sensemaking 0.21.1 → 0.22.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +1 -1
- package/dist/cjs/chunk/group.js +13 -9
- package/dist/cjs/chunk/group.js.map +1 -1
- package/dist/cjs/commands/search.js +1 -1
- package/dist/cjs/commands/search.js.map +1 -1
- package/dist/cjs/features/embed.js +9 -1
- package/dist/cjs/features/embed.js.map +1 -1
- package/dist/cjs/features/links.js +30 -8
- package/dist/cjs/features/links.js.map +1 -1
- package/dist/cjs/features/sections.js +13 -5
- package/dist/cjs/features/sections.js.map +1 -1
- package/dist/cjs/features/tags.js +7 -4
- package/dist/cjs/features/tags.js.map +1 -1
- package/dist/cjs/index.d.cts +1 -1
- package/dist/cjs/index.d.ts +1 -1
- package/dist/cjs/index.js.map +1 -1
- package/dist/cjs/scan/pool.d.cts +1 -0
- package/dist/cjs/scan/pool.d.ts +1 -0
- package/dist/cjs/scan/pool.js +2 -1
- package/dist/cjs/scan/pool.js.map +1 -1
- package/dist/cjs/scan/reparse.d.cts +1 -0
- package/dist/cjs/scan/reparse.d.ts +1 -0
- package/dist/cjs/scan/reparse.js +6 -3
- package/dist/cjs/scan/reparse.js.map +1 -1
- package/dist/cjs/store/builder.d.cts +2 -0
- package/dist/cjs/store/builder.d.ts +2 -0
- package/dist/cjs/store/builder.js.map +1 -1
- package/dist/cjs/store/duckdb/connection.js +120 -0
- package/dist/cjs/store/duckdb/connection.js.map +1 -1
- package/dist/cjs/store/duckdb/reconcile.js +11 -113
- package/dist/cjs/store/duckdb/reconcile.js.map +1 -1
- package/dist/cjs/store/index.d.cts +1 -5
- package/dist/cjs/store/index.d.ts +1 -5
- package/dist/cjs/store/index.js +9 -46
- package/dist/cjs/store/index.js.map +1 -1
- package/dist/cjs/store/lock-wait.d.cts +2 -0
- package/dist/cjs/store/lock-wait.d.ts +2 -0
- package/dist/cjs/store/lock-wait.js +51 -0
- package/dist/cjs/store/lock-wait.js.map +1 -0
- package/dist/cjs/store/open.d.cts +2 -4
- package/dist/cjs/store/open.d.ts +2 -4
- package/dist/cjs/store/open.js +29 -66
- package/dist/cjs/store/open.js.map +1 -1
- package/dist/cjs/store/reconcile.d.cts +2 -0
- package/dist/cjs/store/reconcile.d.ts +2 -0
- package/dist/cjs/store/reconcile.js +384 -255
- package/dist/cjs/store/reconcile.js.map +1 -1
- package/dist/cjs/store/shared.d.cts +1 -0
- package/dist/cjs/store/shared.d.ts +1 -0
- package/dist/cjs/store/shared.js +41 -0
- package/dist/cjs/store/shared.js.map +1 -1
- package/dist/cjs/store/sqlite/lexical.js +1 -1
- package/dist/cjs/store/sqlite/lexical.js.map +1 -1
- package/dist/cjs/store/sqlite/store.js +0 -1
- package/dist/cjs/store/sqlite/store.js.map +1 -1
- package/dist/cjs/store/stages.d.cts +18 -0
- package/dist/cjs/store/stages.d.ts +18 -0
- package/dist/cjs/store/stages.js +372 -0
- package/dist/cjs/store/stages.js.map +1 -0
- package/dist/cjs/store/turso/store.js.map +1 -1
- package/dist/cjs/store/types.d.cts +2 -2
- package/dist/cjs/store/types.d.ts +2 -2
- package/dist/cjs/store/types.js +0 -1
- package/dist/cjs/store/types.js.map +1 -1
- package/dist/cjs/text/segment.js +7 -5
- package/dist/cjs/text/segment.js.map +1 -1
- package/dist/cjs/watch.js +162 -107
- package/dist/cjs/watch.js.map +1 -1
- package/dist/cjs/workers/parse.d.cts +1 -0
- package/dist/cjs/workers/parse.d.ts +1 -0
- package/dist/cjs/workers/parse.js +4 -1
- package/dist/cjs/workers/parse.js.map +1 -1
- package/dist/esm/chunk/group.js +13 -9
- package/dist/esm/chunk/group.js.map +1 -1
- package/dist/esm/commands/search.js +1 -1
- package/dist/esm/commands/search.js.map +1 -1
- package/dist/esm/features/embed.js +9 -1
- package/dist/esm/features/embed.js.map +1 -1
- package/dist/esm/features/links.js +26 -9
- package/dist/esm/features/links.js.map +1 -1
- package/dist/esm/features/sections.js +13 -3
- package/dist/esm/features/sections.js.map +1 -1
- package/dist/esm/features/tags.js +7 -2
- package/dist/esm/features/tags.js.map +1 -1
- package/dist/esm/index.d.ts +1 -1
- package/dist/esm/index.js.map +1 -1
- package/dist/esm/scan/pool.d.ts +1 -0
- package/dist/esm/scan/pool.js +2 -1
- package/dist/esm/scan/pool.js.map +1 -1
- package/dist/esm/scan/reparse.d.ts +1 -0
- package/dist/esm/scan/reparse.js +5 -2
- package/dist/esm/scan/reparse.js.map +1 -1
- package/dist/esm/store/builder.d.ts +2 -0
- package/dist/esm/store/builder.js.map +1 -1
- package/dist/esm/store/duckdb/connection.js +35 -0
- package/dist/esm/store/duckdb/connection.js.map +1 -1
- package/dist/esm/store/duckdb/reconcile.js +12 -37
- package/dist/esm/store/duckdb/reconcile.js.map +1 -1
- package/dist/esm/store/index.d.ts +1 -5
- package/dist/esm/store/index.js +6 -28
- package/dist/esm/store/index.js.map +1 -1
- package/dist/esm/store/lock-wait.d.ts +2 -0
- package/dist/esm/store/lock-wait.js +36 -0
- package/dist/esm/store/lock-wait.js.map +1 -0
- package/dist/esm/store/open.d.ts +2 -4
- package/dist/esm/store/open.js +17 -32
- package/dist/esm/store/open.js.map +1 -1
- package/dist/esm/store/reconcile.d.ts +2 -0
- package/dist/esm/store/reconcile.js +71 -44
- package/dist/esm/store/reconcile.js.map +1 -1
- package/dist/esm/store/shared.d.ts +1 -0
- package/dist/esm/store/shared.js +8 -0
- package/dist/esm/store/shared.js.map +1 -1
- package/dist/esm/store/sqlite/lexical.js +1 -1
- package/dist/esm/store/sqlite/lexical.js.map +1 -1
- package/dist/esm/store/sqlite/store.js +0 -1
- package/dist/esm/store/sqlite/store.js.map +1 -1
- package/dist/esm/store/stages.d.ts +18 -0
- package/dist/esm/store/stages.js +63 -0
- package/dist/esm/store/stages.js.map +1 -0
- package/dist/esm/store/turso/store.js +1 -1
- package/dist/esm/store/turso/store.js.map +1 -1
- package/dist/esm/store/types.d.ts +2 -2
- package/dist/esm/store/types.js +0 -1
- package/dist/esm/store/types.js.map +1 -1
- package/dist/esm/text/segment.js +7 -5
- package/dist/esm/text/segment.js.map +1 -1
- package/dist/esm/watch.js +63 -38
- package/dist/esm/watch.js.map +1 -1
- package/dist/esm/workers/parse.d.ts +1 -0
- package/dist/esm/workers/parse.js +4 -1
- package/dist/esm/workers/parse.js.map +1 -1
- package/package.json +1 -1
- package/skills/sense/SKILL.md +1 -1
- package/skills/sense-setup/SKILL.md +1 -1
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/features/sections.ts"],"sourcesContent":["import type { Block } from '../chunk/index.ts';\nimport { estimateTokens, parse } from '../chunk/index.ts';\nimport { countLines } from '../scan/frontmatter.ts';\nimport type { Feature } from './types.ts';\n\n// sections(path, idx, level, heading, start_line, end_line, tokens): the heading outline,\n// 1-indexed over the raw file so a row is a direct Read range; tokens is estimateTokens (D5), CJK-aware.\n\nexport interface Section {\n level: number;\n heading: string;\n startLine: number;\n endLine: number;\n tokens: number;\n}\n\n// Heading blocks parse() already found (mdast/CommonMark fences), offset back onto the raw file;\n// a section runs to just before the next heading, or EOF.\nfunction sectionsFromBlocks(blocks: Block[], raw: string, body: string): Section[] {\n const rawLines = raw.split('\\n');\n const offset = rawLines.length - countLines(body);\n const found: Section[] = blocks.filter((b) => b.type === 'heading').map((b) => ({ level: b.depth ?? 1, heading: (b.text ?? '').trim(), startLine: b.startLine + offset, endLine: rawLines.length, tokens: 0 }));\n for (let s = 0; s < found.length; s++) {\n if (s + 1 < found.length) found[s].endLine = found[s + 1].startLine - 1;\n const text = rawLines.slice(found[s].startLine - 1, found[s].endLine).join('\\n');\n found[s].tokens = Math.ceil(estimateTokens(text));\n }\n return found;\n}\n\nexport const sections: Feature = {\n name: 'sections',\n async schema(db) {\n await db.exec(`CREATE TABLE IF NOT EXISTS sections (\"path\" TEXT, idx INTEGER, level INTEGER, heading TEXT, start_line INTEGER, end_line INTEGER, tokens INTEGER, PRIMARY KEY (\"path\", idx))`);\n },\n extract(raw, body, _search, _data, _cfg, blocks) {\n return sectionsFromBlocks(blocks ?? parse(body), raw, body);\n },\n async remove(db, paths) {\n if (paths.length === 0) return;\n await db.runBatch(\n 'DELETE FROM sections WHERE \"path\" = ?',\n paths.map((p) => [p])\n );\n },\n async store(db, docs) {\n const rows: unknown[][] = [];\n for (const { path, extracted } of docs) (extracted as Section[]).forEach((s, idx) => rows.push([path, idx, s.level, s.heading, s.startLine, s.endLine, s.tokens]));\n
|
|
1
|
+
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/features/sections.ts"],"sourcesContent":["import type { Block } from '../chunk/index.ts';\nimport { estimateTokens, parse } from '../chunk/index.ts';\nimport { countLines } from '../scan/frontmatter.ts';\nimport { appendRows } from '../store/shared.ts';\nimport type { Feature } from './types.ts';\n\n// sections(path, idx, level, heading, start_line, end_line, tokens): the heading outline,\n// 1-indexed over the raw file so a row is a direct Read range; tokens is estimateTokens (D5), CJK-aware.\n\nexport interface Section {\n level: number;\n heading: string;\n startLine: number;\n endLine: number;\n tokens: number;\n}\n\n// Heading blocks parse() already found (mdast/CommonMark fences), offset back onto the raw file;\n// a section runs to just before the next heading, or EOF.\nfunction sectionsFromBlocks(blocks: Block[], raw: string, body: string): Section[] {\n const rawLines = raw.split('\\n');\n const offset = rawLines.length - countLines(body);\n const found: Section[] = blocks.filter((b) => b.type === 'heading').map((b) => ({ level: b.depth ?? 1, heading: (b.text ?? '').trim(), startLine: b.startLine + offset, endLine: rawLines.length, tokens: 0 }));\n for (let s = 0; s < found.length; s++) {\n if (s + 1 < found.length) found[s].endLine = found[s + 1].startLine - 1;\n const text = rawLines.slice(found[s].startLine - 1, found[s].endLine).join('\\n');\n found[s].tokens = Math.ceil(estimateTokens(text));\n }\n return found;\n}\n\nconst SECTION_COLUMNS = ['path', 'idx', 'level', 'heading', 'start_line', 'end_line', 'tokens'];\n\nexport const sections: Feature = {\n name: 'sections',\n async schema(db) {\n await db.exec(`CREATE TABLE IF NOT EXISTS sections (\"path\" TEXT, idx INTEGER, level INTEGER, heading TEXT, start_line INTEGER, end_line INTEGER, tokens INTEGER, PRIMARY KEY (\"path\", idx))`);\n },\n extract(raw, body, _search, _data, _cfg, blocks) {\n return sectionsFromBlocks(blocks ?? parse(body), raw, body);\n },\n async remove(db, paths) {\n if (paths.length === 0) return;\n await db.runBatch(\n 'DELETE FROM sections WHERE \"path\" = ?',\n paths.map((p) => [p])\n );\n },\n async store(db, docs) {\n const rows: unknown[][] = [];\n for (const { path, extracted } of docs) (extracted as Section[]).forEach((s, idx) => rows.push([path, idx, s.level, s.heading, s.startLine, s.endLine, s.tokens]));\n // DO NOTHING, not a bare INSERT: reconcile's added/touched split is decided before this\n // write's lock, so a path called \"added\" here can already have this row from a concurrent\n // reconcile that committed first -- same file, same parse, same row. A store with an append\n // path has no second writer, and remove() cleared every touched path above, so it skips the guard.