAKO,YAAYE,UAAUU,KAAK;AACxE;AAIO,SAASjC,iBAAiBa,EAAgB,EAAEuB,KAAa,EAAEN,KAAkB;IAClFjB,GAAGwB,IAAI,CAAC,AAAC,wBAA6B,OAAND;IAChCvB,GAAGwB,IAAI,CAAC,AAAC,qBAA0B,OAAND,OAAM;IACnCvB,GAAGI,OAAO,CAAC,AAAC,eAAoB,OAANmB,OAAM,8CAA4CE,GAAG,CAACC,KAAKC,SAAS,CAAE,qBAAGV;AACrG;AAEO,SAASzB,cAAcQ,EAAgB,EAAEiB,KAAkB;IAChE9B,iBAAiBa,IAAI,cAAciB;AACrC"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/commands/scope.ts"],"sourcesContent":["import { matchesGlob } from 'node:path';\nimport type { DatabaseSync } from 'node:sqlite';\nimport { columnHint } from '../column-hint.ts';\nimport type { ResolvedConfig, SearchOverrides } from '../config/index.ts';\nimport { anyPresetEmbeds, resolveSearch } from '../config/index.ts';\n\nexport const INTERNAL_COLUMNS = new Set(['path', '_mtime', '_ctime', '_size', '_rank', '_parse_error']);\n\n// node:path's matchesGlob is experimental (stable behind an unstable-API flag) as of the\n// engines floor (Node >=22.20); scope filtering only ever needs single-pattern matching, so\n// it's used here in JS rather than running a directory walk in the query path.\nexport function inScope(path: string, include: string[], exclude?: string[]): boolean {\n if (!include.some((g) => matchesGlob(path, g))) return false;\n if (exclude?.some((g) => matchesGlob(path, g))) return false;\n return true;\n}\n\nexport function scopeHasEmbeddings(db: DatabaseSync, cfg: ResolvedConfig, scopedPaths: Set<string>): boolean {\n if (!anyPresetEmbeds(cfg)) return false; // the embeddings table doesn't exist at all in this case\n const rows = db.prepare('SELECT DISTINCT \"path\" FROM embeddings').all() as Array<{ path: string }>;\n return rows.some((r) => scopedPaths.has(r.path));\n}\n\n// Scope only (no --where): preset_files for a named preset, JS glob matching for an ad hoc\n// include/exclude override. Shared by scopedPaths() and search(), which also needs the\n// pre-where set to size the candidate-pool filter.\nexport function rawScope(db: DatabaseSync, cfg: ResolvedConfig, overrides: SearchOverrides, allPaths?: string[]): Set<string> {\n const effective = resolveSearch(cfg, overrides);\n const { include, exclude } = effective;\n const adHocScope = overrides.include !== undefined || overrides.exclude !== undefined || overrides.noExclude === true;\n if (!adHocScope) return new Set((db.prepare('SELECT \"path\" FROM preset_files WHERE preset = ?').all(effective.presetName) as Array<{ path: string }>).map((r) => r.path));\n const paths = allPaths ?? (db.prepare('SELECT \"path\" FROM frontmatter').all() as Array<{ path: string }>).map((r) => r.path);\n return new Set(paths.filter((p) => inScope(p, include, exclude)));\n}\n\nexport function narrowByWhere(db: DatabaseSync, paths: Set<string>, where: string | undefined): Set<string> {\n if (!where) return paths;\n let whereRows: Array<{ path: string }>;\n try {\n whereRows = db.prepare(`SELECT \"path\" FROM frontmatter f WHERE (${where})`).all() as Array<{ path: string }>;\n } catch (err) {\n throw columnHint(db, err as Error);\n }\n const wherePaths = new Set(whereRows.map((r) => r.path));\n return new Set([...paths].filter((p) => wherePaths.has(p)));\n}\n\n// The scope resolver for non-search commands (path, peek, map): same coverage rule search()\n// applies, then narrowed by the resolved `where`.\nexport function scopedPaths(db: DatabaseSync, cfg: ResolvedConfig, overrides: SearchOverrides): Set<string> {\n const effective = resolveSearch(cfg, overrides);\n return narrowByWhere(db, rawScope(db, cfg, overrides), effective.where);\n}\n\n// Materializes a path set into a named temp table (same shape as traverse.ts's allowed_nodes)\n// so a query can join/filter against it cheaply instead of binding a parameter per path.\nexport function materializeScope(db: DatabaseSync, table: string, paths: Set<string>): void {\n db.exec(`DROP TABLE IF EXISTS ${table}`);\n db.exec(`CREATE TEMP TABLE ${table} (\"path\" TEXT PRIMARY KEY)`);\n db.prepare(`INSERT INTO ${table} SELECT DISTINCT value FROM json_each(?1)`).run(JSON.stringify([...paths]));\n}\n\nexport function setupMapScope(db: DatabaseSync, paths: Set<string>): void {\n materializeScope(db, '_map_scope', paths);\n}\n"],"names":["INTERNAL_COLUMNS","inScope","materializeScope","narrowByWhere","rawScope","scopeHasEmbeddings","scopedPaths","setupMapScope","Set","path","include","exclude","some","g","matchesGlob","db","cfg","anyPresetEmbeds","rows","prepare","all","r","has","overrides","allPaths","effective","resolveSearch","adHocScope","undefined","noExclude","presetName","map","paths","filter","p","where","whereRows","err","columnHint","wherePaths","table","exec","run","JSON","stringify"],"mappings":";;;;;;;;;;;QAMaA;eAAAA;;QAKGC;eAAAA;;QA6CAC;eAAAA;;QArBAC;eAAAA;;QATAC;eAAAA;;QATAC;eAAAA;;QAgCAC;eAAAA;;QAaAC;eAAAA;;;wBA9DY;4BAED;uBAEoB;;;;;;;;;;;;;;;;;;;;;;;;;;AAExC,IAAMP,mBAAmB,IAAIQ,IAAI;IAAC;IAAQ;IAAU;IAAU;IAAS;IAAS;CAAe;AAK/F,SAASP,QAAQQ,IAAY,EAAEC,OAAiB,EAAEC,OAAkB;IACzE,IAAI,CAACD,QAAQE,IAAI,CAAC,SAACC;eAAMC,IAAAA,qBAAW,EAACL,MAAMI;QAAK,OAAO;IACvD,IAAIF,oBAAAA,8BAAAA,QAASC,IAAI,CAAC,SAACC;eAAMC,IAAAA,qBAAW,EAACL,MAAMI;QAAK,OAAO;IACvD,OAAO;AACT;AAEO,SAASR,mBAAmBU,EAAgB,EAAEC,GAAmB,EAAEV,WAAwB;IAChG,IAAI,CAACW,IAAAA,wBAAe,EAACD,MAAM,OAAO,OAAO,yDAAyD;IAClG,IAAME,OAAOH,GAAGI,OAAO,CAAC,0CAA0CC,GAAG;IACrE,OAAOF,KAAKN,IAAI,CAAC,SAACS;eAAMf,YAAYgB,GAAG,CAACD,EAAEZ,IAAI;;AAChD;AAKO,SAASL,SAASW,EAAgB,EAAEC,GAAmB,EAAEO,SAA0B,EAAEC,QAAmB;IAC7G,IAAMC,YAAYC,IAAAA,sBAAa,EAACV,KAAKO;IACrC,IAAQb,UAAqBe,UAArBf,SAASC,UAAYc,UAAZd;IACjB,IAAMgB,aAAaJ,UAAUb,OAAO,KAAKkB,aAAaL,UAAUZ,OAAO,KAAKiB,aAAaL,UAAUM,SAAS,KAAK;IACjH,IAAI,CAACF,YAAY,OAAO,IAAInB,IAAI,AAACO,GAAGI,OAAO,CAAC,oDAAoDC,GAAG,CAACK,UAAUK,UAAU,EAA8BC,GAAG,CAAC,SAACV;eAAMA,EAAEZ,IAAI;;IACvK,IAAMuB,QAAQR,qBAAAA,sBAAAA,WAAY,AAACT,GAAGI,OAAO,CAAC,kCAAkCC,GAAG,GAA+BW,GAAG,CAAC,SAACV;eAAMA,EAAEZ,IAAI;;IAC3H,OAAO,IAAID,IAAIwB,MAAMC,MAAM,CAAC,SAACC;eAAMjC,QAAQiC,GAAGxB,SAASC;;AACzD;AAEO,SAASR,cAAcY,EAAgB,EAAEiB,KAAkB,EAAEG,KAAyB;IAC3F,IAAI,CAACA,OAAO,OAAOH;IACnB,IAAII;IACJ,IAAI;QACFA,YAAYrB,GAAGI,OAAO,CAAC,AAAC,2CAAgD,OAANgB,OAAM,MAAIf,GAAG;IACjF,EAAE,OAAOiB,KAAK;QACZ,MAAMC,IAAAA,wBAAU,EAACvB,IAAIsB;IACvB;IACA,IAAME,aAAa,IAAI/B,IAAI4B,UAAUL,GAAG,CAAC,SAACV;eAAMA,EAAEZ,IAAI;;IACtD,OAAO,IAAID,IAAI,AAAC,qBAAGwB,OAAOC,MAAM,CAAC,SAACC;eAAMK,WAAWjB,GAAG,CAACY;;AACzD;AAIO,SAAS5B,YAAYS,EAAgB,EAAEC,GAAmB,EAAEO,SAA0B;IAC3F,IAAME,YAAYC,IAAAA,sBAAa,EAACV,KAAKO;IACrC,OAAOpB,cAAcY,IAAIX,SAASW,IAAIC,KAAKO,YAAYE,UAAUU,KAAK;AACxE;AAIO,SAASjC,iBAAiBa,EAAgB,EAAEyB,KAAa,EAAER,KAAkB;IAClFjB,GAAG0B,IAAI,CAAC,AAAC,wBAA6B,OAAND;IAChCzB,GAAG0B,IAAI,CAAC,AAAC,qBAA0B,OAAND,OAAM;IACnCzB,GAAGI,OAAO,CAAC,AAAC,eAAoB,OAANqB,OAAM,8CAA4CE,GAAG,CAACC,KAAKC,SAAS,CAAE,qBAAGZ;AACrG;AAEO,SAASzB,cAAcQ,EAAgB,EAAEiB,KAAkB;IAChE9B,iBAAiBa,IAAI,cAAciB;AACrC"}
@@ -1,7 +1,7 @@
1
1
  import { DatabaseSync } from 'node:sqlite';
2
2
  import type { ResolvedConfig } from '../config/index.js';
3
3
  export declare const DB_FILENAME = "cache.db";
4
- export declare const SCHEMA_VERSION = "13";
4
+ export declare const SCHEMA_VERSION = "14";
5
5
  export interface OpenResult {
6
6
  db: DatabaseSync;
7
7
  cfg: ResolvedConfig;
@@ -1,7 +1,7 @@
1
1
  import { DatabaseSync } from 'node:sqlite';
2
2
  import type { ResolvedConfig } from '../config/index.js';
3
3
  export declare const DB_FILENAME = "cache.db";
4
- export declare const SCHEMA_VERSION = "13";
4
+ export declare const SCHEMA_VERSION = "14";
5
5
  export interface OpenResult {
6
6
  db: DatabaseSync;
7
7
  cfg: ResolvedConfig;
@@ -62,7 +62,7 @@ function _unsupported_iterable_to_array(o, minLen) {
62
62
  if (n === "Arguments" || /^(?:Ui|I)nt(?:8|16|32)(?:Clamped)?Array$/.test(n)) return _array_like_to_array(o, minLen);
63
63
  }
64
64
  var DB_FILENAME = 'cache.db';
65
- var SCHEMA_VERSION = '13';
65
+ var SCHEMA_VERSION = '14';
66
66
  // Stemming is English-only, but the segmentation underneath it is what decides coverage:
67
67
  // unicode61 splits on spaces, so a language written without them (Chinese, Japanese, Thai)
68
68
  // indexes a whole run as one token and word search finds nothing. `content.tokenize` is how
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/db/open.ts"],"sourcesContent":["// The Node floor (>=22.20) is explained here and nowhere else: 22.20 is the first release with\n// both FTS5 and row-returning INSERT ... RETURNING. Raise it only for a load-bearing capability.\nimport { mkdirSync, rmSync } from 'node:fs';\nimport { join } from 'node:path';\nimport { DatabaseSync } from 'node:sqlite';\nimport type { Config, ResolvedConfig } from '../config/index.ts';\nimport { contentTokenize, featureSignature, STATE_DIR } from '../config/index.ts';\nimport { SenseError } from '../errors.ts';\nimport { activeFeatures } from '../features/index.ts';\nimport { registerFunctions } from '../sql-functions.ts';\nimport { changedSignatureKeys, rebuildContentTable, reconcile, signatureDiff } from './reconcile.ts';\nimport { getMeta, setMeta } from './shared.ts';\n\nexport const DB_FILENAME = 'cache.db';\n// Cache shape version, independent of the config's own `version`. Bumping it rebuilds\n// existing trees on first query.\nexport const SCHEMA_VERSION = '13';\n\nexport interface OpenResult {\n db: DatabaseSync;\n cfg: ResolvedConfig;\n dbPath: string;\n parsed: number;\n warnings: string[];\n}\n\n// Stemming is English-only, but the segmentation underneath it is what decides coverage:\n// unicode61 splits on spaces, so a language written without them (Chinese, Japanese, Thai)\n// indexes a whole run as one token and word search finds nothing. `content.tokenize` is how\n// such a tree picks trigram instead.\nconst DEFAULT_TOKENIZE = 'porter unicode61';\n\n// FTS5 takes its tokenizer as a string literal inside DDL, where nothing can bind, so a\n// configured value has to be concatenated. Probing a throwaway table is what makes that safe\n// and is also the whole validation: anything the linked SQLite accepts passes, anything else\n// fails here with SQLite's own message rather than against the real table. It means no table\n// of which version added which tokenizer has to be maintained.\nfunction resolveTokenize(db: DatabaseSync, cfg: Config): string {\n const configured = contentTokenize(cfg);\n if (configured === undefined) return DEFAULT_TOKENIZE;\n const literal = configured.replace(/'/g, \"''\");\n try {\n db.exec('DROP TABLE IF EXISTS temp.sense_tokenize_probe');\n db.exec(`CREATE VIRTUAL TABLE temp.sense_tokenize_probe USING fts5(x, tokenize = '${literal}')`);\n db.exec('DROP TABLE IF EXISTS temp.sense_tokenize_probe');\n } catch (err) {\n throw new SenseError('CONFIG_INVALID', `content.tokenize \"${configured}\" is not a tokenizer this SQLite accepts (${(err as Error).message}); the built-in choices are unicode61, ascii, porter, and trigram, each with their own options`);\n }\n return literal;\n}\n\n// The tokenizer the content table was actually built with, from its own DDL -- the one\n// record that cannot desynchronize from the table. NULL when the table does not exist yet.\nfunction storedTokenize(db: DatabaseSync): string | null {\n const row = db.prepare(`SELECT sql FROM sqlite_master WHERE name = 'content'`).get() as { sql: string } | undefined;\n if (!row) return null;\n const m = row.sql.match(/tokenize = '((?:[^']|'')*)'/);\n return m ? m[1] : null;\n}\n\n// Content is a separate table (not a column on frontmatter) so `SELECT * FROM frontmatter`\n// can't dump file text into context. Features add their own tables after the core ones.\nfunction ensureSchema(db: DatabaseSync, cfg: Config, tokenize: string): void {\n db.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" REAL, \"_ctime\" REAL, \"_size\" INTEGER, \"_parse_error\" TEXT)`);\n // IF NOT EXISTS is safe against a tokenizer change: open() compares the table's own DDL\n // against the resolved tokenizer before this runs, so a stale table is already gone by now.\n // The three `_seg` sidecars are appended after path, never inserted: bm25(content, ...) and\n // snippet(content, 2, ...) are documented against the first three columns and keep working\n // (FTS5 defaults the weights it was not given). Each carries its field's exploded unspaced\n // runs for text that needs it, and an empty string for text that does not.\n db.exec(`CREATE VIRTUAL TABLE IF NOT EXISTS content USING fts5(title, summary, text, path UNINDEXED, title_seg, summary_seg, text_seg, tokenize = '${tokenize}')`);\n // Coverage, not ownership: a path can appear under several presets. path leads the PK so the\n // per-doc delete is an index hit -- keyed the other way, cold builds went quadratic.\n db.exec(`CREATE TABLE IF NOT EXISTS preset_files (\"path\" TEXT, preset TEXT, PRIMARY KEY (\"path\", preset))`);\n db.exec('CREATE INDEX IF NOT EXISTS preset_files_preset ON preset_files(preset)');\n for (const feature of activeFeatures(cfg)) feature.schema(db);\n if (getMeta(db, 'schema_version') === null) setMeta(db, 'schema_version', SCHEMA_VERSION);\n if (getMeta(db, 'features') === null) setMeta(db, 'features', featureSignature(cfg));\n}\n\nexport function docCount(db: DatabaseSync): number {\n const row = db.prepare('SELECT COUNT(*) AS n FROM frontmatter').get() as { n: number };\n return row.n;\n}\n\nexport function open(cfg: ResolvedConfig): OpenResult {\n const stateDir = join(cfg.baseDir, STATE_DIR);\n mkdirSync(stateDir, { recursive: true });\n const dbPath = join(stateDir, DB_FILENAME);\n\n const db = new DatabaseSync(dbPath);\n db.exec('PRAGMA journal_mode = WAL');\n // Covers a concurrent watcher's bulk reconcile (~5s for 500 files at 26k notes). A query\n // that outwaits it still fails loudly.\n db.exec('PRAGMA busy_timeout = 30000');\n registerFunctions(db, contentTokenize(cfg) === undefined);\n\n db.exec('CREATE TABLE IF NOT EXISTS meta (key TEXT PRIMARY KEY, value TEXT)');\n\n // Before the rebuild branch below, never after: that branch deletes the cache, so a typo'd\n // tokenizer validated later would cost a full re-index (and re-embed) to reach its own error.\n const tokenize = resolveTokenize(db, cfg);\n\n // Schema-version or feature-set mismatch: reconcile only reparses changed files, so an\n // old cache can't be patched incrementally -- rebuild instead (cheap: nothing expensive lives here).\n const version = getMeta(db, 'schema_version');\n const features = getMeta(db, 'features');\n const wantFeatures = featureSignature(cfg);\n let tokenizeOnlyRebuild = false;\n if ((version !== null && version !== SCHEMA_VERSION) || (features !== null && features !== wantFeatures)) {\n // Indexing derives from presets, so a config edit rebuilding the cache must say so and\n // name what changed -- silent rebuilds make derived indexing look like a hang or a bug.\n if (version !== null && version !== SCHEMA_VERSION) {\n console.error('sense: cache format changed (new sensemaking version); rebuilding the index');\n db.close();\n clearCache(cfg);\n return open(cfg);\n }\n const changedKeys = changedSignatureKeys(features ?? '', wantFeatures);\n // Only the tokenizer moved: frontmatter, links, sections, and embeddings are file-derived\n // and tokenizer-independent, so they don't need re-deriving. Everything else -- a preset\n // edit, an embed model change -- still takes the full clear/reopen below.\n if (changedKeys.size === 1 && changedKeys.has('tokenize')) {\n console.error('sense: config change (content tokenizer) rebuilds the text index; vectors, links, and sections are kept');\n db.exec('DROP TABLE content');\n tokenizeOnlyRebuild = true;\n } else {\n const changed = signatureDiff(features ?? '', wantFeatures);\n console.error(`sense: config change (${changed}) rebuilds the index`);\n db.close();\n clearCache(cfg);\n return open(cfg);\n }\n }\n\n // Meta can lie after a crash between table creation and the signature write; the table's\n // own DDL cannot. A mismatch here rebuilds no matter what meta says. A tokenize-only rebuild\n // just dropped content above, so storedTokenize sees no table and this guard is skipped\n // naturally rather than needing its own case.\n const stored = storedTokenize(db);\n if (stored !== null && stored !== tokenize) {\n console.error('sense: cache was built with a different content tokenizer; rebuilding the index');\n db.close();\n clearCache(cfg);\n return open(cfg);\n }\n\n ensureSchema(db, cfg, tokenize);\n\n let rebuildWarnings: string[] = [];\n if (tokenizeOnlyRebuild) {\n rebuildWarnings = rebuildContentTable(db, cfg, cfg.baseDir);\n setMeta(db, 'features', wantFeatures);\n }\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(getMeta(db, 'reconcile_max_ms') ?? '0');\n db.exec(`PRAGMA busy_timeout = ${Math.min(Math.max(30000, 3 * recordedMaxMs), 600_000)}`);\n\n const { parsed, warnings } = reconcile(db, cfg, cfg.baseDir);\n\n return { db, cfg, dbPath, parsed, warnings: [...rebuildWarnings, ...warnings] };\n}\n\n// Deletes the cache directory, and only that. The rebuild is not this function's job: the next\n// open() reconciles, which is what running any command already does, so a verb that bundled the\n// two described the half that was not its own. Manual reset for a doubted cache; the schema and\n// config-signature mismatches below reset themselves.\nexport function clearCache(cfg: ResolvedConfig): void {\n rmSync(join(cfg.baseDir, STATE_DIR), { recursive: true, force: true });\n}\n"],"names":["DB_FILENAME","SCHEMA_VERSION","clearCache","docCount","open","DEFAULT_TOKENIZE","resolveTokenize","db","cfg","configured","contentTokenize","undefined","literal","replace","exec","err","SenseError","message","storedTokenize","row","prepare","get","m","sql","match","ensureSchema","tokenize","activeFeatures","feature","schema","getMeta","setMeta","featureSignature","n","stateDir","join","baseDir","STATE_DIR","mkdirSync","recursive","dbPath","DatabaseSync","registerFunctions","version","features","wantFeatures","tokenizeOnlyRebuild","console","error","close","changedKeys","changedSignatureKeys","size","has","changed","signatureDiff","stored","rebuildWarnings","rebuildContentTable","recordedMaxMs","Number","Math","min","max","reconcile","parsed","warnings","rmSync","force"],"mappings":"AAAA,+FAA+F;AAC/F,iGAAiG;;;;;;;;;;;;QAYpFA;eAAAA;;QAGAC;eAAAA;;QAyJGC;eAAAA;;QAzFAC;eAAAA;;QAKAC;eAAAA;;;sBAnFkB;wBACb;0BACQ;uBAEgC;wBAClC;wBACI;8BACG;2BACkD;wBACnD;;;;;;;;;;;;;;;;;;;;;;;;;;AAE1B,IAAMJ,cAAc;AAGpB,IAAMC,iBAAiB;AAU9B,yFAAyF;AACzF,2FAA2F;AAC3F,4FAA4F;AAC5F,qCAAqC;AACrC,IAAMI,mBAAmB;AAEzB,wFAAwF;AACxF,6FAA6F;AAC7F,6FAA6F;AAC7F,6FAA6F;AAC7F,+DAA+D;AAC/D,SAASC,gBAAgBC,EAAgB,EAAEC,GAAW;IACpD,IAAMC,aAAaC,IAAAA,wBAAe,EAACF;IACnC,IAAIC,eAAeE,WAAW,OAAON;IACrC,IAAMO,UAAUH,WAAWI,OAAO,CAAC,MAAM;IACzC,IAAI;QACFN,GAAGO,IAAI,CAAC;QACRP,GAAGO,IAAI,CAAC,AAAC,4EAAmF,OAARF,SAAQ;QAC5FL,GAAGO,IAAI,CAAC;IACV,EAAE,OAAOC,KAAK;QACZ,MAAM,IAAIC,oBAAU,CAAC,kBAAkB,AAAC,qBAA2E,OAAvDP,YAAW,8CAAmE,OAAvB,AAACM,IAAcE,OAAO,EAAC;IAC5I;IACA,OAAOL;AACT;AAEA,uFAAuF;AACvF,2FAA2F;AAC3F,SAASM,eAAeX,EAAgB;IACtC,IAAMY,MAAMZ,GAAGa,OAAO,CAAC,wDAAwDC,GAAG;IAClF,IAAI,CAACF,KAAK,OAAO;IACjB,IAAMG,IAAIH,IAAII,GAAG,CAACC,KAAK,CAAC;IACxB,OAAOF,IAAIA,CAAC,CAAC,EAAE,GAAG;AACpB;AAEA,2FAA2F;AAC3F,wFAAwF;AACxF,SAASG,aAAalB,EAAgB,EAAEC,GAAW,EAAEkB,QAAgB;IACnEnB,GAAGO,IAAI,CAAC;IACR,wFAAwF;IACxF,4FAA4F;IAC5F,4FAA4F;IAC5F,2FAA2F;IAC3F,2FAA2F;IAC3F,2EAA2E;IAC3EP,GAAGO,IAAI,CAAC,AAAC,6IAAqJ,OAATY,UAAS;IAC9J,6FAA6F;IAC7F,qFAAqF;IACrFnB,GAAGO,IAAI,CAAC;IACRP,GAAGO,IAAI,CAAC;QACH,kCAAA,2BAAA;;QAAL,QAAK,YAAiBa,IAAAA,wBAAc,EAACnB,yBAAhC,SAAA,6BAAA,QAAA,yBAAA;YAAA,IAAMoB,UAAN;YAAsCA,QAAQC,MAAM,CAACtB;;;QAArD;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IACL,IAAIuB,IAAAA,iBAAO,EAACvB,IAAI,sBAAsB,MAAMwB,IAAAA,iBAAO,EAACxB,IAAI,kBAAkBN;IAC1E,IAAI6B,IAAAA,iBAAO,EAACvB,IAAI,gBAAgB,MAAMwB,IAAAA,iBAAO,EAACxB,IAAI,YAAYyB,IAAAA,yBAAgB,EAACxB;AACjF;AAEO,SAASL,SAASI,EAAgB;IACvC,IAAMY,MAAMZ,GAAGa,OAAO,CAAC,yCAAyCC,GAAG;IACnE,OAAOF,IAAIc,CAAC;AACd;AAEO,SAAS7B,KAAKI,GAAmB;QAwETsB;IAvE7B,IAAMI,WAAWC,IAAAA,cAAI,EAAC3B,IAAI4B,OAAO,EAAEC,kBAAS;IAC5CC,IAAAA,iBAAS,EAACJ,UAAU;QAAEK,WAAW;IAAK;IACtC,IAAMC,SAASL,IAAAA,cAAI,EAACD,UAAUlC;IAE9B,IAAMO,KAAK,IAAIkC,wBAAY,CAACD;IAC5BjC,GAAGO,IAAI,CAAC;IACR,yFAAyF;IACzF,uCAAuC;IACvCP,GAAGO,IAAI,CAAC;IACR4B,IAAAA,iCAAiB,EAACnC,IAAIG,IAAAA,wBAAe,EAACF,SAASG;IAE/CJ,GAAGO,IAAI,CAAC;IAER,2FAA2F;IAC3F,8FAA8F;IAC9F,IAAMY,WAAWpB,gBAAgBC,IAAIC;IAErC,uFAAuF;IACvF,qGAAqG;IACrG,IAAMmC,UAAUb,IAAAA,iBAAO,EAACvB,IAAI;IAC5B,IAAMqC,WAAWd,IAAAA,iBAAO,EAACvB,IAAI;IAC7B,IAAMsC,eAAeb,IAAAA,yBAAgB,EAACxB;IACtC,IAAIsC,sBAAsB;IAC1B,IAAI,AAACH,YAAY,QAAQA,YAAY1C,kBAAoB2C,aAAa,QAAQA,aAAaC,cAAe;QACxG,uFAAuF;QACvF,wFAAwF;QACxF,IAAIF,YAAY,QAAQA,YAAY1C,gBAAgB;YAClD8C,QAAQC,KAAK,CAAC;YACdzC,GAAG0C,KAAK;YACR/C,WAAWM;YACX,OAAOJ,KAAKI;QACd;QACA,IAAM0C,cAAcC,IAAAA,iCAAoB,EAACP,qBAAAA,sBAAAA,WAAY,IAAIC;QACzD,0FAA0F;QAC1F,yFAAyF;QACzF,0EAA0E;QAC1E,IAAIK,YAAYE,IAAI,KAAK,KAAKF,YAAYG,GAAG,CAAC,aAAa;YACzDN,QAAQC,KAAK,CAAC;YACdzC,GAAGO,IAAI,CAAC;YACRgC,sBAAsB;QACxB,OAAO;YACL,IAAMQ,UAAUC,IAAAA,0BAAa,EAACX,qBAAAA,sBAAAA,WAAY,IAAIC;YAC9CE,QAAQC,KAAK,CAAC,AAAC,yBAAgC,OAARM,SAAQ;YAC/C/C,GAAG0C,KAAK;YACR/C,WAAWM;YACX,OAAOJ,KAAKI;QACd;IACF;IAEA,yFAAyF;IACzF,6FAA6F;IAC7F,wFAAwF;IACxF,8CAA8C;IAC9C,IAAMgD,SAAStC,eAAeX;IAC9B,IAAIiD,WAAW,QAAQA,WAAW9B,UAAU;QAC1CqB,QAAQC,KAAK,CAAC;QACdzC,GAAG0C,KAAK;QACR/C,WAAWM;QACX,OAAOJ,KAAKI;IACd;IAEAiB,aAAalB,IAAIC,KAAKkB;IAEtB,IAAI+B,kBAA4B,EAAE;IAClC,IAAIX,qBAAqB;QACvBW,kBAAkBC,IAAAA,gCAAmB,EAACnD,IAAIC,KAAKA,IAAI4B,OAAO;QAC1DL,IAAAA,iBAAO,EAACxB,IAAI,YAAYsC;IAC1B;IAEA,wFAAwF;IACxF,2FAA2F;IAC3F,IAAMc,gBAAgBC,QAAO9B,WAAAA,IAAAA,iBAAO,EAACvB,IAAI,iCAAZuB,sBAAAA,WAAmC;IAChEvB,GAAGO,IAAI,CAAC,AAAC,yBAA8E,OAAtD+C,KAAKC,GAAG,CAACD,KAAKE,GAAG,CAAC,OAAO,IAAIJ,gBAAgB;IAE9E,IAA6BK,aAAAA,IAAAA,sBAAS,EAACzD,IAAIC,KAAKA,IAAI4B,OAAO,GAAnD6B,SAAqBD,WAArBC,QAAQC,WAAaF,WAAbE;IAEhB,OAAO;QAAE3D,IAAAA;QAAIC,KAAAA;QAAKgC,QAAAA;QAAQyB,QAAAA;QAAQC,UAAU,AAAC,qBAAGT,wBAAiB,qBAAGS;IAAU;AAChF;AAMO,SAAShE,WAAWM,GAAmB;IAC5C2D,IAAAA,cAAM,EAAChC,IAAAA,cAAI,EAAC3B,IAAI4B,OAAO,EAAEC,kBAAS,GAAG;QAAEE,WAAW;QAAM6B,OAAO;IAAK;AACtE"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/db/open.ts"],"sourcesContent":["// The Node floor (>=22.20) is explained here and nowhere else: 22.20 is the first release with\n// both FTS5 and row-returning INSERT ... RETURNING. Raise it only for a load-bearing capability.\nimport { mkdirSync, rmSync } from 'node:fs';\nimport { join } from 'node:path';\nimport { DatabaseSync } from 'node:sqlite';\nimport type { Config, ResolvedConfig } from '../config/index.ts';\nimport { contentTokenize, featureSignature, STATE_DIR } from '../config/index.ts';\nimport { SenseError } from '../errors.ts';\nimport { activeFeatures } from '../features/index.ts';\nimport { registerFunctions } from '../sql-functions.ts';\nimport { changedSignatureKeys, rebuildContentTable, reconcile, signatureDiff } from './reconcile.ts';\nimport { getMeta, setMeta } from './shared.ts';\n\nexport const DB_FILENAME = 'cache.db';\n// Cache shape version, independent of the config's own `version`. Bumping it rebuilds\n// existing trees on first query.\nexport const SCHEMA_VERSION = '14';\n\nexport interface OpenResult {\n db: DatabaseSync;\n cfg: ResolvedConfig;\n dbPath: string;\n parsed: number;\n warnings: string[];\n}\n\n// Stemming is English-only, but the segmentation underneath it is what decides coverage:\n// unicode61 splits on spaces, so a language written without them (Chinese, Japanese, Thai)\n// indexes a whole run as one token and word search finds nothing. `content.tokenize` is how\n// such a tree picks trigram instead.\nconst DEFAULT_TOKENIZE = 'porter unicode61';\n\n// FTS5 takes its tokenizer as a string literal inside DDL, where nothing can bind, so a\n// configured value has to be concatenated. Probing a throwaway table is what makes that safe\n// and is also the whole validation: anything the linked SQLite accepts passes, anything else\n// fails here with SQLite's own message rather than against the real table. It means no table\n// of which version added which tokenizer has to be maintained.\nfunction resolveTokenize(db: DatabaseSync, cfg: Config): string {\n const configured = contentTokenize(cfg);\n if (configured === undefined) return DEFAULT_TOKENIZE;\n const literal = configured.replace(/'/g, \"''\");\n try {\n db.exec('DROP TABLE IF EXISTS temp.sense_tokenize_probe');\n db.exec(`CREATE VIRTUAL TABLE temp.sense_tokenize_probe USING fts5(x, tokenize = '${literal}')`);\n db.exec('DROP TABLE IF EXISTS temp.sense_tokenize_probe');\n } catch (err) {\n throw new SenseError('CONFIG_INVALID', `content.tokenize \"${configured}\" is not a tokenizer this SQLite accepts (${(err as Error).message}); the built-in choices are unicode61, ascii, porter, and trigram, each with their own options`);\n }\n return literal;\n}\n\n// The tokenizer the content table was actually built with, from its own DDL -- the one\n// record that cannot desynchronize from the table. NULL when the table does not exist yet.\nfunction storedTokenize(db: DatabaseSync): string | null {\n const row = db.prepare(`SELECT sql FROM sqlite_master WHERE name = 'content'`).get() as { sql: string } | undefined;\n if (!row) return null;\n const m = row.sql.match(/tokenize = '((?:[^']|'')*)'/);\n return m ? m[1] : null;\n}\n\n// Content is a separate table (not a column on frontmatter) so `SELECT * FROM frontmatter`\n// can't dump file text into context. Features add their own tables after the core ones.\nfunction ensureSchema(db: DatabaseSync, cfg: Config, tokenize: string): void {\n db.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" REAL, \"_ctime\" REAL, \"_size\" INTEGER, \"_parse_error\" TEXT)`);\n // IF NOT EXISTS is safe against a tokenizer change: open() compares the table's own DDL\n // against the resolved tokenizer before this runs, so a stale table is already gone by now.\n // The three `_seg` sidecars are appended after path, never inserted: bm25(content, ...) and\n // snippet(content, 2, ...) are documented against the first three columns and keep working\n // (FTS5 defaults the weights it was not given). Each carries its field's exploded unspaced\n // runs for text that needs it, and an empty string for text that does not.\n db.exec(`CREATE VIRTUAL TABLE IF NOT EXISTS content USING fts5(title, summary, text, path UNINDEXED, title_seg, summary_seg, text_seg, tokenize = '${tokenize}')`);\n // Coverage, not ownership: a path can appear under several presets. path leads the PK so the\n // per-doc delete is an index hit -- keyed the other way, cold builds went quadratic.\n db.exec(`CREATE TABLE IF NOT EXISTS preset_files (\"path\" TEXT, preset TEXT, PRIMARY KEY (\"path\", preset))`);\n db.exec('CREATE INDEX IF NOT EXISTS preset_files_preset ON preset_files(preset)');\n for (const feature of activeFeatures(cfg)) feature.schema(db);\n if (getMeta(db, 'schema_version') === null) setMeta(db, 'schema_version', SCHEMA_VERSION);\n if (getMeta(db, 'features') === null) setMeta(db, 'features', featureSignature(cfg));\n}\n\nexport function docCount(db: DatabaseSync): number {\n const row = db.prepare('SELECT COUNT(*) AS n FROM frontmatter').get() as { n: number };\n return row.n;\n}\n\nexport function open(cfg: ResolvedConfig): OpenResult {\n const stateDir = join(cfg.baseDir, STATE_DIR);\n mkdirSync(stateDir, { recursive: true });\n const dbPath = join(stateDir, DB_FILENAME);\n\n const db = new DatabaseSync(dbPath);\n db.exec('PRAGMA journal_mode = WAL');\n // Covers a concurrent watcher's bulk reconcile (~5s for 500 files at 26k notes). A query\n // that outwaits it still fails loudly.\n db.exec('PRAGMA busy_timeout = 30000');\n registerFunctions(db, contentTokenize(cfg) === undefined);\n\n db.exec('CREATE TABLE IF NOT EXISTS meta (key TEXT PRIMARY KEY, value TEXT)');\n\n // Before the rebuild branch below, never after: that branch deletes the cache, so a typo'd\n // tokenizer validated later would cost a full re-index (and re-embed) to reach its own error.\n const tokenize = resolveTokenize(db, cfg);\n\n // Schema-version or feature-set mismatch: reconcile only reparses changed files, so an\n // old cache can't be patched incrementally -- rebuild instead (cheap: nothing expensive lives here).\n const version = getMeta(db, 'schema_version');\n const features = getMeta(db, 'features');\n const wantFeatures = featureSignature(cfg);\n let tokenizeOnlyRebuild = false;\n if ((version !== null && version !== SCHEMA_VERSION) || (features !== null && features !== wantFeatures)) {\n // Indexing derives from presets, so a config edit rebuilding the cache must say so and\n // name what changed -- silent rebuilds make derived indexing look like a hang or a bug.\n if (version !== null && version !== SCHEMA_VERSION) {\n console.error('sense: cache format changed (new sensemaking version); rebuilding the index');\n db.close();\n clearCache(cfg);\n return open(cfg);\n }\n const changedKeys = changedSignatureKeys(features ?? '', wantFeatures);\n // Only the tokenizer moved: frontmatter, links, sections, and embeddings are file-derived\n // and tokenizer-independent, so they don't need re-deriving. Everything else -- a preset\n // edit, an embed model change -- still takes the full clear/reopen below.\n if (changedKeys.size === 1 && changedKeys.has('tokenize')) {\n console.error('sense: config change (content tokenizer) rebuilds the text index; vectors, links, and sections are kept');\n db.exec('DROP TABLE content');\n tokenizeOnlyRebuild = true;\n } else {\n const changed = signatureDiff(features ?? '', wantFeatures);\n console.error(`sense: config change (${changed}) rebuilds the index`);\n db.close();\n clearCache(cfg);\n return open(cfg);\n }\n }\n\n // Meta can lie after a crash between table creation and the signature write; the table's\n // own DDL cannot. A mismatch here rebuilds no matter what meta says. A tokenize-only rebuild\n // just dropped content above, so storedTokenize sees no table and this guard is skipped\n // naturally rather than needing its own case.\n const stored = storedTokenize(db);\n if (stored !== null && stored !== tokenize) {\n console.error('sense: cache was built with a different content tokenizer; rebuilding the index');\n db.close();\n clearCache(cfg);\n return open(cfg);\n }\n\n ensureSchema(db, cfg, tokenize);\n\n let rebuildWarnings: string[] = [];\n if (tokenizeOnlyRebuild) {\n rebuildWarnings = rebuildContentTable(db, cfg, cfg.baseDir);\n setMeta(db, 'features', wantFeatures);\n }\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(getMeta(db, 'reconcile_max_ms') ?? '0');\n db.exec(`PRAGMA busy_timeout = ${Math.min(Math.max(30000, 3 * recordedMaxMs), 600_000)}`);\n\n const { parsed, warnings } = reconcile(db, cfg, cfg.baseDir);\n\n return { db, cfg, dbPath, parsed, warnings: [...rebuildWarnings, ...warnings] };\n}\n\n// Deletes the cache directory, and only that. The rebuild is not this function's job: the next\n// open() reconciles, which is what running any command already does, so a verb that bundled the\n// two described the half that was not its own. Manual reset for a doubted cache; the schema and\n// config-signature mismatches below reset themselves.\nexport function clearCache(cfg: ResolvedConfig): void {\n rmSync(join(cfg.baseDir, STATE_DIR), { recursive: true, force: true });\n}\n"],"names":["DB_FILENAME","SCHEMA_VERSION","clearCache","docCount","open","DEFAULT_TOKENIZE","resolveTokenize","db","cfg","configured","contentTokenize","undefined","literal","replace","exec","err","SenseError","message","storedTokenize","row","prepare","get","m","sql","match","ensureSchema","tokenize","activeFeatures","feature","schema","getMeta","setMeta","featureSignature","n","stateDir","join","baseDir","STATE_DIR","mkdirSync","recursive","dbPath","DatabaseSync","registerFunctions","version","features","wantFeatures","tokenizeOnlyRebuild","console","error","close","changedKeys","changedSignatureKeys","size","has","changed","signatureDiff","stored","rebuildWarnings","rebuildContentTable","recordedMaxMs","Number","Math","min","max","reconcile","parsed","warnings","rmSync","force"],"mappings":"AAAA,+FAA+F;AAC/F,iGAAiG;;;;;;;;;;;;QAYpFA;eAAAA;;QAGAC;eAAAA;;QAyJGC;eAAAA;;QAzFAC;eAAAA;;QAKAC;eAAAA;;;sBAnFkB;wBACb;0BACQ;uBAEgC;wBAClC;wBACI;8BACG;2BACkD;wBACnD;;;;;;;;;;;;;;;;;;;;;;;;;;AAE1B,IAAMJ,cAAc;AAGpB,IAAMC,iBAAiB;AAU9B,yFAAyF;AACzF,2FAA2F;AAC3F,4FAA4F;AAC5F,qCAAqC;AACrC,IAAMI,mBAAmB;AAEzB,wFAAwF;AACxF,6FAA6F;AAC7F,6FAA6F;AAC7F,6FAA6F;AAC7F,+DAA+D;AAC/D,SAASC,gBAAgBC,EAAgB,EAAEC,GAAW;IACpD,IAAMC,aAAaC,IAAAA,wBAAe,EAACF;IACnC,IAAIC,eAAeE,WAAW,OAAON;IACrC,IAAMO,UAAUH,WAAWI,OAAO,CAAC,MAAM;IACzC,IAAI;QACFN,GAAGO,IAAI,CAAC;QACRP,GAAGO,IAAI,CAAC,AAAC,4EAAmF,OAARF,SAAQ;QAC5FL,GAAGO,IAAI,CAAC;IACV,EAAE,OAAOC,KAAK;QACZ,MAAM,IAAIC,oBAAU,CAAC,kBAAkB,AAAC,qBAA2E,OAAvDP,YAAW,8CAAmE,OAAvB,AAACM,IAAcE,OAAO,EAAC;IAC5I;IACA,OAAOL;AACT;AAEA,uFAAuF;AACvF,2FAA2F;AAC3F,SAASM,eAAeX,EAAgB;IACtC,IAAMY,MAAMZ,GAAGa,OAAO,CAAC,wDAAwDC,GAAG;IAClF,IAAI,CAACF,KAAK,OAAO;IACjB,IAAMG,IAAIH,IAAII,GAAG,CAACC,KAAK,CAAC;IACxB,OAAOF,IAAIA,CAAC,CAAC,EAAE,GAAG;AACpB;AAEA,2FAA2F;AAC3F,wFAAwF;AACxF,SAASG,aAAalB,EAAgB,EAAEC,GAAW,EAAEkB,QAAgB;IACnEnB,GAAGO,IAAI,CAAC;IACR,wFAAwF;IACxF,4FAA4F;IAC5F,4FAA4F;IAC5F,2FAA2F;IAC3F,2FAA2F;IAC3F,2EAA2E;IAC3EP,GAAGO,IAAI,CAAC,AAAC,6IAAqJ,OAATY,UAAS;IAC9J,6FAA6F;IAC7F,qFAAqF;IACrFnB,GAAGO,IAAI,CAAC;IACRP,GAAGO,IAAI,CAAC;QACH,kCAAA,2BAAA;;QAAL,QAAK,YAAiBa,IAAAA,wBAAc,EAACnB,yBAAhC,SAAA,6BAAA,QAAA,yBAAA;YAAA,IAAMoB,UAAN;YAAsCA,QAAQC,MAAM,CAACtB;;;QAArD;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IACL,IAAIuB,IAAAA,iBAAO,EAACvB,IAAI,sBAAsB,MAAMwB,IAAAA,iBAAO,EAACxB,IAAI,kBAAkBN;IAC1E,IAAI6B,IAAAA,iBAAO,EAACvB,IAAI,gBAAgB,MAAMwB,IAAAA,iBAAO,EAACxB,IAAI,YAAYyB,IAAAA,yBAAgB,EAACxB;AACjF;AAEO,SAASL,SAASI,EAAgB;IACvC,IAAMY,MAAMZ,GAAGa,OAAO,CAAC,yCAAyCC,GAAG;IACnE,OAAOF,IAAIc,CAAC;AACd;AAEO,SAAS7B,KAAKI,GAAmB;QAwETsB;IAvE7B,IAAMI,WAAWC,IAAAA,cAAI,EAAC3B,IAAI4B,OAAO,EAAEC,kBAAS;IAC5CC,IAAAA,iBAAS,EAACJ,UAAU;QAAEK,WAAW;IAAK;IACtC,IAAMC,SAASL,IAAAA,cAAI,EAACD,UAAUlC;IAE9B,IAAMO,KAAK,IAAIkC,wBAAY,CAACD;IAC5BjC,GAAGO,IAAI,CAAC;IACR,yFAAyF;IACzF,uCAAuC;IACvCP,GAAGO,IAAI,CAAC;IACR4B,IAAAA,iCAAiB,EAACnC,IAAIG,IAAAA,wBAAe,EAACF,SAASG;IAE/CJ,GAAGO,IAAI,CAAC;IAER,2FAA2F;IAC3F,8FAA8F;IAC9F,IAAMY,WAAWpB,gBAAgBC,IAAIC;IAErC,uFAAuF;IACvF,qGAAqG;IACrG,IAAMmC,UAAUb,IAAAA,iBAAO,EAACvB,IAAI;IAC5B,IAAMqC,WAAWd,IAAAA,iBAAO,EAACvB,IAAI;IAC7B,IAAMsC,eAAeb,IAAAA,yBAAgB,EAACxB;IACtC,IAAIsC,sBAAsB;IAC1B,IAAI,AAACH,YAAY,QAAQA,YAAY1C,kBAAoB2C,aAAa,QAAQA,aAAaC,cAAe;QACxG,uFAAuF;QACvF,wFAAwF;QACxF,IAAIF,YAAY,QAAQA,YAAY1C,gBAAgB;YAClD8C,QAAQC,KAAK,CAAC;YACdzC,GAAG0C,KAAK;YACR/C,WAAWM;YACX,OAAOJ,KAAKI;QACd;QACA,IAAM0C,cAAcC,IAAAA,iCAAoB,EAACP,qBAAAA,sBAAAA,WAAY,IAAIC;QACzD,0FAA0F;QAC1F,yFAAyF;QACzF,0EAA0E;QAC1E,IAAIK,YAAYE,IAAI,KAAK,KAAKF,YAAYG,GAAG,CAAC,aAAa;YACzDN,QAAQC,KAAK,CAAC;YACdzC,GAAGO,IAAI,CAAC;YACRgC,sBAAsB;QACxB,OAAO;YACL,IAAMQ,UAAUC,IAAAA,0BAAa,EAACX,qBAAAA,sBAAAA,WAAY,IAAIC;YAC9CE,QAAQC,KAAK,CAAC,AAAC,yBAAgC,OAARM,SAAQ;YAC/C/C,GAAG0C,KAAK;YACR/C,WAAWM;YACX,OAAOJ,KAAKI;QACd;IACF;IAEA,yFAAyF;IACzF,6FAA6F;IAC7F,wFAAwF;IACxF,8CAA8C;IAC9C,IAAMgD,SAAStC,eAAeX;IAC9B,IAAIiD,WAAW,QAAQA,WAAW9B,UAAU;QAC1CqB,QAAQC,KAAK,CAAC;QACdzC,GAAG0C,KAAK;QACR/C,WAAWM;QACX,OAAOJ,KAAKI;IACd;IAEAiB,aAAalB,IAAIC,KAAKkB;IAEtB,IAAI+B,kBAA4B,EAAE;IAClC,IAAIX,qBAAqB;QACvBW,kBAAkBC,IAAAA,gCAAmB,EAACnD,IAAIC,KAAKA,IAAI4B,OAAO;QAC1DL,IAAAA,iBAAO,EAACxB,IAAI,YAAYsC;IAC1B;IAEA,wFAAwF;IACxF,2FAA2F;IAC3F,IAAMc,gBAAgBC,QAAO9B,WAAAA,IAAAA,iBAAO,EAACvB,IAAI,iCAAZuB,sBAAAA,WAAmC;IAChEvB,GAAGO,IAAI,CAAC,AAAC,yBAA8E,OAAtD+C,KAAKC,GAAG,CAACD,KAAKE,GAAG,CAAC,OAAO,IAAIJ,gBAAgB;IAE9E,IAA6BK,aAAAA,IAAAA,sBAAS,EAACzD,IAAIC,KAAKA,IAAI4B,OAAO,GAAnD6B,SAAqBD,WAArBC,QAAQC,WAAaF,WAAbE;IAEhB,OAAO;QAAE3D,IAAAA;QAAIC,KAAAA;QAAKgC,QAAAA;QAAQyB,QAAAA;QAAQC,UAAU,AAAC,qBAAGT,wBAAiB,qBAAGS;IAAU;AAChF;AAMO,SAAShE,WAAWM,GAAmB;IAC5C2D,IAAAA,cAAM,EAAChC,IAAAA,cAAI,EAAC3B,IAAI4B,OAAO,EAAEC,kBAAS,GAAG;QAAEE,WAAW;QAAM6B,OAAO;IAAK;AACtE"}
@@ -66,10 +66,12 @@ function extract(_raw, body) {
66
66
  };
67
67
  var _iteratorNormalCompletion = true, _didIteratorError = false, _iteratorError = undefined;
68
68
  try {
69
- for(var _iterator = body.matchAll(/\[\[([^\]|#]+)(?:#[^\]|]*)?(?:\|[^\]]*)?\]\]/g)[Symbol.iterator](), _step; !(_iteratorNormalCompletion = (_step = _iterator.next()).done); _iteratorNormalCompletion = true){
69
+ // To the first ]], as Obsidian parses it, so a heading or alias holding a lone ] still
70
+ // matches; anchor and alias split off in code, since a class-based regex stops at the ].