\n await appendRows(db, 'sections', SECTION_COLUMNS, 'INSERT INTO sections (\"path\", idx, level, heading, start_line, end_line, tokens) VALUES (?, ?, ?, ?, ?, ?, ?) ON CONFLICT(\"path\", idx) DO NOTHING', rows);\n },\n};\n"],"names":["sections","sectionsFromBlocks","blocks","raw","body","rawLines","split","offset","length","countLines","found","filter","b","type","map","level","depth","heading","text","trim","startLine","endLine","tokens","s","slice","join","Math","ceil","estimateTokens","SECTION_COLUMNS","name","schema","db","exec","extract","_search","_data","_cfg","parse","remove","paths","runBatch","p","store","docs","rows","path","extracted","forEach","idx","push","appendRows"],"mappings":";;;;+BAiCaA;;;eAAAA;;;uBAhCyB;6BACX;wBACA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAc3B,iGAAiG;AACjG,0DAA0D;AAC1D,SAASC,mBAAmBC,MAAe,EAAEC,GAAW,EAAEC,IAAY;IACpE,IAAMC,WAAWF,IAAIG,KAAK,CAAC;IAC3B,IAAMC,SAASF,SAASG,MAAM,GAAGC,IAAAA,yBAAU,EAACL;IAC5C,IAAMM,QAAmBR,OAAOS,MAAM,CAAC,SAACC;eAAMA,EAAEC,IAAI,KAAK;OAAWC,GAAG,CAAC,SAACF;YAAgBA,UAAwBA;eAAjC;YAAEG,KAAK,GAAEH,WAAAA,EAAEI,KAAK,cAAPJ,sBAAAA,WAAW;YAAGK,SAAS,EAACL,UAAAA,EAAEM,IAAI,cAANN,qBAAAA,UAAU,IAAIO,IAAI;YAAIC,WAAWR,EAAEQ,SAAS,GAAGb;YAAQc,SAAShB,SAASG,MAAM;YAAEc,QAAQ;QAAE;;IAC5M,IAAK,IAAIC,IAAI,GAAGA,IAAIb,MAAMF,MAAM,EAAEe,IAAK;QACrC,IAAIA,IAAI,IAAIb,MAAMF,MAAM,EAAEE,KAAK,CAACa,EAAE,CAACF,OAAO,GAAGX,KAAK,CAACa,IAAI,EAAE,CAACH,SAAS,GAAG;QACtE,IAAMF,OAAOb,SAASmB,KAAK,CAACd,KAAK,CAACa,EAAE,CAACH,SAAS,GAAG,GAAGV,KAAK,CAACa,EAAE,CAACF,OAAO,EAAEI,IAAI,CAAC;QAC3Ef,KAAK,CAACa,EAAE,CAACD,MAAM,GAAGI,KAAKC,IAAI,CAACC,IAAAA,uBAAc,EAACV;IAC7C;IACA,OAAOR;AACT;AAEA,IAAMmB,kBAAkB;IAAC;IAAQ;IAAO;IAAS;IAAW;IAAc;IAAY;CAAS;AAExF,IAAM7B,WAAoB;IAC/B8B,MAAM;IACAC,QAAN,SAAMA,OAAOC,EAAE;;;;;wBACb;;4BAAMA,GAAGC,IAAI,CAAC;;;wBAAd;;;;;;QACF;;IACAC,SAAAA,SAAAA,QAAQ/B,GAAG,EAAEC,IAAI,EAAE+B,OAAO,EAAEC,KAAK,EAAEC,IAAI,EAAEnC,MAAM;QAC7C,OAAOD,mBAAmBC,mBAAAA,oBAAAA,SAAUoC,IAAAA,cAAK,EAAClC,OAAOD,KAAKC;IACxD;IACMmC,QAAN,SAAMA,OAAOP,EAAE,EAAEQ,KAAK;;;;;wBACpB,IAAIA,MAAMhC,MAAM,KAAK,GAAG;;;wBACxB;;4BAAMwB,GAAGS,QAAQ,CACf,yCACAD,MAAM1B,GAAG,CAAC,SAAC4B;uCAAM;oCAACA;iCAAE;;;;wBAFtB;;;;;;QAIF;;IACMC,OAAN,SAAMA,MAAMX,EAAE,EAAEY,IAAI;;gBACZC,MACD,2BAAA,mBAAA,uBAAA,WAAA;;;;wBADCA;wBACD,kCAAA,2BAAA;;;gCAAA,kBAAA,aAAQC,mBAAAA,MAAMC,wBAAAA;gCAAsBA,UAAwBC,OAAO,CAAC,SAACzB,GAAG0B;2CAAQJ,KAAKK,IAAI,CAAC;wCAACJ;wCAAMG;wCAAK1B,EAAER,KAAK;wCAAEQ,EAAEN,OAAO;wCAAEM,EAAEH,SAAS;wCAAEG,EAAEF,OAAO;wCAAEE,EAAED,MAAM;qCAAC;;;4BAAhK,IAAK,YAA6BsB,2BAA7B,6BAAA,QAAA,yBAAA;;4BAAA;4BAAA;;;qCAAA,6BAAA;oCAAA;;;oCAAA;0CAAA;;;;wBACL,wFAAwF;wBACxF,0FAA0F;wBAC1F,4FAA4F;wBAC5F,mGAAmG;wBACnG;;4BAAMO,IAAAA,oBAAU,EAACnB,IAAI,YAAYH,iBAAiB,qJAAqJgB;;;wBAAvM;;;;;;QACF;;AACF"}
|
|
@@ -8,6 +8,7 @@ Object.defineProperty(exports, "tags", {
|
|
|
8
8
|
return tags;
|
|
9
9
|
}
|
|
10
10
|
});
|
|
11
|
+
var _sharedts = require("../store/shared.js");
|
|
11
12
|
var _fencests = require("./fences.js");
|
|
12
13
|
function _array_like_to_array(arr, len) {
|
|
13
14
|
if (len == null || len > arr.length) len = arr.length;
|
|
@@ -372,12 +373,14 @@ var tags = {
|
|
|
372
373
|
}
|
|
373
374
|
}
|
|
374
375
|
}
|
|
375
|
-
|
|
376
|
-
|
|
377
|
-
];
|
|
376
|
+
// OR IGNORE guards the same concurrent-reconcile race sections.ts documents; remove() cleared
|
|
377
|
+
// every touched path above, so a store with an append path needs no guard.
|
|
378
378
|
return [
|
|
379
379
|
4,
|
|
380
|
-
|
|
380
|
+
(0, _sharedts.appendRows)(db, 'tags', [
|
|
381
|
+
'path',
|
|
382
|
+
'tag'
|
|
383
|
+
], 'INSERT OR IGNORE INTO tags ("path", tag) VALUES (?, ?)', rows)
|
|
381
384
|
];
|
|
382
385
|
case 1:
|
|
383
386
|
_state.sent();
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/features/tags.ts"],"sourcesContent":["import { maskRegions } from './fences.ts';\nimport type { Feature } from './types.ts';\n\n// tags(path, tag): Obsidian's file.tags grain -- frontmatter list/string tags plus inline\n// #tags, deduplicated, source not distinguished. Nested tags store full (book/scifi); `tag = 'book' OR tag LIKE 'book/%'` matches the parent too.\n\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// A markdown link destination `](...)` -- `[text](#anchor)` is a same-page link, not a tag.\nconst LINK_DEST_RE = /\\]\\((?:[^()]|\\([^()]*\\))*\\)/g; // one paren-nesting level, as CommonMark destinations allow: (https://x/a_(b)#frag)\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 the regions maskRegions() masks (fences, code spans, wikilinks, HTML,\n// comments, link destinations); this loop only does per-line tag extraction.\nfunction inlineTags(body: string): string[] {\n const found: string[] = [];\n for (const line of maskRegions(body).split('\\n')) {\n if (!line.includes('#')) continue; // most lines; skip the regex work\n let cleaned = line;\n if (cleaned.includes('[[')) cleaned = cleaned.replace(WIKILINK_RE, (m) => ' '.repeat(m.length));\n if (cleaned.includes('](')) cleaned = cleaned.replace(LINK_DEST_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 async schema(db) {\n await db.exec('CREATE TABLE IF NOT EXISTS tags (\"path\" TEXT, tag TEXT, PRIMARY KEY (\"path\", tag))');\n await db.exec('CREATE INDEX IF NOT EXISTS tags_tag ON tags(tag)');\n },\n extract,\n // Blanket clear-then-reinsert (store() always writes the full found list back), so remove()\n // covers vanished and reparsed files without per-file NOT IN diffing, which cannot be one statement shared across files.\n async remove(db, paths) {\n if (paths.length === 0) return;\n await db.runBatch(\n 'DELETE FROM tags WHERE \"path\" = ?',\n paths.map((p) => [p])\n );\n },\n async store(db, docs) {\n const rows: unknown[][] = [];\n for (const { path, extracted } of docs) for (const tag of extracted as string[]) rows.push([path, tag]);\n
|
|
1
|
+
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/features/tags.ts"],"sourcesContent":["import { appendRows } from '../store/shared.ts';\nimport { maskRegions } from './fences.ts';\nimport type { Feature } from './types.ts';\n\n// tags(path, tag): Obsidian's file.tags grain -- frontmatter list/string tags plus inline\n// #tags, deduplicated, source not distinguished. Nested tags store full (book/scifi); `tag = 'book' OR tag LIKE 'book/%'` matches the parent too.\n\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// A markdown link destination `](...)` -- `[text](#anchor)` is a same-page link, not a tag.\nconst LINK_DEST_RE = /\\]\\((?:[^()]|\\([^()]*\\))*\\)/g; // one paren-nesting level, as CommonMark destinations allow: (https://x/a_(b)#frag)\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 the regions maskRegions() masks (fences, code spans, wikilinks, HTML,\n// comments, link destinations); this loop only does per-line tag extraction.\nfunction inlineTags(body: string): string[] {\n const found: string[] = [];\n for (const line of maskRegions(body).split('\\n')) {\n if (!line.includes('#')) continue; // most lines; skip the regex work\n let cleaned = line;\n if (cleaned.includes('[[')) cleaned = cleaned.replace(WIKILINK_RE, (m) => ' '.repeat(m.length));\n if (cleaned.includes('](')) cleaned = cleaned.replace(LINK_DEST_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 async schema(db) {\n await db.exec('CREATE TABLE IF NOT EXISTS tags (\"path\" TEXT, tag TEXT, PRIMARY KEY (\"path\", tag))');\n await db.exec('CREATE INDEX IF NOT EXISTS tags_tag ON tags(tag)');\n },\n extract,\n // Blanket clear-then-reinsert (store() always writes the full found list back), so remove()\n // covers vanished and reparsed files without per-file NOT IN diffing, which cannot be one statement shared across files.\n async remove(db, paths) {\n if (paths.length === 0) return;\n await db.runBatch(\n 'DELETE FROM tags WHERE \"path\" = ?',\n paths.map((p) => [p])\n );\n },\n async store(db, docs) {\n const rows: unknown[][] = [];\n for (const { path, extracted } of docs) for (const tag of extracted as string[]) rows.push([path, tag]);\n // OR IGNORE guards the same concurrent-reconcile race sections.ts documents; remove() cleared\n // every touched path above, so a store with an append path needs no guard.\n await appendRows(db, 'tags', ['path', 'tag'], 'INSERT OR IGNORE INTO tags (\"path\", tag) VALUES (?, ?)', rows);\n },\n};\n"],"names":["tags","WIKILINK_RE","HTML_TAG_RE","INLINE_TAG_RE","LINK_DEST_RE","normalizeTag","raw","stripped","replace","test","frontmatterTags","data","list","Array","isArray","found","item","tag","push","inlineTags","body","maskRegions","split","line","includes","cleaned","m","repeat","length","matchAll","extract","_raw","_search","Set","sort","name","schema","db","exec","remove","paths","runBatch","map","p","store","docs","rows","path","extracted","appendRows"],"mappings":";;;;+BA4DaA;;;eAAAA;;;wBA5Dc;wBACC;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAG5B,0FAA0F;AAC1F,kJAAkJ;AAElJ,+DAA+D;AAC/D,IAAMC,cAAc,gBAAgB,6DAA6D;AACjG,iDAAiD;AACjD,IAAMC,cAAc,uBAAuB,6DAA6D;AACxG,iGAAiG;AACjG,IAAMC,gBAAgB;AACtB,4FAA4F;AAC5F,IAAMC,eAAe,gCAAgC,oFAAoF;AAEzI,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,KAAMX,IAAI;IACtB,IAAMY,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,4FAA4F;AAC5F,6EAA6E;AAC7E,SAASI,WAAWC,IAAY;IAC9B,IAAML,QAAkB,EAAE;QACrB,kCAAA,2BAAA;;QAAL,QAAK,YAAcM,IAAAA,qBAAW,EAACD,MAAME,KAAK,CAAC,0BAAtC,SAAA,6BAAA,QAAA,yBAAA,iCAA6C;YAA7C,IAAMC,OAAN;YACH,IAAI,CAACA,KAAKC,QAAQ,CAAC,MAAM,UAAU,kCAAkC;YACrE,IAAIC,UAAUF;YACd,IAAIE,QAAQD,QAAQ,CAAC,OAAOC,UAAUA,QAAQjB,OAAO,CAACP,aAAa,SAACyB;uBAAM,IAAIC,MAAM,CAACD,EAAEE,MAAM;;YAC7F,IAAIH,QAAQD,QAAQ,CAAC,OAAOC,UAAUA,QAAQjB,OAAO,CAACJ,cAAc,SAACsB;uBAAM,IAAIC,MAAM,CAACD,EAAEE,MAAM;;YAC9F,IAAIH,QAAQD,QAAQ,CAAC,MAAMC,UAAUA,QAAQjB,OAAO,CAACN,aAAa,SAACwB;uBAAM,IAAIC,MAAM,CAACD,EAAEE,MAAM;;gBACvF,mCAAA,4BAAA;;gBAAL,QAAK,aAAWH,QAAQI,QAAQ,CAAC1B,mCAA5B,UAAA,8BAAA,SAAA,0BAAA,kCAA4C;oBAA5C,IAAMuB,IAAN;oBACH,IAAMT,MAAMZ,aAAaqB,CAAC,CAAC,EAAE;oBAC7B,IAAIT,KAAKF,MAAMG,IAAI,CAACD;gBACtB;;gBAHK;gBAAA;;;yBAAA,8BAAA;wBAAA;;;wBAAA;8BAAA;;;;QAIP;;QAVK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAWL,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,IAAMlC,OAAgB;IAC3BmC,MAAM;IACAC,QAAN,SAAMA,OAAOC,EAAE;;;;;wBACb;;4BAAMA,GAAGC,IAAI,CAAC;;;wBAAd;wBACA;;4BAAMD,GAAGC,IAAI,CAAC;;;wBAAd;;;;;;QACF;;IACAR,SAAAA;IAGMS,QAFN,4FAA4F;IAC5F,yHAAyH;IACzH,SAAMA,OAAOF,EAAE,EAAEG,KAAK;;;;;wBACpB,IAAIA,MAAMZ,MAAM,KAAK,GAAG;;;wBACxB;;4BAAMS,GAAGI,QAAQ,CACf,qCACAD,MAAME,GAAG,CAAC,SAACC;uCAAM;oCAACA;iCAAE;;;;wBAFtB;;;;;;QAIF;;IACMC,OAAN,SAAMA,MAAMP,EAAE,EAAEQ,IAAI;;gBACZC,MACuC,2BAAA,mBAAA,gBAAxC,4BAAA,oBAAA,iBAAA,WAAA,oBAAQC,MAAMC,WAA0B,YAAA,QAAM/B;;;;wBAD7C6B;wBACuC,kCAAA,2BAAA,4BAAxC,mCAAA,4BAAA;;4BAAL,IAAK,YAA6BD,2BAA7B,8BAAA,QAAA,yBAAA;8CAAA,aAAQE,mBAAAA,MAAMC,wBAAAA;;oCAAqB,IAAK,aAAaA,gCAAb,6BAAA,SAAA,0BAAA;wCAAM/B,MAAN;wCAAoC6B,KAAK5B,IAAI;4CAAE6B;4CAAM9B;;;;oCAArD;oCAAA;;;6CAAA,6BAAA;4CAAA;;;4CAAA;kDAAA;;;;;;4BAAxC;4BAAA;;;qCAAA,8BAAA;oCAAA;;;oCAAA;0CAAA;;;;wBACL,8FAA8F;wBAC9F,2EAA2E;wBAC3E;;4BAAMgC,IAAAA,oBAAU,EAACZ,IAAI;gCAAS;gCAAQ;+BAAQ,0DAA0DS;;;wBAAxG;;;;;;QACF;;AACF"}
|
package/dist/cjs/index.d.cts
CHANGED
|
@@ -6,7 +6,7 @@ export type { SenseErrorCode } from './errors.js';
|
|
|
6
6
|
export { SenseError } from './errors.js';
|
|
7
7
|
export type { Format, Row, RowFormat } from './output/output.js';
|
|
8
8
|
export { printRows } from './output/output.js';
|
|
9
|
-
export type { OpenResult } from './store/index.js';
|
|
9
|
+
export type { OpenResult, Stages } from './store/index.js';
|
|
10
10
|
export { CAPABILITY_NAMES, clearCache, openStore as open } from './store/index.js';
|
|
11
11
|
export type { Capability, DocumentStore, FieldStat, LexicalHit, LexicalIndex, LexicalQueryOptions, RawStatement, RunResult, SqlSession, Statement, Store, VectorCandidate, VectorSimilar, VectorStore, VectorWriteRow } from './store/types.js';
|
|
12
12
|
export type { WatchEvent, WatchOptions } from './watch.js';
|
package/dist/cjs/index.d.ts
CHANGED
|
@@ -6,7 +6,7 @@ export type { SenseErrorCode } from './errors.js';
|
|
|
6
6
|
export { SenseError } from './errors.js';
|
|
7
7
|
export type { Format, Row, RowFormat } from './output/output.js';
|
|
8
8
|
export { printRows } from './output/output.js';
|
|
9
|
-
export type { OpenResult } from './store/index.js';
|
|
9
|
+
export type { OpenResult, Stages } from './store/index.js';
|
|
10
10
|
export { CAPABILITY_NAMES, clearCache, openStore as open } from './store/index.js';
|
|
11
11
|
export type { Capability, DocumentStore, FieldStat, LexicalHit, LexicalIndex, LexicalQueryOptions, RawStatement, RunResult, SqlSession, Statement, Store, VectorCandidate, VectorSimilar, VectorStore, VectorWriteRow } from './store/types.js';
|
|
12
12
|
export type { WatchEvent, WatchOptions } from './watch.js';
|
package/dist/cjs/index.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/index.ts"],"sourcesContent":["// Public library API. Deliberately small: every export is a stability promise;\n// internals (feature registry, graph, scan, meta) stay module-private.\n\nexport type { Peek, PresetCoverage, SearchOptions, TreeMap } from './commands/index.ts';\nexport { mapTree, peek, presetCoverage, search } from './commands/index.ts';\nexport type { Config, EmbedConfig, FeatureName, InitOverrides, Preset, ResolvedConfig, SavedQuery, SavedSearch, SearchOverrides, SignalName, SignalWeights, StoreName } from './config/index.ts';\nexport { CONFIG_FILENAME, initConfig, loadConfig, migrateConfig, SIGNAL_NAMES, STATE_DIR, STORE_NAMES, SUPPORTED_CONFIG_VERSION } from './config/index.ts';\nexport type { SenseErrorCode } from './errors.ts';\nexport { SenseError } from './errors.ts';\nexport type { Format, Row, RowFormat } from './output/output.ts';\nexport { printRows } from './output/output.ts';\nexport type { OpenResult } from './store/index.ts';\nexport { CAPABILITY_NAMES, clearCache, openStore as open } from './store/index.ts';\nexport type { Capability, DocumentStore, FieldStat, LexicalHit, LexicalIndex, LexicalQueryOptions, RawStatement, RunResult, SqlSession, Statement, Store, VectorCandidate, VectorSimilar, VectorStore, VectorWriteRow } from './store/types.ts';\n\nexport type { WatchEvent, WatchOptions } from './watch.ts';\nexport { runWatch } from './watch.ts';\n"],"names":["CAPABILITY_NAMES","CONFIG_FILENAME","SIGNAL_NAMES","STATE_DIR","STORE_NAMES","SUPPORTED_CONFIG_VERSION","SenseError","clearCache","initConfig","loadConfig","mapTree","migrateConfig","open","openStore","peek","presetCoverage","printRows","runWatch","search"],"mappings":"AAAA,+EAA+E;AAC/E,uEAAuE;;;;;;;;;;;;QAW9DA;eAAAA,0BAAgB;;QANhBC;eAAAA,yBAAe;;QAAyCC;eAAAA,sBAAY;;QAAEC;eAAAA,mBAAS;;QAAEC;eAAAA,qBAAW;;QAAEC;eAAAA,kCAAwB;;QAEtHC;eAAAA,oBAAU;;QAIQC;eAAAA,oBAAU;;QANXC;eAAAA,oBAAU;;QAAEC;eAAAA,oBAAU;;QAFvCC;eAAAA,gBAAO;;QAEkCC;eAAAA,uBAAa;;QAMXC;eAAbC,mBAAS;;QAR9BC;eAAAA,aAAI;;QAAEC;eAAAA,uBAAc;;QAM7BC;eAAAA,mBAAS;;QAMTC;eAAAA,iBAAQ;;QAZuBC;eAAAA,eAAM;;;uBAAQ;wBAEiF;wBAE5G;wBAED;wBAEsC;uBAIvC"}
|
|
1
|
+
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/index.ts"],"sourcesContent":["// Public library API. Deliberately small: every export is a stability promise;\n// internals (feature registry, graph, scan, meta) stay module-private.\n\nexport type { Peek, PresetCoverage, SearchOptions, TreeMap } from './commands/index.ts';\nexport { mapTree, peek, presetCoverage, search } from './commands/index.ts';\nexport type { Config, EmbedConfig, FeatureName, InitOverrides, Preset, ResolvedConfig, SavedQuery, SavedSearch, SearchOverrides, SignalName, SignalWeights, StoreName } from './config/index.ts';\nexport { CONFIG_FILENAME, initConfig, loadConfig, migrateConfig, SIGNAL_NAMES, STATE_DIR, STORE_NAMES, SUPPORTED_CONFIG_VERSION } from './config/index.ts';\nexport type { SenseErrorCode } from './errors.ts';\nexport { SenseError } from './errors.ts';\nexport type { Format, Row, RowFormat } from './output/output.ts';\nexport { printRows } from './output/output.ts';\nexport type { OpenResult, Stages } from './store/index.ts';\nexport { CAPABILITY_NAMES, clearCache, openStore as open } from './store/index.ts';\nexport type { Capability, DocumentStore, FieldStat, LexicalHit, LexicalIndex, LexicalQueryOptions, RawStatement, RunResult, SqlSession, Statement, Store, VectorCandidate, VectorSimilar, VectorStore, VectorWriteRow } from './store/types.ts';\n\nexport type { WatchEvent, WatchOptions } from './watch.ts';\nexport { runWatch } from './watch.ts';\n"],"names":["CAPABILITY_NAMES","CONFIG_FILENAME","SIGNAL_NAMES","STATE_DIR","STORE_NAMES","SUPPORTED_CONFIG_VERSION","SenseError","clearCache","initConfig","loadConfig","mapTree","migrateConfig","open","openStore","peek","presetCoverage","printRows","runWatch","search"],"mappings":"AAAA,+EAA+E;AAC/E,uEAAuE;;;;;;;;;;;;QAW9DA;eAAAA,0BAAgB;;QANhBC;eAAAA,yBAAe;;QAAyCC;eAAAA,sBAAY;;QAAEC;eAAAA,mBAAS;;QAAEC;eAAAA,qBAAW;;QAAEC;eAAAA,kCAAwB;;QAEtHC;eAAAA,oBAAU;;QAIQC;eAAAA,oBAAU;;QANXC;eAAAA,oBAAU;;QAAEC;eAAAA,oBAAU;;QAFvCC;eAAAA,gBAAO;;QAEkCC;eAAAA,uBAAa;;QAMXC;eAAbC,mBAAS;;QAR9BC;eAAAA,aAAI;;QAAEC;eAAAA,uBAAc;;QAM7BC;eAAAA,mBAAS;;QAMTC;eAAAA,iBAAQ;;QAZuBC;eAAAA,eAAM;;;uBAAQ;wBAEiF;wBAE5G;wBAED;wBAEsC;uBAIvC"}
|
package/dist/cjs/scan/pool.d.cts
CHANGED
package/dist/cjs/scan/pool.d.ts
CHANGED
package/dist/cjs/scan/pool.js
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/scan/pool.ts"],"sourcesContent":["import { createRequire } from 'node:module';\nimport { dirname, join } from 'node:path';\n// Type-only: erased at build, but keeps depcheck's usage check satisfied for the tier-2\n// `_require` below (see coding-standards' deferral tiers).\nimport type * as TinypoolNS from 'tinypool';\nimport type { Config } from '../config/index.ts';\nimport type { Feature } from '../features/types.ts';\nimport type { ParsedDoc } from '../scan/index.ts';\nimport type { ParseTask, ParseTaskResult, ParseWorkerData } from '../workers/parse.ts';\nimport type { FileStat } from './list.ts';\nimport { reviveError } from './worker-error.ts';\n\n// Tinypool is ESM-only; our floor (>=22.20) has native require(esm), so the tier-2 house deferral reaches it.\nconst _require = typeof require === 'undefined' ? createRequire(import.meta.url) : require;\n\n// Worker MUST load from dist/cjs/: a worker_threads thread is a fresh realm with no inherited\n// TS loader hook, so a source path cannot work. Resolved on first pooled dispatch, not at import.\nlet workerFile: string | undefined;\nfunction resolveWorkerFile(): string {\n if (workerFile) return workerFile;\n const load = createRequire(import.meta.url);\n for (const rel of ['..', '../..', '../../..']) {\n try {\n if ((load(`${rel}/package.json`) as { name?: string }).name === 'sensemaking') {\n workerFile = join(dirname(load.resolve(`${rel}/package.json`)), 'dist', 'cjs', 'workers', 'parse.js');\n return workerFile;\n }\n } catch {}\n }\n throw new Error('cannot locate the sensemaking package root, so the parse worker cannot be found; run npm run build');\n}\n\nexport type FileResult = { doc: ParsedDoc; warnings: string[] };\n\n// One Tinypool per instance, created lazily on first dispatch and reused by every later call:\n// a builder owns one of these for its whole lifetime instead of paying pool startup per reconcile.\nexport class ParsePool {\n private pool: TinypoolNS.Tinypool | undefined;\n // Tinypools this instance has constructed. One dispatch or a hundred should leave it at 1;\n // it is how a caller, and the specs, observe that a lifetime reuses its pool rather than churning one.\n poolsCreated = 0;\n\n private ensure(workerData: ParseWorkerData, maxWorkers: number): TinypoolNS.Tinypool {\n if (!this.pool) {\n const { Tinypool } = _require('tinypool') as typeof TinypoolNS;\n this.pool = new Tinypool({ filename: resolveWorkerFile(), minThreads: maxWorkers, maxThreads: maxWorkers, workerData });\n this.poolsCreated++;\n }\n return this.pool;\n }\n\n // Never the tree: a worker task carries one FileStat and returns only what parseFile returns --\n // extracted text and feature values, never the mdast tree.\n async run(files: FileStat[], features: Feature[], cfg: Config, onParsed: ((done: number) => void) | undefined, maxWorkers: number): Promise<FileResult[]> {\n // cfg and the feature selection are constant for the dispatch, so they cross once per worker\n // as workerData. Features carry closures and cannot cross at all; only their names do.\n const workerData: ParseWorkerData = { cfg, featureNames: features.map((feature) => feature.name) };\n const pool = this.ensure(workerData, maxWorkers);\n let done = 0;\n // Promise.all over a mapped array is load-bearing: the resolved array keeps `files` order\n // regardless of task completion order, which first-seen column order depends on downstream.\n return Promise.all(\n files.map(async (file): Promise<FileResult> => {\n const result = (await pool.run(file as ParseTask)) as ParseTaskResult;\n if (!result.ok) throw reviveError(result.error);\n onParsed?.(++done);\n return { doc: result.doc, warnings: result.warnings };\n })\n );\n }\n\n // A pool never created costs nothing to destroy.\n async close(): Promise<void> {\n if (!this.pool) return;\n const pool = this.pool;\n this.pool = undefined;\n await pool.destroy();\n }\n}\n"],"names":["ParsePool","_require","require","createRequire","workerFile","resolveWorkerFile","load","rel","name","join","dirname","resolve","Error","poolsCreated","ensure","workerData","maxWorkers","pool","Tinypool","filename","minThreads","maxThreads","run","files","features","cfg","onParsed","done","featureNames","map","feature","Promise","all","file","result","ok","reviveError","error","doc","warnings","close","undefined","destroy"],"mappings":";;;;+
|
|
1
|
+