71
+ for(var _iterator = body.matchAll(/\[\[(.*?)\]\]/g)[Symbol.iterator](), _step; !(_iteratorNormalCompletion = (_step = _iterator.next()).done); _iteratorNormalCompletion = true){
70
72
  var m = _step.value;
71
73
  var _m_index;
72
- var target = m[1].trim();
74
+ var target = m[1].split('|')[0].split('#')[0].trim();
73
75
  if (target) add(target, body[((_m_index = m.index) !== null && _m_index !== void 0 ? _m_index : 0) - 1] === '!');
74
76
  }
75
77
  } catch (err) {
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/features/links.ts"],"sourcesContent":["import posix from 'node:path/posix';\nimport type { DatabaseSync } from 'node:sqlite';\nimport type { FileStat } from '../scan.ts';\nimport type { Feature, ReconcileDelta } from './types.ts';\n\n// links(src, target, target_base, dst, embed): target as written, target_base its baseKey\n// (indexed, for the incremental resolve below), dst the resolved path or NULL (a\n// queryable dead link), embed whether it's `![[x]]`/`![](x.md)` rather than `[[x]]`/`[](x.md)`\n// -- Obsidian's grain: a target both linked and embedded in the same note is two rows.\n\n// Wikilinks ([[target]], [[target#anchor|alias]], embeds) plus relative markdown links to .md files.\nfunction extract(_raw: string, body: string): Array<{ target: string; embed: boolean }> {\n const seen = new Set<string>();\n const results: Array<{ target: string; embed: boolean }> = [];\n const add = (target: string, embed: boolean) => {\n const key = `${target}\\0${embed ? '1' : '0'}`;\n if (seen.has(key)) return;\n seen.add(key);\n results.push({ target, embed });\n };\n for (const m of body.matchAll(/\\[\\[([^\\]|#]+)(?:#[^\\]|]*)?(?:\\|[^\\]]*)?\\]\\]/g)) {\n const target = m[1].trim();\n if (target) add(target, body[(m.index ?? 0) - 1] === '!');\n }\n for (const m of body.matchAll(/(!)?\\[[^\\]]*\\]\\(([^)]+\\.md)(?:#[^)]*)?\\)/g)) {\n const target = m[2].trim();\n if (target && !/^[a-z]+:\\/\\//i.test(target)) add(target, m[1] === '!');\n }\n return results;\n}\n\nfunction cleanTarget(target: string): string {\n return target.replace(/\\\\/g, '/').replace(/^\\.\\//, '');\n}\n\nfunction baseKey(path: string): string {\n return posix.basename(path).replace(/\\.md$/i, '').toLowerCase();\n}\n\n// Obsidian-style: exact relative path (with/without .md), path relative to the linking\n// note's directory, then basename match (lexicographically first on ties).\nfunction resolveTarget(src: string, target: string, pathSet: Set<string>, byBase: Map<string, string[]>): string | null {\n const clean = cleanTarget(target);\n const fromSrc = posix.normalize(posix.join(posix.dirname(src), clean));\n for (const candidate of [clean, `${clean}.md`, fromSrc, `${fromSrc}.md`]) {\n if (pathSet.has(candidate)) return candidate;\n }\n return byBase.get(baseKey(clean))?.[0] ?? null;\n}\n\nfunction buildByBase(files: FileStat[]): Map<string, string[]> {\n const byBase = new Map<string, string[]>();\n for (const path of files.map((f) => f.relPath).sort()) {\n const key = baseKey(path);\n const list = byBase.get(key);\n if (list) list.push(path);\n else byBase.set(key, [path]);\n }\n return byBase;\n}\n\ninterface LinkRow {\n src: string;\n target: string;\n dst: string | null;\n embed: number;\n}\n\n// Applies resolveTarget to `rows`, writing only rows whose dst actually changes. The UPDATE\n// keys on (src, target): a link/embed sibling pair shares one resolution, so either row's\n// mismatch rewrites both.\nfunction resolveRows(db: DatabaseSync, rows: LinkRow[], pathSet: Set<string>, byBase: Map<string, string[]>): boolean {\n const update = db.prepare('UPDATE links SET dst = ? WHERE src = ? AND target = ?');\n let changed = false;\n for (const row of rows) {\n const dst = resolveTarget(row.src, row.target, pathSet, byBase);\n if (dst !== row.dst) {\n update.run(dst, row.src, row.target);\n changed = true;\n }\n }\n return changed;\n}\n\n// A new or deleted file can change any note's resolution, so re-resolve the whole table.\n// Fallback for cold builds and large deltas -- see afterReconcile's threshold.\nfunction resolveAll(db: DatabaseSync, files: FileStat[]): boolean {\n const pathSet = new Set(files.map((f) => f.relPath));\n const byBase = buildByBase(files);\n const rows = db.prepare('SELECT src, target, dst, embed FROM links').all() as unknown as LinkRow[];\n return resolveRows(db, rows, pathSet, byBase);\n}\n\n// Re-resolve only what this reconcile could have touched: re-stored sources, links whose\n// target basename matches an added/vanished file, and links whose dst just vanished. All\n// three are indexed lookups, so this never reads the whole table.\nfunction resolveIncremental(db: DatabaseSync, delta: ReconcileDelta): boolean {\n const pathSet = new Set(delta.files.map((f) => f.relPath));\n const byBase = buildByBase(delta.files);\n\n const changedBasenames = new Set<string>();\n for (const p of delta.added) changedBasenames.add(baseKey(p));\n for (const p of delta.vanished) changedBasenames.add(baseKey(p));\n\n // \\0 can't appear in a path or target, so the key never collides; a space can.\n const candidates = new Map<string, LinkRow>();\n const collect = (rows: LinkRow[]) => {\n for (const row of rows) candidates.set(`${row.src}\\0${row.target}\\0${row.embed}`, row);\n };\n // Chunked so a large-but-under-threshold delta never exceeds SQLite's bound-variable\n // limit (SQLITE_MAX_VARIABLE_NUMBER, 32766).\n const collectIn = (column: string, keys: string[]) => {\n for (let i = 0; i < keys.length; i += 500) {\n const chunk = keys.slice(i, i + 500);\n const placeholders = chunk.map(() => '?').join(', ');\n collect(db.prepare(`SELECT src, target, dst, embed FROM links WHERE ${column} IN (${placeholders})`).all(...chunk) as unknown as LinkRow[]);\n }\n };\n\n collectIn('src', delta.reparsed);\n collectIn('target_base', [...changedBasenames]);\n collectIn('dst', delta.vanished);\n\n return resolveRows(db, [...candidates.values()], pathSet, byBase);\n}\n\n// Resolved edges, for rank and for search's graph expansion.\nexport function linkEdges(db: DatabaseSync): [string, string][] {\n // One edge per row, as 0.12's grain always had it -- two written targets resolving to one\n // dst stay two edges (a weight the fusion evals were gated on). The only exclusion is an\n // embed whose exact (src, target) also exists as a link: that second row is new with the\n // embed grain and would double an edge that used to be one row. The NOT EXISTS probes the\n // primary key and fires only for embed rows; both branches still scan the table.\n const sql = `SELECT src, dst FROM links WHERE dst IS NOT NULL AND embed = 0\n UNION ALL\n SELECT src, dst FROM links l WHERE dst IS NOT NULL AND embed = 1\n AND NOT EXISTS (SELECT 1 FROM links l0 WHERE l0.src = l.src AND l0.target = l.target AND l0.embed = 0)`;\n return (db.prepare(sql).all() as Array<{ src: string; dst: string }>).map((r) => [r.src, r.dst]);\n}\n\n// remove() has already deleted the rows by the time afterReconcile runs, so it records here\n// whether they carried edges. Keyed on the delta object, so the state dies with the reconcile.\nconst removedWithLinks = new WeakMap<ReconcileDelta, Set<string>>();\n\nfunction recordRemoved(delta: ReconcileDelta, path: string): void {\n let set = removedWithLinks.get(delta);\n if (!set) {\n set = new Set();\n removedWithLinks.set(delta, set);\n }\n set.add(path);\n}\n\n// remove() runs per path; the vanished list is an array, so cache the Set per delta.\nconst vanishedSets = new WeakMap<ReconcileDelta, Set<string>>();\nfunction vanishedSet(delta: ReconcileDelta): Set<string> {\n let set = vanishedSets.get(delta);\n if (!set) {\n set = new Set(delta.vanished);\n vanishedSets.set(delta, set);\n }\n return set;\n}\n\nexport const links: Feature = {\n name: 'links',\n schema(db) {\n db.exec('CREATE TABLE IF NOT EXISTS links (src TEXT, target TEXT, target_base TEXT, dst TEXT, embed INTEGER, PRIMARY KEY (src, target, embed))');\n db.exec('CREATE INDEX IF NOT EXISTS links_dst ON links(dst)');\n db.exec('CREATE INDEX IF NOT EXISTS links_target_base ON links(target_base)');\n },\n extract,\n // Reparsed docs are diffed in store() below instead of wiped here: deleting and\n // re-inserting resets dst to NULL, which made every touch-only reparse read as an edge\n // change and recompute PageRank -- measured as most of the remaining update cost.\n remove(db, path, delta) {\n if (!vanishedSet(delta).has(path)) return;\n const result = db.prepare('DELETE FROM links WHERE src = ?').run(path);\n if (Number(result.changes) > 0) recordRemoved(delta, path);\n },\n store(db, path, extracted, delta) {\n const targets = extracted as Array<{ target: string; embed: boolean }>;\n // Stale rows (target/embed pairs the new parse no longer contains) are real edge removals.\n // \\0 can't appear in a target, so the composite key never collides.\n let stale: ReturnType<ReturnType<DatabaseSync['prepare']>['run']>;\n if (targets.length === 0) {\n stale = db.prepare('DELETE FROM links WHERE src = ?').run(path);\n } else {\n const placeholders = targets.map(() => '?').join(', ');\n const keys = targets.map((t) => `${t.target}\\0${t.embed ? '1' : '0'}`);\n stale = db.prepare(`DELETE FROM links WHERE src = ? AND (target || char(0) || embed) NOT IN (${placeholders})`).run(path, ...keys);\n }\n if (Number(stale.changes) > 0) recordRemoved(delta, path);\n // Upsert preserves dst on surviving rows, so an unchanged link resolves to the same\n // value and reports no change; only genuinely new rows start at NULL.\n const insert = db.prepare('INSERT INTO links (src, target, target_base, dst, embed) VALUES (?, ?, ?, NULL, ?) ON CONFLICT(src, target, embed) DO UPDATE SET target_base = excluded.target_base');\n for (const { target, embed } of targets) insert.run(path, target, baseKey(cleanTarget(target)), embed ? 1 : 0);\n },\n afterReconcile(db, delta) {\n // Any deleted rows that carried edges mean the edge set shrank -- whether the file\n // vanished or was reparsed down to fewer/no links. dstChanged alone misses the\n // reparsed-to-zero-links case (no surviving rows to re-resolve).\n const removeHadLinks = (removedWithLinks.get(delta)?.size ?? 0) > 0;\n\n const churn = delta.reparsed.length + delta.vanished.length;\n // Past this share of the tree, a full pass is the only one guaranteed to match\n // resolveTarget's ambiguity rules. A cold build clears the threshold on its own.\n const large = delta.files.length === 0 || churn > 0.2 * delta.files.length;\n\n const dstChanged = large ? resolveAll(db, delta.files) : resolveIncremental(db, delta);\n delta.linksChanged = dstChanged || removeHadLinks;\n },\n};\n"],"names":["linkEdges","links","extract","_raw","body","seen","Set","results","add","target","embed","key","has","push","matchAll","m","trim","index","test","cleanTarget","replace","baseKey","path","posix","basename","toLowerCase","resolveTarget","src","pathSet","byBase","clean","fromSrc","normalize","join","dirname","candidate","get","buildByBase","files","Map","map","f","relPath","sort","list","set","resolveRows","db","rows","update","prepare","changed","row","dst","run","resolveAll","all","resolveIncremental","delta","changedBasenames","added","p","vanished","candidates","collect","collectIn","column","keys","i","length","chunk","slice","placeholders","reparsed","values","sql","r","removedWithLinks","WeakMap","recordRemoved","vanishedSets","vanishedSet","name","schema","exec","remove","result","Number","changes","store","extracted","targets","stale","t","insert","afterReconcile","removeHadLinks","size","churn","large","dstChanged","linksChanged"],"mappings":";;;;;;;;;;;QA+HgBA;eAAAA;;QAqCHC;eAAAA;;;4DApKK;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAKlB,0FAA0F;AAC1F,iFAAiF;AACjF,+FAA+F;AAC/F,uFAAuF;AAEvF,qGAAqG;AACrG,SAASC,QAAQC,IAAY,EAAEC,IAAY;IACzC,IAAMC,OAAO,IAAIC;IACjB,IAAMC,UAAqD,EAAE;IAC7D,IAAMC,MAAM,aAACC,QAAgBC;QAC3B,IAAMC,MAAM,AAAC,GAAaD,OAAXD,QAAO,MAAsB,OAAlBC,QAAQ,MAAM;QACxC,IAAIL,KAAKO,GAAG,CAACD,MAAM;QACnBN,KAAKG,GAAG,CAACG;QACTJ,QAAQM,IAAI,CAAC;YAAEJ,QAAAA;YAAQC,OAAAA;QAAM;IAC/B;QACK,kCAAA,2BAAA;;QAAL,QAAK,YAAWN,KAAKU,QAAQ,CAAC,qEAAzB,SAAA,6BAAA,QAAA,yBAAA,iCAA2E;YAA3E,IAAMC,IAAN;gBAE2BA;YAD9B,IAAMN,SAASM,CAAC,CAAC,EAAE,CAACC,IAAI;YACxB,IAAIP,QAAQD,IAAIC,QAAQL,IAAI,CAAC,EAACW,WAAAA,EAAEE,KAAK,cAAPF,sBAAAA,WAAW,KAAK,EAAE,KAAK;QACvD;;QAHK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;QAIA,mCAAA,4BAAA;;QAAL,QAAK,aAAWX,KAAKU,QAAQ,CAAC,iEAAzB,UAAA,8BAAA,SAAA,0BAAA,kCAAuE;YAAvE,IAAMC,KAAN;YACH,IAAMN,UAASM,EAAC,CAAC,EAAE,CAACC,IAAI;YACxB,IAAIP,WAAU,CAAC,gBAAgBS,IAAI,CAACT,UAASD,IAAIC,SAAQM,EAAC,CAAC,EAAE,KAAK;QACpE;;QAHK;QAAA;;;iBAAA,8BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAIL,OAAOR;AACT;AAEA,SAASY,YAAYV,MAAc;IACjC,OAAOA,OAAOW,OAAO,CAAC,OAAO,KAAKA,OAAO,CAAC,SAAS;AACrD;AAEA,SAASC,QAAQC,IAAY;IAC3B,OAAOC,cAAK,CAACC,QAAQ,CAACF,MAAMF,OAAO,CAAC,UAAU,IAAIK,WAAW;AAC/D;AAEA,uFAAuF;AACvF,2EAA2E;AAC3E,SAASC,cAAcC,GAAW,EAAElB,MAAc,EAAEmB,OAAoB,EAAEC,MAA6B;;QAM9FA;IALP,IAAMC,QAAQX,YAAYV;IAC1B,IAAMsB,UAAUR,cAAK,CAACS,SAAS,CAACT,cAAK,CAACU,IAAI,CAACV,cAAK,CAACW,OAAO,CAACP,MAAMG;IAC/D,gBAAwB,QAAA;QAACA;QAAQ,GAAQ,OAANA,OAAM;QAAMC;QAAU,GAAU,OAARA,SAAQ;KAAK,OAAhD,mBAAkD;YAA/DI,YAAa;QACtB,IAAIP,QAAQhB,GAAG,CAACuB,YAAY,OAAOA;IACrC;IACA,gBAAON,cAAAA,OAAOO,GAAG,CAACf,QAAQS,qBAAnBD,kCAAAA,WAA4B,CAAC,EAAE,uCAAI;AAC5C;AAEA,SAASQ,YAAYC,KAAiB;IACpC,IAAMT,SAAS,IAAIU;QACd,kCAAA,2BAAA;;QAAL,QAAK,YAAcD,MAAME,GAAG,CAAC,SAACC;mBAAMA,EAAEC,OAAO;WAAEC,IAAI,uBAA9C,SAAA,6BAAA,QAAA,yBAAA,iCAAkD;YAAlD,IAAMrB,OAAN;YACH,IAAMX,MAAMU,QAAQC;YACpB,IAAMsB,OAAOf,OAAOO,GAAG,CAACzB;YACxB,IAAIiC,MAAMA,KAAK/B,IAAI,CAACS;iBACfO,OAAOgB,GAAG,CAAClC,KAAK;gBAACW;aAAK;QAC7B;;QALK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAML,OAAOO;AACT;AASA,4FAA4F;AAC5F,0FAA0F;AAC1F,0BAA0B;AAC1B,SAASiB,YAAYC,EAAgB,EAAEC,IAAe,EAAEpB,OAAoB,EAAEC,MAA6B;IACzG,IAAMoB,SAASF,GAAGG,OAAO,CAAC;IAC1B,IAAIC,UAAU;QACT,kCAAA,2BAAA;;QAAL,QAAK,YAAaH,yBAAb,SAAA,6BAAA,QAAA,yBAAA,iCAAmB;YAAnB,IAAMI,MAAN;YACH,IAAMC,MAAM3B,cAAc0B,IAAIzB,GAAG,EAAEyB,IAAI3C,MAAM,EAAEmB,SAASC;YACxD,IAAIwB,QAAQD,IAAIC,GAAG,EAAE;gBACnBJ,OAAOK,GAAG,CAACD,KAAKD,IAAIzB,GAAG,EAAEyB,IAAI3C,MAAM;gBACnC0C,UAAU;YACZ;QACF;;QANK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAOL,OAAOA;AACT;AAEA,yFAAyF;AACzF,+EAA+E;AAC/E,SAASI,WAAWR,EAAgB,EAAET,KAAiB;IACrD,IAAMV,UAAU,IAAItB,IAAIgC,MAAME,GAAG,CAAC,SAACC;eAAMA,EAAEC,OAAO;;IAClD,IAAMb,SAASQ,YAAYC;IAC3B,IAAMU,OAAOD,GAAGG,OAAO,CAAC,6CAA6CM,GAAG;IACxE,OAAOV,YAAYC,IAAIC,MAAMpB,SAASC;AACxC;AAEA,yFAAyF;AACzF,yFAAyF;AACzF,kEAAkE;AAClE,SAAS4B,mBAAmBV,EAAgB,EAAEW,KAAqB;IACjE,IAAM9B,UAAU,IAAItB,IAAIoD,MAAMpB,KAAK,CAACE,GAAG,CAAC,SAACC;eAAMA,EAAEC,OAAO;;IACxD,IAAMb,SAASQ,YAAYqB,MAAMpB,KAAK;IAEtC,IAAMqB,mBAAmB,IAAIrD;QACxB,kCAAA,2BAAA;;QAAL,QAAK,YAAWoD,MAAME,KAAK,qBAAtB,SAAA,6BAAA,QAAA,yBAAA;YAAA,IAAMC,IAAN;YAAwBF,iBAAiBnD,GAAG,CAACa,QAAQwC;;;QAArD;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;QACA,mCAAA,4BAAA;;QAAL,QAAK,aAAWH,MAAMI,QAAQ,qBAAzB,UAAA,8BAAA,SAAA,0BAAA;YAAA,IAAMD,KAAN;YAA2BF,iBAAiBnD,GAAG,CAACa,QAAQwC;;;QAAxD;QAAA;;;iBAAA,8BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAEL,+EAA+E;IAC/E,IAAME,aAAa,IAAIxB;IACvB,IAAMyB,UAAU,iBAAChB;YACV,kCAAA,2BAAA;;YAAL,QAAK,YAAaA,yBAAb,SAAA,6BAAA,QAAA,yBAAA;gBAAA,IAAMI,MAAN;gBAAmBW,WAAWlB,GAAG,CAAC,AAAC,GAAcO,OAAZA,IAAIzB,GAAG,EAAC,MAAmByB,OAAfA,IAAI3C,MAAM,EAAC,MAAc,OAAV2C,IAAI1C,KAAK,GAAI0C;;;YAA7E;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;IACP;IACA,qFAAqF;IACrF,6CAA6C;IAC7C,IAAMa,YAAY,mBAACC,QAAgBC;QACjC,IAAK,IAAIC,IAAI,GAAGA,IAAID,KAAKE,MAAM,EAAED,KAAK,IAAK;gBAGjCrB;YAFR,IAAMuB,QAAQH,KAAKI,KAAK,CAACH,GAAGA,IAAI;YAChC,IAAMI,eAAeF,MAAM9B,GAAG,CAAC;uBAAM;eAAKP,IAAI,CAAC;YAC/C+B,QAAQjB,CAAAA,cAAAA,GAAGG,OAAO,CAAC,AAAC,mDAAgEsB,OAAdN,QAAO,SAAoB,OAAbM,cAAa,OAAIhB,GAAG,OAAhGT,aAAiG,qBAAGuB;QAC9G;IACF;IAEAL,UAAU,OAAOP,MAAMe,QAAQ;IAC/BR,UAAU,eAAgB,qBAAGN;IAC7BM,UAAU,OAAOP,MAAMI,QAAQ;IAE/B,OAAOhB,YAAYC,IAAK,qBAAGgB,WAAWW,MAAM,KAAK9C,SAASC;AAC5D;AAGO,SAAS7B,UAAU+C,EAAgB;IACxC,0FAA0F;IAC1F,yFAAyF;IACzF,yFAAyF;IACzF,0FAA0F;IAC1F,iFAAiF;IACjF,IAAM4B,MAAM;IAIZ,OAAO,AAAC5B,GAAGG,OAAO,CAACyB,KAAKnB,GAAG,GAA2ChB,GAAG,CAAC,SAACoC;eAAM;YAACA,EAAEjD,GAAG;YAAEiD,EAAEvB,GAAG;SAAC;;AACjG;AAEA,4FAA4F;AAC5F,+FAA+F;AAC/F,IAAMwB,mBAAmB,IAAIC;AAE7B,SAASC,cAAcrB,KAAqB,EAAEpC,IAAY;IACxD,IAAIuB,MAAMgC,iBAAiBzC,GAAG,CAACsB;IAC/B,IAAI,CAACb,KAAK;QACRA,MAAM,IAAIvC;QACVuE,iBAAiBhC,GAAG,CAACa,OAAOb;IAC9B;IACAA,IAAIrC,GAAG,CAACc;AACV;AAEA,qFAAqF;AACrF,IAAM0D,eAAe,IAAIF;AACzB,SAASG,YAAYvB,KAAqB;IACxC,IAAIb,MAAMmC,aAAa5C,GAAG,CAACsB;IAC3B,IAAI,CAACb,KAAK;QACRA,MAAM,IAAIvC,IAAIoD,MAAMI,QAAQ;QAC5BkB,aAAanC,GAAG,CAACa,OAAOb;IAC1B;IACA,OAAOA;AACT;AAEO,IAAM5C,QAAiB;IAC5BiF,MAAM;IACNC,QAAAA,SAAAA,OAAOpC,EAAE;QACPA,GAAGqC,IAAI,CAAC;QACRrC,GAAGqC,IAAI,CAAC;QACRrC,GAAGqC,IAAI,CAAC;IACV;IACAlF,SAAAA;IACA,gFAAgF;IAChF,uFAAuF;IACvF,kFAAkF;IAClFmF,QAAAA,SAAAA,OAAOtC,EAAE,EAAEzB,IAAI,EAAEoC,KAAK;QACpB,IAAI,CAACuB,YAAYvB,OAAO9C,GAAG,CAACU,OAAO;QACnC,IAAMgE,SAASvC,GAAGG,OAAO,CAAC,mCAAmCI,GAAG,CAAChC;QACjE,IAAIiE,OAAOD,OAAOE,OAAO,IAAI,GAAGT,cAAcrB,OAAOpC;IACvD;IACAmE,OAAAA,SAAAA,MAAM1C,EAAE,EAAEzB,IAAI,EAAEoE,SAAS,EAAEhC,KAAK;QAC9B,IAAMiC,UAAUD;QAChB,2FAA2F;QAC3F,oEAAoE;QACpE,IAAIE;QACJ,IAAID,QAAQtB,MAAM,KAAK,GAAG;YACxBuB,QAAQ7C,GAAGG,OAAO,CAAC,mCAAmCI,GAAG,CAAChC;QAC5D,OAAO;gBAGGyB;YAFR,IAAMyB,eAAemB,QAAQnD,GAAG,CAAC;uBAAM;eAAKP,IAAI,CAAC;YACjD,IAAMkC,OAAOwB,QAAQnD,GAAG,CAAC,SAACqD;uBAAM,AAAC,GAAeA,OAAbA,EAAEpF,MAAM,EAAC,MAAwB,OAApBoF,EAAEnF,KAAK,GAAG,MAAM;;YAChEkF,QAAQ7C,CAAAA,cAAAA,GAAGG,OAAO,CAAC,AAAC,4EAAwF,OAAbsB,cAAa,OAAIlB,GAAG,OAA3GP,aAAAA;gBAA4GzB;aAAc,CAA1HyB,OAAkH,qBAAGoB;QAC/H;QACA,IAAIoB,OAAOK,MAAMJ,OAAO,IAAI,GAAGT,cAAcrB,OAAOpC;QACpD,oFAAoF;QACpF,sEAAsE;QACtE,IAAMwE,SAAS/C,GAAGG,OAAO,CAAC;YACrB,kCAAA,2BAAA;;YAAL,QAAK,YAA2ByC,4BAA3B,SAAA,6BAAA,QAAA,yBAAA;gBAAA,kBAAA,aAAQlF,qBAAAA,QAAQC,oBAAAA;gBAAoBoF,OAAOxC,GAAG,CAAChC,MAAMb,QAAQY,QAAQF,YAAYV,UAAUC,QAAQ,IAAI;;;YAAvG;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;IACP;IACAqF,gBAAAA,SAAAA,eAAehD,EAAE,EAAEW,KAAK;;YAIEmB;QAHxB,mFAAmF;QACnF,+EAA+E;QAC/E,iEAAiE;QACjE,IAAMmB,iBAAiB,UAACnB,wBAAAA,iBAAiBzC,GAAG,CAACsB,oBAArBmB,4CAAAA,sBAA6BoB,IAAI,uCAAI,KAAK;QAElE,IAAMC,QAAQxC,MAAMe,QAAQ,CAACJ,MAAM,GAAGX,MAAMI,QAAQ,CAACO,MAAM;QAC3D,+EAA+E;QAC/E,iFAAiF;QACjF,IAAM8B,QAAQzC,MAAMpB,KAAK,CAAC+B,MAAM,KAAK,KAAK6B,QAAQ,MAAMxC,MAAMpB,KAAK,CAAC+B,MAAM;QAE1E,IAAM+B,aAAaD,QAAQ5C,WAAWR,IAAIW,MAAMpB,KAAK,IAAImB,mBAAmBV,IAAIW;QAChFA,MAAM2C,YAAY,GAAGD,cAAcJ;IACrC;AACF"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/features/links.ts"],"sourcesContent":["import posix from 'node:path/posix';\nimport type { DatabaseSync } from 'node:sqlite';\nimport type { FileStat } from '../scan.ts';\nimport type { Feature, ReconcileDelta } from './types.ts';\n\n// links(src, target, target_base, dst, embed): target as written, target_base its baseKey\n// (indexed, for the incremental resolve below), dst the resolved path or NULL (a\n// queryable dead link), embed whether it's `![[x]]`/`![](x.md)` rather than `[[x]]`/`[](x.md)`\n// -- Obsidian's grain: a target both linked and embedded in the same note is two rows.\n\n// Wikilinks ([[target]], [[target#anchor|alias]], embeds) plus relative markdown links to .md files.\nfunction extract(_raw: string, body: string): Array<{ target: string; embed: boolean }> {\n const seen = new Set<string>();\n const results: Array<{ target: string; embed: boolean }> = [];\n const add = (target: string, embed: boolean) => {\n const key = `${target}\\0${embed ? '1' : '0'}`;\n if (seen.has(key)) return;\n seen.add(key);\n results.push({ target, embed });\n };\n // To the first ]], as Obsidian parses it, so a heading or alias holding a lone ] still\n // matches; anchor and alias split off in code, since a class-based regex stops at the ].\n for (const m of body.matchAll(/\\[\\[(.*?)\\]\\]/g)) {\n const target = m[1].split('|')[0].split('#')[0].trim();\n if (target) add(target, body[(m.index ?? 0) - 1] === '!');\n }\n for (const m of body.matchAll(/(!)?\\[[^\\]]*\\]\\(([^)]+\\.md)(?:#[^)]*)?\\)/g)) {\n const target = m[2].trim();\n if (target && !/^[a-z]+:\\/\\//i.test(target)) add(target, m[1] === '!');\n }\n return results;\n}\n\nfunction cleanTarget(target: string): string {\n return target.replace(/\\\\/g, '/').replace(/^\\.\\//, '');\n}\n\nfunction baseKey(path: string): string {\n return posix.basename(path).replace(/\\.md$/i, '').toLowerCase();\n}\n\n// Obsidian-style: exact relative path (with/without .md), path relative to the linking\n// note's directory, then basename match (lexicographically first on ties).\nfunction resolveTarget(src: string, target: string, pathSet: Set<string>, byBase: Map<string, string[]>): string | null {\n const clean = cleanTarget(target);\n const fromSrc = posix.normalize(posix.join(posix.dirname(src), clean));\n for (const candidate of [clean, `${clean}.md`, fromSrc, `${fromSrc}.md`]) {\n if (pathSet.has(candidate)) return candidate;\n }\n return byBase.get(baseKey(clean))?.[0] ?? null;\n}\n\nfunction buildByBase(files: FileStat[]): Map<string, string[]> {\n const byBase = new Map<string, string[]>();\n for (const path of files.map((f) => f.relPath).sort()) {\n const key = baseKey(path);\n const list = byBase.get(key);\n if (list) list.push(path);\n else byBase.set(key, [path]);\n }\n return byBase;\n}\n\ninterface LinkRow {\n src: string;\n target: string;\n dst: string | null;\n embed: number;\n}\n\n// Applies resolveTarget to `rows`, writing only rows whose dst actually changes. The UPDATE\n// keys on (src, target): a link/embed sibling pair shares one resolution, so either row's\n// mismatch rewrites both.\nfunction resolveRows(db: DatabaseSync, rows: LinkRow[], pathSet: Set<string>, byBase: Map<string, string[]>): boolean {\n const update = db.prepare('UPDATE links SET dst = ? WHERE src = ? AND target = ?');\n let changed = false;\n for (const row of rows) {\n const dst = resolveTarget(row.src, row.target, pathSet, byBase);\n if (dst !== row.dst) {\n update.run(dst, row.src, row.target);\n changed = true;\n }\n }\n return changed;\n}\n\n// A new or deleted file can change any note's resolution, so re-resolve the whole table.\n// Fallback for cold builds and large deltas -- see afterReconcile's threshold.\nfunction resolveAll(db: DatabaseSync, files: FileStat[]): boolean {\n const pathSet = new Set(files.map((f) => f.relPath));\n const byBase = buildByBase(files);\n const rows = db.prepare('SELECT src, target, dst, embed FROM links').all() as unknown as LinkRow[];\n return resolveRows(db, rows, pathSet, byBase);\n}\n\n// Re-resolve only what this reconcile could have touched: re-stored sources, links whose\n// target basename matches an added/vanished file, and links whose dst just vanished. All\n// three are indexed lookups, so this never reads the whole table.\nfunction resolveIncremental(db: DatabaseSync, delta: ReconcileDelta): boolean {\n const pathSet = new Set(delta.files.map((f) => f.relPath));\n const byBase = buildByBase(delta.files);\n\n const changedBasenames = new Set<string>();\n for (const p of delta.added) changedBasenames.add(baseKey(p));\n for (const p of delta.vanished) changedBasenames.add(baseKey(p));\n\n // \\0 can't appear in a path or target, so the key never collides; a space can.\n const candidates = new Map<string, LinkRow>();\n const collect = (rows: LinkRow[]) => {\n for (const row of rows) candidates.set(`${row.src}\\0${row.target}\\0${row.embed}`, row);\n };\n // Chunked so a large-but-under-threshold delta never exceeds SQLite's bound-variable\n // limit (SQLITE_MAX_VARIABLE_NUMBER, 32766).\n const collectIn = (column: string, keys: string[]) => {\n for (let i = 0; i < keys.length; i += 500) {\n const chunk = keys.slice(i, i + 500);\n const placeholders = chunk.map(() => '?').join(', ');\n collect(db.prepare(`SELECT src, target, dst, embed FROM links WHERE ${column} IN (${placeholders})`).all(...chunk) as unknown as LinkRow[]);\n }\n };\n\n collectIn('src', delta.reparsed);\n collectIn('target_base', [...changedBasenames]);\n collectIn('dst', delta.vanished);\n\n return resolveRows(db, [...candidates.values()], pathSet, byBase);\n}\n\n// Resolved edges, for rank and for search's graph expansion.\nexport function linkEdges(db: DatabaseSync): [string, string][] {\n // One edge per row, as 0.12's grain always had it -- two written targets resolving to one\n // dst stay two edges (a weight the fusion evals were gated on). The only exclusion is an\n // embed whose exact (src, target) also exists as a link: that second row is new with the\n // embed grain and would double an edge that used to be one row. The NOT EXISTS probes the\n // primary key and fires only for embed rows; both branches still scan the table.\n const sql = `SELECT src, dst FROM links WHERE dst IS NOT NULL AND embed = 0\n UNION ALL\n SELECT src, dst FROM links l WHERE dst IS NOT NULL AND embed = 1\n AND NOT EXISTS (SELECT 1 FROM links l0 WHERE l0.src = l.src AND l0.target = l.target AND l0.embed = 0)`;\n return (db.prepare(sql).all() as Array<{ src: string; dst: string }>).map((r) => [r.src, r.dst]);\n}\n\n// remove() has already deleted the rows by the time afterReconcile runs, so it records here\n// whether they carried edges. Keyed on the delta object, so the state dies with the reconcile.\nconst removedWithLinks = new WeakMap<ReconcileDelta, Set<string>>();\n\nfunction recordRemoved(delta: ReconcileDelta, path: string): void {\n let set = removedWithLinks.get(delta);\n if (!set) {\n set = new Set();\n removedWithLinks.set(delta, set);\n }\n set.add(path);\n}\n\n// remove() runs per path; the vanished list is an array, so cache the Set per delta.\nconst vanishedSets = new WeakMap<ReconcileDelta, Set<string>>();\nfunction vanishedSet(delta: ReconcileDelta): Set<string> {\n let set = vanishedSets.get(delta);\n if (!set) {\n set = new Set(delta.vanished);\n vanishedSets.set(delta, set);\n }\n return set;\n}\n\nexport const links: Feature = {\n name: 'links',\n schema(db) {\n db.exec('CREATE TABLE IF NOT EXISTS links (src TEXT, target TEXT, target_base TEXT, dst TEXT, embed INTEGER, PRIMARY KEY (src, target, embed))');\n db.exec('CREATE INDEX IF NOT EXISTS links_dst ON links(dst)');\n db.exec('CREATE INDEX IF NOT EXISTS links_target_base ON links(target_base)');\n },\n extract,\n // Reparsed docs are diffed in store() below instead of wiped here: deleting and\n // re-inserting resets dst to NULL, which made every touch-only reparse read as an edge\n // change and recompute PageRank -- measured as most of the remaining update cost.\n remove(db, path, delta) {\n if (!vanishedSet(delta).has(path)) return;\n const result = db.prepare('DELETE FROM links WHERE src = ?').run(path);\n if (Number(result.changes) > 0) recordRemoved(delta, path);\n },\n store(db, path, extracted, delta) {\n const targets = extracted as Array<{ target: string; embed: boolean }>;\n // Stale rows (target/embed pairs the new parse no longer contains) are real edge removals.\n // \\0 can't appear in a target, so the composite key never collides.\n let stale: ReturnType<ReturnType<DatabaseSync['prepare']>['run']>;\n if (targets.length === 0) {\n stale = db.prepare('DELETE FROM links WHERE src = ?').run(path);\n } else {\n const placeholders = targets.map(() => '?').join(', ');\n const keys = targets.map((t) => `${t.target}\\0${t.embed ? '1' : '0'}`);\n stale = db.prepare(`DELETE FROM links WHERE src = ? AND (target || char(0) || embed) NOT IN (${placeholders})`).run(path, ...keys);\n }\n if (Number(stale.changes) > 0) recordRemoved(delta, path);\n // Upsert preserves dst on surviving rows, so an unchanged link resolves to the same\n // value and reports no change; only genuinely new rows start at NULL.\n const insert = db.prepare('INSERT INTO links (src, target, target_base, dst, embed) VALUES (?, ?, ?, NULL, ?) ON CONFLICT(src, target, embed) DO UPDATE SET target_base = excluded.target_base');\n for (const { target, embed } of targets) insert.run(path, target, baseKey(cleanTarget(target)), embed ? 1 : 0);\n },\n afterReconcile(db, delta) {\n // Any deleted rows that carried edges mean the edge set shrank -- whether the file\n // vanished or was reparsed down to fewer/no links. dstChanged alone misses the\n // reparsed-to-zero-links case (no surviving rows to re-resolve).\n const removeHadLinks = (removedWithLinks.get(delta)?.size ?? 0) > 0;\n\n const churn = delta.reparsed.length + delta.vanished.length;\n // Past this share of the tree, a full pass is the only one guaranteed to match\n // resolveTarget's ambiguity rules. A cold build clears the threshold on its own.\n const large = delta.files.length === 0 || churn > 0.2 * delta.files.length;\n\n const dstChanged = large ? resolveAll(db, delta.files) : resolveIncremental(db, delta);\n delta.linksChanged = dstChanged || removeHadLinks;\n },\n};\n"],"names":["linkEdges","links","extract","_raw","body","seen","Set","results","add","target","embed","key","has","push","matchAll","m","split","trim","index","test","cleanTarget","replace","baseKey","path","posix","basename","toLowerCase","resolveTarget","src","pathSet","byBase","clean","fromSrc","normalize","join","dirname","candidate","get","buildByBase","files","Map","map","f","relPath","sort","list","set","resolveRows","db","rows","update","prepare","changed","row","dst","run","resolveAll","all","resolveIncremental","delta","changedBasenames","added","p","vanished","candidates","collect","collectIn","column","keys","i","length","chunk","slice","placeholders","reparsed","values","sql","r","removedWithLinks","WeakMap","recordRemoved","vanishedSets","vanishedSet","name","schema","exec","remove","result","Number","changes","store","extracted","targets","stale","t","insert","afterReconcile","removeHadLinks","size","churn","large","dstChanged","linksChanged"],"mappings":";;;;;;;;;;;QAiIgBA;eAAAA;;QAqCHC;eAAAA;;;4DAtKK;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAKlB,0FAA0F;AAC1F,iFAAiF;AACjF,+FAA+F;AAC/F,uFAAuF;AAEvF,qGAAqG;AACrG,SAASC,QAAQC,IAAY,EAAEC,IAAY;IACzC,IAAMC,OAAO,IAAIC;IACjB,IAAMC,UAAqD,EAAE;IAC7D,IAAMC,MAAM,aAACC,QAAgBC;QAC3B,IAAMC,MAAM,AAAC,GAAaD,OAAXD,QAAO,MAAsB,OAAlBC,QAAQ,MAAM;QACxC,IAAIL,KAAKO,GAAG,CAACD,MAAM;QACnBN,KAAKG,GAAG,CAACG;QACTJ,QAAQM,IAAI,CAAC;YAAEJ,QAAAA;YAAQC,OAAAA;QAAM;IAC/B;QAGK,kCAAA,2BAAA;;QAFL,uFAAuF;QACvF,yFAAyF;QACzF,QAAK,YAAWN,KAAKU,QAAQ,CAAC,sCAAzB,SAAA,6BAAA,QAAA,yBAAA,iCAA4C;YAA5C,IAAMC,IAAN;gBAE2BA;YAD9B,IAAMN,SAASM,CAAC,CAAC,EAAE,CAACC,KAAK,CAAC,IAAI,CAAC,EAAE,CAACA,KAAK,CAAC,IAAI,CAAC,EAAE,CAACC,IAAI;YACpD,IAAIR,QAAQD,IAAIC,QAAQL,IAAI,CAAC,EAACW,WAAAA,EAAEG,KAAK,cAAPH,sBAAAA,WAAW,KAAK,EAAE,KAAK;QACvD;;QAHK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;QAIA,mCAAA,4BAAA;;QAAL,QAAK,aAAWX,KAAKU,QAAQ,CAAC,iEAAzB,UAAA,8BAAA,SAAA,0BAAA,kCAAuE;YAAvE,IAAMC,KAAN;YACH,IAAMN,UAASM,EAAC,CAAC,EAAE,CAACE,IAAI;YACxB,IAAIR,WAAU,CAAC,gBAAgBU,IAAI,CAACV,UAASD,IAAIC,SAAQM,EAAC,CAAC,EAAE,KAAK;QACpE;;QAHK;QAAA;;;iBAAA,8BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAIL,OAAOR;AACT;AAEA,SAASa,YAAYX,MAAc;IACjC,OAAOA,OAAOY,OAAO,CAAC,OAAO,KAAKA,OAAO,CAAC,SAAS;AACrD;AAEA,SAASC,QAAQC,IAAY;IAC3B,OAAOC,cAAK,CAACC,QAAQ,CAACF,MAAMF,OAAO,CAAC,UAAU,IAAIK,WAAW;AAC/D;AAEA,uFAAuF;AACvF,2EAA2E;AAC3E,SAASC,cAAcC,GAAW,EAAEnB,MAAc,EAAEoB,OAAoB,EAAEC,MAA6B;;QAM9FA;IALP,IAAMC,QAAQX,YAAYX;IAC1B,IAAMuB,UAAUR,cAAK,CAACS,SAAS,CAACT,cAAK,CAACU,IAAI,CAACV,cAAK,CAACW,OAAO,CAACP,MAAMG;IAC/D,gBAAwB,QAAA;QAACA;QAAQ,GAAQ,OAANA,OAAM;QAAMC;QAAU,GAAU,OAARA,SAAQ;KAAK,OAAhD,mBAAkD;YAA/DI,YAAa;QACtB,IAAIP,QAAQjB,GAAG,CAACwB,YAAY,OAAOA;IACrC;IACA,gBAAON,cAAAA,OAAOO,GAAG,CAACf,QAAQS,qBAAnBD,kCAAAA,WAA4B,CAAC,EAAE,uCAAI;AAC5C;AAEA,SAASQ,YAAYC,KAAiB;IACpC,IAAMT,SAAS,IAAIU;QACd,kCAAA,2BAAA;;QAAL,QAAK,YAAcD,MAAME,GAAG,CAAC,SAACC;mBAAMA,EAAEC,OAAO;WAAEC,IAAI,uBAA9C,SAAA,6BAAA,QAAA,yBAAA,iCAAkD;YAAlD,IAAMrB,OAAN;YACH,IAAMZ,MAAMW,QAAQC;YACpB,IAAMsB,OAAOf,OAAOO,GAAG,CAAC1B;YACxB,IAAIkC,MAAMA,KAAKhC,IAAI,CAACU;iBACfO,OAAOgB,GAAG,CAACnC,KAAK;gBAACY;aAAK;QAC7B;;QALK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAML,OAAOO;AACT;AASA,4FAA4F;AAC5F,0FAA0F;AAC1F,0BAA0B;AAC1B,SAASiB,YAAYC,EAAgB,EAAEC,IAAe,EAAEpB,OAAoB,EAAEC,MAA6B;IACzG,IAAMoB,SAASF,GAAGG,OAAO,CAAC;IAC1B,IAAIC,UAAU;QACT,kCAAA,2BAAA;;QAAL,QAAK,YAAaH,yBAAb,SAAA,6BAAA,QAAA,yBAAA,iCAAmB;YAAnB,IAAMI,MAAN;YACH,IAAMC,MAAM3B,cAAc0B,IAAIzB,GAAG,EAAEyB,IAAI5C,MAAM,EAAEoB,SAASC;YACxD,IAAIwB,QAAQD,IAAIC,GAAG,EAAE;gBACnBJ,OAAOK,GAAG,CAACD,KAAKD,IAAIzB,GAAG,EAAEyB,IAAI5C,MAAM;gBACnC2C,UAAU;YACZ;QACF;;QANK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAOL,OAAOA;AACT;AAEA,yFAAyF;AACzF,+EAA+E;AAC/E,SAASI,WAAWR,EAAgB,EAAET,KAAiB;IACrD,IAAMV,UAAU,IAAIvB,IAAIiC,MAAME,GAAG,CAAC,SAACC;eAAMA,EAAEC,OAAO;;IAClD,IAAMb,SAASQ,YAAYC;IAC3B,IAAMU,OAAOD,GAAGG,OAAO,CAAC,6CAA6CM,GAAG;IACxE,OAAOV,YAAYC,IAAIC,MAAMpB,SAASC;AACxC;AAEA,yFAAyF;AACzF,yFAAyF;AACzF,kEAAkE;AAClE,SAAS4B,mBAAmBV,EAAgB,EAAEW,KAAqB;IACjE,IAAM9B,UAAU,IAAIvB,IAAIqD,MAAMpB,KAAK,CAACE,GAAG,CAAC,SAACC;eAAMA,EAAEC,OAAO;;IACxD,IAAMb,SAASQ,YAAYqB,MAAMpB,KAAK;IAEtC,IAAMqB,mBAAmB,IAAItD;QACxB,kCAAA,2BAAA;;QAAL,QAAK,YAAWqD,MAAME,KAAK,qBAAtB,SAAA,6BAAA,QAAA,yBAAA;YAAA,IAAMC,IAAN;YAAwBF,iBAAiBpD,GAAG,CAACc,QAAQwC;;;QAArD;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;QACA,mCAAA,4BAAA;;QAAL,QAAK,aAAWH,MAAMI,QAAQ,qBAAzB,UAAA,8BAAA,SAAA,0BAAA;YAAA,IAAMD,KAAN;YAA2BF,iBAAiBpD,GAAG,CAACc,QAAQwC;;;QAAxD;QAAA;;;iBAAA,8BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAEL,+EAA+E;IAC/E,IAAME,aAAa,IAAIxB;IACvB,IAAMyB,UAAU,iBAAChB;YACV,kCAAA,2BAAA;;YAAL,QAAK,YAAaA,yBAAb,SAAA,6BAAA,QAAA,yBAAA;gBAAA,IAAMI,MAAN;gBAAmBW,WAAWlB,GAAG,CAAC,AAAC,GAAcO,OAAZA,IAAIzB,GAAG,EAAC,MAAmByB,OAAfA,IAAI5C,MAAM,EAAC,MAAc,OAAV4C,IAAI3C,KAAK,GAAI2C;;;YAA7E;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;IACP;IACA,qFAAqF;IACrF,6CAA6C;IAC7C,IAAMa,YAAY,mBAACC,QAAgBC;QACjC,IAAK,IAAIC,IAAI,GAAGA,IAAID,KAAKE,MAAM,EAAED,KAAK,IAAK;gBAGjCrB;YAFR,IAAMuB,QAAQH,KAAKI,KAAK,CAACH,GAAGA,IAAI;YAChC,IAAMI,eAAeF,MAAM9B,GAAG,CAAC;uBAAM;eAAKP,IAAI,CAAC;YAC/C+B,QAAQjB,CAAAA,cAAAA,GAAGG,OAAO,CAAC,AAAC,mDAAgEsB,OAAdN,QAAO,SAAoB,OAAbM,cAAa,OAAIhB,GAAG,OAAhGT,aAAiG,qBAAGuB;QAC9G;IACF;IAEAL,UAAU,OAAOP,MAAMe,QAAQ;IAC/BR,UAAU,eAAgB,qBAAGN;IAC7BM,UAAU,OAAOP,MAAMI,QAAQ;IAE/B,OAAOhB,YAAYC,IAAK,qBAAGgB,WAAWW,MAAM,KAAK9C,SAASC;AAC5D;AAGO,SAAS9B,UAAUgD,EAAgB;IACxC,0FAA0F;IAC1F,yFAAyF;IACzF,yFAAyF;IACzF,0FAA0F;IAC1F,iFAAiF;IACjF,IAAM4B,MAAM;IAIZ,OAAO,AAAC5B,GAAGG,OAAO,CAACyB,KAAKnB,GAAG,GAA2ChB,GAAG,CAAC,SAACoC;eAAM;YAACA,EAAEjD,GAAG;YAAEiD,EAAEvB,GAAG;SAAC;;AACjG;AAEA,4FAA4F;AAC5F,+FAA+F;AAC/F,IAAMwB,mBAAmB,IAAIC;AAE7B,SAASC,cAAcrB,KAAqB,EAAEpC,IAAY;IACxD,IAAIuB,MAAMgC,iBAAiBzC,GAAG,CAACsB;IAC/B,IAAI,CAACb,KAAK;QACRA,MAAM,IAAIxC;QACVwE,iBAAiBhC,GAAG,CAACa,OAAOb;IAC9B;IACAA,IAAItC,GAAG,CAACe;AACV;AAEA,qFAAqF;AACrF,IAAM0D,eAAe,IAAIF;AACzB,SAASG,YAAYvB,KAAqB;IACxC,IAAIb,MAAMmC,aAAa5C,GAAG,CAACsB;IAC3B,IAAI,CAACb,KAAK;QACRA,MAAM,IAAIxC,IAAIqD,MAAMI,QAAQ;QAC5BkB,aAAanC,GAAG,CAACa,OAAOb;IAC1B;IACA,OAAOA;AACT;AAEO,IAAM7C,QAAiB;IAC5BkF,MAAM;IACNC,QAAAA,SAAAA,OAAOpC,EAAE;QACPA,GAAGqC,IAAI,CAAC;QACRrC,GAAGqC,IAAI,CAAC;QACRrC,GAAGqC,IAAI,CAAC;IACV;IACAnF,SAAAA;IACA,gFAAgF;IAChF,uFAAuF;IACvF,kFAAkF;IAClFoF,QAAAA,SAAAA,OAAOtC,EAAE,EAAEzB,IAAI,EAAEoC,KAAK;QACpB,IAAI,CAACuB,YAAYvB,OAAO/C,GAAG,CAACW,OAAO;QACnC,IAAMgE,SAASvC,GAAGG,OAAO,CAAC,mCAAmCI,GAAG,CAAChC;QACjE,IAAIiE,OAAOD,OAAOE,OAAO,IAAI,GAAGT,cAAcrB,OAAOpC;IACvD;IACAmE,OAAAA,SAAAA,MAAM1C,EAAE,EAAEzB,IAAI,EAAEoE,SAAS,EAAEhC,KAAK;QAC9B,IAAMiC,UAAUD;QAChB,2FAA2F;QAC3F,oEAAoE;QACpE,IAAIE;QACJ,IAAID,QAAQtB,MAAM,KAAK,GAAG;YACxBuB,QAAQ7C,GAAGG,OAAO,CAAC,mCAAmCI,GAAG,CAAChC;QAC5D,OAAO;gBAGGyB;YAFR,IAAMyB,eAAemB,QAAQnD,GAAG,CAAC;uBAAM;eAAKP,IAAI,CAAC;YACjD,IAAMkC,OAAOwB,QAAQnD,GAAG,CAAC,SAACqD;uBAAM,AAAC,GAAeA,OAAbA,EAAErF,MAAM,EAAC,MAAwB,OAApBqF,EAAEpF,KAAK,GAAG,MAAM;;YAChEmF,QAAQ7C,CAAAA,cAAAA,GAAGG,OAAO,CAAC,AAAC,4EAAwF,OAAbsB,cAAa,OAAIlB,GAAG,OAA3GP,aAAAA;gBAA4GzB;aAAc,CAA1HyB,OAAkH,qBAAGoB;QAC/H;QACA,IAAIoB,OAAOK,MAAMJ,OAAO,IAAI,GAAGT,cAAcrB,OAAOpC;QACpD,oFAAoF;QACpF,sEAAsE;QACtE,IAAMwE,SAAS/C,GAAGG,OAAO,CAAC;YACrB,kCAAA,2BAAA;;YAAL,QAAK,YAA2ByC,4BAA3B,SAAA,6BAAA,QAAA,yBAAA;gBAAA,kBAAA,aAAQnF,qBAAAA,QAAQC,oBAAAA;gBAAoBqF,OAAOxC,GAAG,CAAChC,MAAMd,QAAQa,QAAQF,YAAYX,UAAUC,QAAQ,IAAI;;;YAAvG;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;IACP;IACAsF,gBAAAA,SAAAA,eAAehD,EAAE,EAAEW,KAAK;;YAIEmB;QAHxB,mFAAmF;QACnF,+EAA+E;QAC/E,iEAAiE;QACjE,IAAMmB,iBAAiB,UAACnB,wBAAAA,iBAAiBzC,GAAG,CAACsB,oBAArBmB,4CAAAA,sBAA6BoB,IAAI,uCAAI,KAAK;QAElE,IAAMC,QAAQxC,MAAMe,QAAQ,CAACJ,MAAM,GAAGX,MAAMI,QAAQ,CAACO,MAAM;QAC3D,+EAA+E;QAC/E,iFAAiF;QACjF,IAAM8B,QAAQzC,MAAMpB,KAAK,CAAC+B,MAAM,KAAK,KAAK6B,QAAQ,MAAMxC,MAAMpB,KAAK,CAAC+B,MAAM;QAE1E,IAAM+B,aAAaD,QAAQ5C,WAAWR,IAAIW,MAAMpB,KAAK,IAAImB,mBAAmBV,IAAIW;QAChFA,MAAM2C,YAAY,GAAGD,cAAcJ;IACrC;AACF"}
@@ -38,6 +38,10 @@ function _unsupported_iterable_to_array(o, minLen) {
38
38
  // (book/scifi); `tag = 'book' OR tag LIKE 'book/%'` is how a caller matches the parent too.
39
39
  var FENCE_RE = /^(```|~~~)/;
40
40
  var INLINE_CODE_RE = /`[^`]*`/g;
41
+ // Obsidian treats [[#Heading]] as a same-note link, not a tag.
42
+ var WIKILINK_RE = /\[\[.*?\]\]/g; // to the first ]], so a heading holding a lone ] still masks
43
+ // Obsidian doesn't read tags inside HTML markup.
44
+ var HTML_TAG_RE = /<\/?[a-zA-Z][^>]*>/g; // tag-shaped only: a comparison's `< 5` must not open a span
41
45
  // Anchors on start-of-line or a preceding whitespace/(/[ so `a#b` and URL fragments don't count.
42
46
  var INLINE_TAG_RE = RegExp("(?:^|[\\s([])#([\\u0041-\\u005A\\u0061-\\u007A\\u00AA\\u00B5\\u00BA\\u00C0-\\u00D6\\u00D8-\\u00F6\\u00F8-\\u02C1\\u02C6-\\u02D1\\u02E0-\\u02E4\\u02EC\\u02EE\\u0370-\\u0374\\u0376-\\u0377\\u037A-\\u037D\\u037F\\u0386\\u0388-\\u038A\\u038C\\u038E-\\u03A1\\u03A3-\\u03F5\\u03F7-\\u0481\\u048A-\\u052F\\u0531-\\u0556\\u0559\\u0560-\\u0588\\u05D0-\\u05EA\\u05EF-\\u05F2\\u0620-\\u064A\\u066E-\\u066F\\u0671-\\u06D3\\u06D5\\u06E5-\\u06E6\\u06EE-\\u06EF\\u06FA-\\u06FC\\u06FF\\u0710\\u0712-\\u072F\\u074D-\\u07A5\\u07B1\\u07CA-\\u07EA\\u07F4-\\u07F5\\u07FA\\u0800-\\u0815\\u081A\\u0824\\u0828\\u0840-\\u0858\\u0860-\\u086A\\u0870-\\u0887\\u0889-\\u088F\\u08A0-\\u08C9\\u0904-\\u0939\\u093D\\u0950\\u0958-\\u0961\\u0971-\\u0980\\u0985-\\u098C\\u098F-\\u0990\\u0993-\\u09A8\\u09AA-\\u09B0\\u09B2\\u09B6-\\u09B9\\u09BD\\u09CE\\u09DC-\\u09DD\\u09DF-\\u09E1\\u09F0-\\u09F1\\u09FC\\u0A05-\\u0A0A\\u0A0F-\\u0A10\\u0A13-\\u0A28\\u0A2A-\\u0A30\\u0A32-\\u0A33\\u0A35-\\u0A36\\u0A38-\\u0A39\\u0A59-\\u0A5C\\u0A5E\\u0A72-\\u0A74\\u0A85-\\u0A8D\\u0A8F-\\u0A91\\u0A93-\\u0AA8\\u0AAA-\\u0AB0\\u0AB2-\\u0AB3\\u0AB5-\\u0AB9\\u0ABD\\u0AD0\\u0AE0-\\u0AE1\\u0AF9\\u0B05-\\u0B0C\\u0B0F-\\u0B10\\u0B13-\\u0B28\\u0B2A-\\u0B30\\u0B32-\\u0B33\\u0B35-\\u0B39\\u0B3D\\u0B5C-\\u0B5D\\u0B5F-\\u0B61\\u0B71\\u0B83\\u0B85-\\u0B8A\\u0B8E-\\u0B90\\u0B92-\\u0B95\\u0B99-\\u0B9A\\u0B9C\\u0B9E-\\u0B9F\\u0BA3-\\u0BA4\\u0BA8-\\u0BAA\\u0BAE-\\u0BB9\\u0BD0\\u0C05-\\u0C0C\\u0C0E-\\u0C10\\u0C12-\\u0C28\\u0C2A-\\u0C39\\u0C3D\\u0C58-\\u0C5A\\u0C5C-\\u0C5D\\u0C60-\\u0C61\\u0C80\\u0C85-\\u0C8C\\u0C8E-\\u0C90\\u0C92-\\u0CA8\\u0CAA-\\u0CB3\\u0CB5-\\u0CB9\\u0CBD\\u0CDC-\\u0CDE\\u0CE0-\\u0CE1\\u0CF1-\\u0CF2\\u0D04-\\u0D0C\\u0D0E-\\u0D10\\u0D12-\\u0D3A\\u0D3D\\u0D4E\\u0D54-\\u0D56\\u0D5F-\\u0D61\\u0D7A-\\u0D7F\\u0D85-\\u0D96\\u0D9A-\\u0DB1\\u0DB3-\\u0DBB\\u0DBD\\u0DC0-\\u0DC6\\u0E01-\\u0E30\\u0E32-\\u0E33\\u0E40-\\u0E46\\u0E81-\\u0E82\\u0E84\\u0E86-\\u0E8A\\u0E8C-\\u0EA3\\u0EA5\\u0EA7-\\u0EB0\\u0EB2-\\u0EB3\\u0EBD\\u0EC0-\\u0EC4\\u0EC6\\u0EDC-\\u0EDF\\u0F00\\u0F40-\\u0F47\\u0F49-\\u0F6C\\u0F88-\\u0F8C\\u1000-\\u102A\\u103F\\u1050-\\u1055\\u105A-\\u105D\\u1061\\u1065-\\u1066\\u106E-\\u1070\\u1075-\\u1081\\u108E\\u10A0-\\u10C5\\u10C7\\u10CD\\u10D0-\\u10FA\\u10FC-\\u1248\\u124A-\\u124D\\u1250-\\u1256\\u1258\\u125A-\\u125D\\u1260-\\u1288\\u128A-\\u128D\\u1290-\\u12B0\\u12B2-\\u12B5\\u12B8-\\u12BE\\u12C0\\u12C2-\\u12C5\\u12C8-\\u12D6\\u12D8-\\u1310\\u1312-\\u1315\\u1318-\\u135A\\u1380-\\u138F\\u13A0-\\u13F5\\u13F8-\\u13FD\\u1401-\\u166C\\u166F-\\u167F\\u1681-\\u169A\\u16A0-\\u16EA\\u16F1-\\u16F8\\u1700-\\u1711\\u171F-\\u1731\\u1740-\\u1751\\u1760-\\u176C\\u176E-\\u1770\\u1780-\\u17B3\\u17D7\\u17DC\\u1820-\\u1878\\u1880-\\u1884\\u1887-\\u18A8\\u18AA\\u18B0-\\u18F5\\u1900-\\u191E\\u1950-\\u196D\\u1970-\\u1974\\u1980-\\u19AB\\u19B0-\\u19C9\\u1A00-\\u1A16\\u1A20-\\u1A54\\u1AA7\\u1B05-\\u1B33\\u1B45-\\u1B4C\\u1B83-\\u1BA0\\u1BAE-\\u1BAF\\u1BBA-\\u1BE5\\u1C00-\\u1C23\\u1C4D-\\u1C4F\\u1C5A-\\u1C7D\\u1C80-\\u1C8A\\u1C90-\\u1CBA\\u1CBD-\\u1CBF\\u1CE9-\\u1CEC\\u1CEE-\\u1CF3\\u1CF5-\\u1CF6\\u1CFA\\u1D00-\\u1DBF\\u1E00-\\u1F15\\u1F18-\\u1F1D\\u1F20-\\u1F45\\u1F48-\\u1F4D\\u1F50-\\u1F57\\u1F59\\u1F5B\\u1F5D\\u1F5F-\\u1F7D\\u1F80-\\u1FB4\\u1FB6-\\u1FBC\\u1FBE\\u1FC2-\\u1FC4\\u1FC6-\\u1FCC\\u1FD0-\\u1FD3\\u1FD6-\\u1FDB\\u1FE0-\\u1FEC\\u1FF2-\\u1FF4\\u1FF6-\\u1FFC\\u2071\\u207F\\u2090-\\u209C\\u2102\\u2107\\u210A-\\u2113\\u2115\\u2119-\\u211D\\u2124\\u2126\\u2128\\u212A-\\u212D\\u212F-\\u2139\\u213C-\\u213F\\u2145-\\u2149\\u214E\\u2183-\\u2184\\u2C00-\\u2CE4\\u2CEB-\\u2CEE\\u2CF2-\\u2CF3\\u2D00-\\u2D25\\u2D27\\u2D2D\\u2D30-\\u2D67\\u2D6F\\u2D80-\\u2D96\\u2DA0-\\u2DA6\\u2DA8-\\u2DAE\\u2DB0-\\u2DB6\\u2DB8-\\u2DBE\\u2DC0-\\u2DC6\\u2DC8-\\u2DCE\\u2DD0-\\u2DD6\\u2DD8-\\u2DDE\\u2E2F\\u3005-\\u3006\\u3031-\\u3035\\u303B-\\u303C\\u3041-\\u3096\\u309D-\\u309F\\u30A1-\\u30FA\\u30FC-\\u30FF\\u3105-\\u312F\\u3131-\\u318E\\u31A0-\\u31BF\\u31F0-\\u31FF\\u3400-\\u4DBF\\u4E00-\\uA48C\\uA4D0-\\uA4FD\\uA500-\\uA60C\\uA610-\\uA61F\\uA62A-\\uA62B\\uA640-\\uA66E\\uA67F-\\uA69D\\uA6A0-\\uA6E5\\uA717-\\uA71F\\uA722-\\uA788\\uA78B-\\uA7DC\\uA7F1-\\uA801\\uA803-\\uA805\\uA807-\\uA80A\\uA80C-\\uA822\\uA840-\\uA873\\uA882-\\uA8B3\\uA8F2-\\uA8F7\\uA8FB\\uA8FD-\\uA8FE\\uA90A-\\uA925\\uA930-\\uA946\\uA960-\\uA97C\\uA984-\\uA9B2\\uA9CF\\uA9E0-\\uA9E4\\uA9E6-\\uA9EF\\uA9FA-\\uA9FE\\uAA00-\\uAA28\\uAA40-\\uAA42\\uAA44-\\uAA4B\\uAA60-\\uAA76\\uAA7A\\uAA7E-\\uAAAF\\uAAB1\\uAAB5-\\uAAB6\\uAAB9-\\uAABD\\uAAC0\\uAAC2\\uAADB-\\uAADD\\uAAE0-\\uAAEA\\uAAF2-\\uAAF4\\uAB01-\\uAB06\\uAB09-\\uAB0E\\uAB11-\\uAB16\\uAB20-\\uAB26\\uAB28-\\uAB2E\\uAB30-\\uAB5A\\uAB5C-\\uAB69\\uAB70-\\uABE2\\uAC00-\\uD7A3\\uD7B0-\\uD7C6\\uD7CB-\\uD7FB\\uF900-\\uFA6D\\uFA70-\\uFAD9\\uFB00-\\uFB06\\uFB13-\\uFB17\\uFB1D\\uFB1F-\\uFB28\\uFB2A-\\uFB36\\uFB38-\\uFB3C\\uFB3E\\uFB40-\\uFB41\\uFB43-\\uFB44\\uFB46-\\uFBB1\\uFBD3-\\uFD3D\\uFD50-\\uFD8F\\uFD92-\\uFDC7\\uFDF0-\\uFDFB\\uFE70-\\uFE74\\uFE76-\\uFEFC\\uFF21-\\uFF3A\\uFF41-\\uFF5A\\uFF66-\\uFFBE\\uFFC2-\\uFFC7\\uFFCA-\\uFFCF\\uFFD2-\\uFFD7\\uFFDA-\\uFFDC\\u{10000}-\\u{1000B}\\u{1000D}-\\u{10026}\\u{10028}-\\u{1003A}\\u{1003C}-\\u{1003D}\\u{1003F}-\\u{1004D}\\u{10050}-\\u{1005D}\\u{10080}-\\u{100FA}\\u{10280}-\\u{1029C}\\u{102A0}-\\u{102D0}\\u{10300}-\\u{1031F}\\u{1032D}-\\u{10340}\\u{10342}-\\u{10349}\\u{10350}-\\u{10375}\\u{10380}-\\u{1039D}\\u{103A0}-\\u{103C3}\\u{103C8}-\\u{103CF}\\u{10400}-\\u{1049D}\\u{104B0}-\\u{104D3}\\u{104D8}-\\u{104FB}\\u{10500}-\\u{10527}\\u{10530}-\\u{10563}\\u{10570}-\\u{1057A}\\u{1057C}-\\u{1058A}\\u{1058C}-\\u{10592}\\u{10594}-\\u{10595}\\u{10597}-\\u{105A1}\\u{105A3}-\\u{105B1}\\u{105B3}-\\u{105B9}\\u{105BB}-\\u{105BC}\\u{105C0}-\\u{105F3}\\u{10600}-\\u{10736}\\u{10740}-\\u{10755}\\u{10760}-\\u{10767}\\u{10780}-\\u{10785}\\u{10787}-\\u{107B0}\\u{107B2}-\\u{107BA}\\u{10800}-\\u{10805}\\u{10808}\\u{1080A}-\\u{10835}\\u{10837}-\\u{10838}\\u{1083C}\\u{1083F}-\\u{10855}\\u{10860}-\\u{10876}\\u{10880}-\\u{1089E}\\u{108E0}-\\u{108F2}\\u{108F4}-\\u{108F5}\\u{10900}-\\u{10915}\\u{10920}-\\u{10939}\\u{10940}-\\u{10959}\\u{10980}-\\u{109B7}\\u{109BE}-\\u{109BF}\\u{10A00}\\u{10A10}-\\u{10A13}\\u{10A15}-\\u{10A17}\\u{10A19}-\\u{10A35}\\u{10A60}-\\u{10A7C}\\u{10A80}-\\u{10A9C}\\u{10AC0}-\\u{10AC7}\\u{10AC9}-\\u{10AE4}\\u{10B00}-\\u{10B35}\\u{10B40}-\\u{10B55}\\u{10B60}-\\u{10B72}\\u{10B80}-\\u{10B91}\\u{10C00}-\\u{10C48}\\u{10C80}-\\u{10CB2}\\u{10CC0}-\\u{10CF2}\\u{10D00}-\\u{10D23}\\u{10D4A}-\\u{10D65}\\u{10D6F}-\\u{10D85}\\u{10E80}-\\u{10EA9}\\u{10EB0}-\\u{10EB1}\\u{10EC2}-\\u{10EC7}\\u{10F00}-\\u{10F1C}\\u{10F27}\\u{10F30}-\\u{10F45}\\u{10F70}-\\u{10F81}\\u{10FB0}-\\u{10FC4}\\u{10FE0}-\\u{10FF6}\\u{11003}-\\u{11037}\\u{11071}-\\u{11072}\\u{11075}\\u{11083}-\\u{110AF}\\u{110D0}-\\u{110E8}\\u{11103}-\\u{11126}\\u{11144}\\u{11147}\\u{11150}-\\u{11172}\\u{11176}\\u{11183}-\\u{111B2}\\u{111C1}-\\u{111C4}\\u{111DA}\\u{111DC}\\u{11200}-\\u{11211}\\u{11213}-\\u{1122B}\\u{1123F}-\\u{11240}\\u{11280}-\\u{11286}\\u{11288}\\u{1128A}-\\u{1128D}\\u{1128F}-\\u{1129D}\\u{1129F}-\\u{112A8}\\u{112B0}-\\u{112DE}\\u{11305}-\\u{1130C}\\u{1130F}-\\u{11310}\\u{11313}-\\u{11328}\\u{1132A}-\\u{11330}\\u{11332}-\\u{11333}\\u{11335}-\\u{11339}\\u{1133D}\\u{11350}\\u{1135D}-\\u{11361}\\u{11380}-\\u{11389}\\u{1138B}\\u{1138E}\\u{11390}-\\u{113B5}\\u{113B7}\\u{113D1}\\u{113D3}\\u{11400}-\\u{11434}\\u{11447}-\\u{1144A}\\u{1145F}-\\u{11461}\\u{11480}-\\u{114AF}\\u{114C4}-\\u{114C5}\\u{114C7}\\u{11580}-\\u{115AE}\\u{115D8}-\\u{115DB}\\u{11600}-\\u{1162F}\\u{11644}\\u{11680}-\\u{116AA}\\u{116B8}\\u{11700}-\\u{1171A}\\u{11740}-\\u{11746}\\u{11800}-\\u{1182B}\\u{118A0}-\\u{118DF}\\u{118FF}-\\u{11906}\\u{11909}\\u{1190C}-\\u{11913}\\u{11915}-\\u{11916}\\u{11918}-\\u{1192F}\\u{1193F}\\u{11941}\\u{119A0}-\\u{119A7}\\u{119AA}-\\u{119D0}\\u{119E1}\\u{119E3}\\u{11A00}\\u{11A0B}-\\u{11A32}\\u{11A3A}\\u{11A50}\\u{11A5C}-\\u{11A89}\\u{11A9D}\\u{11AB0}-\\u{11AF8}\\u{11BC0}-\\u{11BE0}\\u{11C00}-\\u{11C08}\\u{11C0A}-\\u{11C2E}\\u{11C40}\\u{11C72}-\\u{11C8F}\\u{11D00}-\\u{11D06}\\u{11D08}-\\u{11D09}\\u{11D0B}-\\u{11D30}\\u{11D46}\\u{11D60}-\\u{11D65}\\u{11D67}-\\u{11D68}\\u{11D6A}-\\u{11D89}\\u{11D98}\\u{11DB0}-\\u{11DDB}\\u{11EE0}-\\u{11EF2}\\u{11F02}\\u{11F04}-\\u{11F10}\\u{11F12}-\\u{11F33}\\u{11FB0}\\u{12000}-\\u{12399}\\u{12480}-\\u{12543}\\u{12F90}-\\u{12FF0}\\u{13000}-\\u{1342F}\\u{13441}-\\u{13446}\\u{13460}-\\u{143FA}\\u{14400}-\\u{14646}\\u{16100}-\\u{1611D}\\u{16800}-\\u{16A38}\\u{16A40}-\\u{16A5E}\\u{16A70}-\\u{16ABE}\\u{16AD0}-\\u{16AED}\\u{16B00}-\\u{16B2F}\\u{16B40}-\\u{16B43}\\u{16B63}-\\u{16B77}\\u{16B7D}-\\u{16B8F}\\u{16D40}-\\u{16D6C}\\u{16E40}-\\u{16E7F}\\u{16EA0}-\\u{16EB8}\\u{16EBB}-\\u{16ED3}\\u{16F00}-\\u{16F4A}\\u{16F50}\\u{16F93}-\\u{16F9F}\\u{16FE0}-\\u{16FE1}\\u{16FE3}\\u{16FF2}-\\u{16FF3}\\u{17000}-\\u{18CD5}\\u{18CFF}-\\u{18D1E}\\u{18D80}-\\u{18DF2}\\u{1AFF0}-\\u{1AFF3}\\u{1AFF5}-\\u{1AFFB}\\u{1AFFD}-\\u{1AFFE}\\u{1B000}-\\u{1B122}\\u{1B132}\\u{1B150}-\\u{1B152}\\u{1B155}\\u{1B164}-\\u{1B167}\\u{1B170}-\\u{1B2FB}\\u{1BC00}-\\u{1BC6A}\\u{1BC70}-\\u{1BC7C}\\u{1BC80}-\\u{1BC88}\\u{1BC90}-\\u{1BC99}\\u{1D400}-\\u{1D454}\\u{1D456}-\\u{1D49C}\\u{1D49E}-\\u{1D49F}\\u{1D4A2}\\u{1D4A5}-\\u{1D4A6}\\u{1D4A9}-\\u{1D4AC}\\u{1D4AE}-\\u{1D4B9}\\u{1D4BB}\\u{1D4BD}-\\u{1D4C3}\\u{1D4C5}-\\u{1D505}\\u{1D507}-\\u{1D50A}\\u{1D50D}-\\u{1D514}\\u{1D516}-\\u{1D51C}\\u{1D51E}-\\u{1D539}\\u{1D53B}-\\u{1D53E}\\u{1D540}-\\u{1D544}\\u{1D546}\\u{1D54A}-\\u{1D550}\\u{1D552}-\\u{1D6A5}\\u{1D6A8}-\\u{1D6C0}\\u{1D6C2}-\\u{1D6DA}\\u{1D6DC}-\\u{1D6FA}\\u{1D6FC}-\\u{1D714}\\u{1D716}-\\u{1D734}\\u{1D736}-\\u{1D74E}\\u{1D750}-\\u{1D76E}\\u{1D770}-\\u{1D788}\\u{1D78A}-\\u{1D7A8}\\u{1D7AA}-\\u{1D7C2}\\u{1D7C4}-\\u{1D7CB}\\u{1DF00}-\\u{1DF1E}\\u{1DF25}-\\u{1DF2A}\\u{1E030}-\\u{1E06D}\\u{1E100}-\\u{1E12C}\\u{1E137}-\\u{1E13D}\\u{1E14E}\\u{1E290}-\\u{1E2AD}\\u{1E2C0}-\\u{1E2EB}\\u{1E4D0}-\\u{1E4EB}\\u{1E5D0}-\\u{1E5ED}\\u{1E5F0}\\u{1E6C0}-\\u{1E6DE}\\u{1E6E0}-\\u{1E6E2}\\u{1E6E4}-\\u{1E6E5}\\u{1E6E7}-\\u{1E6ED}\\u{1E6F0}-\\u{1E6F4}\\u{1E6FE}-\\u{1E6FF}\\u{1E7E0}-\\u{1E7E6}\\u{1E7E8}-\\u{1E7EB}\\u{1E7ED}-\\u{1E7EE}\\u{1E7F0}-\\u{1E7FE}\\u{1E800}-\\u{1E8C4}\\u{1E900}-\\u{1E943}\\u{1E94B}\\u{1EE00}-\\u{1EE03}\\u{1EE05}-\\u{1EE1F}\\u{1EE21}-\\u{1EE22}\\u{1EE24}\\u{1EE27}\\u{1EE29}-\\u{1EE32}\\u{1EE34}-\\u{1EE37}\\u{1EE39}\\u{1EE3B}\\u{1EE42}\\u{1EE47}\\u{1EE49}\\u{1EE4B}\\u{1EE4D}-\\u{1EE4F}\\u{1EE51}-\\u{1EE52}\\u{1EE54}\\u{1EE57}\\u{1EE59}\\u{1EE5B}\\u{1EE5D}\\u{1EE5F}\\u{1EE61}-\\u{1EE62}\\u{1EE64}\\u{1EE67}-\\u{1EE6A}\\u{1EE6C}-\\u{1EE72}\\u{1EE74}-\\u{1EE77}\\u{1EE79}-\\u{1EE7C}\\u{1EE7E}\\u{1EE80}-\\u{1EE89}\\u{1EE8B}-\\u{1EE9B}\\u{1EEA1}-\\u{1EEA3}\\u{1EEA5}-\\u{1EEA9}\\u{1EEAB}-\\u{1EEBB}\\u{20000}-\\u{2A6DF}\\u{2A700}-\\u{2B81D}\\u{2B820}-\\u{2CEAD}\\u{2CEB0}-\\u{2EBE0}\\u{2EBF0}-\\u{2EE5D}\\u{2F800}-\\u{2FA1D}\\u{30000}-\\u{3134A}\\u{31350}-\\u{33479}\\u0030-\\u0039\\u00B2-\\u00B3\\u00B9\\u00BC-\\u00BE\\u0660-\\u0669\\u06F0-\\u06F9\\u07C0-\\u07C9\\u0966-\\u096F\\u09E6-\\u09EF\\u09F4-\\u09F9\\u0A66-\\u0A6F\\u0AE6-\\u0AEF\\u0B66-\\u0B6F\\u0B72-\\u0B77\\u0BE6-\\u0BF2\\u0C66-\\u0C6F\\u0C78-\\u0C7E\\u0CE6-\\u0CEF\\u0D58-\\u0D5E\\u0D66-\\u0D78\\u0DE6-\\u0DEF\\u0E50-\\u0E59\\u0ED0-\\u0ED9\\u0F20-\\u0F33\\u1040-\\u1049\\u1090-\\u1099\\u1369-\\u137C\\u16EE-\\u16F0\\u17E0-\\u17E9\\u17F0-\\u17F9\\u1810-\\u1819\\u1946-\\u194F\\u19D0-\\u19DA\\u1A80-\\u1A89\\u1A90-\\u1A99\\u1B50-\\u1B59\\u1BB0-\\u1BB9\\u1C40-\\u1C49\\u1C50-\\u1C59\\u2070\\u2074-\\u2079\\u2080-\\u2089\\u2150-\\u2182\\u2185-\\u2189\\u2460-\\u249B\\u24EA-\\u24FF\\u2776-\\u2793\\u2CFD\\u3007\\u3021-\\u3029\\u3038-\\u303A\\u3192-\\u3195\\u3220-\\u3229\\u3248-\\u324F\\u3251-\\u325F\\u3280-\\u3289\\u32B1-\\u32BF\\uA620-\\uA629\\uA6E6-\\uA6EF\\uA830-\\uA835\\uA8D0-\\uA8D9\\uA900-\\uA909\\uA9D0-\\uA9D9\\uA9F0-\\uA9F9\\uAA50-\\uAA59\\uABF0-\\uABF9\\uFF10-\\uFF19\\u{10107}-\\u{10133}\\u{10140}-\\u{10178}\\u{1018A}-\\u{1018B}\\u{102E1}-\\u{102FB}\\u{10320}-\\u{10323}\\u{10341}\\u{1034A}\\u{103D1}-\\u{103D5}\\u{104A0}-\\u{104A9}\\u{10858}-\\u{1085F}\\u{10879}-\\u{1087F}\\u{108A7}-\\u{108AF}\\u{108FB}-\\u{108FF}\\u{10916}-\\u{1091B}\\u{109BC}-\\u{109BD}\\u{109C0}-\\u{109CF}\\u{109D2}-\\u{109FF}\\u{10A40}-\\u{10A48}\\u{10A7D}-\\u{10A7E}\\u{10A9D}-\\u{10A9F}\\u{10AEB}-\\u{10AEF}\\u{10B58}-\\u{10B5F}\\u{10B78}-\\u{10B7F}\\u{10BA9}-\\u{10BAF}\\u{10CFA}-\\u{10CFF}\\u{10D30}-\\u{10D39}\\u{10D40}-\\u{10D49}\\u{10E60}-\\u{10E7E}\\u{10F1D}-\\u{10F26}\\u{10F51}-\\u{10F54}\\u{10FC5}-\\u{10FCB}\\u{11052}-\\u{1106F}\\u{110F0}-\\u{110F9}\\u{11136}-\\u{1113F}\\u{111D0}-\\u{111D9}\\u{111E1}-\\u{111F4}\\u{112F0}-\\u{112F9}\\u{11450}-\\u{11459}\\u{114D0}-\\u{114D9}\\u{11650}-\\u{11659}\\u{116C0}-\\u{116C9}\\u{116D0}-\\u{116E3}\\u{11730}-\\u{1173B}\\u{118E0}-\\u{118F2}\\u{11950}-\\u{11959}\\u{11BF0}-\\u{11BF9}\\u{11C50}-\\u{11C6C}\\u{11D50}-\\u{11D59}\\u{11DA0}-\\u{11DA9}\\u{11DE0}-\\u{11DE9}\\u{11F50}-\\u{11F59}\\u{11FC0}-\\u{11FD4}\\u{12400}-\\u{1246E}\\u{16130}-\\u{16139}\\u{16A60}-\\u{16A69}\\u{16AC0}-\\u{16AC9}\\u{16B50}-\\u{16B59}\\u{16B5B}-\\u{16B61}\\u{16D70}-\\u{16D79}\\u{16E80}-\\u{16E96}\\u{16FF4}-\\u{16FF6}\\u{1CCF0}-\\u{1CCF9}\\u{1D2C0}-\\u{1D2D3}\\u{1D2E0}-\\u{1D2F3}\\u{1D360}-\\u{1D378}\\u{1D7CE}-\\u{1D7FF}\\u{1E140}-\\u{1E149}\\u{1E2F0}-\\u{1E2F9}\\u{1E4F0}-\\u{1E4F9}\\u{1E5F1}-\\u{1E5FA}\\u{1E8C7}-\\u{1E8CF}\\u{1E950}-\\u{1E959}\\u{1EC71}-\\u{1ECAB}\\u{1ECAD}-\\u{1ECAF}\\u{1ECB1}-\\u{1ECB4}\\u{1ED01}-\\u{1ED2D}\\u{1ED2F}-\\u{1ED3D}\\u{1F100}-\\u{1F10C}\\u{1FBF0}-\\u{1FBF9}_/-]+)", "gu");
43
47
  // Strips a leading # (frontmatter entries may carry one) and a trailing /; rejects an
@@ -79,7 +83,7 @@ function frontmatterTags(data) {
79
83
  }
80
84
  return found;
81
85
  }
82
- // #tag tokens outside fenced code blocks and inline code spans.