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/scan/pool.ts"],"sourcesContent":["import { createRequire } from 'node:module';\nimport { dirname, join } from 'node:path';\n// Type-only: erased at build, but keeps depcheck's usage check satisfied for the tier-2\n// `_require` below (see coding-standards' deferral tiers).\nimport type * as TinypoolNS from 'tinypool';\nimport type { Config } from '../config/index.ts';\nimport type { Feature } from '../features/types.ts';\nimport type { ParsedDoc } from '../scan/index.ts';\nimport type { ParseTask, ParseTaskResult, ParseWorkerData } from '../workers/parse.ts';\nimport type { FileStat } from './list.ts';\nimport { reviveError } from './worker-error.ts';\n\n// Tinypool is ESM-only; our floor (>=22.20) has native require(esm), so the tier-2 house deferral reaches it.\nconst _require = typeof require === 'undefined' ? createRequire(import.meta.url) : require;\n\n// Worker MUST load from dist/cjs/: a worker_threads thread is a fresh realm with no inherited\n// TS loader hook, so a source path cannot work. Resolved on first pooled dispatch, not at import.\nlet workerFile: string | undefined;\nfunction resolveWorkerFile(): string {\n if (workerFile) return workerFile;\n const load = createRequire(import.meta.url);\n for (const rel of ['..', '../..', '../../..']) {\n try {\n if ((load(`${rel}/package.json`) as { name?: string }).name === 'sensemaking') {\n workerFile = join(dirname(load.resolve(`${rel}/package.json`)), 'dist', 'cjs', 'workers', 'parse.js');\n return workerFile;\n }\n } catch {}\n }\n throw new Error('cannot locate the sensemaking package root, so the parse worker cannot be found; run npm run build');\n}\n\n// parseMs is worker-side only (see workers/parse.ts); absent on the serial path.\nexport type FileResult = { doc: ParsedDoc; warnings: string[]; parseMs?: number };\n\n// One Tinypool per instance, created lazily on first dispatch and reused by every later call:\n// a builder owns one of these for its whole lifetime instead of paying pool startup per reconcile.\nexport class ParsePool {\n private pool: TinypoolNS.Tinypool | undefined;\n // Tinypools this instance has constructed. One dispatch or a hundred should leave it at 1;\n // it is how a caller, and the specs, observe that a lifetime reuses its pool rather than churning one.\n poolsCreated = 0;\n\n private ensure(workerData: ParseWorkerData, maxWorkers: number): TinypoolNS.Tinypool {\n if (!this.pool) {\n const { Tinypool } = _require('tinypool') as typeof TinypoolNS;\n this.pool = new Tinypool({ filename: resolveWorkerFile(), minThreads: maxWorkers, maxThreads: maxWorkers, workerData });\n this.poolsCreated++;\n }\n return this.pool;\n }\n\n // Never the tree: a worker task carries one FileStat and returns only what parseFile returns --\n // extracted text and feature values, never the mdast tree.\n async run(files: FileStat[], features: Feature[], cfg: Config, onParsed: ((done: number) => void) | undefined, maxWorkers: number): Promise<FileResult[]> {\n // cfg and the feature selection are constant for the dispatch, so they cross once per worker\n // as workerData. Features carry closures and cannot cross at all; only their names do.\n const workerData: ParseWorkerData = { cfg, featureNames: features.map((feature) => feature.name) };\n const pool = this.ensure(workerData, maxWorkers);\n let done = 0;\n // Promise.all over a mapped array is load-bearing: the resolved array keeps `files` order\n // regardless of task completion order, which first-seen column order depends on downstream.\n return Promise.all(\n files.map(async (file): Promise<FileResult> => {\n const result = (await pool.run(file as ParseTask)) as ParseTaskResult;\n if (!result.ok) throw reviveError(result.error);\n onParsed?.(++done);\n return { doc: result.doc, warnings: result.warnings, parseMs: result.parseMs };\n })\n );\n }\n\n // A pool never created costs nothing to destroy.\n async close(): Promise<void> {\n if (!this.pool) return;\n const pool = this.pool;\n this.pool = undefined;\n await pool.destroy();\n }\n}\n"],"names":["ParsePool","_require","require","createRequire","workerFile","resolveWorkerFile","load","rel","name","join","dirname","resolve","Error","poolsCreated","ensure","workerData","maxWorkers","pool","Tinypool","filename","minThreads","maxThreads","run","files","features","cfg","onParsed","done","featureNames","map","feature","Promise","all","file","result","ok","reviveError","error","doc","warnings","parseMs","close","undefined","destroy"],"mappings":";;;;+BAqCaA;;;eAAAA;;;0BArCiB;wBACA;6BASF;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAE5B,8GAA8G;AAC9G,IAAMC,WAAW,OAAOC,YAAY,cAAcC,IAAAA,yBAAa,EAAC,uDAAmBD;AAEnF,8FAA8F;AAC9F,kGAAkG;AAClG,IAAIE;AACJ,SAASC;IACP,IAAID,YAAY,OAAOA;IACvB,IAAME,OAAOH,IAAAA,yBAAa,EAAC;IAC3B,gBAAkB,QAAA;QAAC;QAAM;QAAS;KAAW,OAA3B,mBAA6B;YAApCI,MAAO;QAChB,IAAI;YACF,IAAI,AAACD,KAAK,AAAC,GAAM,OAAJC,KAAI,kBAAsCC,IAAI,KAAK,eAAe;gBAC7EJ,aAAaK,IAAAA,cAAI,EAACC,IAAAA,iBAAO,EAACJ,KAAKK,OAAO,CAAC,AAAC,GAAM,OAAJJ,KAAI,oBAAkB,QAAQ,OAAO,WAAW;gBAC1F,OAAOH;YACT;QACF,EAAE,eAAM,CAAC;IACX;IACA,MAAM,IAAIQ,MAAM;AAClB;AAOO,IAAA,AAAMZ,0BAAN;;aAAMA;gCAAAA;QAEX,2FAA2F;QAC3F,uGAAuG;aACvGa,eAAe;;iBAJJb;IAMX,OAAQc,MAOP,GAPD,SAAQA,OAAOC,UAA2B,EAAEC,UAAkB;QAC5D,IAAI,CAAC,IAAI,CAACC,IAAI,EAAE;YACd,IAAM,AAAEC,WAAajB,SAAS,YAAtBiB;YACR,IAAI,CAACD,IAAI,GAAG,IAAIC,SAAS;gBAAEC,UAAUd;gBAAqBe,YAAYJ;gBAAYK,YAAYL;gBAAYD,YAAAA;YAAW;YACrH,IAAI,CAACF,YAAY;QACnB;QACA,OAAO,IAAI,CAACI,IAAI;IAClB;IAEA,gGAAgG;IAChG,2DAA2D;IAC3D,OAAMK,GAgBL,GAhBD,SAAMA,IAAIC,KAAiB,EAAEC,QAAmB,EAAEC,GAAW,EAAEC,QAA8C,EAAEV,UAAkB;;gBAGzHD,YACAE,MACFU;;gBAJJ,6FAA6F;gBAC7F,uFAAuF;gBACjFZ,aAA8B;oBAAEU,KAAAA;oBAAKG,cAAcJ,SAASK,GAAG,CAAC,SAACC;+BAAYA,QAAQtB,IAAI;;gBAAE;gBAC3FS,OAAO,IAAI,CAACH,MAAM,CAACC,YAAYC;gBACjCW,OAAO;gBACX,0FAA0F;gBAC1F,4FAA4F;gBAC5F;;oBAAOI,QAAQC,GAAG,CAChBT,MAAMM,GAAG,CAAC,SAAOI;;gCACTC;;;;wCAAU;;4CAAMjB,KAAKK,GAAG,CAACW;;;wCAAzBC,SAAU;wCAChB,IAAI,CAACA,OAAOC,EAAE,EAAE,MAAMC,IAAAA,0BAAW,EAACF,OAAOG,KAAK;wCAC9CX,qBAAAA,+BAAAA,SAAW,EAAEC;wCACb;;4CAAO;gDAAEW,KAAKJ,OAAOI,GAAG;gDAAEC,UAAUL,OAAOK,QAAQ;gDAAEC,SAASN,OAAOM,OAAO;4CAAC;;;;wBAC/E;;;;QAEJ;;IAEA,iDAAiD;IACjD,OAAMC,KAKL,GALD,SAAMA;;gBAEExB;;;;wBADN,IAAI,CAAC,IAAI,CAACA,IAAI,EAAE;;;wBACVA,OAAO,IAAI,CAACA,IAAI;wBACtB,IAAI,CAACA,IAAI,GAAGyB;wBACZ;;4BAAMzB,KAAK0B,OAAO;;;wBAAlB;;;;;;QACF;;WAzCW3C"}
|
package/dist/cjs/scan/reparse.js
CHANGED
|
@@ -265,7 +265,7 @@ function reparsePooled(files, features, cfg, onParsed, maxWorkers, pool) {
|
|
|
265
265
|
}
|
|
266
266
|
function reparseFiles(_0, _1, _2, _3, _4) {
|
|
267
267
|
return _async_to_generator(function(files, features, cfg, knownColumns, onParsed) {
|
|
268
|
-
var options, _options_threshold, _options_maxWorkers, threshold, maxWorkers, results, _tmp, docs, warnings, newColumns, seen, _iteratorNormalCompletion, _didIteratorError, _iteratorError, _iterator, _step, _step_value, doc, fileWarnings, _warnings, _iteratorNormalCompletion1, _didIteratorError1, _iteratorError1, _iterator1, _step1, key;
|
|
268
|
+
var options, _options_threshold, _options_maxWorkers, threshold, maxWorkers, results, _tmp, docs, warnings, newColumns, seen, workerParseMs, _iteratorNormalCompletion, _didIteratorError, _iteratorError, _iterator, _step, _step_value, doc, fileWarnings, parseMs, _warnings, _iteratorNormalCompletion1, _didIteratorError1, _iteratorError1, _iterator1, _step1, key;
|
|
269
269
|
var _arguments = arguments;
|
|
270
270
|
return _ts_generator(this, function(_state) {
|
|
271
271
|
switch(_state.label){
|
|
@@ -296,12 +296,14 @@ function reparseFiles(_0, _1, _2, _3, _4) {
|
|
|
296
296
|
warnings = [];
|
|
297
297
|
newColumns = [];
|
|
298
298
|
seen = new Set(knownColumns);
|
|
299
|
+
workerParseMs = 0;
|
|
299
300
|
_iteratorNormalCompletion = true, _didIteratorError = false, _iteratorError = undefined;
|
|
300
301
|
try {
|
|
301
302
|
for(_iterator = results[Symbol.iterator](); !(_iteratorNormalCompletion = (_step = _iterator.next()).done); _iteratorNormalCompletion = true){
|
|
302
|
-
_step_value = _step.value, doc = _step_value.doc, fileWarnings = _step_value.warnings;
|
|
303
|
+
_step_value = _step.value, doc = _step_value.doc, fileWarnings = _step_value.warnings, parseMs = _step_value.parseMs;
|
|
303
304
|
;
|
|
304
305
|
(_warnings = warnings).push.apply(_warnings, _to_consumable_array(fileWarnings));
|
|
306
|
+
workerParseMs += parseMs !== null && parseMs !== void 0 ? parseMs : 0;
|
|
305
307
|
_iteratorNormalCompletion1 = true, _didIteratorError1 = false, _iteratorError1 = undefined;
|
|
306
308
|
try {
|
|
307
309
|
for(_iterator1 = Object.keys(doc.data)[Symbol.iterator](); !(_iteratorNormalCompletion1 = (_step1 = _iterator1.next()).done); _iteratorNormalCompletion1 = true){
|
|
@@ -346,7 +348,8 @@ function reparseFiles(_0, _1, _2, _3, _4) {
|
|
|
346
348
|
{
|
|
347
349
|
docs: docs,
|
|
348
350
|
warnings: warnings,
|
|
349
|
-
newColumns: newColumns
|
|
351
|
+
newColumns: newColumns,
|
|
352
|
+
workerParseMs: workerParseMs
|
|
350
353
|
}
|
|
351
354
|
];
|
|
352
355
|
}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/scan/reparse.ts"],"sourcesContent":["import { availableParallelism } from 'node:os';\nimport type { Config } from '../config/index.ts';\nimport type { Feature } from '../features/types.ts';\nimport type { ParsedDoc } from './index.ts';\nimport { parseFile } from './index.ts';\nimport type { FileStat } from './list.ts';\nimport type { FileResult } from './pool.ts';\nimport { ParsePool } from './pool.ts';\n\n// Store-agnostic per-file parse pass shared by every store's reconcile(). Index-preserving:\n// one pass over `files`, pushed in order.\nexport interface ReparseResult {\n docs: ParsedDoc[];\n warnings: string[];\n // Frontmatter keys not in `knownColumns`, first-seen order -- the order callers ALTER TABLE ADD COLUMN in.\n newColumns: string[];\n}\n\nexport interface ReparseOptions {\n // Overrides DEFAULT_THRESHOLD below, so tests can force either dispatch mode without\n // needing thousands of fixture files. Internal: no caller in src/ passes it.\n threshold?: number;\n // Overrides DEFAULT_MAX_WORKERS below.\n maxWorkers?: number;\n // A builder-owned pool to dispatch through instead of creating and destroying one for this\n // call alone (see store/builder.ts). Absent, reparseFiles manages its own ephemeral pool.\n pool?: ParsePool;\n}\n\n// Measured 2026-08-29 on a 14-core Apple M4 Pro: serial and pooled cross between 160-180 files\n// (pool fixed cost ~85-100ms), so 200 keeps margin. 8 workers ties 10 in a 6,566-file sweep; 12+ trends worse.\nconst DEFAULT_THRESHOLD = 200;\nconst DEFAULT_MAX_WORKERS = 8;\n\n// Per-file feature filtering, shared by the serial loop (reads `features` from the caller) and\n// the worker path (can only pass `cfg` across the thread boundary, so derives `features` itself).\nexport function featuresForFile(features: Feature[], cfg: Config, file: FileStat): Feature[] {\n return features.filter((feature) => !feature.enabledForFile || feature.enabledForFile(cfg, file));\n}\n\nfunction reparseSerial(files: FileStat[], features: Feature[], cfg: Config, onParsed?: (done: number) => void): FileResult[] {\n const results: FileResult[] = [];\n let done = 0;\n for (const file of files) {\n const fileFeatures = featuresForFile(features, cfg, file);\n results.push(parseFile(file, fileFeatures, cfg));\n onParsed?.(++done);\n }\n return results;\n}\n\n// Dispatches through `pool` when the caller owns one (a builder, kept alive across calls); with\n// none given, opens an ephemeral ParsePool for this call alone and always closes it.