86
+ // #tag tokens outside fenced code blocks, inline code spans, wikilinks, and HTML tags.
83
87
  function inlineTags(body) {
84
88
  var found = [];
85
89
  var inFence = false;
@@ -96,6 +100,12 @@ function inlineTags(body) {
96
100
  var cleaned = line.includes('`') ? line.replace(INLINE_CODE_RE, function(m) {
97
101
  return ' '.repeat(m.length);
98
102
  }) : line;
103
+ if (cleaned.includes('[[')) cleaned = cleaned.replace(WIKILINK_RE, function(m) {
104
+ return ' '.repeat(m.length);
105
+ });
106
+ if (cleaned.includes('<')) cleaned = cleaned.replace(HTML_TAG_RE, function(m) {
107
+ return ' '.repeat(m.length);
108
+ });
99
109
  var _iteratorNormalCompletion1 = true, _didIteratorError1 = false, _iteratorError1 = undefined;
100
110
  try {
101
111
  for(var _iterator1 = cleaned.matchAll(INLINE_TAG_RE)[Symbol.iterator](), _step1; !(_iteratorNormalCompletion1 = (_step1 = _iterator1.next()).done); _iteratorNormalCompletion1 = true){
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/features/tags.ts"],"sourcesContent":["import type { Feature } from './types.ts';\n\n// tags(path, tag): Obsidian's file.tags grain -- frontmatter list/string tags plus inline\n// #tags from the prose, deduplicated, source not distinguished. Nested tags store full\n// (book/scifi); `tag = 'book' OR tag LIKE 'book/%'` is how a caller matches the parent too.\n\nconst FENCE_RE = /^(```|~~~)/;\nconst INLINE_CODE_RE = /`[^`]*`/g;\n// Anchors on start-of-line or a preceding whitespace/(/[ so `a#b` and URL fragments don't count.\nconst INLINE_TAG_RE = /(?:^|[\\s([])#([\\p{L}\\p{N}_/-]+)/gu;\n\n// Strips a leading # (frontmatter entries may carry one) and a trailing /; rejects an\n// all-digit result -- a tag needs at least one non-digit character.\nfunction normalizeTag(raw: string): string | null {\n const stripped = raw.replace(/^#/, '').replace(/\\/+$/, '');\n if (!stripped || /^\\d+$/.test(stripped)) return null;\n return stripped;\n}\n\n// data.tags: a YAML list (Obsidian also accepts a bare string). Null members and non-string\n// members are skipped rather than throwing -- `tags:\\n -` parses to [null].\nfunction frontmatterTags(data?: Record<string, unknown>): string[] {\n const raw = data?.tags;\n const list = Array.isArray(raw) ? raw : typeof raw === 'string' ? [raw] : [];\n const found: string[] = [];\n for (const item of list) {\n if (typeof item !== 'string') continue;\n const tag = normalizeTag(item);\n if (tag) found.push(tag);\n }\n return found;\n}\n\n// #tag tokens outside fenced code blocks and inline code spans.\nfunction inlineTags(body: string): string[] {\n const found: string[] = [];\n let inFence = false;\n for (const line of body.split('\\n')) {\n if (FENCE_RE.test(line)) {\n inFence = !inFence;\n continue;\n }\n if (inFence) continue;\n if (!line.includes('#')) continue; // most lines; skip the regex work\n const cleaned = line.includes('`') ? line.replace(INLINE_CODE_RE, (m) => ' '.repeat(m.length)) : line;\n for (const m of cleaned.matchAll(INLINE_TAG_RE)) {\n const tag = normalizeTag(m[1]);\n if (tag) found.push(tag);\n }\n }\n return found;\n}\n\nfunction extract(_raw: string, body: string, _search?: { title: string; summary: string }, data?: Record<string, unknown>): string[] {\n return [...new Set([...frontmatterTags(data), ...inlineTags(body)])].sort();\n}\n\nexport const tags: Feature = {\n name: 'tags',\n schema(db) {\n db.exec('CREATE TABLE IF NOT EXISTS tags (\"path\" TEXT, tag TEXT, PRIMARY KEY (\"path\", tag))');\n db.exec('CREATE INDEX IF NOT EXISTS tags_tag ON tags(tag)');\n },\n extract,\n // Per-file rows with nothing else to resolve, so only a vanished file needs a delete here;\n // store() below handles a reparse's stale rows itself.\n remove(db, path, delta) {\n if (!delta.vanished.includes(path)) return;\n db.prepare('DELETE FROM tags WHERE \"path\" = ?').run(path);\n },\n store(db, path, extracted) {\n const found = extracted as string[];\n if (found.length === 0) {\n db.prepare('DELETE FROM tags WHERE \"path\" = ?').run(path);\n } else {\n const placeholders = found.map(() => '?').join(', ');\n db.prepare(`DELETE FROM tags WHERE \"path\" = ? AND tag NOT IN (${placeholders})`).run(path, ...found);\n }\n const insert = db.prepare('INSERT OR IGNORE INTO tags (\"path\", tag) VALUES (?, ?)');\n for (const tag of found) insert.run(path, tag);\n },\n};\n"],"names":["tags","FENCE_RE","INLINE_CODE_RE","INLINE_TAG_RE","normalizeTag","raw","stripped","replace","test","frontmatterTags","data","list","Array","isArray","found","item","tag","push","inlineTags","body","inFence","split","line","includes","cleaned","m","repeat","length","matchAll","extract","_raw","_search","Set","sort","name","schema","db","exec","remove","path","delta","vanished","prepare","run","store","extracted","placeholders","map","join","insert"],"mappings":";;;;+BAyDaA;;;eAAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAvDb,0FAA0F;AAC1F,uFAAuF;AACvF,4FAA4F;AAE5F,IAAMC,WAAW;AACjB,IAAMC,iBAAiB;AACvB,iGAAiG;AACjG,IAAMC,gBAAgB;AAEtB,sFAAsF;AACtF,oEAAoE;AACpE,SAASC,aAAaC,GAAW;IAC/B,IAAMC,WAAWD,IAAIE,OAAO,CAAC,MAAM,IAAIA,OAAO,CAAC,QAAQ;IACvD,IAAI,CAACD,YAAY,QAAQE,IAAI,CAACF,WAAW,OAAO;IAChD,OAAOA;AACT;AAEA,4FAA4F;AAC5F,6EAA6E;AAC7E,SAASG,gBAAgBC,IAA8B;IACrD,IAAML,MAAMK,iBAAAA,2BAAAA,KAAMV,IAAI;IACtB,IAAMW,OAAOC,MAAMC,OAAO,CAACR,OAAOA,MAAM,OAAOA,QAAQ,WAAW;QAACA;KAAI,GAAG,EAAE;IAC5E,IAAMS,QAAkB,EAAE;QACrB,kCAAA,2BAAA;;QAAL,QAAK,YAAcH,yBAAd,SAAA,6BAAA,QAAA,yBAAA,iCAAoB;YAApB,IAAMI,OAAN;YACH,IAAI,OAAOA,SAAS,UAAU;YAC9B,IAAMC,MAAMZ,aAAaW;YACzB,IAAIC,KAAKF,MAAMG,IAAI,CAACD;QACtB;;QAJK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAKL,OAAOF;AACT;AAEA,gEAAgE;AAChE,SAASI,WAAWC,IAAY;IAC9B,IAAML,QAAkB,EAAE;IAC1B,IAAIM,UAAU;QACT,kCAAA,2BAAA;;QAAL,QAAK,YAAcD,KAAKE,KAAK,CAAC,0BAAzB,SAAA,6BAAA,QAAA,yBAAA,iCAAgC;YAAhC,IAAMC,OAAN;YACH,IAAIrB,SAASO,IAAI,CAACc,OAAO;gBACvBF,UAAU,CAACA;gBACX;YACF;YACA,IAAIA,SAAS;YACb,IAAI,CAACE,KAAKC,QAAQ,CAAC,MAAM,UAAU,kCAAkC;YACrE,IAAMC,UAAUF,KAAKC,QAAQ,CAAC,OAAOD,KAAKf,OAAO,CAACL,gBAAgB,SAACuB;uBAAM,IAAIC,MAAM,CAACD,EAAEE,MAAM;iBAAKL;gBAC5F,mCAAA,4BAAA;;gBAAL,QAAK,aAAWE,QAAQI,QAAQ,CAACzB,mCAA5B,UAAA,8BAAA,SAAA,0BAAA,kCAA4C;oBAA5C,IAAMsB,IAAN;oBACH,IAAMT,MAAMZ,aAAaqB,CAAC,CAAC,EAAE;oBAC7B,IAAIT,KAAKF,MAAMG,IAAI,CAACD;gBACtB;;gBAHK;gBAAA;;;yBAAA,8BAAA;wBAAA;;;wBAAA;8BAAA;;;;QAIP;;QAZK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAaL,OAAOF;AACT;AAEA,SAASe,QAAQC,IAAY,EAAEX,IAAY,EAAEY,OAA4C,EAAErB,IAA8B;IACvH,OAAO,AAAC,qBAAG,IAAIsB,IAAI,AAAC,qBAAGvB,gBAAgBC,cAAO,qBAAGQ,WAAWC,UAASc,IAAI;AAC3E;AAEO,IAAMjC,OAAgB;IAC3BkC,MAAM;IACNC,QAAAA,SAAAA,OAAOC,EAAE;QACPA,GAAGC,IAAI,CAAC;QACRD,GAAGC,IAAI,CAAC;IACV;IACAR,SAAAA;IACA,2FAA2F;IAC3F,uDAAuD;IACvDS,QAAAA,SAAAA,OAAOF,EAAE,EAAEG,IAAI,EAAEC,KAAK;QACpB,IAAI,CAACA,MAAMC,QAAQ,CAAClB,QAAQ,CAACgB,OAAO;QACpCH,GAAGM,OAAO,CAAC,qCAAqCC,GAAG,CAACJ;IACtD;IACAK,OAAAA,SAAAA,MAAMR,EAAE,EAAEG,IAAI,EAAEM,SAAS;QACvB,IAAM/B,QAAQ+B;QACd,IAAI/B,MAAMa,MAAM,KAAK,GAAG;YACtBS,GAAGM,OAAO,CAAC,qCAAqCC,GAAG,CAACJ;QACtD,OAAO;gBAELH;YADA,IAAMU,eAAehC,MAAMiC,GAAG,CAAC;uBAAM;eAAKC,IAAI,CAAC;YAC/CZ,CAAAA,cAAAA,GAAGM,OAAO,CAAC,AAAC,qDAAiE,OAAbI,cAAa,OAAIH,GAAG,OAApFP,aAAAA;gBAAqFG;aAAe,CAApGH,OAA2F,qBAAGtB;QAChG;QACA,IAAMmC,SAASb,GAAGM,OAAO,CAAC;YACrB,kCAAA,2BAAA;;YAAL,QAAK,YAAa5B,0BAAb,SAAA,6BAAA,QAAA,yBAAA;gBAAA,IAAME,MAAN;gBAAoBiC,OAAON,GAAG,CAACJ,MAAMvB;;;YAArC;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;IACP;AACF"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/features/tags.ts"],"sourcesContent":["import type { Feature } from './types.ts';\n\n// tags(path, tag): Obsidian's file.tags grain -- frontmatter list/string tags plus inline\n// #tags from the prose, deduplicated, source not distinguished. Nested tags store full\n// (book/scifi); `tag = 'book' OR tag LIKE 'book/%'` is how a caller matches the parent too.\n\nconst FENCE_RE = /^(```|~~~)/;\nconst INLINE_CODE_RE = /`[^`]*`/g;\n// Obsidian treats [[#Heading]] as a same-note link, not a tag.\nconst WIKILINK_RE = /\\[\\[.*?\\]\\]/g; // to the first ]], so a heading holding a lone ] still masks\n// Obsidian doesn't read tags inside HTML markup.\nconst HTML_TAG_RE = /<\\/?[a-zA-Z][^>]*>/g; // tag-shaped only: a comparison's `< 5` must not open a span\n// Anchors on start-of-line or a preceding whitespace/(/[ so `a#b` and URL fragments don't count.\nconst INLINE_TAG_RE = /(?:^|[\\s([])#([\\p{L}\\p{N}_/-]+)/gu;\n\n// Strips a leading # (frontmatter entries may carry one) and a trailing /; rejects an\n// all-digit result -- a tag needs at least one non-digit character.\nfunction normalizeTag(raw: string): string | null {\n const stripped = raw.replace(/^#/, '').replace(/\\/+$/, '');\n if (!stripped || /^\\d+$/.test(stripped)) return null;\n return stripped;\n}\n\n// data.tags: a YAML list (Obsidian also accepts a bare string). Null members and non-string\n// members are skipped rather than throwing -- `tags:\\n -` parses to [null].\nfunction frontmatterTags(data?: Record<string, unknown>): string[] {\n const raw = data?.tags;\n const list = Array.isArray(raw) ? raw : typeof raw === 'string' ? [raw] : [];\n const found: string[] = [];\n for (const item of list) {\n if (typeof item !== 'string') continue;\n const tag = normalizeTag(item);\n if (tag) found.push(tag);\n }\n return found;\n}\n\n// #tag tokens outside fenced code blocks, inline code spans, wikilinks, and HTML tags.\nfunction inlineTags(body: string): string[] {\n const found: string[] = [];\n let inFence = false;\n for (const line of body.split('\\n')) {\n if (FENCE_RE.test(line)) {\n inFence = !inFence;\n continue;\n }\n if (inFence) continue;\n if (!line.includes('#')) continue; // most lines; skip the regex work\n let cleaned = line.includes('`') ? line.replace(INLINE_CODE_RE, (m) => ' '.repeat(m.length)) : line;\n if (cleaned.includes('[[')) cleaned = cleaned.replace(WIKILINK_RE, (m) => ' '.repeat(m.length));\n if (cleaned.includes('<')) cleaned = cleaned.replace(HTML_TAG_RE, (m) => ' '.repeat(m.length));\n for (const m of cleaned.matchAll(INLINE_TAG_RE)) {\n const tag = normalizeTag(m[1]);\n if (tag) found.push(tag);\n }\n }\n return found;\n}\n\nfunction extract(_raw: string, body: string, _search?: { title: string; summary: string }, data?: Record<string, unknown>): string[] {\n return [...new Set([...frontmatterTags(data), ...inlineTags(body)])].sort();\n}\n\nexport const tags: Feature = {\n name: 'tags',\n schema(db) {\n db.exec('CREATE TABLE IF NOT EXISTS tags (\"path\" TEXT, tag TEXT, PRIMARY KEY (\"path\", tag))');\n db.exec('CREATE INDEX IF NOT EXISTS tags_tag ON tags(tag)');\n },\n extract,\n // Per-file rows with nothing else to resolve, so only a vanished file needs a delete here;\n // store() below handles a reparse's stale rows itself.\n remove(db, path, delta) {\n if (!delta.vanished.includes(path)) return;\n db.prepare('DELETE FROM tags WHERE \"path\" = ?').run(path);\n },\n store(db, path, extracted) {\n const found = extracted as string[];\n if (found.length === 0) {\n db.prepare('DELETE FROM tags WHERE \"path\" = ?').run(path);\n } else {\n const placeholders = found.map(() => '?').join(', ');\n db.prepare(`DELETE FROM tags WHERE \"path\" = ? AND tag NOT IN (${placeholders})`).run(path, ...found);\n }\n const insert = db.prepare('INSERT OR IGNORE INTO tags (\"path\", tag) VALUES (?, ?)');\n for (const tag of found) insert.run(path, tag);\n },\n};\n"],"names":["tags","FENCE_RE","INLINE_CODE_RE","WIKILINK_RE","HTML_TAG_RE","INLINE_TAG_RE","normalizeTag","raw","stripped","replace","test","frontmatterTags","data","list","Array","isArray","found","item","tag","push","inlineTags","body","inFence","split","line","includes","cleaned","m","repeat","length","matchAll","extract","_raw","_search","Set","sort","name","schema","db","exec","remove","path","delta","vanished","prepare","run","store","extracted","placeholders","map","join","insert"],"mappings":";;;;+BA+DaA;;;eAAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;AA7Db,0FAA0F;AAC1F,uFAAuF;AACvF,4FAA4F;AAE5F,IAAMC,WAAW;AACjB,IAAMC,iBAAiB;AACvB,+DAA+D;AAC/D,IAAMC,cAAc,gBAAgB,6DAA6D;AACjG,iDAAiD;AACjD,IAAMC,cAAc,uBAAuB,6DAA6D;AACxG,iGAAiG;AACjG,IAAMC,gBAAgB;AAEtB,sFAAsF;AACtF,oEAAoE;AACpE,SAASC,aAAaC,GAAW;IAC/B,IAAMC,WAAWD,IAAIE,OAAO,CAAC,MAAM,IAAIA,OAAO,CAAC,QAAQ;IACvD,IAAI,CAACD,YAAY,QAAQE,IAAI,CAACF,WAAW,OAAO;IAChD,OAAOA;AACT;AAEA,4FAA4F;AAC5F,6EAA6E;AAC7E,SAASG,gBAAgBC,IAA8B;IACrD,IAAML,MAAMK,iBAAAA,2BAAAA,KAAMZ,IAAI;IACtB,IAAMa,OAAOC,MAAMC,OAAO,CAACR,OAAOA,MAAM,OAAOA,QAAQ,WAAW;QAACA;KAAI,GAAG,EAAE;IAC5E,IAAMS,QAAkB,EAAE;QACrB,kCAAA,2BAAA;;QAAL,QAAK,YAAcH,yBAAd,SAAA,6BAAA,QAAA,yBAAA,iCAAoB;YAApB,IAAMI,OAAN;YACH,IAAI,OAAOA,SAAS,UAAU;YAC9B,IAAMC,MAAMZ,aAAaW;YACzB,IAAIC,KAAKF,MAAMG,IAAI,CAACD;QACtB;;QAJK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAKL,OAAOF;AACT;AAEA,uFAAuF;AACvF,SAASI,WAAWC,IAAY;IAC9B,IAAML,QAAkB,EAAE;IAC1B,IAAIM,UAAU;QACT,kCAAA,2BAAA;;QAAL,QAAK,YAAcD,KAAKE,KAAK,CAAC,0BAAzB,SAAA,6BAAA,QAAA,yBAAA,iCAAgC;YAAhC,IAAMC,OAAN;YACH,IAAIvB,SAASS,IAAI,CAACc,OAAO;gBACvBF,UAAU,CAACA;gBACX;YACF;YACA,IAAIA,SAAS;YACb,IAAI,CAACE,KAAKC,QAAQ,CAAC,MAAM,UAAU,kCAAkC;YACrE,IAAIC,UAAUF,KAAKC,QAAQ,CAAC,OAAOD,KAAKf,OAAO,CAACP,gBAAgB,SAACyB;uBAAM,IAAIC,MAAM,CAACD,EAAEE,MAAM;iBAAKL;YAC/F,IAAIE,QAAQD,QAAQ,CAAC,OAAOC,UAAUA,QAAQjB,OAAO,CAACN,aAAa,SAACwB;uBAAM,IAAIC,MAAM,CAACD,EAAEE,MAAM;;YAC7F,IAAIH,QAAQD,QAAQ,CAAC,MAAMC,UAAUA,QAAQjB,OAAO,CAACL,aAAa,SAACuB;uBAAM,IAAIC,MAAM,CAACD,EAAEE,MAAM;;gBACvF,mCAAA,4BAAA;;gBAAL,QAAK,aAAWH,QAAQI,QAAQ,CAACzB,mCAA5B,UAAA,8BAAA,SAAA,0BAAA,kCAA4C;oBAA5C,IAAMsB,IAAN;oBACH,IAAMT,MAAMZ,aAAaqB,CAAC,CAAC,EAAE;oBAC7B,IAAIT,KAAKF,MAAMG,IAAI,CAACD;gBACtB;;gBAHK;gBAAA;;;yBAAA,8BAAA;wBAAA;;;wBAAA;8BAAA;;;;QAIP;;QAdK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAeL,OAAOF;AACT;AAEA,SAASe,QAAQC,IAAY,EAAEX,IAAY,EAAEY,OAA4C,EAAErB,IAA8B;IACvH,OAAO,AAAC,qBAAG,IAAIsB,IAAI,AAAC,qBAAGvB,gBAAgBC,cAAO,qBAAGQ,WAAWC,UAASc,IAAI;AAC3E;AAEO,IAAMnC,OAAgB;IAC3BoC,MAAM;IACNC,QAAAA,SAAAA,OAAOC,EAAE;QACPA,GAAGC,IAAI,CAAC;QACRD,GAAGC,IAAI,CAAC;IACV;IACAR,SAAAA;IACA,2FAA2F;IAC3F,uDAAuD;IACvDS,QAAAA,SAAAA,OAAOF,EAAE,EAAEG,IAAI,EAAEC,KAAK;QACpB,IAAI,CAACA,MAAMC,QAAQ,CAAClB,QAAQ,CAACgB,OAAO;QACpCH,GAAGM,OAAO,CAAC,qCAAqCC,GAAG,CAACJ;IACtD;IACAK,OAAAA,SAAAA,MAAMR,EAAE,EAAEG,IAAI,EAAEM,SAAS;QACvB,IAAM/B,QAAQ+B;QACd,IAAI/B,MAAMa,MAAM,KAAK,GAAG;YACtBS,GAAGM,OAAO,CAAC,qCAAqCC,GAAG,CAACJ;QACtD,OAAO;gBAELH;YADA,IAAMU,eAAehC,MAAMiC,GAAG,CAAC;uBAAM;eAAKC,IAAI,CAAC;YAC/CZ,CAAAA,cAAAA,GAAGM,OAAO,CAAC,AAAC,qDAAiE,OAAbI,cAAa,OAAIH,GAAG,OAApFP,aAAAA;gBAAqFG;aAAe,CAApGH,OAA2F,qBAAGtB;QAChG;QACA,IAAMmC,SAASb,GAAGM,OAAO,CAAC;YACrB,kCAAA,2BAAA;;YAAL,QAAK,YAAa5B,0BAAb,SAAA,6BAAA,QAAA,yBAAA;gBAAA,IAAME,MAAN;gBAAoBiC,OAAON,GAAG,CAACJ,MAAMvB;;;YAArC;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;IACP;AACF"}
@@ -1,4 +1,5 @@
1
1
  import { parseArgs } from 'node:util';
2
+ import { columnHint } from '../column-hint.js';
2
3
  import { resolvePreset } from '../config/index.js';
3
4
  import { open } from '../db/index.js';
4
5
  import { printRowStream } from '../output.js';
@@ -176,12 +177,13 @@ export function runSql(cfg, sql, params, format, label, preset) {
176
177
  const columns = statement.columns().map((c)=>c.name);
177
178
  printRowStream(statement.iterate(...params), format, columns);
178
179
  } catch (err) {
180
+ const hinted = columnHint(db, err); // pragma needs the db still open
179
181
  db.close();
180
182
  // A saved query or ad-hoc SQL can carry `content MATCH ?` too, so the same FTS5
181
183
  // punctuation trap applies -- the bound parameters are the search terms there. SQL
182
184
  // without MATCH gets its error verbatim; search advice on a plain typo would mislead.