\nasync function reparsePooled(files: FileStat[], features: Feature[], cfg: Config, onParsed: ((done: number) => void) | undefined, maxWorkers: number, pool: ParsePool | undefined): Promise<FileResult[]> {\n if (pool) return pool.run(files, features, cfg, onParsed, maxWorkers);\n const ephemeral = new ParsePool();\n try {\n const results = await ephemeral.run(files, features, cfg, onParsed, maxWorkers);\n await ephemeral.close();\n return results;\n } catch (err) {\n // Tear the pool down before rethrowing, but a close failure must not mask the parse\n // failure that caused it -- not a `finally`, for that reason.\n await ephemeral.close().catch(() => {});\n throw err;\n }\n}\n\n// `onParsed` receives the running count (1-based) after each file, mirroring a Progress.tick\n// call; pass one in to keep progress reporting working without this module owning a reporter.\nexport async function reparseFiles(files: FileStat[], features: Feature[], cfg: Config, knownColumns: ReadonlySet<string>, onParsed?: (done: number) => void, options: ReparseOptions = {}): Promise<ReparseResult> {\n const threshold = options.threshold ?? DEFAULT_THRESHOLD;\n const maxWorkers = options.maxWorkers ?? Math.min(DEFAULT_MAX_WORKERS, availableParallelism());\n const results = files.length >= threshold ? await reparsePooled(files, features, cfg, onParsed, maxWorkers, options.pool) : reparseSerial(files, features, cfg, onParsed);\n\n const docs: ParsedDoc[] = [];\n const warnings: string[] = [];\n const newColumns: string[] = [];\n const seen = new Set(knownColumns);\n for (const { doc, warnings: fileWarnings } of results) {\n warnings.push(...fileWarnings);\n for (const key of Object.keys(doc.data)) {\n if (!seen.has(key)) {\n seen.add(key);\n newColumns.push(key);\n }\n }\n docs.push(doc);\n }\n\n return { docs, warnings, newColumns };\n}\n"],"names":["featuresForFile","reparseFiles","DEFAULT_THRESHOLD","DEFAULT_MAX_WORKERS","features","cfg","file","filter","feature","enabledForFile","reparseSerial","files","onParsed","results","done","fileFeatures","push","parseFile","reparsePooled","maxWorkers","pool","ephemeral","err","run","ParsePool","close","catch","knownColumns","options","threshold","docs","warnings","newColumns","seen","doc","fileWarnings","key","Math","min","availableParallelism","length","Set","Object","keys","data","has","add"],"mappings":";;;;;;;;;;;
|
|
1
|
+
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/scan/reparse.ts"],"sourcesContent":["import { availableParallelism } from 'node:os';\nimport type { Config } from '../config/index.ts';\nimport type { Feature } from '../features/types.ts';\nimport type { ParsedDoc } from './index.ts';\nimport { parseFile } from './index.ts';\nimport type { FileStat } from './list.ts';\nimport type { FileResult } from './pool.ts';\nimport { ParsePool } from './pool.ts';\n\n// Store-agnostic per-file parse pass shared by every store's reconcile(). Index-preserving:\n// one pass over `files`, pushed in order.\nexport interface ReparseResult {\n docs: ParsedDoc[];\n warnings: string[];\n // Frontmatter keys not in `knownColumns`, first-seen order -- the order callers ALTER TABLE ADD COLUMN in.\n newColumns: string[];\n // Sum of worker-side parseFile time (workers/parse.ts) across files; 0 on the serial path.\n workerParseMs: number;\n}\n\nexport interface ReparseOptions {\n // Overrides DEFAULT_THRESHOLD below, so tests can force either dispatch mode without\n // needing thousands of fixture files. Internal: no caller in src/ passes it.\n threshold?: number;\n // Overrides DEFAULT_MAX_WORKERS below.\n maxWorkers?: number;\n // A builder-owned pool to dispatch through instead of creating and destroying one for this\n // call alone (see store/builder.ts). Absent, reparseFiles manages its own ephemeral pool.\n pool?: ParsePool;\n}\n\n// Measured 2026-08-29 on a 14-core Apple M4 Pro: serial and pooled cross between 160-180 files\n// (pool fixed cost ~85-100ms), so 200 keeps margin. 8 workers ties 10 in a 6,566-file sweep; 12+ trends worse.\nconst DEFAULT_THRESHOLD = 200;\nconst DEFAULT_MAX_WORKERS = 8;\n\n// Per-file feature filtering, shared by the serial loop (reads `features` from the caller) and\n// the worker path (can only pass `cfg` across the thread boundary, so derives `features` itself).\nexport function featuresForFile(features: Feature[], cfg: Config, file: FileStat): Feature[] {\n return features.filter((feature) => !feature.enabledForFile || feature.enabledForFile(cfg, file));\n}\n\nfunction reparseSerial(files: FileStat[], features: Feature[], cfg: Config, onParsed?: (done: number) => void): FileResult[] {\n const results: FileResult[] = [];\n let done = 0;\n for (const file of files) {\n const fileFeatures = featuresForFile(features, cfg, file);\n results.push(parseFile(file, fileFeatures, cfg));\n onParsed?.(++done);\n }\n return results;\n}\n\n// Dispatches through `pool` when the caller owns one (a builder, kept alive across calls); with\n// none given, opens an ephemeral ParsePool for this call alone and always closes it.\nasync function reparsePooled(files: FileStat[], features: Feature[], cfg: Config, onParsed: ((done: number) => void) | undefined, maxWorkers: number, pool: ParsePool | undefined): Promise<FileResult[]> {\n if (pool) return pool.run(files, features, cfg, onParsed, maxWorkers);\n const ephemeral = new ParsePool();\n try {\n const results = await ephemeral.run(files, features, cfg, onParsed, maxWorkers);\n await ephemeral.close();\n return results;\n } catch (err) {\n // Tear the pool down before rethrowing, but a close failure must not mask the parse\n // failure that caused it -- not a `finally`, for that reason.\n await ephemeral.close().catch(() => {});\n throw err;\n }\n}\n\n// `onParsed` receives the running count (1-based) after each file, mirroring a Progress.tick\n// call; pass one in to keep progress reporting working without this module owning a reporter.\nexport async function reparseFiles(files: FileStat[], features: Feature[], cfg: Config, knownColumns: ReadonlySet<string>, onParsed?: (done: number) => void, options: ReparseOptions = {}): Promise<ReparseResult> {\n const threshold = options.threshold ?? DEFAULT_THRESHOLD;\n const maxWorkers = options.maxWorkers ?? Math.min(DEFAULT_MAX_WORKERS, availableParallelism());\n const results = files.length >= threshold ? await reparsePooled(files, features, cfg, onParsed, maxWorkers, options.pool) : reparseSerial(files, features, cfg, onParsed);\n\n const docs: ParsedDoc[] = [];\n const warnings: string[] = [];\n const newColumns: string[] = [];\n const seen = new Set(knownColumns);\n let workerParseMs = 0;\n for (const { doc, warnings: fileWarnings, parseMs } of results) {\n warnings.push(...fileWarnings);\n workerParseMs += parseMs ?? 0;\n for (const key of Object.keys(doc.data)) {\n if (!seen.has(key)) {\n seen.add(key);\n newColumns.push(key);\n }\n }\n docs.push(doc);\n }\n\n return { docs, warnings, newColumns, workerParseMs };\n}\n"],"names":["featuresForFile","reparseFiles","DEFAULT_THRESHOLD","DEFAULT_MAX_WORKERS","features","cfg","file","filter","feature","enabledForFile","reparseSerial","files","onParsed","results","done","fileFeatures","push","parseFile","reparsePooled","maxWorkers","pool","ephemeral","err","run","ParsePool","close","catch","knownColumns","options","threshold","docs","warnings","newColumns","seen","workerParseMs","doc","fileWarnings","parseMs","key","Math","min","availableParallelism","length","Set","Object","keys","data","has","add"],"mappings":";;;;;;;;;;;QAsCgBA;eAAAA;;QAkCMC;eAAAA;;;sBAxEe;uBAIX;sBAGA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAwB1B,+FAA+F;AAC/F,+GAA+G;AAC/G,IAAMC,oBAAoB;AAC1B,IAAMC,sBAAsB;AAIrB,SAASH,gBAAgBI,QAAmB,EAAEC,GAAW,EAAEC,IAAc;IAC9E,OAAOF,SAASG,MAAM,CAAC,SAACC;eAAY,CAACA,QAAQC,cAAc,IAAID,QAAQC,cAAc,CAACJ,KAAKC;;AAC7F;AAEA,SAASI,cAAcC,KAAiB,EAAEP,QAAmB,EAAEC,GAAW,EAAEO,QAAiC;IAC3G,IAAMC,UAAwB,EAAE;IAChC,IAAIC,OAAO;QACN,kCAAA,2BAAA;;QAAL,QAAK,YAAcH,0BAAd,SAAA,6BAAA,QAAA,yBAAA,iCAAqB;YAArB,IAAML,OAAN;YACH,IAAMS,eAAef,gBAAgBI,UAAUC,KAAKC;YACpDO,QAAQG,IAAI,CAACC,IAAAA,kBAAS,EAACX,MAAMS,cAAcV;YAC3CO,qBAAAA,+BAAAA,SAAW,EAAEE;QACf;;QAJK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAKL,OAAOD;AACT;AAEA,gGAAgG;AAChG,qFAAqF;AACrF,SAAeK,cAAcP,KAAiB,EAAEP,QAAmB,EAAEC,GAAW,EAAEO,QAA8C,EAAEO,UAAkB,EAAEC,IAA2B;;YAEzKC,WAEER,SAGCS;;;;oBANT,IAAIF,MAAM;;wBAAOA,KAAKG,GAAG,CAACZ,OAAOP,UAAUC,KAAKO,UAAUO;;oBACpDE,YAAY,IAAIG,iBAAS;;;;;;;;;oBAEb;;wBAAMH,UAAUE,GAAG,CAACZ,OAAOP,UAAUC,KAAKO,UAAUO;;;oBAA9DN,UAAU;oBAChB;;wBAAMQ,UAAUI,KAAK;;;oBAArB;oBACA;;wBAAOZ;;;oBACAS;oBACP,oFAAoF;oBACpF,8DAA8D;oBAC9D;;wBAAMD,UAAUI,KAAK,GAAGC,KAAK,CAAC,YAAO;;;oBAArC;oBACA,MAAMJ;;;;;;;IAEV;;AAIO,SAAerB;wCAAaU,KAAiB,EAAEP,QAAmB,EAAEC,GAAW,EAAEsB,YAAiC,EAAEf,QAAiC;YAAEgB,SAC1IA,oBACCA,qBADbC,WACAV,YACAN,eAEAiB,MACAC,UACAC,YACAC,MACFC,eACC,2BAAA,mBAAA,gBAAA,WAAA,oBAAQC,KAAeC,cAAcC,SACxCN,WAEK,4BAAA,oBAAA,iBAAA,YAAA,QAAMO;;;;;oBAb+IV,UAAAA,oEAA0B,CAAC;oBACjLC,aAAYD,qBAAAA,QAAQC,SAAS,cAAjBD,gCAAAA,qBAAqB1B;oBACjCiB,cAAaS,sBAAAA,QAAQT,UAAU,cAAlBS,iCAAAA,sBAAsBW,KAAKC,GAAG,CAACrC,qBAAqBsC,IAAAA,4BAAoB;yBAC3E9B,CAAAA,MAAM+B,MAAM,IAAIb,SAAQ,GAAxBlB;;;;oBAA4B;;wBAAMO,cAAcP,OAAOP,UAAUC,KAAKO,UAAUO,YAAYS,QAAQR,IAAI;;;2BAA5E;;;;;;2BAAgFV,cAAcC,OAAOP,UAAUC,KAAKO;;;oBAA1JC;oBAEAiB;oBACAC;oBACAC;oBACAC,OAAO,IAAIU,IAAIhB;oBACjBO,gBAAgB;oBACf,kCAAA,2BAAA;;wBAAL,IAAK,YAAkDrB,8BAAlD,6BAAA,QAAA,yBAAA,iCAA2D;0CAA3D,aAAQsB,kBAAAA,KAAeC,2BAAVL,UAAwBM,sBAAAA;;4BACxCN,CAAAA,YAAAA,UAASf,IAAI,OAAbe,WAAc,qBAAGK;4BACjBF,iBAAiBG,oBAAAA,qBAAAA,UAAW;4BACvB,mCAAA,4BAAA;;gCAAL,IAAK,aAAaO,OAAOC,IAAI,CAACV,IAAIW,IAAI,wBAAjC,8BAAA,SAAA,0BAAA,kCAAoC;oCAA9BR,MAAN;oCACH,IAAI,CAACL,KAAKc,GAAG,CAACT,MAAM;wCAClBL,KAAKe,GAAG,CAACV;wCACTN,WAAWhB,IAAI,CAACsB;oCAClB;gCACF;;gCALK;gCAAA;;;yCAAA,8BAAA;wCAAA;;;wCAAA;8CAAA;;;;4BAMLR,KAAKd,IAAI,CAACmB;wBACZ;;wBAVK;wBAAA;;;iCAAA,6BAAA;gCAAA;;;gCAAA;sCAAA;;;;oBAYL;;wBAAO;4BAAEL,MAAAA;4BAAMC,UAAAA;4BAAUC,YAAAA;4BAAYE,eAAAA;wBAAc;;;;IACrD"}
|
|
@@ -1,11 +1,13 @@
|
|
|
1
1
|
import type { Config } from '../config/index.js';
|
|
2
2
|
import type { EmbedChangeKind } from './embed-scope.js';
|
|
3
3
|
import type { FeatureToggle } from './feature-scope.js';
|
|
4
|
+
import type { Stages } from './stages.js';
|
|
4
5
|
import type { Connection, ReconcileDialect } from './types.js';
|
|
5
6
|
export interface Builder {
|
|
6
7
|
build(forcedPaths?: ReadonlySet<string>): Promise<{
|
|
7
8
|
parsed: number;
|
|
8
9
|
warnings: string[];
|
|
10
|
+
stages: Stages;
|
|
9
11
|
}>;
|
|
10
12
|
invalidate(kind: EmbedChangeKind): Promise<{
|
|
11
13
|
parsed: number;
|
|
@@ -1,11 +1,13 @@
|
|
|
1
1
|
import type { Config } from '../config/index.js';
|
|
2
2
|
import type { EmbedChangeKind } from './embed-scope.js';
|
|
3
3
|
import type { FeatureToggle } from './feature-scope.js';
|
|
4
|
+
import type { Stages } from './stages.js';
|
|
4
5
|
import type { Connection, ReconcileDialect } from './types.js';
|
|
5
6
|
export interface Builder {
|
|
6
7
|
build(forcedPaths?: ReadonlySet<string>): Promise<{
|
|
7
8
|
parsed: number;
|
|
8
9
|
warnings: string[];
|
|
10
|
+
stages: Stages;
|
|
9
11
|
}>;
|
|
10
12
|
invalidate(kind: EmbedChangeKind): Promise<{
|
|
11
13
|
parsed: number;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/builder.ts"],"sourcesContent":["import type { Config } from '../config/index.ts';\nimport { embed } from '../features/embed.ts';\nimport { FEATURES } from '../features/index.ts';\nimport { rank } from '../features/rank.ts';\nimport type { ReconcileDelta } from '../features/types.ts';\nimport { listFiles } from '../scan/index.ts';\nimport { ParsePool } from '../scan/pool.ts';\nimport { reparseFiles } from '../scan/reparse.ts';\nimport type { EmbedChangeKind } from './embed-scope.ts';\nimport type { FeatureToggle } from './feature-scope.ts';\nimport { NARROW_FEATURE_TABLE } from './feature-scope.ts';\nimport { reconcile } from './reconcile.ts';\nimport { getColumns } from './shared.ts';\nimport { withTransaction } from './transaction.ts';\nimport type { Connection, ReconcileDialect } from './types.ts';\n\n// Owns bringing a store's index current: the write half `Store` (types.ts) deliberately does not\n// carry. A one-shot open calls build() once; a watcher calls it repeatedly on the same instance.\nexport interface Builder {\n // forcedPaths (open.ts's preset-only narrow rebuild) applies to this call alone, not future ones.\n build(forcedPaths?: ReadonlySet<string>): Promise<{ parsed: number; warnings: string[] }>;\n // Narrow embed invalidation (open.ts, embed-scope.ts's classifyEmbedChange): 'model' nulls every\n // vector/scale in place; 'chunk' rebuilds every embedding row. Neither touches another table.\n invalidate(kind: EmbedChangeKind): Promise<{ parsed: number }>;\n // Narrow feature-toggle invalidation (open.ts, feature-scope.ts's classifyFeatureToggles): a\n // toggle drops (or, turning back on, fully re-derives) that feature's own table; rank instead\n // nulls or recomputes its `_rank` column, since rows written while a feature was off are\n // untrustworthy (reconcile.ts's activeFeatures skips a disabled feature's hooks entirely).\n invalidateFeatures(toggles: FeatureToggle[]): Promise<{ parsed: number }>;\n // Releases the parse worker pool, if this lifetime ever created one; never the connection.\n close(): Promise<void>;\n // Pools this lifetime has constructed, so reuse is observable rather than inferred.\n readonly poolsCreated: number;\n}\n\n// 'model': the provider or model name moved but chunk boundaries did not, so only the vector\n// values are stale. No reparse, and no table but embeddings is touched.\nasync function nullEmbedVectors(conn: Connection, dialect: ReconcileDialect): Promise<{ parsed: number }> {\n await withTransaction(conn, () => conn.exec('UPDATE embeddings SET vector = NULL, scale = NULL'), dialect.beginMode());\n return { parsed: 0 };\n}\n\n// 'chunk': chunkTokens or the chunk version moved, so chunk boundaries genuinely changed. Rebuilds\n// every embedding row through the embed feature's own remove/store, touching no other table.\nasync function rebuildEmbeddings(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect, pool: ParsePool): Promise<{ parsed: number }> {\n const existingStmt = await conn.prepare('SELECT \"path\" FROM frontmatter');\n const existingPaths = new Set((await existingStmt.all()).map((r) => (r as { path: string }).path));\n // A file added since the last build has no row yet; leaving it for build() to insert avoids a\n // primary-key collision between this INSERT and build()'s own insert for the same new file.\n const files = listFiles(cfg, baseDir).filter((f) => f.embed && existingPaths.has(f.relPath));\n if (files.length === 0) return { parsed: 0 };\n\n const { docs } = await reparseFiles(files, [embed], cfg, new Set(), undefined, { pool });\n const paths = docs.map((d) => d.relPath);\n const delta: ReconcileDelta = { files, reparsed: paths, added: [], vanished: [] };\n await withTransaction(\n conn,\n async () => {\n await embed.remove?.(conn, paths, delta);\n await embed.store?.(\n conn,\n docs.map((d) => ({ path: d.relPath, extracted: d.extracted.embed })),\n delta\n );\n },\n dialect.beginMode()\n );\n return { parsed: docs.length };\n}\n\n// One feature's own table, dropped and (turning on) fully re-derived across every indexed file.\n// Rows left over from before this feature was disabled are untrustworthy by construction (files\n// added, edited, or deleted got no remove/store for it), so this never diffs against them --\n// dropping every row and reparsing the whole tree is the only safe path back to consistent state.\nasync function invalidateFeatureTable(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect, pool: ParsePool, toggle: FeatureToggle): Promise<{ parsed: number }> {\n const table = NARROW_FEATURE_TABLE[toggle.name];\n const feature = FEATURES.find((f) => f.name === toggle.name);\n if (!feature) throw new Error(`invalidateFeatureTable: no registered feature named \"${toggle.name}\"`);\n\n if (!toggle.turnedOn) {\n await withTransaction(conn, () => conn.exec(`DELETE FROM ${table}`), dialect.beginMode());\n return { parsed: 0 };\n }\n\n const files = listFiles(cfg, baseDir);\n const { docs } = await reparseFiles(files, [feature], cfg, new Set(), undefined, { pool });\n const docsForFeature = docs.map((d) => ({ path: d.relPath, extracted: d.extracted[feature.name] }));\n const paths = docs.map((d) => d.relPath);\n // Every path counts as added, not just touched: the table was just dropped, so there is nothing\n // stale to diff against, and links' own afterReconcile takes its cold-build (resolveAll) path\n // only when added.length equals files.length.\n const delta: ReconcileDelta = { files, reparsed: paths, added: paths, vanished: [] };\n await withTransaction(\n conn,\n async () => {\n await conn.exec(`DELETE FROM ${table}`);\n await feature.store?.(conn, docsForFeature, delta);\n await feature.afterReconcile?.(conn, delta);\n },\n dialect.beginMode()\n );\n return { parsed: docs.length };\n}\n\n// rank has no table of its own: frontmatter._rank is the only state it owns. Guarded on column\n// presence, since links can toggle off while rank has never been enabled for this tree.\nasync function nullRank(conn: Connection, dialect: ReconcileDialect): Promise<void> {\n const columns = await getColumns(conn);\n if (!columns.has('_rank')) return;\n await withTransaction(conn, () => conn.exec('UPDATE frontmatter SET \"_rank\" = NULL'), dialect.beginMode());\n}\n\n// Whole-tree and reparse-free: PageRank is computed from the links table, not from files, so\n// turning rank back on needs no reparseFiles pass at all.\nasync function rerunRank(conn: Connection, dialect: ReconcileDialect): Promise<void> {\n const delta: ReconcileDelta = { files: [], reparsed: [], added: [], vanished: [] };\n await withTransaction(\n conn,\n async () => {\n await rank.afterReconcile?.(conn, delta);\n },\n dialect.beginMode()\n );\n}\n\nasync function invalidateFeatureToggles(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect, pool: ParsePool, toggles: FeatureToggle[]): Promise<{ parsed: number }> {\n const byName = new Map(toggles.map((toggle) => [toggle.name, toggle]));\n let parsed = 0;\n\n // links first, if present: rank's rerun below (whether from its own toggle or the co-change\n // rankToggle carries) must read the links table after it is back in its new state.\n const linksToggle = byName.get('links');\n if (linksToggle) {\n parsed += (await invalidateFeatureTable(conn, cfg, baseDir, dialect, pool, linksToggle)).parsed;\n // Unconditional, per rank's dependency on links: nulling is a no-op (guarded) when rank was\n // never enabled for this tree, so this is safe even when no feature:rank segment changed.\n if (!linksToggle.turnedOn) await nullRank(conn, dialect);\n }\n\n // rank's effective on/off (featureEnabled's links dependency) only appears as a changed\n // `feature:rank` segment when it actually flips -- whether that's rank's own config or a\n // links toggle taking it along -- so honoring it here is complete on its own.\n const rankToggle = byName.get('rank');\n if (rankToggle) {\n if (rankToggle.turnedOn) await rerunRank(conn, dialect);\n else await nullRank(conn, dialect);\n }\n\n for (const toggle of toggles) {\n if (toggle.name === 'links' || toggle.name === 'rank') continue;\n parsed += (await invalidateFeatureTable(conn, cfg, baseDir, dialect, pool, toggle)).parsed;\n }\n return { parsed };\n}\n\n// The pool is created at most once, lazily, on whichever build() or invalidate() call first\n// needs it, and reused by every later call on this instance.\nexport function createBuilder(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect): Builder {\n const pool = new ParsePool();\n return {\n build: (forcedPaths) => reconcile(conn, cfg, baseDir, dialect, pool, forcedPaths),\n invalidate: (kind) => (kind === 'model' ? nullEmbedVectors(conn, dialect) : rebuildEmbeddings(conn, cfg, baseDir, dialect, pool)),\n invalidateFeatures: (toggles) => invalidateFeatureToggles(conn, cfg, baseDir, dialect, pool, toggles),\n close: () => pool.close(),\n get poolsCreated() {\n return pool.poolsCreated;\n },\n };\n}\n"],"names":["createBuilder","nullEmbedVectors","conn","dialect","withTransaction","exec","beginMode","parsed","rebuildEmbeddings","cfg","baseDir","pool","existingStmt","existingPaths","files","docs","paths","delta","prepare","Set","all","map","r","path","listFiles","filter","f","embed","has","relPath","length","reparseFiles","undefined","d","reparsed","added","vanished","remove","store","extracted","invalidateFeatureTable","toggle","table","feature","docsForFeature","NARROW_FEATURE_TABLE","name","FEATURES","find","Error","turnedOn","afterReconcile","nullRank","columns","getColumns","rerunRank","rank","invalidateFeatureToggles","toggles","byName","linksToggle","rankToggle","Map","get","ParsePool","build","forcedPaths","reconcile","invalidate","kind","invalidateFeatures","close","poolsCreated"],"mappings":";;;;+BA6JgBA;;;eAAAA;;;uBA5JM;uBACG;sBACJ;wBAEK;sBACA;yBACG;8BAGQ;2BACX;wBACC;6BACK;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAsBhC,6FAA6F;AAC7F,wEAAwE;AACxE,SAAeC,iBAAiBC,IAAgB,EAAEC,OAAyB;;;;;oBACzE;;wBAAMC,IAAAA,8BAAe,EAACF,MAAM;mCAAMA,KAAKG,IAAI,CAAC;2BAAsDF,QAAQG,SAAS;;;oBAAnH;oBACA;;wBAAO;4BAAEC,QAAQ;wBAAE;;;;IACrB;;AAEA,mGAAmG;AACnG,6FAA6F;AAC7F,SAAeC,kBAAkBN,IAAgB,EAAEO,GAAW,EAAEC,OAAe,EAAEP,OAAyB,EAAEQ,IAAe;;YACnHC,cACAC,kBAGAC,OAGEC,MACFC,OACAC;;;;oBATe;;wBAAMf,KAAKgB,OAAO,CAAC;;;oBAAlCN,eAAe;wBACKO;oBAAK;;wBAAMP,aAAaQ,GAAG;;;oBAA/CP,gBAAgB,IAAA,CAAA,EAAA,MAAIM;;wBAAK,cAA0BE,GAAG,CAAC,SAACC;mCAAM,AAACA,EAAuBC,IAAI;;sBAAC;oBACjG,8FAA8F;oBAC9F,4FAA4F;oBACtFT,QAAQU,IAAAA,mBAAS,EAACf,KAAKC,SAASe,MAAM,CAAC,SAACC;+BAAMA,EAAEC,KAAK,IAAId,cAAce,GAAG,CAACF,EAAEG,OAAO;;oBAC1F,IAAIf,MAAMgB,MAAM,KAAK,GAAG;;wBAAO;4BAAEvB,QAAQ;wBAAE;;oBAE1B;;wBAAMwB,IAAAA,uBAAY,EAACjB;4BAAQa,cAAK;2BAAGlB,KAAK,IAAIU,OAAOa,WAAW;4BAAErB,MAAAA;wBAAK;;;oBAA9EI,OAAS,cAATA;oBACFC,QAAQD,KAAKM,