183
- if (/\bMATCH\b/i.test(sql)) throw searchError(err, params.join(' '));
184
- throw err;
185
+ if (/\bMATCH\b/i.test(sql)) throw searchError(hinted, params.join(' '));
186
+ throw hinted;
185
187
  }
186
188
  db.close();
187
189
  }
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/cli/shared.ts"],"sourcesContent":["import type { ParseArgsOptionsConfig } from 'node:util';\nimport { parseArgs } from 'node:util';\nimport type { ResolvedConfig, SearchOverrides } from '../config/index.ts';\nimport { resolvePreset } from '../config/index.ts';\nimport type { OpenResult } from '../db/index.ts';\nimport { open } from '../db/index.ts';\nimport type { Row, RowFormat } from '../output.ts';\nimport { printRowStream } from '../output.ts';\nimport { searchError } from '../search-error.ts';\nimport type { Ctx } from './types.ts';\n\n// Spreadable option fragments -- one flag name keeps one meaning across every command's table.\nexport const FORMAT: ParseArgsOptionsConfig = { format: { type: 'string', default: 'table' } };\nexport const CONFIG: ParseArgsOptionsConfig = { config: { type: 'string' } };\n// The scope vocabulary every scoped command shares: named preset, ad hoc include/exclude\n// globs, a where SQL condition. search adds k on top.\nexport const SCOPE: ParseArgsOptionsConfig = {\n where: { type: 'string' },\n preset: { type: 'string' },\n include: { type: 'string', multiple: true },\n exclude: { type: 'string', multiple: true },\n // Widening, which no other flag can express: --include and --exclude each override their\n // own side of the preset, so neither can drop an exclusion the preset declares.\n 'no-exclude': { type: 'boolean', default: false },\n};\nexport const SEARCH_FLAGS: ParseArgsOptionsConfig = { ...SCOPE, k: { type: 'string' } };\n\ntype Values = Record<string, string | boolean | string[] | undefined>;\n\n// The SCOPE fragment's parsed values as the overrides every scoped command passes down. One\n// place to read them, so a new scope field is added to the fragment and here, not in each\n// command's call.\nexport function scopeOf(values: Values): SearchOverrides {\n return {\n preset: values.preset as string | undefined,\n include: values.include as string[] | undefined,\n exclude: values.exclude as string[] | undefined,\n where: values.where as string | undefined,\n noExclude: values['no-exclude'] === true,\n };\n}\n\n// An unrecognised value used to fall through to `table`, so a typo looked like it worked.\n// parseArgs is strict about flag names; this is the same strictness for the value.\nfunction pickFormat<T extends string>(values: Values, allowed: readonly T[]): T {\n const format = String(values.format);\n if ((allowed as readonly string[]).includes(format)) return format as T;\n console.error(`unknown --format \"${format}\"; expected ${allowed.join(', ')}`);\n process.exit(2);\n}\n\nexport function formatOf(values: Values): 'table' | 'json' {\n return pickFormat(values, ['table', 'json'] as const);\n}\n\n// csv is a row set rendered as rows; map, peek, status, and path render structures instead,\n// and take formatOf.\nexport function rowFormatOf(values: Values): RowFormat {\n return pickFormat(values, ['table', 'json', 'csv'] as const);\n}\n\n// Per-command parseArgs: strict (a foreign flag exits 2), and every table gets --help for free.\nexport function parse(argv: string[], usage: string, options: ParseArgsOptionsConfig): { values: Values; positionals: string[] } {\n let values: Values;\n let positionals: string[];\n try {\n ({ values, positionals } = parseArgs({\n args: argv,\n options: { ...options, help: { type: 'boolean', default: false, short: 'h' } },\n strict: true,\n allowPositionals: true,\n }));\n } catch (err) {\n console.error((err as Error).message);\n console.error(usage);\n process.exit(2);\n }\n if (values.help) {\n console.log(usage);\n process.exit(0);\n }\n return { values, positionals };\n}\n\n// --k must be a positive integer: SQLite reads a bound LIMIT of -1 as \"unlimited\" and 0 as\n// \"nothing\", and parseInt would silently truncate \"5.9\" -- all three are caller mistakes\n// worth a usage error, matching the config-level SavedSearch validation.\nexport function parseK(k: string | undefined, usageError: (message: string) => never): number | undefined {\n if (k === undefined) return undefined;\n const parsed = Number(k);\n if (!Number.isInteger(parsed) || parsed <= 0) usageError(`--k expects a positive integer, got \"${k}\"`);\n return parsed;\n}\n\n// Shared open-query-close envelope for commands that touch the tree.\n\nexport function printWarnings(warnings: string[]): void {\n for (const w of warnings) console.warn(w);\n}\n\nexport async function withDb(ctx: Ctx, configPath: string | undefined, fn: (db: OpenResult['db'], cfg: ResolvedConfig) => void | Promise<void>): Promise<void> {\n const cfg = ctx.resolveConfig(configPath);\n const { db, warnings } = open(cfg);\n printWarnings(warnings);\n try {\n await fn(db, cfg);\n } finally {\n db.close();\n }\n}\n\n// `--preset` on SQL binds the scope rather than applying it: the statement joins `scope`\n// itself. Filtering behind the query's back is not available -- the obvious version, temp views\n// shadowing the base tables, cannot cover `content`, because FTS5 `MATCH` uses the table name\n// as a hidden column and a view has none. A flag that silently scoped three tables of four\n// would look scoped and not be.\nfunction bindScope(db: OpenResult['db'], cfg: ResolvedConfig, preset: string): void {\n const { name } = resolvePreset(cfg, preset); // unknown names throw, listing what is declared\n db.exec('DROP TABLE IF EXISTS temp.scope');\n db.exec('CREATE TEMP TABLE scope (\"path\" TEXT PRIMARY KEY)');\n db.prepare('INSERT INTO temp.scope (\"path\") SELECT \"path\" FROM preset_files WHERE preset = ?').run(name);\n}\n\n// An unbound `?` silently binds NULL, so mismatched param counts fail loudly instead.\nexport function runSql(cfg: ResolvedConfig, sql: string, params: string[], format: RowFormat, label: string, preset?: string): void {\n const placeholderCount = (sql.match(/\\?/g) ?? []).length;\n if (params.length !== placeholderCount) {\n console.error(`${label} expects ${placeholderCount} parameter(s), got ${params.length}`);\n process.exit(2);\n }\n // Naming a preset and never joining it would return the whole index while reading as scoped,\n // which is the one hazard of binding rather than applying. Refuse instead.\n if (preset !== undefined && !/\\bscope\\b/i.test(sql)) {\n console.error(`--preset binds a temporary \"scope\" table of the preset's paths, and ${label} never joins it, so the preset would have no effect`);\n console.error(`add: JOIN scope ON scope.\"path\" = <table>.\"path\"`);\n process.exit(2);\n }\n const { db, warnings } = open(cfg);\n printWarnings(warnings);\n if (preset !== undefined) bindScope(db, cfg, preset);\n // Streamed rather than collected: `sql` is the one caller whose result size is the\n // statement's business, not this package's, so the rows are never held here in bulk.\n // columns() reads the statement's own metadata, so a 0-row csv still has its header.\n // A mid-stream SQLite error (e.g. SQLITE_BUSY outlasting busy_timeout) can still leave\n // partial output; the nonzero exit code is the caller's signal, output completeness is not\n // guaranteed on failure.\n try {\n const statement = db.prepare(sql);\n statement.setReadBigInts(true); // int64 past 2^53 arrives as BigInt instead of throwing at step time\n const columns = statement.columns().map((c) => c.name);\n printRowStream(statement.iterate(...params) as Iterable<Row>, format, columns);\n } catch (err) {\n db.close();\n // A saved query or ad-hoc SQL can carry `content MATCH ?` too, so the same FTS5\n // punctuation trap applies -- the bound parameters are the search terms there. SQL\n // without MATCH gets its error verbatim; search advice on a plain typo would mislead.\n if (/\\bMATCH\\b/i.test(sql)) throw searchError(err as Error, params.join(' '));\n throw err;\n }\n db.close();\n}\n"],"names":["parseArgs","resolvePreset","open","printRowStream","searchError","FORMAT","format","type","default","CONFIG","config","SCOPE","where","preset","include","multiple","exclude","SEARCH_FLAGS","k","scopeOf","values","noExclude","pickFormat","allowed","String","includes","console","error","join","process","exit","formatOf","rowFormatOf","parse","argv","usage","options","positionals","args","help","short","strict","allowPositionals","err","message","log","parseK","usageError","undefined","parsed","Number","isInteger","printWarnings","warnings","w","warn","withDb","ctx","configPath","fn","cfg","resolveConfig","db","close","bindScope","name","exec","prepare","run","runSql","sql","params","label","placeholderCount","match","length","test","statement","setReadBigInts","columns","map","c","iterate"],"mappings":"AACA,SAASA,SAAS,QAAQ,YAAY;AAEtC,SAASC,aAAa,QAAQ,qBAAqB;AAEnD,SAASC,IAAI,QAAQ,iBAAiB;AAEtC,SAASC,cAAc,QAAQ,eAAe;AAC9C,SAASC,WAAW,QAAQ,qBAAqB;AAGjD,+FAA+F;AAC/F,OAAO,MAAMC,SAAiC;IAAEC,QAAQ;QAAEC,MAAM;QAAUC,SAAS;IAAQ;AAAE,EAAE;AAC/F,OAAO,MAAMC,SAAiC;IAAEC,QAAQ;QAAEH,MAAM;IAAS;AAAE,EAAE;AAC7E,yFAAyF;AACzF,sDAAsD;AACtD,OAAO,MAAMI,QAAgC;IAC3CC,OAAO;QAAEL,MAAM;IAAS;IACxBM,QAAQ;QAAEN,MAAM;IAAS;IACzBO,SAAS;QAAEP,MAAM;QAAUQ,UAAU;IAAK;IAC1CC,SAAS;QAAET,MAAM;QAAUQ,UAAU;IAAK;IAC1C,yFAAyF;IACzF,gFAAgF;IAChF,cAAc;QAAER,MAAM;QAAWC,SAAS;IAAM;AAClD,EAAE;AACF,OAAO,MAAMS,eAAuC;IAAE,GAAGN,KAAK;IAAEO,GAAG;QAAEX,MAAM;IAAS;AAAE,EAAE;AAIxF,4FAA4F;AAC5F,0FAA0F;AAC1F,kBAAkB;AAClB,OAAO,SAASY,QAAQC,MAAc;IACpC,OAAO;QACLP,QAAQO,OAAOP,MAAM;QACrBC,SAASM,OAAON,OAAO;QACvBE,SAASI,OAAOJ,OAAO;QACvBJ,OAAOQ,OAAOR,KAAK;QACnBS,WAAWD,MAAM,CAAC,aAAa,KAAK;IACtC;AACF;AAEA,0FAA0F;AAC1F,mFAAmF;AACnF,SAASE,WAA6BF,MAAc,EAAEG,OAAqB;IACzE,MAAMjB,SAASkB,OAAOJ,OAAOd,MAAM;IACnC,IAAI,AAACiB,QAA8BE,QAAQ,CAACnB,SAAS,OAAOA;IAC5DoB,QAAQC,KAAK,CAAC,CAAC,kBAAkB,EAAErB,OAAO,YAAY,EAAEiB,QAAQK,IAAI,CAAC,OAAO;IAC5EC,QAAQC,IAAI,CAAC;AACf;AAEA,OAAO,SAASC,SAASX,MAAc;IACrC,OAAOE,WAAWF,QAAQ;QAAC;QAAS;KAAO;AAC7C;AAEA,4FAA4F;AAC5F,qBAAqB;AACrB,OAAO,SAASY,YAAYZ,MAAc;IACxC,OAAOE,WAAWF,QAAQ;QAAC;QAAS;QAAQ;KAAM;AACpD;AAEA,gGAAgG;AAChG,OAAO,SAASa,MAAMC,IAAc,EAAEC,KAAa,EAAEC,OAA+B;IAClF,IAAIhB;IACJ,IAAIiB;IACJ,IAAI;QACD,CAAA,EAAEjB,MAAM,EAAEiB,WAAW,EAAE,GAAGrC,UAAU;YACnCsC,MAAMJ;YACNE,SAAS;gBAAE,GAAGA,OAAO;gBAAEG,MAAM;oBAAEhC,MAAM;oBAAWC,SAAS;oBAAOgC,OAAO;gBAAI;YAAE;YAC7EC,QAAQ;YACRC,kBAAkB;QACpB,EAAC;IACH,EAAE,OAAOC,KAAK;QACZjB,QAAQC,KAAK,CAAC,AAACgB,IAAcC,OAAO;QACpClB,QAAQC,KAAK,CAACQ;QACdN,QAAQC,IAAI,CAAC;IACf;IACA,IAAIV,OAAOmB,IAAI,EAAE;QACfb,QAAQmB,GAAG,CAACV;QACZN,QAAQC,IAAI,CAAC;IACf;IACA,OAAO;QAAEV;QAAQiB;IAAY;AAC/B;AAEA,2FAA2F;AAC3F,yFAAyF;AACzF,yEAAyE;AACzE,OAAO,SAASS,OAAO5B,CAAqB,EAAE6B,UAAsC;IAClF,IAAI7B,MAAM8B,WAAW,OAAOA;IAC5B,MAAMC,SAASC,OAAOhC;IACtB,IAAI,CAACgC,OAAOC,SAAS,CAACF,WAAWA,UAAU,GAAGF,WAAW,CAAC,qCAAqC,EAAE7B,EAAE,CAAC,CAAC;IACrG,OAAO+B;AACT;AAEA,qEAAqE;AAErE,OAAO,SAASG,cAAcC,QAAkB;IAC9C,KAAK,MAAMC,KAAKD,SAAU3B,QAAQ6B,IAAI,CAACD;AACzC;AAEA,OAAO,eAAeE,OAAOC,GAAQ,EAAEC,UAA8B,EAAEC,EAAuE;IAC5I,MAAMC,MAAMH,IAAII,aAAa,CAACH;IAC9B,MAAM,EAAEI,EAAE,EAAET,QAAQ,EAAE,GAAGnD,KAAK0D;IAC9BR,cAAcC;IACd,IAAI;QACF,MAAMM,GAAGG,IAAIF;IACf,SAAU;QACRE,GAAGC,KAAK;IACV;AACF;AAEA,yFAAyF;AACzF,gGAAgG;AAChG,8FAA8F;AAC9F,2FAA2F;AAC3F,gCAAgC;AAChC,SAASC,UAAUF,EAAoB,EAAEF,GAAmB,EAAE/C,MAAc;IAC1E,MAAM,EAAEoD,IAAI,EAAE,GAAGhE,cAAc2D,KAAK/C,SAAS,gDAAgD;IAC7FiD,GAAGI,IAAI,CAAC;IACRJ,GAAGI,IAAI,CAAC;IACRJ,GAAGK,OAAO,CAAC,oFAAoFC,GAAG,CAACH;AACrG;AAEA,sFAAsF;AACtF,OAAO,SAASI,OAAOT,GAAmB,EAAEU,GAAW,EAAEC,MAAgB,EAAEjE,MAAiB,EAAEkE,KAAa,EAAE3D,MAAe;QAChGyD;IAA1B,MAAMG,mBAAmB,EAACH,aAAAA,IAAII,KAAK,CAAC,oBAAVJ,wBAAAA,aAAoB,EAAE,EAAEK,MAAM;IACxD,IAAIJ,OAAOI,MAAM,KAAKF,kBAAkB;QACtC/C,QAAQC,KAAK,CAAC,GAAG6C,MAAM,SAAS,EAAEC,iBAAiB,mBAAmB,EAAEF,OAAOI,MAAM,EAAE;QACvF9C,QAAQC,IAAI,CAAC;IACf;IACA,6FAA6F;IAC7F,2EAA2E;IAC3E,IAAIjB,WAAWmC,aAAa,CAAC,aAAa4B,IAAI,CAACN,MAAM;QACnD5C,QAAQC,KAAK,CAAC,CAAC,oEAAoE,EAAE6C,MAAM,mDAAmD,CAAC;QAC/I9C,QAAQC,KAAK,CAAC,CAAC,gDAAgD,CAAC;QAChEE,QAAQC,IAAI,CAAC;IACf;IACA,MAAM,EAAEgC,EAAE,EAAET,QAAQ,EAAE,GAAGnD,KAAK0D;IAC9BR,cAAcC;IACd,IAAIxC,WAAWmC,WAAWgB,UAAUF,IAAIF,KAAK/C;IAC7C,mFAAmF;IACnF,qFAAqF;IACrF,qFAAqF;IACrF,uFAAuF;IACvF,2FAA2F;IAC3F,yBAAyB;IACzB,IAAI;QACF,MAAMgE,YAAYf,GAAGK,OAAO,CAACG;QAC7BO,UAAUC,cAAc,CAAC,OAAO,qEAAqE;QACrG,MAAMC,UAAUF,UAAUE,OAAO,GAAGC,GAAG,CAAC,CAACC,IAAMA,EAAEhB,IAAI;QACrD9D,eAAe0E,UAAUK,OAAO,IAAIX,SAA0BjE,QAAQyE;IACxE,EAAE,OAAOpC,KAAK;QACZmB,GAAGC,KAAK;QACR,gFAAgF;QAChF,mFAAmF;QACnF,sFAAsF;QACtF,IAAI,aAAaa,IAAI,CAACN,MAAM,MAAMlE,YAAYuC,KAAc4B,OAAO3C,IAAI,CAAC;QACxE,MAAMe;IACR;IACAmB,GAAGC,KAAK;AACV"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/cli/shared.ts"],"sourcesContent":["import type { ParseArgsOptionsConfig } from 'node:util';\nimport { parseArgs } from 'node:util';\nimport { columnHint } from '../column-hint.ts';\nimport type { ResolvedConfig, SearchOverrides } from '../config/index.ts';\nimport { resolvePreset } from '../config/index.ts';\nimport type { OpenResult } from '../db/index.ts';\nimport { open } from '../db/index.ts';\nimport type { Row, RowFormat } from '../output.ts';\nimport { printRowStream } from '../output.ts';\nimport { searchError } from '../search-error.ts';\nimport type { Ctx } from './types.ts';\n\n// Spreadable option fragments -- one flag name keeps one meaning across every command's table.\nexport const FORMAT: ParseArgsOptionsConfig = { format: { type: 'string', default: 'table' } };\nexport const CONFIG: ParseArgsOptionsConfig = { config: { type: 'string' } };\n// The scope vocabulary every scoped command shares: named preset, ad hoc include/exclude\n// globs, a where SQL condition. search adds k on top.\nexport const SCOPE: ParseArgsOptionsConfig = {\n where: { type: 'string' },\n preset: { type: 'string' },\n include: { type: 'string', multiple: true },\n exclude: { type: 'string', multiple: true },\n // Widening, which no other flag can express: --include and --exclude each override their\n // own side of the preset, so neither can drop an exclusion the preset declares.\n 'no-exclude': { type: 'boolean', default: false },\n};\nexport const SEARCH_FLAGS: ParseArgsOptionsConfig = { ...SCOPE, k: { type: 'string' } };\n\ntype Values = Record<string, string | boolean | string[] | undefined>;\n\n// The SCOPE fragment's parsed values as the overrides every scoped command passes down. One\n// place to read them, so a new scope field is added to the fragment and here, not in each\n// command's call.\nexport function scopeOf(values: Values): SearchOverrides {\n return {\n preset: values.preset as string | undefined,\n include: values.include as string[] | undefined,\n exclude: values.exclude as string[] | undefined,\n where: values.where as string | undefined,\n noExclude: values['no-exclude'] === true,\n };\n}\n\n// An unrecognised value used to fall through to `table`, so a typo looked like it worked.\n// parseArgs is strict about flag names; this is the same strictness for the value.\nfunction pickFormat<T extends string>(values: Values, allowed: readonly T[]): T {\n const format = String(values.format);\n if ((allowed as readonly string[]).includes(format)) return format as T;\n console.error(`unknown --format \"${format}\"; expected ${allowed.join(', ')}`);\n process.exit(2);\n}\n\nexport function formatOf(values: Values): 'table' | 'json' {\n return pickFormat(values, ['table', 'json'] as const);\n}\n\n// csv is a row set rendered as rows; map, peek, status, and path render structures instead,\n// and take formatOf.\nexport function rowFormatOf(values: Values): RowFormat {\n return pickFormat(values, ['table', 'json', 'csv'] as const);\n}\n\n// Per-command parseArgs: strict (a foreign flag exits 2), and every table gets --help for free.\nexport function parse(argv: string[], usage: string, options: ParseArgsOptionsConfig): { values: Values; positionals: string[] } {\n let values: Values;\n let positionals: string[];\n try {\n ({ values, positionals } = parseArgs({\n args: argv,\n options: { ...options, help: { type: 'boolean', default: false, short: 'h' } },\n strict: true,\n allowPositionals: true,\n }));\n } catch (err) {\n console.error((err as Error).message);\n console.error(usage);\n process.exit(2);\n }\n if (values.help) {\n console.log(usage);\n process.exit(0);\n }\n return { values, positionals };\n}\n\n// --k must be a positive integer: SQLite reads a bound LIMIT of -1 as \"unlimited\" and 0 as\n// \"nothing\", and parseInt would silently truncate \"5.9\" -- all three are caller mistakes\n// worth a usage error, matching the config-level SavedSearch validation.\nexport function parseK(k: string | undefined, usageError: (message: string) => never): number | undefined {\n if (k === undefined) return undefined;\n const parsed = Number(k);\n if (!Number.isInteger(parsed) || parsed <= 0) usageError(`--k expects a positive integer, got \"${k}\"`);\n return parsed;\n}\n\n// Shared open-query-close envelope for commands that touch the tree.\n\nexport function printWarnings(warnings: string[]): void {\n for (const w of warnings) console.warn(w);\n}\n\nexport async function withDb(ctx: Ctx, configPath: string | undefined, fn: (db: OpenResult['db'], cfg: ResolvedConfig) => void | Promise<void>): Promise<void> {\n const cfg = ctx.resolveConfig(configPath);\n const { db, warnings } = open(cfg);\n printWarnings(warnings);\n try {\n await fn(db, cfg);\n } finally {\n db.close();\n }\n}\n\n// `--preset` on SQL binds the scope rather than applying it: the statement joins `scope`\n// itself. Filtering behind the query's back is not available -- the obvious version, temp views\n// shadowing the base tables, cannot cover `content`, because FTS5 `MATCH` uses the table name\n// as a hidden column and a view has none. A flag that silently scoped three tables of four\n// would look scoped and not be.\nfunction bindScope(db: OpenResult['db'], cfg: ResolvedConfig, preset: string): void {\n const { name } = resolvePreset(cfg, preset); // unknown names throw, listing what is declared\n db.exec('DROP TABLE IF EXISTS temp.scope');\n db.exec('CREATE TEMP TABLE scope (\"path\" TEXT PRIMARY KEY)');\n db.prepare('INSERT INTO temp.scope (\"path\") SELECT \"path\" FROM preset_files WHERE preset = ?').run(name);\n}\n\n// An unbound `?` silently binds NULL, so mismatched param counts fail loudly instead.\nexport function runSql(cfg: ResolvedConfig, sql: string, params: string[], format: RowFormat, label: string, preset?: string): void {\n const placeholderCount = (sql.match(/\\?/g) ?? []).length;\n if (params.length !== placeholderCount) {\n console.error(`${label} expects ${placeholderCount} parameter(s), got ${params.length}`);\n process.exit(2);\n }\n // Naming a preset and never joining it would return the whole index while reading as scoped,\n // which is the one hazard of binding rather than applying. Refuse instead.\n if (preset !== undefined && !/\\bscope\\b/i.test(sql)) {\n console.error(`--preset binds a temporary \"scope\" table of the preset's paths, and ${label} never joins it, so the preset would have no effect`);\n console.error(`add: JOIN scope ON scope.\"path\" = <table>.\"path\"`);\n process.exit(2);\n }\n const { db, warnings } = open(cfg);\n printWarnings(warnings);\n if (preset !== undefined) bindScope(db, cfg, preset);\n // Streamed rather than collected: `sql` is the one caller whose result size is the\n // statement's business, not this package's, so the rows are never held here in bulk.\n // columns() reads the statement's own metadata, so a 0-row csv still has its header.\n // A mid-stream SQLite error (e.g. SQLITE_BUSY outlasting busy_timeout) can still leave\n // partial output; the nonzero exit code is the caller's signal, output completeness is not\n // guaranteed on failure.\n try {\n const statement = db.prepare(sql);\n statement.setReadBigInts(true); // int64 past 2^53 arrives as BigInt instead of throwing at step time\n const columns = statement.columns().map((c) => c.name);\n printRowStream(statement.iterate(...params) as Iterable<Row>, format, columns);\n } catch (err) {\n const hinted = columnHint(db, err as Error); // pragma needs the db still open\n db.close();\n // A saved query or ad-hoc SQL can carry `content MATCH ?` too, so the same FTS5\n // punctuation trap applies -- the bound parameters are the search terms there. SQL\n // without MATCH gets its error verbatim; search advice on a plain typo would mislead.\n if (/\\bMATCH\\b/i.test(sql)) throw searchError(hinted, params.join(' '));\n throw hinted;\n }\n db.close();\n}\n"],"names":["parseArgs","columnHint","resolvePreset","open","printRowStream","searchError","FORMAT","format","type","default","CONFIG","config","SCOPE","where","preset","include","multiple","exclude","SEARCH_FLAGS","k","scopeOf","values","noExclude","pickFormat","allowed","String","includes","console","error","join","process","exit","formatOf","rowFormatOf","parse","argv","usage","options","positionals","args","help","short","strict","allowPositionals","err","message","log","parseK","usageError","undefined","parsed","Number","isInteger","printWarnings","warnings","w","warn","withDb","ctx","configPath","fn","cfg","resolveConfig","db","close","bindScope","name","exec","prepare","run","runSql","sql","params","label","placeholderCount","match","length","test","statement","setReadBigInts","columns","map","c","iterate","hinted"],"mappings":"AACA,SAASA,SAAS,QAAQ,YAAY;AACtC,SAASC,UAAU,QAAQ,oBAAoB;AAE/C,SAASC,aAAa,QAAQ,qBAAqB;AAEnD,SAASC,IAAI,QAAQ,iBAAiB;AAEtC,SAASC,cAAc,QAAQ,eAAe;AAC9C,SAASC,WAAW,QAAQ,qBAAqB;AAGjD,+FAA+F;AAC/F,OAAO,MAAMC,SAAiC;IAAEC,QAAQ;QAAEC,MAAM;QAAUC,SAAS;IAAQ;AAAE,EAAE;AAC/F,OAAO,MAAMC,SAAiC;IAAEC,QAAQ;QAAEH,MAAM;IAAS;AAAE,EAAE;AAC7E,yFAAyF;AACzF,sDAAsD;AACtD,OAAO,MAAMI,QAAgC;IAC3CC,OAAO;QAAEL,MAAM;IAAS;IACxBM,QAAQ;QAAEN,MAAM;IAAS;IACzBO,SAAS;QAAEP,MAAM;QAAUQ,UAAU;IAAK;IAC1CC,SAAS;QAAET,MAAM;QAAUQ,UAAU;IAAK;IAC1C,yFAAyF;IACzF,gFAAgF;IAChF,cAAc;QAAER,MAAM;QAAWC,SAAS;IAAM;AAClD,EAAE;AACF,OAAO,MAAMS,eAAuC;IAAE,GAAGN,KAAK;IAAEO,GAAG;QAAEX,MAAM;IAAS;AAAE,EAAE;AAIxF,4FAA4F;AAC5F,0FAA0F;AAC1F,kBAAkB;AAClB,OAAO,SAASY,QAAQC,MAAc;IACpC,OAAO;QACLP,QAAQO,OAAOP,MAAM;QACrBC,SAASM,OAAON,OAAO;QACvBE,SAASI,OAAOJ,OAAO;QACvBJ,OAAOQ,OAAOR,KAAK;QACnBS,WAAWD,MAAM,CAAC,aAAa,KAAK;IACtC;AACF;AAEA,0FAA0F;AAC1F,mFAAmF;AACnF,SAASE,WAA6BF,MAAc,EAAEG,OAAqB;IACzE,MAAMjB,SAASkB,OAAOJ,OAAOd,MAAM;IACnC,IAAI,AAACiB,QAA8BE,QAAQ,CAACnB,SAAS,OAAOA;IAC5DoB,QAAQC,KAAK,CAAC,CAAC,kBAAkB,EAAErB,OAAO,YAAY,EAAEiB,QAAQK,IAAI,CAAC,OAAO;IAC5EC,QAAQC,IAAI,CAAC;AACf;AAEA,OAAO,SAASC,SAASX,MAAc;IACrC,OAAOE,WAAWF,QAAQ;QAAC;QAAS;KAAO;AAC7C;AAEA,4FAA4F;AAC5F,qBAAqB;AACrB,OAAO,SAASY,YAAYZ,MAAc;IACxC,OAAOE,WAAWF,QAAQ;QAAC;QAAS;QAAQ;KAAM;AACpD;AAEA,gGAAgG;AAChG,OAAO,SAASa,MAAMC,IAAc,EAAEC,KAAa,EAAEC,OAA+B;IAClF,IAAIhB;IACJ,IAAIiB;IACJ,IAAI;QACD,CAAA,EAAEjB,MAAM,EAAEiB,WAAW,EAAE,GAAGtC,UAAU;YACnCuC,MAAMJ;YACNE,SAAS;gBAAE,GAAGA,OAAO;gBAAEG,MAAM;oBAAEhC,MAAM;oBAAWC,SAAS;oBAAOgC,OAAO;gBAAI;YAAE;YAC7EC,QAAQ;YACRC,kBAAkB;QACpB,EAAC;IACH,EAAE,OAAOC,KAAK;QACZjB,QAAQC,KAAK,CAAC,AAACgB,IAAcC,OAAO;QACpClB,QAAQC,KAAK,CAACQ;QACdN,QAAQC,IAAI,CAAC;IACf;IACA,IAAIV,OAAOmB,IAAI,EAAE;QACfb,QAAQmB,GAAG,CAACV;QACZN,QAAQC,IAAI,CAAC;IACf;IACA,OAAO;QAAEV;QAAQiB;IAAY;AAC/B;AAEA,2FAA2F;AAC3F,yFAAyF;AACzF,yEAAyE;AACzE,OAAO,SAASS,OAAO5B,CAAqB,EAAE6B,UAAsC;IAClF,IAAI7B,MAAM8B,WAAW,OAAOA;IAC5B,MAAMC,SAASC,OAAOhC;IACtB,IAAI,CAACgC,OAAOC,SAAS,CAACF,WAAWA,UAAU,GAAGF,WAAW,CAAC,qCAAqC,EAAE7B,EAAE,CAAC,CAAC;IACrG,OAAO+B;AACT;AAEA,qEAAqE;AAErE,OAAO,SAASG,cAAcC,QAAkB;IAC9C,KAAK,MAAMC,KAAKD,SAAU3B,QAAQ6B,IAAI,CAACD;AACzC;AAEA,OAAO,eAAeE,OAAOC,GAAQ,EAAEC,UAA8B,EAAEC,EAAuE;IAC5I,MAAMC,MAAMH,IAAII,aAAa,CAACH;IAC9B,MAAM,EAAEI,EAAE,EAAET,QAAQ,EAAE,GAAGnD,KAAK0D;IAC9BR,cAAcC;IACd,IAAI;QACF,MAAMM,GAAGG,IAAIF;IACf,SAAU;QACRE,GAAGC,KAAK;IACV;AACF;AAEA,yFAAyF;AACzF,gGAAgG;AAChG,8FAA8F;AAC9F,2FAA2F;AAC3F,gCAAgC;AAChC,SAASC,UAAUF,EAAoB,EAAEF,GAAmB,EAAE/C,MAAc;IAC1E,MAAM,EAAEoD,IAAI,EAAE,GAAGhE,cAAc2D,KAAK/C,SAAS,gDAAgD;IAC7FiD,GAAGI,IAAI,CAAC;IACRJ,GAAGI,IAAI,CAAC;IACRJ,GAAGK,OAAO,CAAC,oFAAoFC,GAAG,CAACH;AACrG;AAEA,sFAAsF;AACtF,OAAO,SAASI,OAAOT,GAAmB,EAAEU,GAAW,EAAEC,MAAgB,EAAEjE,MAAiB,EAAEkE,KAAa,EAAE3D,MAAe;QAChGyD;IAA1B,MAAMG,mBAAmB,EAACH,aAAAA,IAAII,KAAK,CAAC,oBAAVJ,wBAAAA,aAAoB,EAAE,EAAEK,MAAM;IACxD,IAAIJ,OAAOI,MAAM,KAAKF,kBAAkB;QACtC/C,QAAQC,KAAK,CAAC,GAAG6C,MAAM,SAAS,EAAEC,iBAAiB,mBAAmB,EAAEF,OAAOI,MAAM,EAAE;QACvF9C,QAAQC,IAAI,CAAC;IACf;IACA,6FAA6F;IAC7F,2EAA2E;IAC3E,IAAIjB,WAAWmC,aAAa,CAAC,aAAa4B,IAAI,CAACN,MAAM;QACnD5C,QAAQC,KAAK,CAAC,CAAC,oEAAoE,EAAE6C,MAAM,mDAAmD,CAAC;QAC/I9C,QAAQC,KAAK,CAAC,CAAC,gDAAgD,CAAC;QAChEE,QAAQC,IAAI,CAAC;IACf;IACA,MAAM,EAAEgC,EAAE,EAAET,QAAQ,EAAE,GAAGnD,KAAK0D;IAC9BR,cAAcC;IACd,IAAIxC,WAAWmC,WAAWgB,UAAUF,IAAIF,KAAK/C;IAC7C,mFAAmF;IACnF,qFAAqF;IACrF,qFAAqF;IACrF,uFAAuF;IACvF,2FAA2F;IAC3F,yBAAyB;IACzB,IAAI;QACF,MAAMgE,YAAYf,GAAGK,OAAO,CAACG;QAC7BO,UAAUC,cAAc,CAAC,OAAO,qEAAqE;QACrG,MAAMC,UAAUF,UAAUE,OAAO,GAAGC,GAAG,CAAC,CAACC,IAAMA,EAAEhB,IAAI;QACrD9D,eAAe0E,UAAUK,OAAO,IAAIX,SAA0BjE,QAAQyE;IACxE,EAAE,OAAOpC,KAAK;QACZ,MAAMwC,SAASnF,WAAW8D,IAAInB,MAAe,iCAAiC;QAC9EmB,GAAGC,KAAK;QACR,gFAAgF;QAChF,mFAAmF;QACnF,sFAAsF;QACtF,IAAI,aAAaa,IAAI,CAACN,MAAM,MAAMlE,YAAY+E,QAAQZ,OAAO3C,IAAI,CAAC;QAClE,MAAMuD;IACR;IACArB,GAAGC,KAAK;AACV"}
@@ -0,0 +1,2 @@
1
+ import type { DatabaseSync } from 'node:sqlite';
2
+ export declare function columnHint(db: DatabaseSync, err: Error): Error;
@@ -0,0 +1,32 @@
1
+ // A frontmatter key with punctuation (`plugin-id`, `a.b`) is a real column, but naming it
2
+ // unquoted in SQL parses as an expression: `plugin-id` is `plugin - id`, and SQLite's error
3
+ // names a fragment (`plugin`) the user never wrote. This maps that fragment back to the
4
+ // columns it could have come from, so the error names the actual fix instead.
5
+ function startsWithBoundary(name, prefix) {
6
+ return name.length > prefix.length && name.startsWith(prefix) && /\W/.test(name[prefix.length]);
7
+ }
8
+ export function columnHint(db, err) {
9
+ const match = /no such column: (\S+)/.exec(err.message);
10
+ if (!match) return err;
11
+ const token = match[1];
12
+ // `f.plugin` -> `plugin`; left as-is when there's no leading alias segment, which also
13
+ // covers a column literally named `a.b` -- the full-token check below matches that case
14
+ // directly, so stripping here never needs to special-case it.
15
+ const aliasStripped = token.replace(/^[A-Za-z_]\w*\./, '');
16
+ const forms = new Set([
17
+ token,
18
+ aliasStripped
19
+ ]);
20
+ const columns = db.prepare('PRAGMA table_info(frontmatter)').all().map((c)=>c.name);
21
+ const candidates = new Set();
22
+ for (const name of columns){
23
+ for (const form of forms){
24
+ if (name === form || startsWithBoundary(name, form)) candidates.add(name);
25
+ }
26
+ }
27
+ if (candidates.size === 0) return err;
28
+ const quoted = [
29
+ ...candidates
30
+ ].map((name)=>`"${name.split('"').join('""')}"`).join(', ');
31
+ return new Error(`${err.message} -- ${quoted} needs double quotes in SQL: unquoted, SQLite parses the punctuation as an operator instead of column syntax.`);
32
+ }
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/column-hint.ts"],"sourcesContent":["import type { DatabaseSync } from 'node:sqlite';\n\n// A frontmatter key with punctuation (`plugin-id`, `a.b`) is a real column, but naming it\n// unquoted in SQL parses as an expression: `plugin-id` is `plugin - id`, and SQLite's error\n// names a fragment (`plugin`) the user never wrote. This maps that fragment back to the\n// columns it could have come from, so the error names the actual fix instead.\n\nfunction startsWithBoundary(name: string, prefix: string): boolean {\n return name.length > prefix.length && name.startsWith(prefix) && /\\W/.test(name[prefix.length]);\n}\n\nexport function columnHint(db: DatabaseSync, err: Error): Error {\n const match = /no such column: (\\S+)/.exec(err.message);\n if (!match) return err;\n const token = match[1];\n // `f.plugin` -> `plugin`; left as-is when there's no leading alias segment, which also\n // covers a column literally named `a.b` -- the full-token check below matches that case\n // directly, so stripping here never needs to special-case it.\n const aliasStripped = token.replace(/^[A-Za-z_]\\w*\\./, '');\n const forms = new Set([token, aliasStripped]);\n const columns = (db.prepare('PRAGMA table_info(frontmatter)').all() as Array<{ name: string }>).map((c) => c.name);\n const candidates = new Set<string>();\n for (const name of columns) {\n for (const form of forms) {\n if (name === form || startsWithBoundary(name, form)) candidates.add(name);\n }\n }\n if (candidates.size === 0) return err;\n const quoted = [...candidates].map((name) => `\"${name.split('\"').join('\"\"')}\"`).join(', ');\n return new Error(`${err.message} -- ${quoted} needs double quotes in SQL: unquoted, SQLite parses the punctuation as an operator instead of column syntax.`);\n}\n"],"names":["startsWithBoundary","name","prefix","length","startsWith","test","columnHint","db","err","match","exec","message","token","aliasStripped","replace","forms","Set","columns","prepare","all","map","c","candidates","form","add","size","quoted","split","join","Error"],"mappings":"AAEA,0FAA0F;AAC1F,4FAA4F;AAC5F,wFAAwF;AACxF,8EAA8E;AAE9E,SAASA,mBAAmBC,IAAY,EAAEC,MAAc;IACtD,OAAOD,KAAKE,MAAM,GAAGD,OAAOC,MAAM,IAAIF,KAAKG,UAAU,CAACF,WAAW,KAAKG,IAAI,CAACJ,IAAI,CAACC,OAAOC,MAAM,CAAC;AAChG;AAEA,OAAO,SAASG,WAAWC,EAAgB,EAAEC,GAAU;IACrD,MAAMC,QAAQ,wBAAwBC,IAAI,CAACF,IAAIG,OAAO;IACtD,IAAI,CAACF,OAAO,OAAOD;IACnB,MAAMI,QAAQH,KAAK,CAAC,EAAE;IACtB,uFAAuF;IACvF,wFAAwF;IACxF,8DAA8D;IAC9D,MAAMI,gBAAgBD,MAAME,OAAO,CAAC,mBAAmB;IACvD,MAAMC,QAAQ,IAAIC,IAAI;QAACJ;QAAOC;KAAc;IAC5C,MAAMI,UAAU,AAACV,GAAGW,OAAO,CAAC,kCAAkCC,GAAG,GAA+BC,GAAG,CAAC,CAACC,IAAMA,EAAEpB,IAAI;IACjH,MAAMqB,aAAa,IAAIN;IACvB,KAAK,MAAMf,QAAQgB,QAAS;QAC1B,KAAK,MAAMM,QAAQR,MAAO;YACxB,IAAId,SAASsB,QAAQvB,mBAAmBC,MAAMsB,OAAOD,WAAWE,GAAG,CAACvB;QACtE;IACF;IACA,IAAIqB,WAAWG,IAAI,KAAK,GAAG,OAAOjB;IAClC,MAAMkB,SAAS;WAAIJ;KAAW,CAACF,GAAG,CAAC,CAACnB,OAAS,CAAC,CAAC,EAAEA,KAAK0B,KAAK,CAAC,KAAKC,IAAI,CAAC,MAAM,CAAC,CAAC,EAAEA,IAAI,CAAC;IACrF,OAAO,IAAIC,MAAM,GAAGrB,IAAIG,OAAO,CAAC,IAAI,EAAEe,OAAO,6GAA6G,CAAC;AAC7J"}
@@ -1,4 +1,5 @@
1
1
  import { matchesGlob } from 'node:path';
2
+ import { columnHint } from '../column-hint.js';
2
3
  import { anyPresetEmbeds, resolveSearch } from '../config/index.js';
3
4
  export const INTERNAL_COLUMNS = new Set([
4
5
  'path',
@@ -34,7 +35,12 @@ export function rawScope(db, cfg, overrides, allPaths) {
34
35
  }
35
36
  export function narrowByWhere(db, paths, where) {
36
37
  if (!where) return paths;
37
- const whereRows = db.prepare(`SELECT "path" FROM frontmatter f WHERE (${where})`).all();
38
+ let whereRows;
39
+ try {
40
+ whereRows = db.prepare(`SELECT "path" FROM frontmatter f WHERE (${where})`).all();
41
+ } catch (err) {
42
+ throw columnHint(db, err);
43
+ }
38
44
  const wherePaths = new Set(whereRows.map((r)=>r.path));
39
45
  return new Set([
40
46
  ...paths
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/commands/scope.ts"],"sourcesContent":["import { matchesGlob } from 'node:path';\nimport type { DatabaseSync } from 'node:sqlite';\nimport type { ResolvedConfig, SearchOverrides } from '../config/index.ts';\nimport { anyPresetEmbeds, resolveSearch } from '../config/index.ts';\n\nexport const INTERNAL_COLUMNS = new Set(['path', '_mtime', '_ctime', '_size', '_rank', '_parse_error']);\n\n// node:path's matchesGlob is experimental (stable behind an unstable-API flag) as of the\n// engines floor (Node >=22.20); scope filtering only ever needs single-pattern matching, so\n// it's used here in JS rather than running a directory walk in the query path.\nexport function inScope(path: string, include: string[], exclude?: string[]): boolean {\n if (!include.some((g) => matchesGlob(path, g))) return false;\n if (exclude?.some((g) => matchesGlob(path, g))) return false;\n return true;\n}\n\nexport function scopeHasEmbeddings(db: DatabaseSync, cfg: ResolvedConfig, scopedPaths: Set<string>): boolean {\n if (!anyPresetEmbeds(cfg)) return false; // the embeddings table doesn't exist at all in this case\n const rows = db.prepare('SELECT DISTINCT \"path\" FROM embeddings').all() as Array<{ path: string }>;\n return rows.some((r) => scopedPaths.has(r.path));\n}\n\n// Scope only (no --where): preset_files for a named preset, JS glob matching for an ad hoc\n// include/exclude override. Shared by scopedPaths() and search(), which also needs the\n// pre-where set to size the candidate-pool filter.\nexport function rawScope(db: DatabaseSync, cfg: ResolvedConfig, overrides: SearchOverrides, allPaths?: string[]): Set<string> {\n const effective = resolveSearch(cfg, overrides);\n const { include, exclude } = effective;\n const adHocScope = overrides.include !== undefined || overrides.exclude !== undefined || overrides.noExclude === true;\n if (!adHocScope) return new Set((db.prepare('SELECT \"path\" FROM preset_files WHERE preset = ?').all(effective.presetName) as Array<{ path: string }>).map((r) => r.path));\n const paths = allPaths ?? (db.prepare('SELECT \"path\" FROM frontmatter').all() as Array<{ path: string }>).map((r) => r.path);\n return new Set(paths.filter((p) => inScope(p, include, exclude)));\n}\n\nexport function narrowByWhere(db: DatabaseSync, paths: Set<string>, where: string | undefined): Set<string> {\n if (!where) return paths;\n const whereRows = db.prepare(`SELECT \"path\" FROM frontmatter f WHERE (${where})`).all() as Array<{ path: string }>;\n const wherePaths = new Set(whereRows.map((r) => r.path));\n return new Set([...paths].filter((p) => wherePaths.has(p)));\n}\n\n// The scope resolver for non-search commands (path, peek, map): same coverage rule search()\n// applies, then narrowed by the resolved `where`.\nexport function scopedPaths(db: DatabaseSync, cfg: ResolvedConfig, overrides: SearchOverrides): Set<string> {\n const effective = resolveSearch(cfg, overrides);\n return narrowByWhere(db, rawScope(db, cfg, overrides), effective.where);\n}\n\n// Materializes a path set into a named temp table (same shape as traverse.ts's allowed_nodes)\n// so a query can join/filter against it cheaply instead of binding a parameter per path.\nexport function materializeScope(db: DatabaseSync, table: string, paths: Set<string>): void {\n db.exec(`DROP TABLE IF EXISTS ${table}`);\n db.exec(`CREATE TEMP TABLE ${table} (\"path\" TEXT PRIMARY KEY)`);\n db.prepare(`INSERT INTO ${table} SELECT DISTINCT value FROM json_each(?1)`).run(JSON.stringify([...paths]));\n}\n\nexport function setupMapScope(db: DatabaseSync, paths: Set<string>): void {\n materializeScope(db, '_map_scope', paths);\n}\n"],"names":["matchesGlob","anyPresetEmbeds","resolveSearch","INTERNAL_COLUMNS","Set","inScope","path","include","exclude","some","g","scopeHasEmbeddings","db","cfg","scopedPaths","rows","prepare","all","r","has","rawScope","overrides","allPaths","effective","adHocScope","undefined","noExclude","presetName","map","paths","filter","p","narrowByWhere","where","whereRows","wherePaths","materializeScope","table","exec","run","JSON","stringify","setupMapScope"],"mappings":"AAAA,SAASA,WAAW,QAAQ,YAAY;AAGxC,SAASC,eAAe,EAAEC,aAAa,QAAQ,qBAAqB;AAEpE,OAAO,MAAMC,mBAAmB,IAAIC,IAAI;IAAC;IAAQ;IAAU;IAAU;IAAS;IAAS;CAAe,EAAE;AAExG,yFAAyF;AACzF,4FAA4F;AAC5F,+EAA+E;AAC/E,OAAO,SAASC,QAAQC,IAAY,EAAEC,OAAiB,EAAEC,OAAkB;IACzE,IAAI,CAACD,QAAQE,IAAI,CAAC,CAACC,IAAMV,YAAYM,MAAMI,KAAK,OAAO;IACvD,IAAIF,oBAAAA,8BAAAA,QAASC,IAAI,CAAC,CAACC,IAAMV,YAAYM,MAAMI,KAAK,OAAO;IACvD,OAAO;AACT;AAEA,OAAO,SAASC,mBAAmBC,EAAgB,EAAEC,GAAmB,EAAEC,WAAwB;IAChG,IAAI,CAACb,gBAAgBY,MAAM,OAAO,OAAO,yDAAyD;IAClG,MAAME,OAAOH,GAAGI,OAAO,CAAC,0CAA0CC,GAAG;IACrE,OAAOF,KAAKN,IAAI,CAAC,CAACS,IAAMJ,YAAYK,GAAG,CAACD,EAAEZ,IAAI;AAChD;AAEA,2FAA2F;AAC3F,uFAAuF;AACvF,mDAAmD;AACnD,OAAO,SAASc,SAASR,EAAgB,EAAEC,GAAmB,EAAEQ,SAA0B,EAAEC,QAAmB;IAC7G,MAAMC,YAAYrB,cAAcW,KAAKQ;IACrC,MAAM,EAAEd,OAAO,EAAEC,OAAO,EAAE,GAAGe;IAC7B,MAAMC,aAAaH,UAAUd,OAAO,KAAKkB,aAAaJ,UAAUb,OAAO,KAAKiB,aAAaJ,UAAUK,SAAS,KAAK;IACjH,IAAI,CAACF,YAAY,OAAO,IAAIpB,IAAI,AAACQ,GAAGI,OAAO,CAAC,oDAAoDC,GAAG,CAACM,UAAUI,UAAU,EAA8BC,GAAG,CAAC,CAACV,IAAMA,EAAEZ,IAAI;IACvK,MAAMuB,QAAQP,qBAAAA,sBAAAA,WAAY,AAACV,GAAGI,OAAO,CAAC,kCAAkCC,GAAG,GAA+BW,GAAG,CAAC,CAACV,IAAMA,EAAEZ,IAAI;IAC3H,OAAO,IAAIF,IAAIyB,MAAMC,MAAM,CAAC,CAACC,IAAM1B,QAAQ0B,GAAGxB,SAASC;AACzD;AAEA,OAAO,SAASwB,cAAcpB,EAAgB,EAAEiB,KAAkB,EAAEI,KAAyB;IAC3F,IAAI,CAACA,OAAO,OAAOJ;IACnB,MAAMK,YAAYtB,GAAGI,OAAO,CAAC,CAAC,wCAAwC,EAAEiB,MAAM,CAAC,CAAC,EAAEhB,GAAG;IACrF,MAAMkB,aAAa,IAAI/B,IAAI8B,UAAUN,GAAG,CAAC,CAACV,IAAMA,EAAEZ,IAAI;IACtD,OAAO,IAAIF,IAAI;WAAIyB;KAAM,CAACC,MAAM,CAAC,CAACC,IAAMI,WAAWhB,GAAG,CAACY;AACzD;AAEA,4FAA4F;AAC5F,kDAAkD;AAClD,OAAO,SAASjB,YAAYF,EAAgB,EAAEC,GAAmB,EAAEQ,SAA0B;IAC3F,MAAME,YAAYrB,cAAcW,KAAKQ;IACrC,OAAOW,cAAcpB,IAAIQ,SAASR,IAAIC,KAAKQ,YAAYE,UAAUU,KAAK;AACxE;AAEA,8FAA8F;AAC9F,yFAAyF;AACzF,OAAO,SAASG,iBAAiBxB,EAAgB,EAAEyB,KAAa,EAAER,KAAkB;IAClFjB,GAAG0B,IAAI,CAAC,CAAC,qBAAqB,EAAED,OAAO;IACvCzB,GAAG0B,IAAI,CAAC,CAAC,kBAAkB,EAAED,MAAM,0BAA0B,CAAC;IAC9DzB,GAAGI,OAAO,CAAC,CAAC,YAAY,EAAEqB,MAAM,yCAAyC,CAAC,EAAEE,GAAG,CAACC,KAAKC,SAAS,CAAC;WAAIZ;KAAM;AAC3G;AAEA,OAAO,SAASa,cAAc9B,EAAgB,EAAEiB,KAAkB;IAChEO,iBAAiBxB,IAAI,cAAciB;AACrC"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/commands/scope.ts"],"sourcesContent":["import { matchesGlob } from 'node:path';\nimport type { DatabaseSync } from 'node:sqlite';\nimport { columnHint } from '../column-hint.ts';\nimport type { ResolvedConfig, SearchOverrides } from '../config/index.ts';\nimport { anyPresetEmbeds, resolveSearch } from '../config/index.ts';\n\nexport const INTERNAL_COLUMNS = new Set(['path', '_mtime', '_ctime', '_size', '_rank', '_parse_error']);\n\n// node:path's matchesGlob is experimental (stable behind an unstable-API flag) as of the\n// engines floor (Node >=22.20); scope filtering only ever needs single-pattern matching, so\n// it's used here in JS rather than running a directory walk in the query path.\nexport function inScope(path: string, include: string[], exclude?: string[]): boolean {\n if (!include.some((g) => matchesGlob(path, g))) return false;\n if (exclude?.some((g) => matchesGlob(path, g))) return false;\n return true;\n}\n\nexport function scopeHasEmbeddings(db: DatabaseSync, cfg: ResolvedConfig, scopedPaths: Set<string>): boolean {\n if (!anyPresetEmbeds(cfg)) return false; // the embeddings table doesn't exist at all in this case\n const rows = db.prepare('SELECT DISTINCT \"path\" FROM embeddings').all() as Array<{ path: string }>;\n return rows.some((r) => scopedPaths.has(r.path));\n}\n\n// Scope only (no --where): preset_files for a named preset, JS glob matching for an ad hoc\n// include/exclude override. Shared by scopedPaths() and search(), which also needs the\n// pre-where set to size the candidate-pool filter.\nexport function rawScope(db: DatabaseSync, cfg: ResolvedConfig, overrides: SearchOverrides, allPaths?: string[]): Set<string> {\n const effective = resolveSearch(cfg, overrides);\n const { include, exclude } = effective;\n const adHocScope = overrides.include !== undefined || overrides.exclude !== undefined || overrides.noExclude === true;\n if (!adHocScope) return new Set((db.prepare('SELECT \"path\" FROM preset_files WHERE preset = ?').all(effective.presetName) as Array<{ path: string }>).map((r) => r.path));\n const paths = allPaths ?? (db.prepare('SELECT \"path\" FROM frontmatter').all() as Array<{ path: string }>).map((r) => r.path);\n return new Set(paths.filter((p) => inScope(p, include, exclude)));\n}\n\nexport function narrowByWhere(db: DatabaseSync, paths: Set<string>, where: string | undefined): Set<string> {\n if (!where) return paths;\n let whereRows: Array<{ path: string }>;\n try {\n whereRows = db.prepare(`SELECT \"path\" FROM frontmatter f WHERE (${where})`).all() as Array<{ path: string }>;\n } catch (err) {\n throw columnHint(db, err as Error);\n }\n const wherePaths = new Set(whereRows.map((r) => r.path));\n return new Set([...paths].filter((p) => wherePaths.has(p)));\n}\n\n// The scope resolver for non-search commands (path, peek, map): same coverage rule search()\n// applies, then narrowed by the resolved `where`.\nexport function scopedPaths(db: DatabaseSync, cfg: ResolvedConfig, overrides: SearchOverrides): Set<string> {\n const effective = resolveSearch(cfg, overrides);\n return narrowByWhere(db, rawScope(db, cfg, overrides), effective.where);\n}\n\n// Materializes a path set into a named temp table (same shape as traverse.ts's allowed_nodes)\n// so a query can join/filter against it cheaply instead of binding a parameter per path.\nexport function materializeScope(db: DatabaseSync, table: string, paths: Set<string>): void {\n db.exec(`DROP TABLE IF EXISTS ${table}`);\n db.exec(`CREATE TEMP TABLE ${table} (\"path\" TEXT PRIMARY KEY)`);\n db.prepare(`INSERT INTO ${table} SELECT DISTINCT value FROM json_each(?1)`).run(JSON.stringify([...paths]));\n}\n\nexport function setupMapScope(db: DatabaseSync, paths: Set<string>): void {\n materializeScope(db, '_map_scope', paths);\n}\n"],"names":["matchesGlob","columnHint","anyPresetEmbeds","resolveSearch","INTERNAL_COLUMNS","Set","inScope","path","include","exclude","some","g","scopeHasEmbeddings","db","cfg","scopedPaths","rows","prepare","all","r","has","rawScope","overrides","allPaths","effective","adHocScope","undefined","noExclude","presetName","map","paths","filter","p","narrowByWhere","where","whereRows","err","wherePaths","materializeScope","table","exec","run","JSON","stringify","setupMapScope"],"mappings":"AAAA,SAASA,WAAW,QAAQ,YAAY;AAExC,SAASC,UAAU,QAAQ,oBAAoB;AAE/C,SAASC,eAAe,EAAEC,aAAa,QAAQ,qBAAqB;AAEpE,OAAO,MAAMC,mBAAmB,IAAIC,IAAI;IAAC;IAAQ;IAAU;IAAU;IAAS;IAAS;CAAe,EAAE;AAExG,yFAAyF;AACzF,4FAA4F;AAC5F,+EAA+E;AAC/E,OAAO,SAASC,QAAQC,IAAY,EAAEC,OAAiB,EAAEC,OAAkB;IACzE,IAAI,CAACD,QAAQE,IAAI,CAAC,CAACC,IAAMX,YAAYO,MAAMI,KAAK,OAAO;IACvD,IAAIF,oBAAAA,8BAAAA,QAASC,IAAI,CAAC,CAACC,IAAMX,YAAYO,MAAMI,KAAK,OAAO;IACvD,OAAO;AACT;AAEA,OAAO,SAASC,mBAAmBC,EAAgB,EAAEC,GAAmB,EAAEC,WAAwB;IAChG,IAAI,CAACb,gBAAgBY,MAAM,OAAO,OAAO,yDAAyD;IAClG,MAAME,OAAOH,GAAGI,OAAO,CAAC,0CAA0CC,GAAG;IACrE,OAAOF,KAAKN,IAAI,CAAC,CAACS,IAAMJ,YAAYK,GAAG,CAACD,EAAEZ,IAAI;AAChD;AAEA,2FAA2F;AAC3F,uFAAuF;AACvF,mDAAmD;AACnD,OAAO,SAASc,SAASR,EAAgB,EAAEC,GAAmB,EAAEQ,SAA0B,EAAEC,QAAmB;IAC7G,MAAMC,YAAYrB,cAAcW,KAAKQ;IACrC,MAAM,EAAEd,OAAO,EAAEC,OAAO,EAAE,GAAGe;IAC7B,MAAMC,aAAaH,UAAUd,OAAO,KAAKkB,aAAaJ,UAAUb,OAAO,KAAKiB,aAAaJ,UAAUK,SAAS,KAAK;IACjH,IAAI,CAACF,YAAY,OAAO,IAAIpB,IAAI,AAACQ,GAAGI,OAAO,CAAC,oDAAoDC,GAAG,CAACM,UAAUI,UAAU,EAA8BC,GAAG,CAAC,CAACV,IAAMA,EAAEZ,IAAI;IACvK,MAAMuB,QAAQP,qBAAAA,sBAAAA,WAAY,AAACV,GAAGI,OAAO,CAAC,kCAAkCC,GAAG,GAA+BW,GAAG,CAAC,CAACV,IAAMA,EAAEZ,IAAI;IAC3H,OAAO,IAAIF,IAAIyB,MAAMC,MAAM,CAAC,CAACC,IAAM1B,QAAQ0B,GAAGxB,SAASC;AACzD;AAEA,OAAO,SAASwB,cAAcpB,EAAgB,EAAEiB,KAAkB,EAAEI,KAAyB;IAC3F,IAAI,CAACA,OAAO,OAAOJ;IACnB,IAAIK;IACJ,IAAI;QACFA,YAAYtB,GAAGI,OAAO,CAAC,CAAC,wCAAwC,EAAEiB,MAAM,CAAC,CAAC,EAAEhB,GAAG;IACjF,EAAE,OAAOkB,KAAK;QACZ,MAAMnC,WAAWY,IAAIuB;IACvB;IACA,MAAMC,aAAa,IAAIhC,IAAI8B,UAAUN,GAAG,CAAC,CAACV,IAAMA,EAAEZ,IAAI;IACtD,OAAO,IAAIF,IAAI;WAAIyB;KAAM,CAACC,MAAM,CAAC,CAACC,IAAMK,WAAWjB,GAAG,CAACY;AACzD;AAEA,4FAA4F;AAC5F,kDAAkD;AAClD,OAAO,SAASjB,YAAYF,EAAgB,EAAEC,GAAmB,EAAEQ,SAA0B;IAC3F,MAAME,YAAYrB,cAAcW,KAAKQ;IACrC,OAAOW,cAAcpB,IAAIQ,SAASR,IAAIC,KAAKQ,YAAYE,UAAUU,KAAK;AACxE;AAEA,8FAA8F;AAC9F,yFAAyF;AACzF,OAAO,SAASI,iBAAiBzB,EAAgB,EAAE0B,KAAa,EAAET,KAAkB;IAClFjB,GAAG2B,IAAI,CAAC,CAAC,qBAAqB,EAAED,OAAO;IACvC1B,GAAG2B,IAAI,CAAC,CAAC,kBAAkB,EAAED,MAAM,0BAA0B,CAAC;IAC9D1B,GAAGI,OAAO,CAAC,CAAC,YAAY,EAAEsB,MAAM,yCAAyC,CAAC,EAAEE,GAAG,CAACC,KAAKC,SAAS,CAAC;WAAIb;KAAM;AAC3G;AAEA,OAAO,SAASc,cAAc/B,EAAgB,EAAEiB,KAAkB;IAChEQ,iBAAiBzB,IAAI,cAAciB;AACrC"}
@@ -1,7 +1,7 @@
1
1
  import { DatabaseSync } from 'node:sqlite';
2
2
  import type { ResolvedConfig } from '../config/index.js';
3
3
  export declare const DB_FILENAME = "cache.db";
4
- export declare const SCHEMA_VERSION = "13";
4
+ export declare const SCHEMA_VERSION = "14";
5
5
  export interface OpenResult {
6
6
  db: DatabaseSync;
7
7
  cfg: ResolvedConfig;
@@ -12,7 +12,7 @@ import { getMeta, setMeta } from './shared.js';
12
12
  export const DB_FILENAME = 'cache.db';
13
13
  // Cache shape version, independent of the config's own `version`. Bumping it rebuilds
14
14
  // existing trees on first query.
15
- export const SCHEMA_VERSION = '13';
15
+ export const SCHEMA_VERSION = '14';
16
16
  // Stemming is English-only, but the segmentation underneath it is what decides coverage:
17
17
  // unicode61 splits on spaces, so a language written without them (Chinese, Japanese, Thai)
18
18
  // indexes a whole run as one token and word search finds nothing. `content.tokenize` is how
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/db/open.ts"],"sourcesContent":["// The Node floor (>=22.20) is explained here and nowhere else: 22.20 is the first release with\n// both FTS5 and row-returning INSERT ... RETURNING. Raise it only for a load-bearing capability.\nimport { mkdirSync, rmSync } from 'node:fs';\nimport { join } from 'node:path';\nimport { DatabaseSync } from 'node:sqlite';\nimport type { Config, ResolvedConfig } from '../config/index.ts';\nimport { contentTokenize, featureSignature, STATE_DIR } from '../config/index.ts';\nimport { SenseError } from '../errors.ts';\nimport { activeFeatures } from '../features/index.ts';\nimport { registerFunctions } from '../sql-functions.ts';\nimport { changedSignatureKeys, rebuildContentTable, reconcile, signatureDiff } from './reconcile.ts';\nimport { getMeta, setMeta } from './shared.ts';\n\nexport const DB_FILENAME = 'cache.db';\n// Cache shape version, independent of the config's own `version`. Bumping it rebuilds\n// existing trees on first query.\nexport const SCHEMA_VERSION = '13';\n\nexport interface OpenResult {\n db: DatabaseSync;\n cfg: ResolvedConfig;\n dbPath: string;\n parsed: number;\n warnings: string[];\n}\n\n// Stemming is English-only, but the segmentation underneath it is what decides coverage:\n// unicode61 splits on spaces, so a language written without them (Chinese, Japanese, Thai)\n// indexes a whole run as one token and word search finds nothing. `content.tokenize` is how\n// such a tree picks trigram instead.\nconst DEFAULT_TOKENIZE = 'porter unicode61';\n\n// FTS5 takes its tokenizer as a string literal inside DDL, where nothing can bind, so a\n// configured value has to be concatenated. Probing a throwaway table is what makes that safe\n// and is also the whole validation: anything the linked SQLite accepts passes, anything else\n// fails here with SQLite's own message rather than against the real table. It means no table\n// of which version added which tokenizer has to be maintained.\nfunction resolveTokenize(db: DatabaseSync, cfg: Config): string {\n const configured = contentTokenize(cfg);\n if (configured === undefined) return DEFAULT_TOKENIZE;\n const literal = configured.replace(/'/g, \"''\");\n try {\n db.exec('DROP TABLE IF EXISTS temp.sense_tokenize_probe');\n db.exec(`CREATE VIRTUAL TABLE temp.sense_tokenize_probe USING fts5(x, tokenize = '${literal}')`);\n db.exec('DROP TABLE IF EXISTS temp.sense_tokenize_probe');\n } catch (err) {\n throw new SenseError('CONFIG_INVALID', `content.tokenize \"${configured}\" is not a tokenizer this SQLite accepts (${(err as Error).message}); the built-in choices are unicode61, ascii, porter, and trigram, each with their own options`);\n }\n return literal;\n}\n\n// The tokenizer the content table was actually built with, from its own DDL -- the one\n// record that cannot desynchronize from the table. NULL when the table does not exist yet.\nfunction storedTokenize(db: DatabaseSync): string | null {\n const row = db.prepare(`SELECT sql FROM sqlite_master WHERE name = 'content'`).get() as { sql: string } | undefined;\n if (!row) return null;\n const m = row.sql.match(/tokenize = '((?:[^']|'')*)'/);\n return m ? m[1] : null;\n}\n\n// Content is a separate table (not a column on frontmatter) so `SELECT * FROM frontmatter`\n// can't dump file text into context. Features add their own tables after the core ones.\nfunction ensureSchema(db: DatabaseSync, cfg: Config, tokenize: string): void {\n db.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" REAL, \"_ctime\" REAL, \"_size\" INTEGER, \"_parse_error\" TEXT)`);\n // IF NOT EXISTS is safe against a tokenizer change: open() compares the table's own DDL\n // against the resolved tokenizer before this runs, so a stale table is already gone by now.\n // The three `_seg` sidecars are appended after path, never inserted: bm25(content, ...) and\n // snippet(content, 2, ...) are documented against the first three columns and keep working\n // (FTS5 defaults the weights it was not given). Each carries its field's exploded unspaced\n // runs for text that needs it, and an empty string for text that does not.\n db.exec(`CREATE VIRTUAL TABLE IF NOT EXISTS content USING fts5(title, summary, text, path UNINDEXED, title_seg, summary_seg, text_seg, tokenize = '${tokenize}')`);\n // Coverage, not ownership: a path can appear under several presets. path leads the PK so the\n // per-doc delete is an index hit -- keyed the other way, cold builds went quadratic.\n db.exec(`CREATE TABLE IF NOT EXISTS preset_files (\"path\" TEXT, preset TEXT, PRIMARY KEY (\"path\", preset))`);\n db.exec('CREATE INDEX IF NOT EXISTS preset_files_preset ON preset_files(preset)');\n for (const feature of activeFeatures(cfg)) feature.schema(db);\n if (getMeta(db, 'schema_version') === null) setMeta(db, 'schema_version', SCHEMA_VERSION);\n if (getMeta(db, 'features') === null) setMeta(db, 'features', featureSignature(cfg));\n}\n\nexport function docCount(db: DatabaseSync): number {\n const row = db.prepare('SELECT COUNT(*) AS n FROM frontmatter').get() as { n: number };\n return row.n;\n}\n\nexport function open(cfg: ResolvedConfig): OpenResult {\n const stateDir = join(cfg.baseDir, STATE_DIR);\n mkdirSync(stateDir, { recursive: true });\n const dbPath = join(stateDir, DB_FILENAME);\n\n const db = new DatabaseSync(dbPath);\n db.exec('PRAGMA journal_mode = WAL');\n // Covers a concurrent watcher's bulk reconcile (~5s for 500 files at 26k notes). A query\n // that outwaits it still fails loudly.\n db.exec('PRAGMA busy_timeout = 30000');\n registerFunctions(db, contentTokenize(cfg) === undefined);\n\n db.exec('CREATE TABLE IF NOT EXISTS meta (key TEXT PRIMARY KEY, value TEXT)');\n\n // Before the rebuild branch below, never after: that branch deletes the cache, so a typo'd\n // tokenizer validated later would cost a full re-index (and re-embed) to reach its own error.\n const tokenize = resolveTokenize(db, cfg);\n\n // Schema-version or feature-set mismatch: reconcile only reparses changed files, so an\n // old cache can't be patched incrementally -- rebuild instead (cheap: nothing expensive lives here).\n const version = getMeta(db, 'schema_version');\n const features = getMeta(db, 'features');\n const wantFeatures = featureSignature(cfg);\n let tokenizeOnlyRebuild = false;\n if ((version !== null && version !== SCHEMA_VERSION) || (features !== null && features !== wantFeatures)) {\n // Indexing derives from presets, so a config edit rebuilding the cache must say so and\n // name what changed -- silent rebuilds make derived indexing look like a hang or a bug.\n if (version !== null && version !== SCHEMA_VERSION) {\n console.error('sense: cache format changed (new sensemaking version); rebuilding the index');\n db.close();\n clearCache(cfg);\n return open(cfg);\n }\n const changedKeys = changedSignatureKeys(features ?? '', wantFeatures);\n // Only the tokenizer moved: frontmatter, links, sections, and embeddings are file-derived\n // and tokenizer-independent, so they don't need re-deriving. Everything else -- a preset\n // edit, an embed model change -- still takes the full clear/reopen below.\n if (changedKeys.size === 1 && changedKeys.has('tokenize')) {\n console.error('sense: config change (content tokenizer) rebuilds the text index; vectors, links, and sections are kept');\n db.exec('DROP TABLE content');\n tokenizeOnlyRebuild = true;\n } else {\n const changed = signatureDiff(features ?? '', wantFeatures);\n console.error(`sense: config change (${changed}) rebuilds the index`);\n db.close();\n clearCache(cfg);\n return open(cfg);\n }\n }\n\n // Meta can lie after a crash between table creation and the signature write; the table's\n // own DDL cannot. A mismatch here rebuilds no matter what meta says. A tokenize-only rebuild\n // just dropped content above, so storedTokenize sees no table and this guard is skipped\n // naturally rather than needing its own case.\n const stored = storedTokenize(db);\n if (stored !== null && stored !== tokenize) {\n console.error('sense: cache was built with a different content tokenizer; rebuilding the index');\n db.close();\n clearCache(cfg);\n return open(cfg);\n }\n\n ensureSchema(db, cfg, tokenize);\n\n let rebuildWarnings: string[] = [];\n if (tokenizeOnlyRebuild) {\n rebuildWarnings = rebuildContentTable(db, cfg, cfg.baseDir);\n setMeta(db, 'features', wantFeatures);\n }\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(getMeta(db, 'reconcile_max_ms') ?? '0');\n db.exec(`PRAGMA busy_timeout = ${Math.min(Math.max(30000, 3 * recordedMaxMs), 600_000)}`);\n\n const { parsed, warnings } = reconcile(db, cfg, cfg.baseDir);\n\n return { db, cfg, dbPath, parsed, warnings: [...rebuildWarnings, ...warnings] };\n}\n\n// Deletes the cache directory, and only that. The rebuild is not this function's job: the next\n// open() reconciles, which is what running any command already does, so a verb that bundled the\n// two described the half that was not its own. Manual reset for a doubted cache; the schema and\n// config-signature mismatches below reset themselves.\nexport function clearCache(cfg: ResolvedConfig): void {\n rmSync(join(cfg.baseDir, STATE_DIR), { recursive: true, force: true });\n}\n"],"names":["mkdirSync","rmSync","join","DatabaseSync","contentTokenize","featureSignature","STATE_DIR","SenseError","activeFeatures","registerFunctions","changedSignatureKeys","rebuildContentTable","reconcile","signatureDiff","getMeta","setMeta","DB_FILENAME","SCHEMA_VERSION","DEFAULT_TOKENIZE","resolveTokenize","db","cfg","configured","undefined","literal","replace","exec","err","message","storedTokenize","row","prepare","get","m","sql","match","ensureSchema","tokenize","feature","schema","docCount","n","open","stateDir","baseDir","recursive","dbPath","version","features","wantFeatures","tokenizeOnlyRebuild","console","error","close","clearCache","changedKeys","size","has","changed","stored","rebuildWarnings","recordedMaxMs","Number","Math","min","max","parsed","warnings","force"],"mappings":"AAAA,+FAA+F;AAC/F,iGAAiG;AACjG,SAASA,SAAS,EAAEC,MAAM,QAAQ,UAAU;AAC5C,SAASC,IAAI,QAAQ,YAAY;AACjC,SAASC,YAAY,QAAQ,cAAc;AAE3C,SAASC,eAAe,EAAEC,gBAAgB,EAAEC,SAAS,QAAQ,qBAAqB;AAClF,SAASC,UAAU,QAAQ,eAAe;AAC1C,SAASC,cAAc,QAAQ,uBAAuB;AACtD,SAASC,iBAAiB,QAAQ,sBAAsB;AACxD,SAASC,oBAAoB,EAAEC,mBAAmB,EAAEC,SAAS,EAAEC,aAAa,QAAQ,iBAAiB;AACrG,SAASC,OAAO,EAAEC,OAAO,QAAQ,cAAc;AAE/C,OAAO,MAAMC,cAAc,WAAW;AACtC,sFAAsF;AACtF,iCAAiC;AACjC,OAAO,MAAMC,iBAAiB,KAAK;AAUnC,yFAAyF;AACzF,2FAA2F;AAC3F,4FAA4F;AAC5F,qCAAqC;AACrC,MAAMC,mBAAmB;AAEzB,wFAAwF;AACxF,6FAA6F;AAC7F,6FAA6F;AAC7F,6FAA6F;AAC7F,+DAA+D;AAC/D,SAASC,gBAAgBC,EAAgB,EAAEC,GAAW;IACpD,MAAMC,aAAalB,gBAAgBiB;IACnC,IAAIC,eAAeC,WAAW,OAAOL;IACrC,MAAMM,UAAUF,WAAWG,OAAO,CAAC,MAAM;IACzC,IAAI;QACFL,GAAGM,IAAI,CAAC;QACRN,GAAGM,IAAI,CAAC,CAAC,yEAAyE,EAAEF,QAAQ,EAAE,CAAC;QAC/FJ,GAAGM,IAAI,CAAC;IACV,EAAE,OAAOC,KAAK;QACZ,MAAM,IAAIpB,WAAW,kBAAkB,CAAC,kBAAkB,EAAEe,WAAW,0CAA0C,EAAE,AAACK,IAAcC,OAAO,CAAC,8FAA8F,CAAC;IAC3O;IACA,OAAOJ;AACT;AAEA,uFAAuF;AACvF,2FAA2F;AAC3F,SAASK,eAAeT,EAAgB;IACtC,MAAMU,MAAMV,GAAGW,OAAO,CAAC,CAAC,oDAAoD,CAAC,EAAEC,GAAG;IAClF,IAAI,CAACF,KAAK,OAAO;IACjB,MAAMG,IAAIH,IAAII,GAAG,CAACC,KAAK,CAAC;IACxB,OAAOF,IAAIA,CAAC,CAAC,EAAE,GAAG;AACpB;AAEA,2FAA2F;AAC3F,wFAAwF;AACxF,SAASG,aAAahB,EAAgB,EAAEC,GAAW,EAAEgB,QAAgB;IACnEjB,GAAGM,IAAI,CAAC,CAAC,oIAAoI,CAAC;IAC9I,wFAAwF;IACxF,4FAA4F;IAC5F,4FAA4F;IAC5F,2FAA2F;IAC3F,2FAA2F;IAC3F,2EAA2E;IAC3EN,GAAGM,IAAI,CAAC,CAAC,0IAA0I,EAAEW,SAAS,EAAE,CAAC;IACjK,6FAA6F;IAC7F,qFAAqF;IACrFjB,GAAGM,IAAI,CAAC,CAAC,gGAAgG,CAAC;IAC1GN,GAAGM,IAAI,CAAC;IACR,KAAK,MAAMY,WAAW9B,eAAea,KAAMiB,QAAQC,MAAM,CAACnB;IAC1D,IAAIN,QAAQM,IAAI,sBAAsB,MAAML,QAAQK,IAAI,kBAAkBH;IAC1E,IAAIH,QAAQM,IAAI,gBAAgB,MAAML,QAAQK,IAAI,YAAYf,iBAAiBgB;AACjF;AAEA,OAAO,SAASmB,SAASpB,EAAgB;IACvC,MAAMU,MAAMV,GAAGW,OAAO,CAAC,yCAAyCC,GAAG;IACnE,OAAOF,IAAIW,CAAC;AACd;AAEA,OAAO,SAASC,KAAKrB,GAAmB;QAwETP;IAvE7B,MAAM6B,WAAWzC,KAAKmB,IAAIuB,OAAO,EAAEtC;IACnCN,UAAU2C,UAAU;QAAEE,WAAW;IAAK;IACtC,MAAMC,SAAS5C,KAAKyC,UAAU3B;IAE9B,MAAMI,KAAK,IAAIjB,aAAa2C;IAC5B1B,GAAGM,IAAI,CAAC;IACR,yFAAyF;IACzF,uCAAuC;IACvCN,GAAGM,IAAI,CAAC;IACRjB,kBAAkBW,IAAIhB,gBAAgBiB,SAASE;IAE/CH,GAAGM,IAAI,CAAC;IAER,2FAA2F;IAC3F,8FAA8F;IAC9F,MAAMW,WAAWlB,gBAAgBC,IAAIC;IAErC,uFAAuF;IACvF,qGAAqG;IACrG,MAAM0B,UAAUjC,QAAQM,IAAI;IAC5B,MAAM4B,WAAWlC,QAAQM,IAAI;IAC7B,MAAM6B,eAAe5C,iBAAiBgB;IACtC,IAAI6B,sBAAsB;IAC1B,IAAI,AAACH,YAAY,QAAQA,YAAY9B,kBAAoB+B,aAAa,QAAQA,aAAaC,cAAe;QACxG,uFAAuF;QACvF,wFAAwF;QACxF,IAAIF,YAAY,QAAQA,YAAY9B,gBAAgB;YAClDkC,QAAQC,KAAK,CAAC;YACdhC,GAAGiC,KAAK;YACRC,WAAWjC;YACX,OAAOqB,KAAKrB;QACd;QACA,MAAMkC,cAAc7C,qBAAqBsC,qBAAAA,sBAAAA,WAAY,IAAIC;QACzD,0FAA0F;QAC1F,yFAAyF;QACzF,0EAA0E;QAC1E,IAAIM,YAAYC,IAAI,KAAK,KAAKD,YAAYE,GAAG,CAAC,aAAa;YACzDN,QAAQC,KAAK,CAAC;YACdhC,GAAGM,IAAI,CAAC;YACRwB,sBAAsB;QACxB,OAAO;YACL,MAAMQ,UAAU7C,cAAcmC,qBAAAA,sBAAAA,WAAY,IAAIC;YAC9CE,QAAQC,KAAK,CAAC,CAAC,sBAAsB,EAAEM,QAAQ,oBAAoB,CAAC;YACpEtC,GAAGiC,KAAK;YACRC,WAAWjC;YACX,OAAOqB,KAAKrB;QACd;IACF;IAEA,yFAAyF;IACzF,6FAA6F;IAC7F,wFAAwF;IACxF,8CAA8C;IAC9C,MAAMsC,SAAS9B,eAAeT;IAC9B,IAAIuC,WAAW,QAAQA,WAAWtB,UAAU;QAC1Cc,QAAQC,KAAK,CAAC;QACdhC,GAAGiC,KAAK;QACRC,WAAWjC;QACX,OAAOqB,KAAKrB;IACd;IAEAe,aAAahB,IAAIC,KAAKgB;IAEtB,IAAIuB,kBAA4B,EAAE;IAClC,IAAIV,qBAAqB;QACvBU,kBAAkBjD,oBAAoBS,IAAIC,KAAKA,IAAIuB,OAAO;QAC1D7B,QAAQK,IAAI,YAAY6B;IAC1B;IAEA,wFAAwF;IACxF,2FAA2F;IAC3F,MAAMY,gBAAgBC,QAAOhD,WAAAA,QAAQM,IAAI,iCAAZN,sBAAAA,WAAmC;IAChEM,GAAGM,IAAI,CAAC,CAAC,sBAAsB,EAAEqC,KAAKC,GAAG,CAACD,KAAKE,GAAG,CAAC,OAAO,IAAIJ,gBAAgB,SAAU;IAExF,MAAM,EAAEK,MAAM,EAAEC,QAAQ,EAAE,GAAGvD,UAAUQ,IAAIC,KAAKA,IAAIuB,OAAO;IAE3D,OAAO;QAAExB;QAAIC;QAAKyB;QAAQoB;QAAQC,UAAU;eAAIP;eAAoBO;SAAS;IAAC;AAChF;AAEA,+FAA+F;AAC/F,gGAAgG;AAChG,gGAAgG;AAChG,sDAAsD;AACtD,OAAO,SAASb,WAAWjC,GAAmB;IAC5CpB,OAAOC,KAAKmB,IAAIuB,OAAO,EAAEtC,YAAY;QAAEuC,WAAW;QAAMuB,OAAO;IAAK;AACtE"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/db/open.ts"],"sourcesContent":["// The Node floor (>=22.20) is explained here and nowhere else: 22.20 is the first release with\n// both FTS5 and row-returning INSERT ... RETURNING. Raise it only for a load-bearing capability.\nimport { mkdirSync, rmSync } from 'node:fs';\nimport { join } from 'node:path';\nimport { DatabaseSync } from 'node:sqlite';\nimport type { Config, ResolvedConfig } from '../config/index.ts';\nimport { contentTokenize, featureSignature, STATE_DIR } from '../config/index.ts';\nimport { SenseError } from '../errors.ts';\nimport { activeFeatures } from '../features/index.ts';\nimport { registerFunctions } from '../sql-functions.ts';\nimport { changedSignatureKeys, rebuildContentTable, reconcile, signatureDiff } from './reconcile.ts';\nimport { getMeta, setMeta } from './shared.ts';\n\nexport const DB_FILENAME = 'cache.db';\n// Cache shape version, independent of the config's own `version`. Bumping it rebuilds\n// existing trees on first query.\nexport const SCHEMA_VERSION = '14';\n\nexport interface OpenResult {\n db: DatabaseSync;\n cfg: ResolvedConfig;\n dbPath: string;\n parsed: number;\n warnings: string[];\n}\n\n// Stemming is English-only, but the segmentation underneath it is what decides coverage:\n// unicode61 splits on spaces, so a language written without them (Chinese, Japanese, Thai)\n// indexes a whole run as one token and word search finds nothing. `content.tokenize` is how\n// such a tree picks trigram instead.\nconst DEFAULT_TOKENIZE = 'porter unicode61';\n\n// FTS5 takes its tokenizer as a string literal inside DDL, where nothing can bind, so a\n// configured value has to be concatenated. Probing a throwaway table is what makes that safe\n// and is also the whole validation: anything the linked SQLite accepts passes, anything else\n// fails here with SQLite's own message rather than against the real table. It means no table\n// of which version added which tokenizer has to be maintained.\nfunction resolveTokenize(db: DatabaseSync, cfg: Config): string {\n const configured = contentTokenize(cfg);\n if (configured === undefined) return DEFAULT_TOKENIZE;\n const literal = configured.replace(/'/g, \"''\");\n try {\n db.exec('DROP TABLE IF EXISTS temp.sense_tokenize_probe');\n db.exec(`CREATE VIRTUAL TABLE temp.sense_tokenize_probe USING fts5(x, tokenize = '${literal}')`);\n db.exec('DROP TABLE IF EXISTS temp.sense_tokenize_probe');\n } catch (err) {\n throw new SenseError('CONFIG_INVALID', `content.tokenize \"${configured}\" is not a tokenizer this SQLite accepts (${(err as Error).message}); the built-in choices are unicode61, ascii, porter, and trigram, each with their own options`);\n }\n return literal;\n}\n\n// The tokenizer the content table was actually built with, from its own DDL -- the one\n// record that cannot desynchronize from the table. NULL when the table does not exist yet.\nfunction storedTokenize(db: DatabaseSync): string | null {\n const row = db.prepare(`SELECT sql FROM sqlite_master WHERE name = 'content'`).get() as { sql: string } | undefined;\n if (!row) return null;\n const m = row.sql.match(/tokenize = '((?:[^']|'')*)'/);\n return m ? m[1] : null;\n}\n\n// Content is a separate table (not a column on frontmatter) so `SELECT * FROM frontmatter`\n// can't dump file text into context. Features add their own tables after the core ones.\nfunction ensureSchema(db: DatabaseSync, cfg: Config, tokenize: string): void {\n db.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" REAL, \"_ctime\" REAL, \"_size\" INTEGER, \"_parse_error\" TEXT)`);\n // IF NOT EXISTS is safe against a tokenizer change: open() compares the table's own DDL\n // against the resolved tokenizer before this runs, so a stale table is already gone by now.\n // The three `_seg` sidecars are appended after path, never inserted: bm25(content, ...) and\n // snippet(content, 2, ...) are documented against the first three columns and keep working\n // (FTS5 defaults the weights it was not given). Each carries its field's exploded unspaced\n // runs for text that needs it, and an empty string for text that does not.\n db.exec(`CREATE VIRTUAL TABLE IF NOT EXISTS content USING fts5(title, summary, text, path UNINDEXED, title_seg, summary_seg, text_seg, tokenize = '${tokenize}')`);\n // Coverage, not ownership: a path can appear under several presets. path leads the PK so the\n // per-doc delete is an index hit -- keyed the other way, cold builds went quadratic.\n db.exec(`CREATE TABLE IF NOT EXISTS preset_files (\"path\" TEXT, preset TEXT, PRIMARY KEY (\"path\", preset))`);\n db.exec('CREATE INDEX IF NOT EXISTS preset_files_preset ON preset_files(preset)');\n for (const feature of activeFeatures(cfg)) feature.schema(db);\n if (getMeta(db, 'schema_version') === null) setMeta(db, 'schema_version', SCHEMA_VERSION);\n if (getMeta(db, 'features') === null) setMeta(db, 'features', featureSignature(cfg));\n}\n\nexport function docCount(db: DatabaseSync): number {\n const row = db.prepare('SELECT COUNT(*) AS n FROM frontmatter').get() as { n: number };\n return row.n;\n}\n\nexport function open(cfg: ResolvedConfig): OpenResult {\n const stateDir = join(cfg.baseDir, STATE_DIR);\n mkdirSync(stateDir, { recursive: true });\n const dbPath = join(stateDir, DB_FILENAME);\n\n const db = new DatabaseSync(dbPath);\n db.exec('PRAGMA journal_mode = WAL');\n // Covers a concurrent watcher's bulk reconcile (~5s for 500 files at 26k notes). A query\n // that outwaits it still fails loudly.\n db.exec('PRAGMA busy_timeout = 30000');\n registerFunctions(db, contentTokenize(cfg) === undefined);\n\n db.exec('CREATE TABLE IF NOT EXISTS meta (key TEXT PRIMARY KEY, value TEXT)');\n\n // Before the rebuild branch below, never after: that branch deletes the cache, so a typo'd\n // tokenizer validated later would cost a full re-index (and re-embed) to reach its own error.\n const tokenize = resolveTokenize(db, cfg);\n\n // Schema-version or feature-set mismatch: reconcile only reparses changed files, so an\n // old cache can't be patched incrementally -- rebuild instead (cheap: nothing expensive lives here).\n const version = getMeta(db, 'schema_version');\n const features = getMeta(db, 'features');\n const wantFeatures = featureSignature(cfg);\n let tokenizeOnlyRebuild = false;\n if ((version !== null && version !== SCHEMA_VERSION) || (features !== null && features !== wantFeatures)) {\n // Indexing derives from presets, so a config edit rebuilding the cache must say so and\n // name what changed -- silent rebuilds make derived indexing look like a hang or a bug.\n if (version !== null && version !== SCHEMA_VERSION) {\n console.error('sense: cache format changed (new sensemaking version); rebuilding the index');\n db.close();\n clearCache(cfg);\n return open(cfg);\n }\n const changedKeys = changedSignatureKeys(features ?? '', wantFeatures);\n // Only the tokenizer moved: frontmatter, links, sections, and embeddings are file-derived\n // and tokenizer-independent, so they don't need re-deriving. Everything else -- a preset\n // edit, an embed model change -- still takes the full clear/reopen below.\n if (changedKeys.size === 1 && changedKeys.has('tokenize')) {\n console.error('sense: config change (content tokenizer) rebuilds the text index; vectors, links, and sections are kept');\n db.exec('DROP TABLE content');\n tokenizeOnlyRebuild = true;\n } else {\n const changed = signatureDiff(features ?? '', wantFeatures);\n console.error(`sense: config change (${changed}) rebuilds the index`);\n db.close();\n clearCache(cfg);\n return open(cfg);\n }\n }\n\n // Meta can lie after a crash between table creation and the signature write; the table's\n // own DDL cannot. A mismatch here rebuilds no matter what meta says. A tokenize-only rebuild\n // just dropped content above, so storedTokenize sees no table and this guard is skipped\n // naturally rather than needing its own case.\n const stored = storedTokenize(db);\n if (stored !== null && stored !== tokenize) {\n console.error('sense: cache was built with a different content tokenizer; rebuilding the index');\n db.close();\n clearCache(cfg);\n return open(cfg);\n }\n\n ensureSchema(db, cfg, tokenize);\n\n let rebuildWarnings: string[] = [];\n if (tokenizeOnlyRebuild) {\n rebuildWarnings = rebuildContentTable(db, cfg, cfg.baseDir);\n setMeta(db, 'features', wantFeatures);\n }\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(getMeta(db, 'reconcile_max_ms') ?? '0');\n db.exec(`PRAGMA busy_timeout = ${Math.min(Math.max(30000, 3 * recordedMaxMs), 600_000)}`);\n\n const { parsed, warnings } = reconcile(db, cfg, cfg.baseDir);\n\n return { db, cfg, dbPath, parsed, warnings: [...rebuildWarnings, ...warnings] };\n}\n\n// Deletes the cache directory, and only that. The rebuild is not this function's job: the next\n// open() reconciles, which is what running any command already does, so a verb that bundled the\n// two described the half that was not its own. Manual reset for a doubted cache; the schema and\n// config-signature mismatches below reset themselves.\nexport function clearCache(cfg: ResolvedConfig): void {\n rmSync(join(cfg.baseDir, STATE_DIR), { recursive: true, force: true });\n}\n"],"names":["mkdirSync","rmSync","join","DatabaseSync","contentTokenize","featureSignature","STATE_DIR","SenseError","activeFeatures","registerFunctions","changedSignatureKeys","rebuildContentTable","reconcile","signatureDiff","getMeta","setMeta","DB_FILENAME","SCHEMA_VERSION","DEFAULT_TOKENIZE","resolveTokenize","db","cfg","configured","undefined","literal","replace","exec","err","message","storedTokenize","row","prepare","get","m","sql","match","ensureSchema","tokenize","feature","schema","docCount","n","open","stateDir","baseDir","recursive","dbPath","version","features","wantFeatures","tokenizeOnlyRebuild","console","error","close","clearCache","changedKeys","size","has","changed","stored","rebuildWarnings","recordedMaxMs","Number","Math","min","max","parsed","warnings","force"],"mappings":"AAAA,+FAA+F;AAC/F,iGAAiG;AACjG,SAASA,SAAS,EAAEC,MAAM,QAAQ,UAAU;AAC5C,SAASC,IAAI,QAAQ,YAAY;AACjC,SAASC,YAAY,QAAQ,cAAc;AAE3C,SAASC,eAAe,EAAEC,gBAAgB,EAAEC,SAAS,QAAQ,qBAAqB;AAClF,SAASC,UAAU,QAAQ,eAAe;AAC1C,SAASC,cAAc,QAAQ,uBAAuB;AACtD,SAASC,iBAAiB,QAAQ,sBAAsB;AACxD,SAASC,oBAAoB,EAAEC,mBAAmB,EAAEC,SAAS,EAAEC,aAAa,QAAQ,iBAAiB;AACrG,SAASC,OAAO,EAAEC,OAAO,QAAQ,cAAc;AAE/C,OAAO,MAAMC,cAAc,WAAW;AACtC,sFAAsF;AACtF,iCAAiC;AACjC,OAAO,MAAMC,iBAAiB,KAAK;AAUnC,yFAAyF;AACzF,2FAA2F;AAC3F,4FAA4F;AAC5F,qCAAqC;AACrC,MAAMC,mBAAmB;AAEzB,wFAAwF;AACxF,6FAA6F;AAC7F,6FAA6F;AAC7F,6FAA6F;AAC7F,+DAA+D;AAC/D,SAASC,gBAAgBC,EAAgB,EAAEC,GAAW;IACpD,MAAMC,aAAalB,gBAAgBiB;IACnC,IAAIC,eAAeC,WAAW,OAAOL;IACrC,MAAMM,UAAUF,WAAWG,OAAO,CAAC,MAAM;IACzC,IAAI;QACFL,GAAGM,IAAI,CAAC;QACRN,GAAGM,IAAI,CAAC,CAAC,yEAAyE,EAAEF,QAAQ,EAAE,CAAC;QAC/FJ,GAAGM,IAAI,CAAC;IACV,EAAE,OAAOC,KAAK;QACZ,MAAM,IAAIpB,WAAW,kBAAkB,CAAC,kBAAkB,EAAEe,WAAW,0CAA0C,EAAE,AAACK,IAAcC,OAAO,CAAC,8FAA8F,CAAC;IAC3O;IACA,OAAOJ;AACT;AAEA,uFAAuF;AACvF,2FAA2F;AAC3F,SAASK,eAAeT,EAAgB;IACtC,MAAMU,MAAMV,GAAGW,OAAO,CAAC,CAAC,oDAAoD,CAAC,EAAEC,GAAG;IAClF,IAAI,CAACF,KAAK,OAAO;IACjB,MAAMG,IAAIH,IAAII,GAAG,CAACC,KAAK,CAAC;IACxB,OAAOF,IAAIA,CAAC,CAAC,EAAE,GAAG;AACpB;AAEA,2FAA2F;AAC3F,wFAAwF;AACxF,SAASG,aAAahB,EAAgB,EAAEC,GAAW,EAAEgB,QAAgB;IACnEjB,GAAGM,IAAI,CAAC,CAAC,oIAAoI,CAAC;IAC9I,wFAAwF;IACxF,4FAA4F;IAC5F,4FAA4F;IAC5F,2FAA2F;IAC3F,2FAA2F;IAC3F,2EAA2E;IAC3EN,GAAGM,IAAI,CAAC,CAAC,0IAA0I,EAAEW,SAAS,EAAE,CAAC;IACjK,6FAA6F;IAC7F,qFAAqF;IACrFjB,GAAGM,IAAI,CAAC,CAAC,gGAAgG,CAAC;IAC1GN,GAAGM,IAAI,CAAC;IACR,KAAK,MAAMY,WAAW9B,eAAea,KAAMiB,QAAQC,MAAM,CAACnB;IAC1D,IAAIN,QAAQM,IAAI,sBAAsB,MAAML,QAAQK,IAAI,kBAAkBH;IAC1E,IAAIH,QAAQM,IAAI,gBAAgB,MAAML,QAAQK,IAAI,YAAYf,iBAAiBgB;AACjF;AAEA,OAAO,SAASmB,SAASpB,EAAgB;IACvC,MAAMU,MAAMV,GAAGW,OAAO,CAAC,yCAAyCC,GAAG;IACnE,OAAOF,IAAIW,CAAC;AACd;AAEA,OAAO,SAASC,KAAKrB,GAAmB;QAwETP;IAvE7B,MAAM6B,WAAWzC,KAAKmB,IAAIuB,OAAO,EAAEtC;IACnCN,UAAU2C,UAAU;QAAEE,WAAW;IAAK;IACtC,MAAMC,SAAS5C,KAAKyC,UAAU3B;IAE9B,MAAMI,KAAK,IAAIjB,aAAa2C;IAC5B1B,GAAGM,IAAI,CAAC;IACR,yFAAyF;IACzF,uCAAuC;IACvCN,GAAGM,IAAI,CAAC;IACRjB,kBAAkBW,IAAIhB,gBAAgBiB,SAASE;IAE/CH,GAAGM,IAAI,CAAC;IAER,2FAA2F;IAC3F,8FAA8F;IAC9F,MAAMW,WAAWlB,gBAAgBC,IAAIC;IAErC,uFAAuF;IACvF,qGAAqG;IACrG,MAAM0B,UAAUjC,QAAQM,IAAI;IAC5B,MAAM4B,WAAWlC,QAAQM,IAAI;IAC7B,MAAM6B,eAAe5C,iBAAiBgB;IACtC,IAAI6B,sBAAsB;IAC1B,IAAI,AAACH,YAAY,QAAQA,YAAY9B,kBAAoB+B,aAAa,QAAQA,aAAaC,cAAe;QACxG,uFAAuF;QACvF,wFAAwF;QACxF,IAAIF,YAAY,QAAQA,YAAY9B,gBAAgB;YAClDkC,QAAQC,KAAK,CAAC;YACdhC,GAAGiC,KAAK;YACRC,WAAWjC;YACX,OAAOqB,KAAKrB;QACd;QACA,MAAMkC,cAAc7C,qBAAqBsC,qBAAAA,sBAAAA,WAAY,IAAIC;QACzD,0FAA0F;QAC1F,yFAAyF;QACzF,0EAA0E;QAC1E,IAAIM,YAAYC,IAAI,KAAK,KAAKD,YAAYE,GAAG,CAAC,aAAa;YACzDN,QAAQC,KAAK,CAAC;YACdhC,GAAGM,IAAI,CAAC;YACRwB,sBAAsB;QACxB,OAAO;YACL,MAAMQ,UAAU7C,cAAcmC,qBAAAA,sBAAAA,WAAY,IAAIC;YAC9CE,QAAQC,KAAK,CAAC,CAAC,sBAAsB,EAAEM,QAAQ,oBAAoB,CAAC;YACpEtC,GAAGiC,KAAK;YACRC,WAAWjC;YACX,OAAOqB,KAAKrB;QACd;IACF;IAEA,yFAAyF;IACzF,6FAA6F;IAC7F,wFAAwF;IACxF,8CAA8C;IAC9C,MAAMsC,SAAS9B,eAAeT;IAC9B,IAAIuC,WAAW,QAAQA,WAAWtB,UAAU;QAC1Cc,QAAQC,KAAK,CAAC;QACdhC,GAAGiC,KAAK;QACRC,WAAWjC;QACX,OAAOqB,KAAKrB;IACd;IAEAe,aAAahB,IAAIC,KAAKgB;IAEtB,IAAIuB,kBAA4B,EAAE;IAClC,IAAIV,qBAAqB;QACvBU,kBAAkBjD,oBAAoBS,IAAIC,KAAKA,IAAIuB,OAAO;QAC1D7B,QAAQK,IAAI,YAAY6B;IAC1B;IAEA,wFAAwF;IACxF,2FAA2F;IAC3F,MAAMY,gBAAgBC,QAAOhD,WAAAA,QAAQM,IAAI,iCAAZN,sBAAAA,WAAmC;IAChEM,GAAGM,IAAI,CAAC,CAAC,sBAAsB,EAAEqC,KAAKC,GAAG,CAACD,KAAKE,GAAG,CAAC,OAAO,IAAIJ,gBAAgB,SAAU;IAExF,MAAM,EAAEK,MAAM,EAAEC,QAAQ,EAAE,GAAGvD,UAAUQ,IAAIC,KAAKA,IAAIuB,OAAO;IAE3D,OAAO;QAAExB;QAAIC;QAAKyB;QAAQoB;QAAQC,UAAU;eAAIP;eAAoBO;SAAS;IAAC;AAChF;AAEA,+FAA+F;AAC/F,gGAAgG;AAChG,gGAAgG;AAChG,sDAAsD;AACtD,OAAO,SAASb,WAAWjC,GAAmB;IAC5CpB,OAAOC,KAAKmB,IAAIuB,OAAO,EAAEtC,YAAY;QAAEuC,WAAW;QAAMuB,OAAO;IAAK;AACtE"}
@@ -16,9 +16,11 @@ function extract(_raw, body) {
16
16
  embed
17
17
  });
18
18
  };
19
- for (const m of body.matchAll(/\[\[([^\]|#]+)(?:#[^\]|]*)?(?:\|[^\]]*)?\]\]/g)){
19
+ // To the first ]], as Obsidian parses it, so a heading or alias holding a lone ] still
20
+ // matches; anchor and alias split off in code, since a class-based regex stops at the ].