GAAG,CAAC,SAACY;+BAAMA,EAAEJ,OAAO;;oBACjCZ,QAAwB;wBAAEH,OAAAA;wBAAOoB,UAAUlB;wBAAOmB,KAAK;wBAAMC,QAAQ;oBAAK;oBAChF;;wBAAMhC,IAAAA,8BAAe,EACnBF,MACA;;oCACQyB,eACAA;;;;4CADN;;iDAAMA,gBAAAA,cAAK,CAACU,MAAM,cAAZV,oCAAAA,mBAAAA,cAAK,EAAUzB,MAAMc,OAAOC;;;4CAAlC;4CACA;;iDAAMU,eAAAA,cAAK,CAACW,KAAK,cAAXX,mCAAAA,kBAAAA,cAAK,EACTzB,MACAa,KAAKM,GAAG,CAAC,SAACY;2DAAO;wDAAEV,MAAMU,EAAEJ,OAAO;wDAAEU,WAAWN,EAAEM,SAAS,CAACZ,KAAK;oDAAC;oDACjEV;;;4CAHF;;;;;;4BAKF;2BACAd,QAAQG,SAAS;;;oBAVnB;oBAYA;;wBAAO;4BAAEC,QAAQQ,KAAKe,MAAM;wBAAC;;;;IAC/B;;AAEA,gGAAgG;AAChG,gGAAgG;AAChG,6FAA6F;AAC7F,kGAAkG;AAClG,SAAeU,uBAAuBtC,IAAgB,EAAEO,GAAW,EAAEC,OAAe,EAAEP,OAAyB,EAAEQ,IAAe,EAAE8B,MAAqB;;YAC/IC,OACAC,SAQA7B,OACEC,MACF6B,gBACA5B,OAIAC;;;;oBAhBAyB,QAAQG,oCAAoB,CAACJ,OAAOK,IAAI,CAAC;oBACzCH,UAAUI,iBAAQ,CAACC,IAAI,CAAC,SAACtB;+BAAMA,EAAEoB,IAAI,KAAKL,OAAOK,IAAI;;oBAC3D,IAAI,CAACH,SAAS,MAAM,IAAIM,MAAM,AAAC,wDAAmE,OAAZR,OAAOK,IAAI,EAAC;yBAE9F,CAACL,OAAOS,QAAQ,EAAhB;;;;oBACF;;wBAAM9C,IAAAA,8BAAe,EAACF,MAAM;mCAAMA,KAAKG,IAAI,CAAC,AAAC,eAAoB,OAANqC;2BAAUvC,QAAQG,SAAS;;;oBAAtF;oBACA;;wBAAO;4BAAEC,QAAQ;wBAAE;;;oBAGfO,QAAQU,IAAAA,mBAAS,EAACf,KAAKC;oBACZ;;wBAAMqB,IAAAA,uBAAY,EAACjB;4BAAQ6B;2BAAUlC,KAAK,IAAIU,OAAOa,WAAW;4BAAErB,MAAAA;wBAAK;;;oBAAhFI,OAAS,cAATA;oBACF6B,iBAAiB7B,KAAKM,GAAG,CAAC,SAACY;+BAAO;4BAAEV,MAAMU,EAAEJ,OAAO;4BAAEU,WAAWN,EAAEM,SAAS,CAACI,QAAQG,IAAI,CAAC;wBAAC;;oBAC1F9B,QAAQD,KAAKM,GAAG,CAAC,SAACY;+BAAMA,EAAEJ,OAAO;;oBACvC,gGAAgG;oBAChG,8FAA8F;oBAC9F,8CAA8C;oBACxCZ,QAAwB;wBAAEH,OAAAA;wBAAOoB,UAAUlB;wBAAOmB,OAAOnB;wBAAOoB,QAAQ;oBAAK;oBACnF;;wBAAMhC,IAAAA,8BAAe,EACnBF,MACA;;oCAEQyC,gBACAA;;;;4CAFN;;gDAAMzC,KAAKG,IAAI,CAAC,AAAC,eAAoB,OAANqC;;;4CAA/B;4CACA;;iDAAMC,iBAAAA,QAAQL,KAAK,cAAbK,qCAAAA,oBAAAA,SAAgBzC,MAAM0C,gBAAgB3B;;;4CAA5C;4CACA;;iDAAM0B,0BAAAA,QAAQQ,cAAc,cAAtBR,8CAAAA,6BAAAA,SAAyBzC,MAAMe;;;4CAArC;;;;;;4BACF;2BACAd,QAAQG,SAAS;;;oBAPnB;oBASA;;wBAAO;4BAAEC,QAAQQ,KAAKe,MAAM;wBAAC;;;;IAC/B;;AAEA,+FAA+F;AAC/F,wFAAwF;AACxF,SAAesB,SAASlD,IAAgB,EAAEC,OAAyB;;YAC3DkD;;;;oBAAU;;wBAAMC,IAAAA,oBAAU,EAACpD;;;oBAA3BmD,UAAU;oBAChB,IAAI,CAACA,QAAQzB,GAAG,CAAC,UAAU;;;oBAC3B;;wBAAMxB,IAAAA,8BAAe,EAACF,MAAM;mCAAMA,KAAKG,IAAI,CAAC;2BAA0CF,QAAQG,SAAS;;;oBAAvG;;;;;;IACF;;AAEA,6FAA6F;AAC7F,0DAA0D;AAC1D,SAAeiD,UAAUrD,IAAgB,EAAEC,OAAyB;;YAC5Dc;;;;oBAAAA,QAAwB;wBAAEH,KAAK;wBAAMoB,QAAQ;wBAAMC,KAAK;wBAAMC,QAAQ;oBAAK;oBACjF;;wBAAMhC,IAAAA,8BAAe,EACnBF,MACA;;oCACQsD;;;;4CAAN;;iDAAMA,uBAAAA,YAAI,CAACL,cAAc,cAAnBK,2CAAAA,0BAAAA,YAAI,EAAkBtD,MAAMe;;;4CAAlC;;;;;;4BACF;2BACAd,QAAQG,SAAS;;;oBALnB;;;;;;IAOF;;AAEA,SAAemD,yBAAyBvD,IAAgB,EAAEO,GAAW,EAAEC,OAAe,EAAEP,OAAyB,EAAEQ,IAAe,EAAE+C,OAAwB;;YACpJC,QACFpD,QAIEqD,aAWAC,YAMD,2BAAA,mBAAA,gBAAA,WAAA,OAAMpB;;;;oBAtBLkB,SAAS,IAAIG,IAAIJ,QAAQrC,GAAG,CAAC,SAACoB;+BAAW;4BAACA,OAAOK,IAAI;4BAAEL;yBAAO;;oBAChElC,SAAS;oBAEb,4FAA4F;oBAC5F,mFAAmF;oBAC7EqD,cAAcD,OAAOI,GAAG,CAAC;yBAC3BH,aAAAA;;;;oBACS;;wBAAMpB,uBAAuBtC,MAAMO,KAAKC,SAASP,SAASQ,MAAMiD;;;oBAA3ErD,SAAAA,SAAU,AAAC,cAA8EA,MAAM;yBAG3F,CAACqD,YAAYV,QAAQ,EAArB;;;;oBAAuB;;wBAAME,SAASlD,MAAMC;;;oBAArB;;;oBAG7B,wFAAwF;oBACxF,yFAAyF;oBACzF,8EAA8E;oBACxE0D,aAAaF,OAAOI,GAAG,CAAC;yBAC1BF,YAAAA;;;;yBACEA,WAAWX,QAAQ,EAAnBW;;;;oBAAqB;;wBAAMN,UAAUrD,MAAMC;;;oBAAtB;;;;;;oBACpB;;wBAAMiD,SAASlD,MAAMC;;;oBAArB;;;oBAGF,kCAAA,2BAAA;;;;;;;;;oBAAA,YAAgBuD;;;2BAAhB,6BAAA,QAAA;;;;oBAAMjB,SAAN;oBACH,IAAIA,OAAOK,IAAI,KAAK,WAAWL,OAAOK,IAAI,KAAK,QAAQ;;;;oBAC5C;;wBAAMN,uBAAuBtC,MAAMO,KAAKC,SAASP,SAASQ,MAAM8B;;;oBAA3ElC,SAAAA,SAAU,AAAC,cAAyEA,MAAM;;;oBAFvF;;;;;;;;;;;;oBAAA;oBAAA;;;;;;;6BAAA,6BAAA;4BAAA;;;4BAAA;kCAAA;;;;;;;oBAIL;;wBAAO;4BAAEA,QAAAA;wBAAO;;;;IAClB;;AAIO,SAASP,cAAcE,IAAgB,EAAEO,GAAW,EAAEC,OAAe,EAAEP,OAAyB;IACrG,IAAMQ,OAAO,IAAIqD,iBAAS;IAC1B,OAAO;QACLC,OAAO,SAAPA,MAAQC;mBAAgBC,IAAAA,sBAAS,EAACjE,MAAMO,KAAKC,SAASP,SAASQ,MAAMuD;;QACrEE,YAAY,SAAZA,WAAaC;mBAAUA,SAAS,UAAUpE,iBAAiBC,MAAMC,WAAWK,kBAAkBN,MAAMO,KAAKC,SAASP,SAASQ;;QAC3H2D,oBAAoB,SAApBA,mBAAqBZ;mBAAYD,yBAAyBvD,MAAMO,KAAKC,SAASP,SAASQ,MAAM+C;;QAC7Fa,OAAO,SAAPA;mBAAa5D,KAAK4D,KAAK;;QACvB,IAAIC;YACF,OAAO7D,KAAK6D,YAAY;QAC1B;IACF;AACF"}
|
|
1
|
+
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/builder.ts"],"sourcesContent":["import type { Config } from '../config/index.ts';\nimport { embed } from '../features/embed.ts';\nimport { FEATURES } from '../features/index.ts';\nimport { rank } from '../features/rank.ts';\nimport type { ReconcileDelta } from '../features/types.ts';\nimport { listFiles } from '../scan/index.ts';\nimport { ParsePool } from '../scan/pool.ts';\nimport { reparseFiles } from '../scan/reparse.ts';\nimport type { EmbedChangeKind } from './embed-scope.ts';\nimport type { FeatureToggle } from './feature-scope.ts';\nimport { NARROW_FEATURE_TABLE } from './feature-scope.ts';\nimport { reconcile } from './reconcile.ts';\nimport { getColumns } from './shared.ts';\nimport type { Stages } from './stages.ts';\nimport { withTransaction } from './transaction.ts';\nimport type { Connection, ReconcileDialect } from './types.ts';\n\n// Owns bringing a store's index current: the write half `Store` (types.ts) deliberately does not\n// carry. A one-shot open calls build() once; a watcher calls it repeatedly on the same instance.\nexport interface Builder {\n // forcedPaths (open.ts's preset-only narrow rebuild) applies to this call alone, not future ones.\n build(forcedPaths?: ReadonlySet<string>): Promise<{ parsed: number; warnings: string[]; stages: Stages }>;\n // Narrow embed invalidation (open.ts, embed-scope.ts's classifyEmbedChange): 'model' nulls every\n // vector/scale in place; 'chunk' rebuilds every embedding row. Neither touches another table.\n invalidate(kind: EmbedChangeKind): Promise<{ parsed: number }>;\n // Narrow feature-toggle invalidation (open.ts, feature-scope.ts's classifyFeatureToggles): a\n // toggle drops (or, turning back on, fully re-derives) that feature's own table; rank instead\n // nulls or recomputes its `_rank` column, since rows written while a feature was off are\n // untrustworthy (reconcile.ts's activeFeatures skips a disabled feature's hooks entirely).\n invalidateFeatures(toggles: FeatureToggle[]): Promise<{ parsed: number }>;\n // Releases the parse worker pool, if this lifetime ever created one; never the connection.\n close(): Promise<void>;\n // Pools this lifetime has constructed, so reuse is observable rather than inferred.\n readonly poolsCreated: number;\n}\n\n// 'model': the provider or model name moved but chunk boundaries did not, so only the vector\n// values are stale. No reparse, and no table but embeddings is touched.\nasync function nullEmbedVectors(conn: Connection, dialect: ReconcileDialect): Promise<{ parsed: number }> {\n await withTransaction(conn, () => conn.exec('UPDATE embeddings SET vector = NULL, scale = NULL'), dialect.beginMode());\n return { parsed: 0 };\n}\n\n// 'chunk': chunkTokens or the chunk version moved, so chunk boundaries genuinely changed. Rebuilds\n// every embedding row through the embed feature's own remove/store, touching no other table.\nasync function rebuildEmbeddings(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect, pool: ParsePool): Promise<{ parsed: number }> {\n const existingStmt = await conn.prepare('SELECT \"path\" FROM frontmatter');\n const existingPaths = new Set((await existingStmt.all()).map((r) => (r as { path: string }).path));\n // A file added since the last build has no row yet; leaving it for build() to insert avoids a\n // primary-key collision between this INSERT and build()'s own insert for the same new file.\n const files = listFiles(cfg, baseDir).filter((f) => f.embed && existingPaths.has(f.relPath));\n if (files.length === 0) return { parsed: 0 };\n\n const { docs } = await reparseFiles(files, [embed], cfg, new Set(), undefined, { pool });\n const paths = docs.map((d) => d.relPath);\n const delta: ReconcileDelta = { files, reparsed: paths, added: [], vanished: [] };\n await withTransaction(\n conn,\n async () => {\n await embed.remove?.(conn, paths, delta);\n await embed.store?.(\n conn,\n docs.map((d) => ({ path: d.relPath, extracted: d.extracted.embed })),\n delta\n );\n },\n dialect.beginMode()\n );\n return { parsed: docs.length };\n}\n\n// One feature's own table, dropped and (turning on) fully re-derived across every indexed file.\n// Rows left over from before this feature was disabled are untrustworthy by construction (files\n// added, edited, or deleted got no remove/store for it), so this never diffs against them --\n// dropping every row and reparsing the whole tree is the only safe path back to consistent state.\nasync function invalidateFeatureTable(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect, pool: ParsePool, toggle: FeatureToggle): Promise<{ parsed: number }> {\n const table = NARROW_FEATURE_TABLE[toggle.name];\n const feature = FEATURES.find((f) => f.name === toggle.name);\n if (!feature) throw new Error(`invalidateFeatureTable: no registered feature named \"${toggle.name}\"`);\n\n if (!toggle.turnedOn) {\n await withTransaction(conn, () => conn.exec(`DELETE FROM ${table}`), dialect.beginMode());\n return { parsed: 0 };\n }\n\n const files = listFiles(cfg, baseDir);\n const { docs } = await reparseFiles(files, [feature], cfg, new Set(), undefined, { pool });\n const docsForFeature = docs.map((d) => ({ path: d.relPath, extracted: d.extracted[feature.name] }));\n const paths = docs.map((d) => d.relPath);\n // Every path counts as added, not just touched: the table was just dropped, so there is nothing\n // stale to diff against, and links' own afterReconcile takes its cold-build (resolveAll) path\n // only when added.length equals files.length.\n const delta: ReconcileDelta = { files, reparsed: paths, added: paths, vanished: [] };\n await withTransaction(\n conn,\n async () => {\n await conn.exec(`DELETE FROM ${table}`);\n await feature.store?.(conn, docsForFeature, delta);\n await feature.afterReconcile?.(conn, delta);\n },\n dialect.beginMode()\n );\n return { parsed: docs.length };\n}\n\n// rank has no table of its own: frontmatter._rank is the only state it owns. Guarded on column\n// presence, since links can toggle off while rank has never been enabled for this tree.\nasync function nullRank(conn: Connection, dialect: ReconcileDialect): Promise<void> {\n const columns = await getColumns(conn);\n if (!columns.has('_rank')) return;\n await withTransaction(conn, () => conn.exec('UPDATE frontmatter SET \"_rank\" = NULL'), dialect.beginMode());\n}\n\n// Whole-tree and reparse-free: PageRank is computed from the links table, not from files, so\n// turning rank back on needs no reparseFiles pass at all.\nasync function rerunRank(conn: Connection, dialect: ReconcileDialect): Promise<void> {\n const delta: ReconcileDelta = { files: [], reparsed: [], added: [], vanished: [] };\n await withTransaction(\n conn,\n async () => {\n await rank.afterReconcile?.(conn, delta);\n },\n dialect.beginMode()\n );\n}\n\nasync function invalidateFeatureToggles(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect, pool: ParsePool, toggles: FeatureToggle[]): Promise<{ parsed: number }> {\n const byName = new Map(toggles.map((toggle) => [toggle.name, toggle]));\n let parsed = 0;\n\n // links first, if present: rank's rerun below (whether from its own toggle or the co-change\n // rankToggle carries) must read the links table after it is back in its new state.