21
+ for (const m of body.matchAll(/\[\[(.*?)\]\]/g)){
20
22
  var _m_index;
21
- const target = m[1].trim();
23
+ const target = m[1].split('|')[0].split('#')[0].trim();
22
24
  if (target) add(target, body[((_m_index = m.index) !== null && _m_index !== void 0 ? _m_index : 0) - 1] === '!');
23
25
  }
24
26
  for (const m of body.matchAll(/(!)?\[[^\]]*\]\(([^)]+\.md)(?:#[^)]*)?\)/g)){
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/features/links.ts"],"sourcesContent":["import posix from 'node:path/posix';\nimport type { DatabaseSync } from 'node:sqlite';\nimport type { FileStat } from '../scan.ts';\nimport type { Feature, ReconcileDelta } from './types.ts';\n\n// links(src, target, target_base, dst, embed): target as written, target_base its baseKey\n// (indexed, for the incremental resolve below), dst the resolved path or NULL (a\n// queryable dead link), embed whether it's `![[x]]`/`![](x.md)` rather than `[[x]]`/`[](x.md)`\n// -- Obsidian's grain: a target both linked and embedded in the same note is two rows.\n\n// Wikilinks ([[target]], [[target#anchor|alias]], embeds) plus relative markdown links to .md files.\nfunction extract(_raw: string, body: string): Array<{ target: string; embed: boolean }> {\n const seen = new Set<string>();\n const results: Array<{ target: string; embed: boolean }> = [];\n const add = (target: string, embed: boolean) => {\n const key = `${target}\\0${embed ? '1' : '0'}`;\n if (seen.has(key)) return;\n seen.add(key);\n results.push({ target, embed });\n };\n for (const m of body.matchAll(/\\[\\[([^\\]|#]+)(?:#[^\\]|]*)?(?:\\|[^\\]]*)?\\]\\]/g)) {\n const target = m[1].trim();\n if (target) add(target, body[(m.index ?? 0) - 1] === '!');\n }\n for (const m of body.matchAll(/(!)?\\[[^\\]]*\\]\\(([^)]+\\.md)(?:#[^)]*)?\\)/g)) {\n const target = m[2].trim();\n if (target && !/^[a-z]+:\\/\\//i.test(target)) add(target, m[1] === '!');\n }\n return results;\n}\n\nfunction cleanTarget(target: string): string {\n return target.replace(/\\\\/g, '/').replace(/^\\.\\//, '');\n}\n\nfunction baseKey(path: string): string {\n return posix.basename(path).replace(/\\.md$/i, '').toLowerCase();\n}\n\n// Obsidian-style: exact relative path (with/without .md), path relative to the linking\n// note's directory, then basename match (lexicographically first on ties).\nfunction resolveTarget(src: string, target: string, pathSet: Set<string>, byBase: Map<string, string[]>): string | null {\n const clean = cleanTarget(target);\n const fromSrc = posix.normalize(posix.join(posix.dirname(src), clean));\n for (const candidate of [clean, `${clean}.md`, fromSrc, `${fromSrc}.md`]) {\n if (pathSet.has(candidate)) return candidate;\n }\n return byBase.get(baseKey(clean))?.[0] ?? null;\n}\n\nfunction buildByBase(files: FileStat[]): Map<string, string[]> {\n const byBase = new Map<string, string[]>();\n for (const path of files.map((f) => f.relPath).sort()) {\n const key = baseKey(path);\n const list = byBase.get(key);\n if (list) list.push(path);\n else byBase.set(key, [path]);\n }\n return byBase;\n}\n\ninterface LinkRow {\n src: string;\n target: string;\n dst: string | null;\n embed: number;\n}\n\n// Applies resolveTarget to `rows`, writing only rows whose dst actually changes. The UPDATE\n// keys on (src, target): a link/embed sibling pair shares one resolution, so either row's\n// mismatch rewrites both.\nfunction resolveRows(db: DatabaseSync, rows: LinkRow[], pathSet: Set<string>, byBase: Map<string, string[]>): boolean {\n const update = db.prepare('UPDATE links SET dst = ? WHERE src = ? AND target = ?');\n let changed = false;\n for (const row of rows) {\n const dst = resolveTarget(row.src, row.target, pathSet, byBase);\n if (dst !== row.dst) {\n update.run(dst, row.src, row.target);\n changed = true;\n }\n }\n return changed;\n}\n\n// A new or deleted file can change any note's resolution, so re-resolve the whole table.\n// Fallback for cold builds and large deltas -- see afterReconcile's threshold.\nfunction resolveAll(db: DatabaseSync, files: FileStat[]): boolean {\n const pathSet = new Set(files.map((f) => f.relPath));\n const byBase = buildByBase(files);\n const rows = db.prepare('SELECT src, target, dst, embed FROM links').all() as unknown as LinkRow[];\n return resolveRows(db, rows, pathSet, byBase);\n}\n\n// Re-resolve only what this reconcile could have touched: re-stored sources, links whose\n// target basename matches an added/vanished file, and links whose dst just vanished. All\n// three are indexed lookups, so this never reads the whole table.\nfunction resolveIncremental(db: DatabaseSync, delta: ReconcileDelta): boolean {\n const pathSet = new Set(delta.files.map((f) => f.relPath));\n const byBase = buildByBase(delta.files);\n\n const changedBasenames = new Set<string>();\n for (const p of delta.added) changedBasenames.add(baseKey(p));\n for (const p of delta.vanished) changedBasenames.add(baseKey(p));\n\n // \\0 can't appear in a path or target, so the key never collides; a space can.\n const candidates = new Map<string, LinkRow>();\n const collect = (rows: LinkRow[]) => {\n for (const row of rows) candidates.set(`${row.src}\\0${row.target}\\0${row.embed}`, row);\n };\n // Chunked so a large-but-under-threshold delta never exceeds SQLite's bound-variable\n // limit (SQLITE_MAX_VARIABLE_NUMBER, 32766).\n const collectIn = (column: string, keys: string[]) => {\n for (let i = 0; i < keys.length; i += 500) {\n const chunk = keys.slice(i, i + 500);\n const placeholders = chunk.map(() => '?').join(', ');\n collect(db.prepare(`SELECT src, target, dst, embed FROM links WHERE ${column} IN (${placeholders})`).all(...chunk) as unknown as LinkRow[]);\n }\n };\n\n collectIn('src', delta.reparsed);\n collectIn('target_base', [...changedBasenames]);\n collectIn('dst', delta.vanished);\n\n return resolveRows(db, [...candidates.values()], pathSet, byBase);\n}\n\n// Resolved edges, for rank and for search's graph expansion.\nexport function linkEdges(db: DatabaseSync): [string, string][] {\n // One edge per row, as 0.12's grain always had it -- two written targets resolving to one\n // dst stay two edges (a weight the fusion evals were gated on). The only exclusion is an\n // embed whose exact (src, target) also exists as a link: that second row is new with the\n // embed grain and would double an edge that used to be one row. The NOT EXISTS probes the\n // primary key and fires only for embed rows; both branches still scan the table.\n const sql = `SELECT src, dst FROM links WHERE dst IS NOT NULL AND embed = 0\n UNION ALL\n SELECT src, dst FROM links l WHERE dst IS NOT NULL AND embed = 1\n AND NOT EXISTS (SELECT 1 FROM links l0 WHERE l0.src = l.src AND l0.target = l.target AND l0.embed = 0)`;\n return (db.prepare(sql).all() as Array<{ src: string; dst: string }>).map((r) => [r.src, r.dst]);\n}\n\n// remove() has already deleted the rows by the time afterReconcile runs, so it records here\n// whether they carried edges. Keyed on the delta object, so the state dies with the reconcile.\nconst removedWithLinks = new WeakMap<ReconcileDelta, Set<string>>();\n\nfunction recordRemoved(delta: ReconcileDelta, path: string): void {\n let set = removedWithLinks.get(delta);\n if (!set) {\n set = new Set();\n removedWithLinks.set(delta, set);\n }\n set.add(path);\n}\n\n// remove() runs per path; the vanished list is an array, so cache the Set per delta.\nconst vanishedSets = new WeakMap<ReconcileDelta, Set<string>>();\nfunction vanishedSet(delta: ReconcileDelta): Set<string> {\n let set = vanishedSets.get(delta);\n if (!set) {\n set = new Set(delta.vanished);\n vanishedSets.set(delta, set);\n }\n return set;\n}\n\nexport const links: Feature = {\n name: 'links',\n schema(db) {\n db.exec('CREATE TABLE IF NOT EXISTS links (src TEXT, target TEXT, target_base TEXT, dst TEXT, embed INTEGER, PRIMARY KEY (src, target, embed))');\n db.exec('CREATE INDEX IF NOT EXISTS links_dst ON links(dst)');\n db.exec('CREATE INDEX IF NOT EXISTS links_target_base ON links(target_base)');\n },\n extract,\n // Reparsed docs are diffed in store() below instead of wiped here: deleting and\n // re-inserting resets dst to NULL, which made every touch-only reparse read as an edge\n // change and recompute PageRank -- measured as most of the remaining update cost.\n remove(db, path, delta) {\n if (!vanishedSet(delta).has(path)) return;\n const result = db.prepare('DELETE FROM links WHERE src = ?').run(path);\n if (Number(result.changes) > 0) recordRemoved(delta, path);\n },\n store(db, path, extracted, delta) {\n const targets = extracted as Array<{ target: string; embed: boolean }>;\n // Stale rows (target/embed pairs the new parse no longer contains) are real edge removals.\n // \\0 can't appear in a target, so the composite key never collides.\n let stale: ReturnType<ReturnType<DatabaseSync['prepare']>['run']>;\n if (targets.length === 0) {\n stale = db.prepare('DELETE FROM links WHERE src = ?').run(path);\n } else {\n const placeholders = targets.map(() => '?').join(', ');\n const keys = targets.map((t) => `${t.target}\\0${t.embed ? '1' : '0'}`);\n stale = db.prepare(`DELETE FROM links WHERE src = ? AND (target || char(0) || embed) NOT IN (${placeholders})`).run(path, ...keys);\n }\n if (Number(stale.changes) > 0) recordRemoved(delta, path);\n // Upsert preserves dst on surviving rows, so an unchanged link resolves to the same\n // value and reports no change; only genuinely new rows start at NULL.\n const insert = db.prepare('INSERT INTO links (src, target, target_base, dst, embed) VALUES (?, ?, ?, NULL, ?) ON CONFLICT(src, target, embed) DO UPDATE SET target_base = excluded.target_base');\n for (const { target, embed } of targets) insert.run(path, target, baseKey(cleanTarget(target)), embed ? 1 : 0);\n },\n afterReconcile(db, delta) {\n // Any deleted rows that carried edges mean the edge set shrank -- whether the file\n // vanished or was reparsed down to fewer/no links. dstChanged alone misses the\n // reparsed-to-zero-links case (no surviving rows to re-resolve).\n const removeHadLinks = (removedWithLinks.get(delta)?.size ?? 0) > 0;\n\n const churn = delta.reparsed.length + delta.vanished.length;\n // Past this share of the tree, a full pass is the only one guaranteed to match\n // resolveTarget's ambiguity rules. A cold build clears the threshold on its own.\n const large = delta.files.length === 0 || churn > 0.2 * delta.files.length;\n\n const dstChanged = large ? resolveAll(db, delta.files) : resolveIncremental(db, delta);\n delta.linksChanged = dstChanged || removeHadLinks;\n },\n};\n"],"names":["posix","extract","_raw","body","seen","Set","results","add","target","embed","key","has","push","m","matchAll","trim","index","test","cleanTarget","replace","baseKey","path","basename","toLowerCase","resolveTarget","src","pathSet","byBase","clean","fromSrc","normalize","join","dirname","candidate","get","buildByBase","files","Map","map","f","relPath","sort","list","set","resolveRows","db","rows","update","prepare","changed","row","dst","run","resolveAll","all","resolveIncremental","delta","changedBasenames","p","added","vanished","candidates","collect","collectIn","column","keys","i","length","chunk","slice","placeholders","reparsed","values","linkEdges","sql","r","removedWithLinks","WeakMap","recordRemoved","vanishedSets","vanishedSet","links","name","schema","exec","remove","result","Number","changes","store","extracted","targets","stale","t","insert","afterReconcile","removeHadLinks","size","churn","large","dstChanged","linksChanged"],"mappings":"AAAA,OAAOA,WAAW,kBAAkB;AAKpC,0FAA0F;AAC1F,iFAAiF;AACjF,+FAA+F;AAC/F,uFAAuF;AAEvF,qGAAqG;AACrG,SAASC,QAAQC,IAAY,EAAEC,IAAY;IACzC,MAAMC,OAAO,IAAIC;IACjB,MAAMC,UAAqD,EAAE;IAC7D,MAAMC,MAAM,CAACC,QAAgBC;QAC3B,MAAMC,MAAM,GAAGF,OAAO,EAAE,EAAEC,QAAQ,MAAM,KAAK;QAC7C,IAAIL,KAAKO,GAAG,CAACD,MAAM;QACnBN,KAAKG,GAAG,CAACG;QACTJ,QAAQM,IAAI,CAAC;YAAEJ;YAAQC;QAAM;IAC/B;IACA,KAAK,MAAMI,KAAKV,KAAKW,QAAQ,CAAC,iDAAkD;YAEhDD;QAD9B,MAAML,SAASK,CAAC,CAAC,EAAE,CAACE,IAAI;QACxB,IAAIP,QAAQD,IAAIC,QAAQL,IAAI,CAAC,EAACU,WAAAA,EAAEG,KAAK,cAAPH,sBAAAA,WAAW,KAAK,EAAE,KAAK;IACvD;IACA,KAAK,MAAMA,KAAKV,KAAKW,QAAQ,CAAC,6CAA8C;QAC1E,MAAMN,SAASK,CAAC,CAAC,EAAE,CAACE,IAAI;QACxB,IAAIP,UAAU,CAAC,gBAAgBS,IAAI,CAACT,SAASD,IAAIC,QAAQK,CAAC,CAAC,EAAE,KAAK;IACpE;IACA,OAAOP;AACT;AAEA,SAASY,YAAYV,MAAc;IACjC,OAAOA,OAAOW,OAAO,CAAC,OAAO,KAAKA,OAAO,CAAC,SAAS;AACrD;AAEA,SAASC,QAAQC,IAAY;IAC3B,OAAOrB,MAAMsB,QAAQ,CAACD,MAAMF,OAAO,CAAC,UAAU,IAAII,WAAW;AAC/D;AAEA,uFAAuF;AACvF,2EAA2E;AAC3E,SAASC,cAAcC,GAAW,EAAEjB,MAAc,EAAEkB,OAAoB,EAAEC,MAA6B;;QAM9FA;IALP,MAAMC,QAAQV,YAAYV;IAC1B,MAAMqB,UAAU7B,MAAM8B,SAAS,CAAC9B,MAAM+B,IAAI,CAAC/B,MAAMgC,OAAO,CAACP,MAAMG;IAC/D,KAAK,MAAMK,aAAa;QAACL;QAAO,GAAGA,MAAM,GAAG,CAAC;QAAEC;QAAS,GAAGA,QAAQ,GAAG,CAAC;KAAC,CAAE;QACxE,IAAIH,QAAQf,GAAG,CAACsB,YAAY,OAAOA;IACrC;IACA,gBAAON,cAAAA,OAAOO,GAAG,CAACd,QAAQQ,qBAAnBD,kCAAAA,WAA4B,CAAC,EAAE,uCAAI;AAC5C;AAEA,SAASQ,YAAYC,KAAiB;IACpC,MAAMT,SAAS,IAAIU;IACnB,KAAK,MAAMhB,QAAQe,MAAME,GAAG,CAAC,CAACC,IAAMA,EAAEC,OAAO,EAAEC,IAAI,GAAI;QACrD,MAAM/B,MAAMU,QAAQC;QACpB,MAAMqB,OAAOf,OAAOO,GAAG,CAACxB;QACxB,IAAIgC,MAAMA,KAAK9B,IAAI,CAACS;aACfM,OAAOgB,GAAG,CAACjC,KAAK;YAACW;SAAK;IAC7B;IACA,OAAOM;AACT;AASA,4FAA4F;AAC5F,0FAA0F;AAC1F,0BAA0B;AAC1B,SAASiB,YAAYC,EAAgB,EAAEC,IAAe,EAAEpB,OAAoB,EAAEC,MAA6B;IACzG,MAAMoB,SAASF,GAAGG,OAAO,CAAC;IAC1B,IAAIC,UAAU;IACd,KAAK,MAAMC,OAAOJ,KAAM;QACtB,MAAMK,MAAM3B,cAAc0B,IAAIzB,GAAG,EAAEyB,IAAI1C,MAAM,EAAEkB,SAASC;QACxD,IAAIwB,QAAQD,IAAIC,GAAG,EAAE;YACnBJ,OAAOK,GAAG,CAACD,KAAKD,IAAIzB,GAAG,EAAEyB,IAAI1C,MAAM;YACnCyC,UAAU;QACZ;IACF;IACA,OAAOA;AACT;AAEA,yFAAyF;AACzF,+EAA+E;AAC/E,SAASI,WAAWR,EAAgB,EAAET,KAAiB;IACrD,MAAMV,UAAU,IAAIrB,IAAI+B,MAAME,GAAG,CAAC,CAACC,IAAMA,EAAEC,OAAO;IAClD,MAAMb,SAASQ,YAAYC;IAC3B,MAAMU,OAAOD,GAAGG,OAAO,CAAC,6CAA6CM,GAAG;IACxE,OAAOV,YAAYC,IAAIC,MAAMpB,SAASC;AACxC;AAEA,yFAAyF;AACzF,yFAAyF;AACzF,kEAAkE;AAClE,SAAS4B,mBAAmBV,EAAgB,EAAEW,KAAqB;IACjE,MAAM9B,UAAU,IAAIrB,IAAImD,MAAMpB,KAAK,CAACE,GAAG,CAAC,CAACC,IAAMA,EAAEC,OAAO;IACxD,MAAMb,SAASQ,YAAYqB,MAAMpB,KAAK;IAEtC,MAAMqB,mBAAmB,IAAIpD;IAC7B,KAAK,MAAMqD,KAAKF,MAAMG,KAAK,CAAEF,iBAAiBlD,GAAG,CAACa,QAAQsC;IAC1D,KAAK,MAAMA,KAAKF,MAAMI,QAAQ,CAAEH,iBAAiBlD,GAAG,CAACa,QAAQsC;IAE7D,+EAA+E;IAC/E,MAAMG,aAAa,IAAIxB;IACvB,MAAMyB,UAAU,CAAChB;QACf,KAAK,MAAMI,OAAOJ,KAAMe,WAAWlB,GAAG,CAAC,GAAGO,IAAIzB,GAAG,CAAC,EAAE,EAAEyB,IAAI1C,MAAM,CAAC,EAAE,EAAE0C,IAAIzC,KAAK,EAAE,EAAEyC;IACpF;IACA,qFAAqF;IACrF,6CAA6C;IAC7C,MAAMa,YAAY,CAACC,QAAgBC;QACjC,IAAK,IAAIC,IAAI,GAAGA,IAAID,KAAKE,MAAM,EAAED,KAAK,IAAK;YACzC,MAAME,QAAQH,KAAKI,KAAK,CAACH,GAAGA,IAAI;YAChC,MAAMI,eAAeF,MAAM9B,GAAG,CAAC,IAAM,KAAKP,IAAI,CAAC;YAC/C+B,QAAQjB,GAAGG,OAAO,CAAC,CAAC,gDAAgD,EAAEgB,OAAO,KAAK,EAAEM,aAAa,CAAC,CAAC,EAAEhB,GAAG,IAAIc;QAC9G;IACF;IAEAL,UAAU,OAAOP,MAAMe,QAAQ;IAC/BR,UAAU,eAAe;WAAIN;KAAiB;IAC9CM,UAAU,OAAOP,MAAMI,QAAQ;IAE/B,OAAOhB,YAAYC,IAAI;WAAIgB,WAAWW,MAAM;KAAG,EAAE9C,SAASC;AAC5D;AAEA,6DAA6D;AAC7D,OAAO,SAAS8C,UAAU5B,EAAgB;IACxC,0FAA0F;IAC1F,yFAAyF;IACzF,yFAAyF;IACzF,0FAA0F;IAC1F,iFAAiF;IACjF,MAAM6B,MAAM,CAAC;;;4GAG6F,CAAC;IAC3G,OAAO,AAAC7B,GAAGG,OAAO,CAAC0B,KAAKpB,GAAG,GAA2ChB,GAAG,CAAC,CAACqC,IAAM;YAACA,EAAElD,GAAG;YAAEkD,EAAExB,GAAG;SAAC;AACjG;AAEA,4FAA4F;AAC5F,+FAA+F;AAC/F,MAAMyB,mBAAmB,IAAIC;AAE7B,SAASC,cAActB,KAAqB,EAAEnC,IAAY;IACxD,IAAIsB,MAAMiC,iBAAiB1C,GAAG,CAACsB;IAC/B,IAAI,CAACb,KAAK;QACRA,MAAM,IAAItC;QACVuE,iBAAiBjC,GAAG,CAACa,OAAOb;IAC9B;IACAA,IAAIpC,GAAG,CAACc;AACV;AAEA,qFAAqF;AACrF,MAAM0D,eAAe,IAAIF;AACzB,SAASG,YAAYxB,KAAqB;IACxC,IAAIb,MAAMoC,aAAa7C,GAAG,CAACsB;IAC3B,IAAI,CAACb,KAAK;QACRA,MAAM,IAAItC,IAAImD,MAAMI,QAAQ;QAC5BmB,aAAapC,GAAG,CAACa,OAAOb;IAC1B;IACA,OAAOA;AACT;AAEA,OAAO,MAAMsC,QAAiB;IAC5BC,MAAM;IACNC,QAAOtC,EAAE;QACPA,GAAGuC,IAAI,CAAC;QACRvC,GAAGuC,IAAI,CAAC;QACRvC,GAAGuC,IAAI,CAAC;IACV;IACAnF;IACA,gFAAgF;IAChF,uFAAuF;IACvF,kFAAkF;IAClFoF,QAAOxC,EAAE,EAAExB,IAAI,EAAEmC,KAAK;QACpB,IAAI,CAACwB,YAAYxB,OAAO7C,GAAG,CAACU,OAAO;QACnC,MAAMiE,SAASzC,GAAGG,OAAO,CAAC,mCAAmCI,GAAG,CAAC/B;QACjE,IAAIkE,OAAOD,OAAOE,OAAO,IAAI,GAAGV,cAActB,OAAOnC;IACvD;IACAoE,OAAM5C,EAAE,EAAExB,IAAI,EAAEqE,SAAS,EAAElC,KAAK;QAC9B,MAAMmC,UAAUD;QAChB,2FAA2F;QAC3F,oEAAoE;QACpE,IAAIE;QACJ,IAAID,QAAQxB,MAAM,KAAK,GAAG;YACxByB,QAAQ/C,GAAGG,OAAO,CAAC,mCAAmCI,GAAG,CAAC/B;QAC5D,OAAO;YACL,MAAMiD,eAAeqB,QAAQrD,GAAG,CAAC,IAAM,KAAKP,IAAI,CAAC;YACjD,MAAMkC,OAAO0B,QAAQrD,GAAG,CAAC,CAACuD,IAAM,GAAGA,EAAErF,MAAM,CAAC,EAAE,EAAEqF,EAAEpF,KAAK,GAAG,MAAM,KAAK;YACrEmF,QAAQ/C,GAAGG,OAAO,CAAC,CAAC,yEAAyE,EAAEsB,aAAa,CAAC,CAAC,EAAElB,GAAG,CAAC/B,SAAS4C;QAC/H;QACA,IAAIsB,OAAOK,MAAMJ,OAAO,IAAI,GAAGV,cAActB,OAAOnC;QACpD,oFAAoF;QACpF,sEAAsE;QACtE,MAAMyE,SAASjD,GAAGG,OAAO,CAAC;QAC1B,KAAK,MAAM,EAAExC,MAAM,EAAEC,KAAK,EAAE,IAAIkF,QAASG,OAAO1C,GAAG,CAAC/B,MAAMb,QAAQY,QAAQF,YAAYV,UAAUC,QAAQ,IAAI;IAC9G;IACAsF,gBAAelD,EAAE,EAAEW,KAAK;;YAIEoB;QAHxB,mFAAmF;QACnF,+EAA+E;QAC/E,iEAAiE;QACjE,MAAMoB,iBAAiB,UAACpB,wBAAAA,iBAAiB1C,GAAG,CAACsB,oBAArBoB,4CAAAA,sBAA6BqB,IAAI,uCAAI,KAAK;QAElE,MAAMC,QAAQ1C,MAAMe,QAAQ,CAACJ,MAAM,GAAGX,MAAMI,QAAQ,CAACO,MAAM;QAC3D,+EAA+E;QAC/E,iFAAiF;QACjF,MAAMgC,QAAQ3C,MAAMpB,KAAK,CAAC+B,MAAM,KAAK,KAAK+B,QAAQ,MAAM1C,MAAMpB,KAAK,CAAC+B,MAAM;QAE1E,MAAMiC,aAAaD,QAAQ9C,WAAWR,IAAIW,MAAMpB,KAAK,IAAImB,mBAAmBV,IAAIW;QAChFA,MAAM6C,YAAY,GAAGD,cAAcJ;IACrC;AACF,EAAE"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/features/links.ts"],"sourcesContent":["import posix from 'node:path/posix';\nimport type { DatabaseSync } from 'node:sqlite';\nimport type { FileStat } from '../scan.ts';\nimport type { Feature, ReconcileDelta } from './types.ts';\n\n// links(src, target, target_base, dst, embed): target as written, target_base its baseKey\n// (indexed, for the incremental resolve below), dst the resolved path or NULL (a\n// queryable dead link), embed whether it's `![[x]]`/`![](x.md)` rather than `[[x]]`/`[](x.md)`\n// -- Obsidian's grain: a target both linked and embedded in the same note is two rows.\n\n// Wikilinks ([[target]], [[target#anchor|alias]], embeds) plus relative markdown links to .md files.\nfunction extract(_raw: string, body: string): Array<{ target: string; embed: boolean }> {\n const seen = new Set<string>();\n const results: Array<{ target: string; embed: boolean }> = [];\n const add = (target: string, embed: boolean) => {\n const key = `${target}\\0${embed ? '1' : '0'}`;\n if (seen.has(key)) return;\n seen.add(key);\n results.push({ target, embed });\n };\n // To the first ]], as Obsidian parses it, so a heading or alias holding a lone ] still\n // matches; anchor and alias split off in code, since a class-based regex stops at the ].\n for (const m of body.matchAll(/\\[\\[(.*?)\\]\\]/g)) {\n const target = m[1].split('|')[0].split('#')[0].trim();\n if (target) add(target, body[(m.index ?? 0) - 1] === '!');\n }\n for (const m of body.matchAll(/(!)?\\[[^\\]]*\\]\\(([^)]+\\.md)(?:#[^)]*)?\\)/g)) {\n const target = m[2].trim();\n if (target && !/^[a-z]+:\\/\\//i.test(target)) add(target, m[1] === '!');\n }\n return results;\n}\n\nfunction cleanTarget(target: string): string {\n return target.replace(/\\\\/g, '/').replace(/^\\.\\//, '');\n}\n\nfunction baseKey(path: string): string {\n return posix.basename(path).replace(/\\.md$/i, '').toLowerCase();\n}\n\n// Obsidian-style: exact relative path (with/without .md), path relative to the linking\n// note's directory, then basename match (lexicographically first on ties).\nfunction resolveTarget(src: string, target: string, pathSet: Set<string>, byBase: Map<string, string[]>): string | null {\n const clean = cleanTarget(target);\n const fromSrc = posix.normalize(posix.join(posix.dirname(src), clean));\n for (const candidate of [clean, `${clean}.md`, fromSrc, `${fromSrc}.md`]) {\n if (pathSet.has(candidate)) return candidate;\n }\n return byBase.get(baseKey(clean))?.[0] ?? null;\n}\n\nfunction buildByBase(files: FileStat[]): Map<string, string[]> {\n const byBase = new Map<string, string[]>();\n for (const path of files.map((f) => f.relPath).sort()) {\n const key = baseKey(path);\n const list = byBase.get(key);\n if (list) list.push(path);\n else byBase.set(key, [path]);\n }\n return byBase;\n}\n\ninterface LinkRow {\n src: string;\n target: string;\n dst: string | null;\n embed: number;\n}\n\n// Applies resolveTarget to `rows`, writing only rows whose dst actually changes. The UPDATE\n// keys on (src, target): a link/embed sibling pair shares one resolution, so either row's\n// mismatch rewrites both.\nfunction resolveRows(db: DatabaseSync, rows: LinkRow[], pathSet: Set<string>, byBase: Map<string, string[]>): boolean {\n const update = db.prepare('UPDATE links SET dst = ? WHERE src = ? AND target = ?');\n let changed = false;\n for (const row of rows) {\n const dst = resolveTarget(row.src, row.target, pathSet, byBase);\n if (dst !== row.dst) {\n update.run(dst, row.src, row.target);\n changed = true;\n }\n }\n return changed;\n}\n\n// A new or deleted file can change any note's resolution, so re-resolve the whole table.\n// Fallback for cold builds and large deltas -- see afterReconcile's threshold.\nfunction resolveAll(db: DatabaseSync, files: FileStat[]): boolean {\n const pathSet = new Set(files.map((f) => f.relPath));\n const byBase = buildByBase(files);\n const rows = db.prepare('SELECT src, target, dst, embed FROM links').all() as unknown as LinkRow[];\n return resolveRows(db, rows, pathSet, byBase);\n}\n\n// Re-resolve only what this reconcile could have touched: re-stored sources, links whose\n// target basename matches an added/vanished file, and links whose dst just vanished. All\n// three are indexed lookups, so this never reads the whole table.\nfunction resolveIncremental(db: DatabaseSync, delta: ReconcileDelta): boolean {\n const pathSet = new Set(delta.files.map((f) => f.relPath));\n const byBase = buildByBase(delta.files);\n\n const changedBasenames = new Set<string>();\n for (const p of delta.added) changedBasenames.add(baseKey(p));\n for (const p of delta.vanished) changedBasenames.add(baseKey(p));\n\n // \\0 can't appear in a path or target, so the key never collides; a space can.\n const candidates = new Map<string, LinkRow>();\n const collect = (rows: LinkRow[]) => {\n for (const row of rows) candidates.set(`${row.src}\\0${row.target}\\0${row.embed}`, row);\n };\n // Chunked so a large-but-under-threshold delta never exceeds SQLite's bound-variable\n // limit (SQLITE_MAX_VARIABLE_NUMBER, 32766).\n const collectIn = (column: string, keys: string[]) => {\n for (let i = 0; i < keys.length; i += 500) {\n const chunk = keys.slice(i, i + 500);\n const placeholders = chunk.map(() => '?').join(', ');\n collect(db.prepare(`SELECT src, target, dst, embed FROM links WHERE ${column} IN (${placeholders})`).all(...chunk) as unknown as LinkRow[]);\n }\n };\n\n collectIn('src', delta.reparsed);\n collectIn('target_base', [...changedBasenames]);\n collectIn('dst', delta.vanished);\n\n return resolveRows(db, [...candidates.values()], pathSet, byBase);\n}\n\n// Resolved edges, for rank and for search's graph expansion.\nexport function linkEdges(db: DatabaseSync): [string, string][] {\n // One edge per row, as 0.12's grain always had it -- two written targets resolving to one\n // dst stay two edges (a weight the fusion evals were gated on). The only exclusion is an\n // embed whose exact (src, target) also exists as a link: that second row is new with the\n // embed grain and would double an edge that used to be one row. The NOT EXISTS probes the\n // primary key and fires only for embed rows; both branches still scan the table.\n const sql = `SELECT src, dst FROM links WHERE dst IS NOT NULL AND embed = 0\n UNION ALL\n SELECT src, dst FROM links l WHERE dst IS NOT NULL AND embed = 1\n AND NOT EXISTS (SELECT 1 FROM links l0 WHERE l0.src = l.src AND l0.target = l.target AND l0.embed = 0)`;\n return (db.prepare(sql).all() as Array<{ src: string; dst: string }>).map((r) => [r.src, r.dst]);\n}\n\n// remove() has already deleted the rows by the time afterReconcile runs, so it records here\n// whether they carried edges. Keyed on the delta object, so the state dies with the reconcile.\nconst removedWithLinks = new WeakMap<ReconcileDelta, Set<string>>();\n\nfunction recordRemoved(delta: ReconcileDelta, path: string): void {\n let set = removedWithLinks.get(delta);\n if (!set) {\n set = new Set();\n removedWithLinks.set(delta, set);\n }\n set.add(path);\n}\n\n// remove() runs per path; the vanished list is an array, so cache the Set per delta.\nconst vanishedSets = new WeakMap<ReconcileDelta, Set<string>>();\nfunction vanishedSet(delta: ReconcileDelta): Set<string> {\n let set = vanishedSets.get(delta);\n if (!set) {\n set = new Set(delta.vanished);\n vanishedSets.set(delta, set);\n }\n return set;\n}\n\nexport const links: Feature = {\n name: 'links',\n schema(db) {\n db.exec('CREATE TABLE IF NOT EXISTS links (src TEXT, target TEXT, target_base TEXT, dst TEXT, embed INTEGER, PRIMARY KEY (src, target, embed))');\n db.exec('CREATE INDEX IF NOT EXISTS links_dst ON links(dst)');\n db.exec('CREATE INDEX IF NOT EXISTS links_target_base ON links(target_base)');\n },\n extract,\n // Reparsed docs are diffed in store() below instead of wiped here: deleting and\n // re-inserting resets dst to NULL, which made every touch-only reparse read as an edge\n // change and recompute PageRank -- measured as most of the remaining update cost.\n remove(db, path, delta) {\n if (!vanishedSet(delta).has(path)) return;\n const result = db.prepare('DELETE FROM links WHERE src = ?').run(path);\n if (Number(result.changes) > 0) recordRemoved(delta, path);\n },\n store(db, path, extracted, delta) {\n const targets = extracted as Array<{ target: string; embed: boolean }>;\n // Stale rows (target/embed pairs the new parse no longer contains) are real edge removals.\n // \\0 can't appear in a target, so the composite key never collides.\n let stale: ReturnType<ReturnType<DatabaseSync['prepare']>['run']>;\n if (targets.length === 0) {\n stale = db.prepare('DELETE FROM links WHERE src = ?').run(path);\n } else {\n const placeholders = targets.map(() => '?').join(', ');\n const keys = targets.map((t) => `${t.target}\\0${t.embed ? '1' : '0'}`);\n stale = db.prepare(`DELETE FROM links WHERE src = ? AND (target || char(0) || embed) NOT IN (${placeholders})`).run(path, ...keys);\n }\n if (Number(stale.changes) > 0) recordRemoved(delta, path);\n // Upsert preserves dst on surviving rows, so an unchanged link resolves to the same\n // value and reports no change; only genuinely new rows start at NULL.\n const insert = db.prepare('INSERT INTO links (src, target, target_base, dst, embed) VALUES (?, ?, ?, NULL, ?) ON CONFLICT(src, target, embed) DO UPDATE SET target_base = excluded.target_base');\n for (const { target, embed } of targets) insert.run(path, target, baseKey(cleanTarget(target)), embed ? 1 : 0);\n },\n afterReconcile(db, delta) {\n // Any deleted rows that carried edges mean the edge set shrank -- whether the file\n // vanished or was reparsed down to fewer/no links. dstChanged alone misses the\n // reparsed-to-zero-links case (no surviving rows to re-resolve).\n const removeHadLinks = (removedWithLinks.get(delta)?.size ?? 0) > 0;\n\n const churn = delta.reparsed.length + delta.vanished.length;\n // Past this share of the tree, a full pass is the only one guaranteed to match\n // resolveTarget's ambiguity rules. A cold build clears the threshold on its own.\n const large = delta.files.length === 0 || churn > 0.2 * delta.files.length;\n\n const dstChanged = large ? resolveAll(db, delta.files) : resolveIncremental(db, delta);\n delta.linksChanged = dstChanged || removeHadLinks;\n },\n};\n"],"names":["posix","extract","_raw","body","seen","Set","results","add","target","embed","key","has","push","m","matchAll","split","trim","index","test","cleanTarget","replace","baseKey","path","basename","toLowerCase","resolveTarget","src","pathSet","byBase","clean","fromSrc","normalize","join","dirname","candidate","get","buildByBase","files","Map","map","f","relPath","sort","list","set","resolveRows","db","rows","update","prepare","changed","row","dst","run","resolveAll","all","resolveIncremental","delta","changedBasenames","p","added","vanished","candidates","collect","collectIn","column","keys","i","length","chunk","slice","placeholders","reparsed","values","linkEdges","sql","r","removedWithLinks","WeakMap","recordRemoved","vanishedSets","vanishedSet","links","name","schema","exec","remove","result","Number","changes","store","extracted","targets","stale","t","insert","afterReconcile","removeHadLinks","size","churn","large","dstChanged","linksChanged"],"mappings":"AAAA,OAAOA,WAAW,kBAAkB;AAKpC,0FAA0F;AAC1F,iFAAiF;AACjF,+FAA+F;AAC/F,uFAAuF;AAEvF,qGAAqG;AACrG,SAASC,QAAQC,IAAY,EAAEC,IAAY;IACzC,MAAMC,OAAO,IAAIC;IACjB,MAAMC,UAAqD,EAAE;IAC7D,MAAMC,MAAM,CAACC,QAAgBC;QAC3B,MAAMC,MAAM,GAAGF,OAAO,EAAE,EAAEC,QAAQ,MAAM,KAAK;QAC7C,IAAIL,KAAKO,GAAG,CAACD,MAAM;QACnBN,KAAKG,GAAG,CAACG;QACTJ,QAAQM,IAAI,CAAC;YAAEJ;YAAQC;QAAM;IAC/B;IACA,uFAAuF;IACvF,yFAAyF;IACzF,KAAK,MAAMI,KAAKV,KAAKW,QAAQ,CAAC,kBAAmB;YAEjBD;QAD9B,MAAML,SAASK,CAAC,CAAC,EAAE,CAACE,KAAK,CAAC,IAAI,CAAC,EAAE,CAACA,KAAK,CAAC,IAAI,CAAC,EAAE,CAACC,IAAI;QACpD,IAAIR,QAAQD,IAAIC,QAAQL,IAAI,CAAC,EAACU,WAAAA,EAAEI,KAAK,cAAPJ,sBAAAA,WAAW,KAAK,EAAE,KAAK;IACvD;IACA,KAAK,MAAMA,KAAKV,KAAKW,QAAQ,CAAC,6CAA8C;QAC1E,MAAMN,SAASK,CAAC,CAAC,EAAE,CAACG,IAAI;QACxB,IAAIR,UAAU,CAAC,gBAAgBU,IAAI,CAACV,SAASD,IAAIC,QAAQK,CAAC,CAAC,EAAE,KAAK;IACpE;IACA,OAAOP;AACT;AAEA,SAASa,YAAYX,MAAc;IACjC,OAAOA,OAAOY,OAAO,CAAC,OAAO,KAAKA,OAAO,CAAC,SAAS;AACrD;AAEA,SAASC,QAAQC,IAAY;IAC3B,OAAOtB,MAAMuB,QAAQ,CAACD,MAAMF,OAAO,CAAC,UAAU,IAAII,WAAW;AAC/D;AAEA,uFAAuF;AACvF,2EAA2E;AAC3E,SAASC,cAAcC,GAAW,EAAElB,MAAc,EAAEmB,OAAoB,EAAEC,MAA6B;;QAM9FA;IALP,MAAMC,QAAQV,YAAYX;IAC1B,MAAMsB,UAAU9B,MAAM+B,SAAS,CAAC/B,MAAMgC,IAAI,CAAChC,MAAMiC,OAAO,CAACP,MAAMG;IAC/D,KAAK,MAAMK,aAAa;QAACL;QAAO,GAAGA,MAAM,GAAG,CAAC;QAAEC;QAAS,GAAGA,QAAQ,GAAG,CAAC;KAAC,CAAE;QACxE,IAAIH,QAAQhB,GAAG,CAACuB,YAAY,OAAOA;IACrC;IACA,gBAAON,cAAAA,OAAOO,GAAG,CAACd,QAAQQ,qBAAnBD,kCAAAA,WAA4B,CAAC,EAAE,uCAAI;AAC5C;AAEA,SAASQ,YAAYC,KAAiB;IACpC,MAAMT,SAAS,IAAIU;IACnB,KAAK,MAAMhB,QAAQe,MAAME,GAAG,CAAC,CAACC,IAAMA,EAAEC,OAAO,EAAEC,IAAI,GAAI;QACrD,MAAMhC,MAAMW,QAAQC;QACpB,MAAMqB,OAAOf,OAAOO,GAAG,CAACzB;QACxB,IAAIiC,MAAMA,KAAK/B,IAAI,CAACU;aACfM,OAAOgB,GAAG,CAAClC,KAAK;YAACY;SAAK;IAC7B;IACA,OAAOM;AACT;AASA,4FAA4F;AAC5F,0FAA0F;AAC1F,0BAA0B;AAC1B,SAASiB,YAAYC,EAAgB,EAAEC,IAAe,EAAEpB,OAAoB,EAAEC,MAA6B;IACzG,MAAMoB,SAASF,GAAGG,OAAO,CAAC;IAC1B,IAAIC,UAAU;IACd,KAAK,MAAMC,OAAOJ,KAAM;QACtB,MAAMK,MAAM3B,cAAc0B,IAAIzB,GAAG,EAAEyB,IAAI3C,MAAM,EAAEmB,SAASC;QACxD,IAAIwB,QAAQD,IAAIC,GAAG,EAAE;YACnBJ,OAAOK,GAAG,CAACD,KAAKD,IAAIzB,GAAG,EAAEyB,IAAI3C,MAAM;YACnC0C,UAAU;QACZ;IACF;IACA,OAAOA;AACT;AAEA,yFAAyF;AACzF,+EAA+E;AAC/E,SAASI,WAAWR,EAAgB,EAAET,KAAiB;IACrD,MAAMV,UAAU,IAAItB,IAAIgC,MAAME,GAAG,CAAC,CAACC,IAAMA,EAAEC,OAAO;IAClD,MAAMb,SAASQ,YAAYC;IAC3B,MAAMU,OAAOD,GAAGG,OAAO,CAAC,6CAA6CM,GAAG;IACxE,OAAOV,YAAYC,IAAIC,MAAMpB,SAASC;AACxC;AAEA,yFAAyF;AACzF,yFAAyF;AACzF,kEAAkE;AAClE,SAAS4B,mBAAmBV,EAAgB,EAAEW,KAAqB;IACjE,MAAM9B,UAAU,IAAItB,IAAIoD,MAAMpB,KAAK,CAACE,GAAG,CAAC,CAACC,IAAMA,EAAEC,OAAO;IACxD,MAAMb,SAASQ,YAAYqB,MAAMpB,KAAK;IAEtC,MAAMqB,mBAAmB,IAAIrD;IAC7B,KAAK,MAAMsD,KAAKF,MAAMG,KAAK,CAAEF,iBAAiBnD,GAAG,CAACc,QAAQsC;IAC1D,KAAK,MAAMA,KAAKF,MAAMI,QAAQ,CAAEH,iBAAiBnD,GAAG,CAACc,QAAQsC;IAE7D,+EAA+E;IAC/E,MAAMG,aAAa,IAAIxB;IACvB,MAAMyB,UAAU,CAAChB;QACf,KAAK,MAAMI,OAAOJ,KAAMe,WAAWlB,GAAG,CAAC,GAAGO,IAAIzB,GAAG,CAAC,EAAE,EAAEyB,IAAI3C,MAAM,CAAC,EAAE,EAAE2C,IAAI1C,KAAK,EAAE,EAAE0C;IACpF;IACA,qFAAqF;IACrF,6CAA6C;IAC7C,MAAMa,YAAY,CAACC,QAAgBC;QACjC,IAAK,IAAIC,IAAI,GAAGA,IAAID,KAAKE,MAAM,EAAED,KAAK,IAAK;YACzC,MAAME,QAAQH,KAAKI,KAAK,CAACH,GAAGA,IAAI;YAChC,MAAMI,eAAeF,MAAM9B,GAAG,CAAC,IAAM,KAAKP,IAAI,CAAC;YAC/C+B,QAAQjB,GAAGG,OAAO,CAAC,CAAC,gDAAgD,EAAEgB,OAAO,KAAK,EAAEM,aAAa,CAAC,CAAC,EAAEhB,GAAG,IAAIc;QAC9G;IACF;IAEAL,UAAU,OAAOP,MAAMe,QAAQ;IAC/BR,UAAU,eAAe;WAAIN;KAAiB;IAC9CM,UAAU,OAAOP,MAAMI,QAAQ;IAE/B,OAAOhB,YAAYC,IAAI;WAAIgB,WAAWW,MAAM;KAAG,EAAE9C,SAASC;AAC5D;AAEA,6DAA6D;AAC7D,OAAO,SAAS8C,UAAU5B,EAAgB;IACxC,0FAA0F;IAC1F,yFAAyF;IACzF,yFAAyF;IACzF,0FAA0F;IAC1F,iFAAiF;IACjF,MAAM6B,MAAM,CAAC;;;4GAG6F,CAAC;IAC3G,OAAO,AAAC7B,GAAGG,OAAO,CAAC0B,KAAKpB,GAAG,GAA2ChB,GAAG,CAAC,CAACqC,IAAM;YAACA,EAAElD,GAAG;YAAEkD,EAAExB,GAAG;SAAC;AACjG;AAEA,4FAA4F;AAC5F,+FAA+F;AAC/F,MAAMyB,mBAAmB,IAAIC;AAE7B,SAASC,cAActB,KAAqB,EAAEnC,IAAY;IACxD,IAAIsB,MAAMiC,iBAAiB1C,GAAG,CAACsB;IAC/B,IAAI,CAACb,KAAK;QACRA,MAAM,IAAIvC;QACVwE,iBAAiBjC,GAAG,CAACa,OAAOb;IAC9B;IACAA,IAAIrC,GAAG,CAACe;AACV;AAEA,qFAAqF;AACrF,MAAM0D,eAAe,IAAIF;AACzB,SAASG,YAAYxB,KAAqB;IACxC,IAAIb,MAAMoC,aAAa7C,GAAG,CAACsB;IAC3B,IAAI,CAACb,KAAK;QACRA,MAAM,IAAIvC,IAAIoD,MAAMI,QAAQ;QAC5BmB,aAAapC,GAAG,CAACa,OAAOb;IAC1B;IACA,OAAOA;AACT;AAEA,OAAO,MAAMsC,QAAiB;IAC5BC,MAAM;IACNC,QAAOtC,EAAE;QACPA,GAAGuC,IAAI,CAAC;QACRvC,GAAGuC,IAAI,CAAC;QACRvC,GAAGuC,IAAI,CAAC;IACV;IACApF;IACA,gFAAgF;IAChF,uFAAuF;IACvF,kFAAkF;IAClFqF,QAAOxC,EAAE,EAAExB,IAAI,EAAEmC,KAAK;QACpB,IAAI,CAACwB,YAAYxB,OAAO9C,GAAG,CAACW,OAAO;QACnC,MAAMiE,SAASzC,GAAGG,OAAO,CAAC,mCAAmCI,GAAG,CAAC/B;QACjE,IAAIkE,OAAOD,OAAOE,OAAO,IAAI,GAAGV,cAActB,OAAOnC;IACvD;IACAoE,OAAM5C,EAAE,EAAExB,IAAI,EAAEqE,SAAS,EAAElC,KAAK;QAC9B,MAAMmC,UAAUD;QAChB,2FAA2F;QAC3F,oEAAoE;QACpE,IAAIE;QACJ,IAAID,QAAQxB,MAAM,KAAK,GAAG;YACxByB,QAAQ/C,GAAGG,OAAO,CAAC,mCAAmCI,GAAG,CAAC/B;QAC5D,OAAO;YACL,MAAMiD,eAAeqB,QAAQrD,GAAG,CAAC,IAAM,KAAKP,IAAI,CAAC;YACjD,MAAMkC,OAAO0B,QAAQrD,GAAG,CAAC,CAACuD,IAAM,GAAGA,EAAEtF,MAAM,CAAC,EAAE,EAAEsF,EAAErF,KAAK,GAAG,MAAM,KAAK;YACrEoF,QAAQ/C,GAAGG,OAAO,CAAC,CAAC,yEAAyE,EAAEsB,aAAa,CAAC,CAAC,EAAElB,GAAG,CAAC/B,SAAS4C;QAC/H;QACA,IAAIsB,OAAOK,MAAMJ,OAAO,IAAI,GAAGV,cAActB,OAAOnC;QACpD,oFAAoF;QACpF,sEAAsE;QACtE,MAAMyE,SAASjD,GAAGG,OAAO,CAAC;QAC1B,KAAK,MAAM,EAAEzC,MAAM,EAAEC,KAAK,EAAE,IAAImF,QAASG,OAAO1C,GAAG,CAAC/B,MAAMd,QAAQa,QAAQF,YAAYX,UAAUC,QAAQ,IAAI;IAC9G;IACAuF,gBAAelD,EAAE,EAAEW,KAAK;;YAIEoB;QAHxB,mFAAmF;QACnF,+EAA+E;QAC/E,iEAAiE;QACjE,MAAMoB,iBAAiB,UAACpB,wBAAAA,iBAAiB1C,GAAG,CAACsB,oBAArBoB,4CAAAA,sBAA6BqB,IAAI,uCAAI,KAAK;QAElE,MAAMC,QAAQ1C,MAAMe,QAAQ,CAACJ,MAAM,GAAGX,MAAMI,QAAQ,CAACO,MAAM;QAC3D,+EAA+E;QAC/E,iFAAiF;QACjF,MAAMgC,QAAQ3C,MAAMpB,KAAK,CAAC+B,MAAM,KAAK,KAAK+B,QAAQ,MAAM1C,MAAMpB,KAAK,CAAC+B,MAAM;QAE1E,MAAMiC,aAAaD,QAAQ9C,WAAWR,IAAIW,MAAMpB,KAAK,IAAImB,mBAAmBV,IAAIW;QAChFA,MAAM6C,YAAY,GAAGD,cAAcJ;IACrC;AACF,EAAE"}
@@ -3,6 +3,10 @@
3
3
  // (book/scifi); `tag = 'book' OR tag LIKE 'book/%'` is how a caller matches the parent too.
4
4
  const FENCE_RE = /^(```|~~~)/;
5
5
  const INLINE_CODE_RE = /`[^`]*`/g;
6
+ // Obsidian treats [[#Heading]] as a same-note link, not a tag.
7
+ const WIKILINK_RE = /\[\[.*?\]\]/g; // to the first ]], so a heading holding a lone ] still masks
8
+ // Obsidian doesn't read tags inside HTML markup.
9
+ const HTML_TAG_RE = /<\/?[a-zA-Z][^>]*>/g; // tag-shaped only: a comparison's `< 5` must not open a span
6
10
  // Anchors on start-of-line or a preceding whitespace/(/[ so `a#b` and URL fragments don't count.
7
11
  const INLINE_TAG_RE = /(?:^|[\s([])#([\p{L}\p{N}_/-]+)/gu;
8
12
  // Strips a leading # (frontmatter entries may carry one) and a trailing /; rejects an
@@ -27,7 +31,7 @@ function frontmatterTags(data) {
27
31
  }
28
32
  return found;
29
33
  }
30
- // #tag tokens outside fenced code blocks and inline code spans.
34
+ // #tag tokens outside fenced code blocks, inline code spans, wikilinks, and HTML tags.
31
35
  function inlineTags(body) {
32
36
  const found = [];
33
37
  let inFence = false;
@@ -38,7 +42,9 @@ function inlineTags(body) {
38
42
  }
39
43
  if (inFence) continue;
40
44
  if (!line.includes('#')) continue; // most lines; skip the regex work
41
- const cleaned = line.includes('`') ? line.replace(INLINE_CODE_RE, (m)=>' '.repeat(m.length)) : line;
45
+ let cleaned = line.includes('`') ? line.replace(INLINE_CODE_RE, (m)=>' '.repeat(m.length)) : line;
46
+ if (cleaned.includes('[[')) cleaned = cleaned.replace(WIKILINK_RE, (m)=>' '.repeat(m.length));
47
+ if (cleaned.includes('<')) cleaned = cleaned.replace(HTML_TAG_RE, (m)=>' '.repeat(m.length));
42
48
  for (const m of cleaned.matchAll(INLINE_TAG_RE)){
43
49
  const tag = normalizeTag(m[1]);
44
50
  if (tag) found.push(tag);
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/features/tags.ts"],"sourcesContent":["import type { Feature } from './types.ts';\n\n// tags(path, tag): Obsidian's file.tags grain -- frontmatter list/string tags plus inline\n// #tags from the prose, deduplicated, source not distinguished. Nested tags store full\n// (book/scifi); `tag = 'book' OR tag LIKE 'book/%'` is how a caller matches the parent too.\n\nconst FENCE_RE = /^(```|~~~)/;\nconst INLINE_CODE_RE = /`[^`]*`/g;\n// Anchors on start-of-line or a preceding whitespace/(/[ so `a#b` and URL fragments don't count.\nconst INLINE_TAG_RE = /(?:^|[\\s([])#([\\p{L}\\p{N}_/-]+)/gu;\n\n// Strips a leading # (frontmatter entries may carry one) and a trailing /; rejects an\n// all-digit result -- a tag needs at least one non-digit character.\nfunction normalizeTag(raw: string): string | null {\n const stripped = raw.replace(/^#/, '').replace(/\\/+$/, '');\n if (!stripped || /^\\d+$/.test(stripped)) return null;\n return stripped;\n}\n\n// data.tags: a YAML list (Obsidian also accepts a bare string). Null members and non-string\n// members are skipped rather than throwing -- `tags:\\n -` parses to [null].\nfunction frontmatterTags(data?: Record<string, unknown>): string[] {\n const raw = data?.tags;\n const list = Array.isArray(raw) ? raw : typeof raw === 'string' ? [raw] : [];\n const found: string[] = [];\n for (const item of list) {\n if (typeof item !== 'string') continue;\n const tag = normalizeTag(item);\n if (tag) found.push(tag);\n }\n return found;\n}\n\n// #tag tokens outside fenced code blocks and inline code spans.\nfunction inlineTags(body: string): string[] {\n const found: string[] = [];\n let inFence = false;\n for (const line of body.split('\\n')) {\n if (FENCE_RE.test(line)) {\n inFence = !inFence;\n continue;\n }\n if (inFence) continue;\n if (!line.includes('#')) continue; // most lines; skip the regex work\n const cleaned = line.includes('`') ? line.replace(INLINE_CODE_RE, (m) => ' '.repeat(m.length)) : line;\n for (const m of cleaned.matchAll(INLINE_TAG_RE)) {\n const tag = normalizeTag(m[1]);\n if (tag) found.push(tag);\n }\n }\n return found;\n}\n\nfunction extract(_raw: string, body: string, _search?: { title: string; summary: string }, data?: Record<string, unknown>): string[] {\n return [...new Set([...frontmatterTags(data), ...inlineTags(body)])].sort();\n}\n\nexport const tags: Feature = {\n name: 'tags',\n schema(db) {\n db.exec('CREATE TABLE IF NOT EXISTS tags (\"path\" TEXT, tag TEXT, PRIMARY KEY (\"path\", tag))');\n db.exec('CREATE INDEX IF NOT EXISTS tags_tag ON tags(tag)');\n },\n extract,\n // Per-file rows with nothing else to resolve, so only a vanished file needs a delete here;\n // store() below handles a reparse's stale rows itself.\n remove(db, path, delta) {\n if (!delta.vanished.includes(path)) return;\n db.prepare('DELETE FROM tags WHERE \"path\" = ?').run(path);\n },\n store(db, path, extracted) {\n const found = extracted as string[];\n if (found.length === 0) {\n db.prepare('DELETE FROM tags WHERE \"path\" = ?').run(path);\n } else {\n const placeholders = found.map(() => '?').join(', ');\n db.prepare(`DELETE FROM tags WHERE \"path\" = ? AND tag NOT IN (${placeholders})`).run(path, ...found);\n }\n const insert = db.prepare('INSERT OR IGNORE INTO tags (\"path\", tag) VALUES (?, ?)');\n for (const tag of found) insert.run(path, tag);\n },\n};\n"],"names":["FENCE_RE","INLINE_CODE_RE","INLINE_TAG_RE","normalizeTag","raw","stripped","replace","test","frontmatterTags","data","tags","list","Array","isArray","found","item","tag","push","inlineTags","body","inFence","line","split","includes","cleaned","m","repeat","length","matchAll","extract","_raw","_search","Set","sort","name","schema","db","exec","remove","path","delta","vanished","prepare","run","store","extracted","placeholders","map","join","insert"],"mappings":"AAEA,0FAA0F;AAC1F,uFAAuF;AACvF,4FAA4F;AAE5F,MAAMA,WAAW;AACjB,MAAMC,iBAAiB;AACvB,iGAAiG;AACjG,MAAMC,gBAAgB;AAEtB,sFAAsF;AACtF,oEAAoE;AACpE,SAASC,aAAaC,GAAW;IAC/B,MAAMC,WAAWD,IAAIE,OAAO,CAAC,MAAM,IAAIA,OAAO,CAAC,QAAQ;IACvD,IAAI,CAACD,YAAY,QAAQE,IAAI,CAACF,WAAW,OAAO;IAChD,OAAOA;AACT;AAEA,4FAA4F;AAC5F,6EAA6E;AAC7E,SAASG,gBAAgBC,IAA8B;IACrD,MAAML,MAAMK,iBAAAA,2BAAAA,KAAMC,IAAI;IACtB,MAAMC,OAAOC,MAAMC,OAAO,CAACT,OAAOA,MAAM,OAAOA,QAAQ,WAAW;QAACA;KAAI,GAAG,EAAE;IAC5E,MAAMU,QAAkB,EAAE;IAC1B,KAAK,MAAMC,QAAQJ,KAAM;QACvB,IAAI,OAAOI,SAAS,UAAU;QAC9B,MAAMC,MAAMb,aAAaY;QACzB,IAAIC,KAAKF,MAAMG,IAAI,CAACD;IACtB;IACA,OAAOF;AACT;AAEA,gEAAgE;AAChE,SAASI,WAAWC,IAAY;IAC9B,MAAML,QAAkB,EAAE;IAC1B,IAAIM,UAAU;IACd,KAAK,MAAMC,QAAQF,KAAKG,KAAK,CAAC,MAAO;QACnC,IAAItB,SAASO,IAAI,CAACc,OAAO;YACvBD,UAAU,CAACA;YACX;QACF;QACA,IAAIA,SAAS;QACb,IAAI,CAACC,KAAKE,QAAQ,CAAC,MAAM,UAAU,kCAAkC;QACrE,MAAMC,UAAUH,KAAKE,QAAQ,CAAC,OAAOF,KAAKf,OAAO,CAACL,gBAAgB,CAACwB,IAAM,IAAIC,MAAM,CAACD,EAAEE,MAAM,KAAKN;QACjG,KAAK,MAAMI,KAAKD,QAAQI,QAAQ,CAAC1B,eAAgB;YAC/C,MAAMc,MAAMb,aAAasB,CAAC,CAAC,EAAE;YAC7B,IAAIT,KAAKF,MAAMG,IAAI,CAACD;QACtB;IACF;IACA,OAAOF;AACT;AAEA,SAASe,QAAQC,IAAY,EAAEX,IAAY,EAAEY,OAA4C,EAAEtB,IAA8B;IACvH,OAAO;WAAI,IAAIuB,IAAI;eAAIxB,gBAAgBC;eAAUS,WAAWC;SAAM;KAAE,CAACc,IAAI;AAC3E;AAEA,OAAO,MAAMvB,OAAgB;IAC3BwB,MAAM;IACNC,QAAOC,EAAE;QACPA,GAAGC,IAAI,CAAC;QACRD,GAAGC,IAAI,CAAC;IACV;IACAR;IACA,2FAA2F;IAC3F,uDAAuD;IACvDS,QAAOF,EAAE,EAAEG,IAAI,EAAEC,KAAK;QACpB,IAAI,CAACA,MAAMC,QAAQ,CAAClB,QAAQ,CAACgB,OAAO;QACpCH,GAAGM,OAAO,CAAC,qCAAqCC,GAAG,CAACJ;IACtD;IACAK,OAAMR,EAAE,EAAEG,IAAI,EAAEM,SAAS;QACvB,MAAM/B,QAAQ+B;QACd,IAAI/B,MAAMa,MAAM,KAAK,GAAG;YACtBS,GAAGM,OAAO,CAAC,qCAAqCC,GAAG,CAACJ;QACtD,OAAO;YACL,MAAMO,eAAehC,MAAMiC,GAAG,CAAC,IAAM,KAAKC,IAAI,CAAC;YAC/CZ,GAAGM,OAAO,CAAC,CAAC,kDAAkD,EAAEI,aAAa,CAAC,CAAC,EAAEH,GAAG,CAACJ,SAASzB;QAChG;QACA,MAAMmC,SAASb,GAAGM,OAAO,CAAC;QAC1B,KAAK,MAAM1B,OAAOF,MAAOmC,OAAON,GAAG,CAACJ,MAAMvB;IAC5C;AACF,EAAE"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/features/tags.ts"],"sourcesContent":["import type { Feature } from './types.ts';\n\n// tags(path, tag): Obsidian's file.tags grain -- frontmatter list/string tags plus inline\n// #tags from the prose, deduplicated, source not distinguished. Nested tags store full\n// (book/scifi); `tag = 'book' OR tag LIKE 'book/%'` is how a caller matches the parent too.\n\nconst FENCE_RE = /^(```|~~~)/;\nconst INLINE_CODE_RE = /`[^`]*`/g;\n// Obsidian treats [[#Heading]] as a same-note link, not a tag.\nconst WIKILINK_RE = /\\[\\[.*?\\]\\]/g; // to the first ]], so a heading holding a lone ] still masks\n// Obsidian doesn't read tags inside HTML markup.\nconst HTML_TAG_RE = /<\\/?[a-zA-Z][^>]*>/g; // tag-shaped only: a comparison's `< 5` must not open a span\n// Anchors on start-of-line or a preceding whitespace/(/[ so `a#b` and URL fragments don't count.\nconst INLINE_TAG_RE = /(?:^|[\\s([])#([\\p{L}\\p{N}_/-]+)/gu;\n\n// Strips a leading # (frontmatter entries may carry one) and a trailing /; rejects an\n// all-digit result -- a tag needs at least one non-digit character.\nfunction normalizeTag(raw: string): string | null {\n const stripped = raw.replace(/^#/, '').replace(/\\/+$/, '');\n if (!stripped || /^\\d+$/.test(stripped)) return null;\n return stripped;\n}\n\n// data.tags: a YAML list (Obsidian also accepts a bare string). Null members and non-string\n// members are skipped rather than throwing -- `tags:\\n -` parses to [null].\nfunction frontmatterTags(data?: Record<string, unknown>): string[] {\n const raw = data?.tags;\n const list = Array.isArray(raw) ? raw : typeof raw === 'string' ? [raw] : [];\n const found: string[] = [];\n for (const item of list) {\n if (typeof item !== 'string') continue;\n const tag = normalizeTag(item);\n if (tag) found.push(tag);\n }\n return found;\n}\n\n// #tag tokens outside fenced code blocks, inline code spans, wikilinks, and HTML tags.\nfunction inlineTags(body: string): string[] {\n const found: string[] = [];\n let inFence = false;\n for (const line of body.split('\\n')) {\n if (FENCE_RE.test(line)) {\n inFence = !inFence;\n continue;\n }\n if (inFence) continue;\n if (!line.includes('#')) continue; // most lines; skip the regex work\n let cleaned = line.includes('`') ? line.replace(INLINE_CODE_RE, (m) => ' '.repeat(m.length)) : line;\n if (cleaned.includes('[[')) cleaned = cleaned.replace(WIKILINK_RE, (m) => ' '.repeat(m.length));\n if (cleaned.includes('<')) cleaned = cleaned.replace(HTML_TAG_RE, (m) => ' '.repeat(m.length));\n for (const m of cleaned.matchAll(INLINE_TAG_RE)) {\n const tag = normalizeTag(m[1]);\n if (tag) found.push(tag);\n }\n }\n return found;\n}\n\nfunction extract(_raw: string, body: string, _search?: { title: string; summary: string }, data?: Record<string, unknown>): string[] {\n return [...new Set([...frontmatterTags(data), ...inlineTags(body)])].sort();\n}\n\nexport const tags: Feature = {\n name: 'tags',\n schema(db) {\n db.exec('CREATE TABLE IF NOT EXISTS tags (\"path\" TEXT, tag TEXT, PRIMARY KEY (\"path\", tag))');\n db.exec('CREATE INDEX IF NOT EXISTS tags_tag ON tags(tag)');\n },\n extract,\n // Per-file rows with nothing else to resolve, so only a vanished file needs a delete here;\n // store() below handles a reparse's stale rows itself.\n remove(db, path, delta) {\n if (!delta.vanished.includes(path)) return;\n db.prepare('DELETE FROM tags WHERE \"path\" = ?').run(path);\n },\n store(db, path, extracted) {\n const found = extracted as string[];\n if (found.length === 0) {\n db.prepare('DELETE FROM tags WHERE \"path\" = ?').run(path);\n } else {\n const placeholders = found.map(() => '?').join(', ');\n db.prepare(`DELETE FROM tags WHERE \"path\" = ? AND tag NOT IN (${placeholders})`).run(path, ...found);\n }\n const insert = db.prepare('INSERT OR IGNORE INTO tags (\"path\", tag) VALUES (?, ?)');\n for (const tag of found) insert.run(path, tag);\n },\n};\n"],"names":["FENCE_RE","INLINE_CODE_RE","WIKILINK_RE","HTML_TAG_RE","INLINE_TAG_RE","normalizeTag","raw","stripped","replace","test","frontmatterTags","data","tags","list","Array","isArray","found","item","tag","push","inlineTags","body","inFence","line","split","includes","cleaned","m","repeat","length","matchAll","extract","_raw","_search","Set","sort","name","schema","db","exec","remove","path","delta","vanished","prepare","run","store","extracted","placeholders","map","join","insert"],"mappings":"AAEA,0FAA0F;AAC1F,uFAAuF;AACvF,4FAA4F;AAE5F,MAAMA,WAAW;AACjB,MAAMC,iBAAiB;AACvB,+DAA+D;AAC/D,MAAMC,cAAc,gBAAgB,6DAA6D;AACjG,iDAAiD;AACjD,MAAMC,cAAc,uBAAuB,6DAA6D;AACxG,iGAAiG;AACjG,MAAMC,gBAAgB;AAEtB,sFAAsF;AACtF,oEAAoE;AACpE,SAASC,aAAaC,GAAW;IAC/B,MAAMC,WAAWD,IAAIE,OAAO,CAAC,MAAM,IAAIA,OAAO,CAAC,QAAQ;IACvD,IAAI,CAACD,YAAY,QAAQE,IAAI,CAACF,WAAW,OAAO;IAChD,OAAOA;AACT;AAEA,4FAA4F;AAC5F,6EAA6E;AAC7E,SAASG,gBAAgBC,IAA8B;IACrD,MAAML,MAAMK,iBAAAA,2BAAAA,KAAMC,IAAI;IACtB,MAAMC,OAAOC,MAAMC,OAAO,CAACT,OAAOA,MAAM,OAAOA,QAAQ,WAAW;QAACA;KAAI,GAAG,EAAE;IAC5E,MAAMU,QAAkB,EAAE;IAC1B,KAAK,MAAMC,QAAQJ,KAAM;QACvB,IAAI,OAAOI,SAAS,UAAU;QAC9B,MAAMC,MAAMb,aAAaY;QACzB,IAAIC,KAAKF,MAAMG,IAAI,CAACD;IACtB;IACA,OAAOF;AACT;AAEA,uFAAuF;AACvF,SAASI,WAAWC,IAAY;IAC9B,MAAML,QAAkB,EAAE;IAC1B,IAAIM,UAAU;IACd,KAAK,MAAMC,QAAQF,KAAKG,KAAK,CAAC,MAAO;QACnC,IAAIxB,SAASS,IAAI,CAACc,OAAO;YACvBD,UAAU,CAACA;YACX;QACF;QACA,IAAIA,SAAS;QACb,IAAI,CAACC,KAAKE,QAAQ,CAAC,MAAM,UAAU,kCAAkC;QACrE,IAAIC,UAAUH,KAAKE,QAAQ,CAAC,OAAOF,KAAKf,OAAO,CAACP,gBAAgB,CAAC0B,IAAM,IAAIC,MAAM,CAACD,EAAEE,MAAM,KAAKN;QAC/F,IAAIG,QAAQD,QAAQ,CAAC,OAAOC,UAAUA,QAAQlB,OAAO,CAACN,aAAa,CAACyB,IAAM,IAAIC,MAAM,CAACD,EAAEE,MAAM;QAC7F,IAAIH,QAAQD,QAAQ,CAAC,MAAMC,UAAUA,QAAQlB,OAAO,CAACL,aAAa,CAACwB,IAAM,IAAIC,MAAM,CAACD,EAAEE,MAAM;QAC5F,KAAK,MAAMF,KAAKD,QAAQI,QAAQ,CAAC1B,eAAgB;YAC/C,MAAMc,MAAMb,aAAasB,CAAC,CAAC,EAAE;YAC7B,IAAIT,KAAKF,MAAMG,IAAI,CAACD;QACtB;IACF;IACA,OAAOF;AACT;AAEA,SAASe,QAAQC,IAAY,EAAEX,IAAY,EAAEY,OAA4C,EAAEtB,IAA8B;IACvH,OAAO;WAAI,IAAIuB,IAAI;eAAIxB,gBAAgBC;eAAUS,WAAWC;SAAM;KAAE,CAACc,IAAI;AAC3E;AAEA,OAAO,MAAMvB,OAAgB;IAC3BwB,MAAM;IACNC,QAAOC,EAAE;QACPA,GAAGC,IAAI,CAAC;QACRD,GAAGC,IAAI,CAAC;IACV;IACAR;IACA,2FAA2F;IAC3F,uDAAuD;IACvDS,QAAOF,EAAE,EAAEG,IAAI,EAAEC,KAAK;QACpB,IAAI,CAACA,MAAMC,QAAQ,CAAClB,QAAQ,CAACgB,OAAO;QACpCH,GAAGM,OAAO,CAAC,qCAAqCC,GAAG,CAACJ;IACtD;IACAK,OAAMR,EAAE,EAAEG,IAAI,EAAEM,SAAS;QACvB,MAAM/B,QAAQ+B;QACd,IAAI/B,MAAMa,MAAM,KAAK,GAAG;YACtBS,GAAGM,OAAO,CAAC,qCAAqCC,GAAG,CAACJ;QACtD,OAAO;YACL,MAAMO,eAAehC,MAAMiC,GAAG,CAAC,IAAM,KAAKC,IAAI,CAAC;YAC/CZ,GAAGM,OAAO,CAAC,CAAC,kDAAkD,EAAEI,aAAa,CAAC,CAAC,EAAEH,GAAG,CAACJ,SAASzB;QAChG;QACA,MAAMmC,SAASb,GAAGM,OAAO,CAAC;QAC1B,KAAK,MAAM1B,OAAOF,MAAOmC,OAAON,GAAG,CAACJ,MAAMvB;IAC5C;AACF,EAAE"}
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "sensemaking",
3
- "version": "0.13.2",
3
+ "version": "0.14.0",
4
4
  "description": "Query and search your markdown notes with context-aware progressive disclosure: SQL over frontmatter, links, and text, plus semantic search and link-graph ranking. No server, no build step",
5
5
  "keywords": [
6
6
  "markdown",
@@ -61,6 +61,7 @@ The commands are shorthands over those tables; anything they don't express, SQL
61
61
 
62
62
  ```
63
63
  sense sql "SELECT name FROM pragma_table_info('frontmatter')" # what fields exist
64
+ sense sql 'SELECT path FROM frontmatter WHERE "plugin-id" IS NOT NULL' # punctuated keys need double quotes: unquoted, plugin-id reads as subtraction
64
65
  sense sql "SELECT DISTINCT status FROM frontmatter" # what values a field takes
65
66
  sense sql "SELECT src FROM links WHERE dst = ?" notes/pricing-model.md # backlinks
66
67
  sense sql "SELECT f.path FROM frontmatter f JOIN scope ON scope.path = f.path" --preset default # scope SQL to a preset
@@ -111,7 +112,7 @@ Worked traces: [EXAMPLES.md](EXAMPLES.md).
111
112
  - `map` and `status` report each preset's coverage (files matched, embedded count). Indexing derives from presets, so the coverage numbers are how you see what a config actually indexes and embeds. A scope with fewer signals just uses fewer (a semantic-off preset searches lexically); a saved search naming an unknown preset errors when run, listing the declared ones.
112
113
  - Save a query into `sense.config.json` only when it will be reused; run ad-hoc otherwise.
113
114
  - A one-line `summary:` per note is optional and pays twice: it appears in result rows and is a weighted search field. Date comparisons work for dates written as ISO 8601 (`2026-08-12`, or with time and offset); other formats do not compare. Field names in examples (`status`, `tags`, `created`) are illustrative; your tree defines its own.
114
- - Reserved frontmatter keys (dropped with a warning): `path`, `_mtime`, `_ctime`, `_size`, `_rank`, `_parse_error`, `content`, `links`, `sections`. The `tags` frontmatter column and the `tags` table coexist: the column holds the raw YAML list, the table the merged frontmatter+inline set.
115
+ - Reserved frontmatter keys (dropped with a warning): `path`, `_mtime`, `_ctime`, `_size`, `_rank`, `_parse_error`, `content`, `links`, `sections`. The `tags` frontmatter column and the `tags` table coexist, mirroring Obsidian's own split: the column is the raw YAML list one note's frontmatter declares (Obsidian's `tags` property), the table is the merged, deduplicated frontmatter+inline set per note (what Obsidian's tag pane and Bases' `file.tags` read). "What is tagged X" is a table query; the column answers only what a note's frontmatter literally says. Inline tags inside `%%...%%` comments are indexed -- some trees run their whole maintenance-tag system in comments.
115
116
  - A note whose frontmatter does not parse is indexed with **no** frontmatter columns and `_parse_error` set to the YAML message, which carries the line. Nothing is half-recovered: a non-NULL value is a value the author wrote. So a NULL column means the key was absent *or* the note did not parse, and `_parse_error` is how you tell: `WHERE status IS NULL AND _parse_error IS NULL` is "genuinely missing status". List what needs fixing with `sense sql "SELECT path, _parse_error FROM frontmatter WHERE _parse_error IS NOT NULL"`; fixing a file clears it on the next command. `sense status` reports the count.
116
117
  - Exit codes: `0` ok, `1` error (SQLite message verbatim), `2` usage (unknown query, wrong param count).
117
118
  - Doubted cache: delete the directory `sense status` prints on its `cache:` line. Rarely needed; every query reconciles first.
@@ -30,7 +30,7 @@ A preset is a named, self-contained bundle of settings. `default` (required) is
30
30
  - Presets may overlap; they are views, not partitions.
31
31
  - Global `features` (`links`, `sections`, `rank`) still toggle tree-wide; most trees never touch them.
32
32
 
33
- **Vectors take two decisions, in two places.** The top-level `"embed": { "model", "type": "static"|"api", "url", "key" }` block names the model and says whether the tree has vectors at all. A preset's `semantic` says whether that scope uses them: a layer searched for exact wording (ingested sources, archives, generated output) sets `semantic: false`, costs no embedding, and its searches run on words and links. That is the main scale lever, and it is the llm-wiki split: compiled pages searched by meaning, raw sources searched for the phrasing you are citing. `static` is the built-in pure-JS Model2Vec loader and handles paraphrase and reworded concepts; tight domain jargon ("heart attack" for "myocardial infarction") is where an `api` transformer model tends to do better, measured in BENCHMARKING.md, "Retrieval quality". Nothing downloads the model implicitly: `sense download` fetches it once per machine into the cache (`$XDG_CACHE_HOME/sensemaking/models`, else `~/.cache/...`), one directory per model, so several models coexist and switching between them rebuilds the index rather than mixing vector spaces. A `model` holding a path instead of a Hugging Face id points at a local directory, which `sense download` reports as nothing to fetch. The first search after that embeds the tree (progress on stderr; minutes on tens of thousands of notes, seconds on small trees).
33
+ **Vectors take two decisions, in two places.** The top-level `"embed": { "model", "type": "static"|"api", "url", "key" }` block names the model and says whether the tree has vectors at all. A preset's `semantic` says whether that scope uses them: a layer searched for exact wording (ingested sources, archives, generated output) sets `semantic: false`, costs no embedding, and its searches run on words and links. That is the main scale lever, and it is the llm-wiki split: compiled pages searched by meaning, raw sources searched for the phrasing you are citing. `static` is the built-in pure-JS Model2Vec loader and handles paraphrase and reworded concepts; tight domain jargon ("heart attack" for "myocardial infarction") is where an `api` transformer model tends to do better, measured in BENCHMARKING.md, "Retrieval quality". Nothing downloads the model implicitly: `sense download` fetches it once per machine into the cache (`$XDG_CACHE_HOME/sensemaking/models`, else `~/.cache/...`), one directory per model, so several models coexist and switching between them rebuilds the index rather than mixing vector spaces. A `model` holding a path instead of a Hugging Face id points at a local directory, which `sense download` reports as nothing to fetch. The first search after that embeds the tree (progress on stderr; minutes on tens of thousands of notes, seconds on small trees). A config edit that changes coverage or features rebuilds the cache, vectors included, so settle presets before the first semantic search on a large tree or that embedding run is paid twice.
34
34
 
35
35
  **Large vaults**: everything except the vector build is measured linear to 100k notes with no tuning (BENCHMARKING.md). The knobs that matter are `k` (more, smaller results; rows carry `lines` section ranges, so agents read sections, not files) and `semantic: false` on the layers that do not earn vectors.
36
36