\n const linksToggle = byName.get('links');\n if (linksToggle) {\n parsed += (await invalidateFeatureTable(conn, cfg, baseDir, dialect, pool, linksToggle)).parsed;\n // Unconditional, per rank's dependency on links: nulling is a no-op (guarded) when rank was\n // never enabled for this tree, so this is safe even when no feature:rank segment changed.\n if (!linksToggle.turnedOn) await nullRank(conn, dialect);\n }\n\n // rank's effective on/off (featureEnabled's links dependency) only appears as a changed\n // `feature:rank` segment when it actually flips -- whether that's rank's own config or a\n // links toggle taking it along -- so honoring it here is complete on its own.\n const rankToggle = byName.get('rank');\n if (rankToggle) {\n if (rankToggle.turnedOn) await rerunRank(conn, dialect);\n else await nullRank(conn, dialect);\n }\n\n for (const toggle of toggles) {\n if (toggle.name === 'links' || toggle.name === 'rank') continue;\n parsed += (await invalidateFeatureTable(conn, cfg, baseDir, dialect, pool, toggle)).parsed;\n }\n return { parsed };\n}\n\n// The pool is created at most once, lazily, on whichever build() or invalidate() call first\n// needs it, and reused by every later call on this instance.\nexport function createBuilder(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect): Builder {\n const pool = new ParsePool();\n return {\n build: (forcedPaths) => reconcile(conn, cfg, baseDir, dialect, pool, forcedPaths),\n invalidate: (kind) => (kind === 'model' ? nullEmbedVectors(conn, dialect) : rebuildEmbeddings(conn, cfg, baseDir, dialect, pool)),\n invalidateFeatures: (toggles) => invalidateFeatureToggles(conn, cfg, baseDir, dialect, pool, toggles),\n close: () => pool.close(),\n get poolsCreated() {\n return pool.poolsCreated;\n },\n };\n}\n"],"names":["createBuilder","nullEmbedVectors","conn","dialect","withTransaction","exec","beginMode","parsed","rebuildEmbeddings","cfg","baseDir","pool","existingStmt","existingPaths","files","docs","paths","delta","prepare","Set","all","map","r","path","listFiles","filter","f","embed","has","relPath","length","reparseFiles","undefined","d","reparsed","added","vanished","remove","store","extracted","invalidateFeatureTable","toggle","table","feature","docsForFeature","NARROW_FEATURE_TABLE","name","FEATURES","find","Error","turnedOn","afterReconcile","nullRank","columns","getColumns","rerunRank","rank","invalidateFeatureToggles","toggles","byName","linksToggle","rankToggle","Map","get","ParsePool","build","forcedPaths","reconcile","invalidate","kind","invalidateFeatures","close","poolsCreated"],"mappings":";;;;+BA8JgBA;;;eAAAA;;;uBA7JM;uBACG;sBACJ;wBAEK;sBACA;yBACG;8BAGQ;2BACX;wBACC;6BAEK;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAsBhC,6FAA6F;AAC7F,wEAAwE;AACxE,SAAeC,iBAAiBC,IAAgB,EAAEC,OAAyB;;;;;oBACzE;;wBAAMC,IAAAA,8BAAe,EAACF,MAAM;mCAAMA,KAAKG,IAAI,CAAC;2BAAsDF,QAAQG,SAAS;;;oBAAnH;oBACA;;wBAAO;4BAAEC,QAAQ;wBAAE;;;;IACrB;;AAEA,mGAAmG;AACnG,6FAA6F;AAC7F,SAAeC,kBAAkBN,IAAgB,EAAEO,GAAW,EAAEC,OAAe,EAAEP,OAAyB,EAAEQ,IAAe;;YACnHC,cACAC,kBAGAC,OAGEC,MACFC,OACAC;;;;oBATe;;wBAAMf,KAAKgB,OAAO,CAAC;;;oBAAlCN,eAAe;wBACKO;oBAAK;;wBAAMP,aAAaQ,GAAG;;;oBAA/CP,gBAAgB,IAAA,CAAA,EAAA,MAAIM;;wBAAK,cAA0BE,GAAG,CAAC,SAACC;mCAAM,AAACA,EAAuBC,IAAI;;sBAAC;oBACjG,8FAA8F;oBAC9F,4FAA4F;oBACtFT,QAAQU,IAAAA,mBAAS,EAACf,KAAKC,SAASe,MAAM,CAAC,SAACC;+BAAMA,EAAEC,KAAK,IAAId,cAAce,GAAG,CAACF,EAAEG,OAAO;;oBAC1F,IAAIf,MAAMgB,MAAM,KAAK,GAAG;;wBAAO;4BAAEvB,QAAQ;wBAAE;;oBAE1B;;wBAAMwB,IAAAA,uBAAY,EAACjB;4BAAQa,cAAK;2BAAGlB,KAAK,IAAIU,OAAOa,WAAW;4BAAErB,MAAAA;wBAAK;;;oBAA9EI,OAAS,cAATA;oBACFC,QAAQD,KAAKM,GAAG,CAAC,SAACY;+BAAMA,EAAEJ,OAAO;;oBACjCZ,QAAwB;wBAAEH,OAAAA;wBAAOoB,UAAUlB;wBAAOmB,KAAK;wBAAMC,QAAQ;oBAAK;oBAChF;;wBAAMhC,IAAAA,8BAAe,EACnBF,MACA;;oCACQyB,eACAA;;;;4CADN;;iDAAMA,gBAAAA,cAAK,CAACU,MAAM,cAAZV,oCAAAA,mBAAAA,cAAK,EAAUzB,MAAMc,OAAOC;;;4CAAlC;4CACA;;iDAAMU,eAAAA,cAAK,CAACW,KAAK,cAAXX,mCAAAA,kBAAAA,cAAK,EACTzB,MACAa,KAAKM,GAAG,CAAC,SAACY;2DAAO;wDAAEV,MAAMU,EAAEJ,OAAO;wDAAEU,WAAWN,EAAEM,SAAS,CAACZ,KAAK;oDAAC;oDACjEV;;;4CAHF;;;;;;4BAKF;2BACAd,QAAQG,SAAS;;;oBAVnB;oBAYA;;wBAAO;4BAAEC,QAAQQ,KAAKe,MAAM;wBAAC;;;;IAC/B;;AAEA,gGAAgG;AAChG,gGAAgG;AAChG,6FAA6F;AAC7F,kGAAkG;AAClG,SAAeU,uBAAuBtC,IAAgB,EAAEO,GAAW,EAAEC,OAAe,EAAEP,OAAyB,EAAEQ,IAAe,EAAE8B,MAAqB;;YAC/IC,OACAC,SAQA7B,OACEC,MACF6B,gBACA5B,OAIAC;;;;oBAhBAyB,QAAQG,oCAAoB,CAACJ,OAAOK,IAAI,CAAC;oBACzCH,UAAUI,iBAAQ,CAACC,IAAI,CAAC,SAACtB;+BAAMA,EAAEoB,IAAI,KAAKL,OAAOK,IAAI;;oBAC3D,IAAI,CAACH,SAAS,MAAM,IAAIM,MAAM,AAAC,wDAAmE,OAAZR,OAAOK,IAAI,EAAC;yBAE9F,CAACL,OAAOS,QAAQ,EAAhB;;;;oBACF;;wBAAM9C,IAAAA,8BAAe,EAACF,MAAM;mCAAMA,KAAKG,IAAI,CAAC,AAAC,eAAoB,OAANqC;2BAAUvC,QAAQG,SAAS;;;oBAAtF;oBACA;;wBAAO;4BAAEC,QAAQ;wBAAE;;;oBAGfO,QAAQU,IAAAA,mBAAS,EAACf,KAAKC;oBACZ;;wBAAMqB,IAAAA,uBAAY,EAACjB;4BAAQ6B;2BAAUlC,KAAK,IAAIU,OAAOa,WAAW;4BAAErB,MAAAA;wBAAK;;;oBAAhFI,OAAS,cAATA;oBACF6B,iBAAiB7B,KAAKM,GAAG,CAAC,SAACY;+BAAO;4BAAEV,MAAMU,EAAEJ,OAAO;4BAAEU,WAAWN,EAAEM,SAAS,CAACI,QAAQG,IAAI,CAAC;wBAAC;;oBAC1F9B,QAAQD,KAAKM,GAAG,CAAC,SAACY;+BAAMA,EAAEJ,OAAO;;oBACvC,gGAAgG;oBAChG,8FAA8F;oBAC9F,8CAA8C;oBACxCZ,QAAwB;wBAAEH,OAAAA;wBAAOoB,UAAUlB;wBAAOmB,OAAOnB;wBAAOoB,QAAQ;oBAAK;oBACnF;;wBAAMhC,IAAAA,8BAAe,EACnBF,MACA;;oCAEQyC,gBACAA;;;;4CAFN;;gDAAMzC,KAAKG,IAAI,CAAC,AAAC,eAAoB,OAANqC;;;4CAA/B;4CACA;;iDAAMC,iBAAAA,QAAQL,KAAK,cAAbK,qCAAAA,oBAAAA,SAAgBzC,MAAM0C,gBAAgB3B;;;4CAA5C;4CACA;;iDAAM0B,0BAAAA,QAAQQ,cAAc,cAAtBR,8CAAAA,6BAAAA,SAAyBzC,MAAMe;;;4CAArC;;;;;;4BACF;2BACAd,QAAQG,SAAS;;;oBAPnB;oBASA;;wBAAO;4BAAEC,QAAQQ,KAAKe,MAAM;wBAAC;;;;IAC/B;;AAEA,+FAA+F;AAC/F,wFAAwF;AACxF,SAAesB,SAASlD,IAAgB,EAAEC,OAAyB;;YAC3DkD;;;;oBAAU;;wBAAMC,IAAAA,oBAAU,EAACpD;;;oBAA3BmD,UAAU;oBAChB,IAAI,CAACA,QAAQzB,GAAG,CAAC,UAAU;;;oBAC3B;;wBAAMxB,IAAAA,8BAAe,EAACF,MAAM;mCAAMA,KAAKG,IAAI,CAAC;2BAA0CF,QAAQG,SAAS;;;oBAAvG;;;;;;IACF;;AAEA,6FAA6F;AAC7F,0DAA0D;AAC1D,SAAeiD,UAAUrD,IAAgB,EAAEC,OAAyB;;YAC5Dc;;;;oBAAAA,QAAwB;wBAAEH,KAAK;wBAAMoB,QAAQ;wBAAMC,KAAK;wBAAMC,QAAQ;oBAAK;oBACjF;;wBAAMhC,IAAAA,8BAAe,EACnBF,MACA;;oCACQsD;;;;4CAAN;;iDAAMA,uBAAAA,YAAI,CAACL,cAAc,cAAnBK,2CAAAA,0BAAAA,YAAI,EAAkBtD,MAAMe;;;4CAAlC;;;;;;4BACF;2BACAd,QAAQG,SAAS;;;oBALnB;;;;;;IAOF;;AAEA,SAAemD,yBAAyBvD,IAAgB,EAAEO,GAAW,EAAEC,OAAe,EAAEP,OAAyB,EAAEQ,IAAe,EAAE+C,OAAwB;;YACpJC,QACFpD,QAIEqD,aAWAC,YAMD,2BAAA,mBAAA,gBAAA,WAAA,OAAMpB;;;;oBAtBLkB,SAAS,IAAIG,IAAIJ,QAAQrC,GAAG,CAAC,SAACoB;+BAAW;4BAACA,OAAOK,IAAI;4BAAEL;yBAAO;;oBAChElC,SAAS;oBAEb,4FAA4F;oBAC5F,mFAAmF;oBAC7EqD,cAAcD,OAAOI,GAAG,CAAC;yBAC3BH,aAAAA;;;;oBACS;;wBAAMpB,uBAAuBtC,MAAMO,KAAKC,SAASP,SAASQ,MAAMiD;;;oBAA3ErD,SAAAA,SAAU,AAAC,cAA8EA,MAAM;yBAG3F,CAACqD,YAAYV,QAAQ,EAArB;;;;oBAAuB;;wBAAME,SAASlD,MAAMC;;;oBAArB;;;oBAG7B,wFAAwF;oBACxF,yFAAyF;oBACzF,8EAA8E;oBACxE0D,aAAaF,OAAOI,GAAG,CAAC;yBAC1BF,YAAAA;;;;yBACEA,WAAWX,QAAQ,EAAnBW;;;;oBAAqB;;wBAAMN,UAAUrD,MAAMC;;;oBAAtB;;;;;;oBACpB;;wBAAMiD,SAASlD,MAAMC;;;oBAArB;;;oBAGF,kCAAA,2BAAA;;;;;;;;;oBAAA,YAAgBuD;;;2BAAhB,6BAAA,QAAA;;;;oBAAMjB,SAAN;oBACH,IAAIA,OAAOK,IAAI,KAAK,WAAWL,OAAOK,IAAI,KAAK,QAAQ;;;;oBAC5C;;wBAAMN,uBAAuBtC,MAAMO,KAAKC,SAASP,SAASQ,MAAM8B;;;oBAA3ElC,SAAAA,SAAU,AAAC,cAAyEA,MAAM;;;oBAFvF;;;;;;;;;;;;oBAAA;oBAAA;;;;;;;6BAAA,6BAAA;4BAAA;;;4BAAA;kCAAA;;;;;;;oBAIL;;wBAAO;4BAAEA,QAAAA;wBAAO;;;;IAClB;;AAIO,SAASP,cAAcE,IAAgB,EAAEO,GAAW,EAAEC,OAAe,EAAEP,OAAyB;IACrG,IAAMQ,OAAO,IAAIqD,iBAAS;IAC1B,OAAO;QACLC,OAAO,SAAPA,MAAQC;mBAAgBC,IAAAA,sBAAS,EAACjE,MAAMO,KAAKC,SAASP,SAASQ,MAAMuD;;QACrEE,YAAY,SAAZA,WAAaC;mBAAUA,SAAS,UAAUpE,iBAAiBC,MAAMC,WAAWK,kBAAkBN,MAAMO,KAAKC,SAASP,SAASQ;;QAC3H2D,oBAAoB,SAApBA,mBAAqBZ;mBAAYD,yBAAyBvD,MAAMO,KAAKC,SAASP,SAASQ,MAAM+C;;QAC7Fa,OAAO,SAAPA;mBAAa5D,KAAK4D,KAAK;;QACvB,IAAIC;YACF,OAAO7D,KAAK6D,YAAY;QAC1B;IACF;AACF"}
|
|
@@ -19,8 +19,10 @@ _export(exports, {
|
|
|
19
19
|
return storeValueToJs;
|
|
20
20
|
}
|
|
21
21
|
});
|
|
22
|
+
var _sharedts = require("../shared.js");
|
|
22
23
|
var _transactionts = require("../transaction.js");
|
|
23
24
|
var _batchts = require("./batch.js");
|
|
25
|
+
var _nativets = require("./native.js");
|
|
24
26
|
function _array_like_to_array(arr, len) {
|
|
25
27
|
if (len == null || len > arr.length) len = arr.length;
|
|
26
28
|
for(var i = 0, arr2 = new Array(len); i < len; i++)arr2[i] = arr[i];
|
|
@@ -610,6 +612,124 @@ function createConnection(duckdb) {
|
|
|
610
612
|
});
|
|
611
613
|
})();
|
|
612
614
|
},
|
|
615
|
+
appendRows: // The appender writes columnar vectors with no per-parameter binding, which is the cost
|
|
616
|
+
// runBatch pays: measured 7.2x on frontmatter at 6,566 rows. Alignment is against the table's
|
|
617
|
+
// own physical column order, read fresh, so a column the caller does not write (a
|
|
618
|
+
// feature-owned "_rank", say) takes appendDefault() rather than shifting every value one slot.
|
|
619
|
+
function appendRows(table, columns, rows) {
|
|
620
|
+
return _async_to_generator(function() {
|
|
621
|
+
var variantValue, infoStmt, physical, rowIndexOf;
|
|
622
|
+
return _ts_generator(this, function(_state) {
|
|
623
|
+
switch(_state.label){
|
|
624
|
+
case 0:
|
|
625
|
+
if (rows.length === 0) return [
|
|
626
|
+
2
|
|
627
|
+
];
|
|
628
|
+
return [
|
|
629
|
+
4,
|
|
630
|
+
(0, _nativets.duckdbApi)()
|
|
631
|
+
];
|
|
632
|
+
case 1:
|
|
633
|
+
variantValue = _state.sent().variantValue;
|
|
634
|
+
return [
|
|
635
|
+
4,
|
|
636
|
+
conn.prepare("PRAGMA table_info(".concat((0, _sharedts.quoteIdent)(table), ")"))
|
|
637
|
+
];
|
|
638
|
+
case 2:
|
|
639
|
+
infoStmt = _state.sent();
|
|
640
|
+
return [
|
|
641
|
+
4,
|
|
642
|
+
infoStmt.all()
|
|
643
|
+
];
|
|
644
|
+
case 3:
|
|
645
|
+
physical = _state.sent();
|
|
646
|
+
rowIndexOf = new Map(columns.map(function(name, i) {
|
|
647
|
+
return [
|
|
648
|
+
name,
|
|
649
|
+
i
|
|
650
|
+
];
|
|
651
|
+
}));
|
|
652
|
+
return [
|
|
653
|
+
4,
|
|
654
|
+
(0, _transactionts.withTransaction)(conn, function() {
|
|
655
|
+
return _async_to_generator(function() {
|
|
656
|
+
var appender, _iteratorNormalCompletion, _didIteratorError, _iteratorError, _iterator, _step, row, _iteratorNormalCompletion1, _didIteratorError1, _iteratorError1, _iterator1, _step1, column, idx, value;
|
|
657
|
+
return _ts_generator(this, function(_state) {
|
|
658
|
+
switch(_state.label){
|
|
659
|
+
case 0:
|
|
660
|
+
return [
|
|
661
|
+
4,
|
|
662
|
+
duckdb.createAppender(table)
|
|
663
|
+
];
|
|
664
|
+
case 1:
|
|
665
|
+
appender = _state.sent();
|
|
666
|
+
try {
|
|
667
|
+
_iteratorNormalCompletion = true, _didIteratorError = false, _iteratorError = undefined;
|
|
668
|
+
try {
|
|
669
|
+
for(_iterator = rows[Symbol.iterator](); !(_iteratorNormalCompletion = (_step = _iterator.next()).done); _iteratorNormalCompletion = true){
|
|
670
|
+
row = _step.value;
|
|
671
|
+
_iteratorNormalCompletion1 = true, _didIteratorError1 = false, _iteratorError1 = undefined;
|
|
672
|
+
try {
|
|
673
|
+
for(_iterator1 = physical[Symbol.iterator](); !(_iteratorNormalCompletion1 = (_step1 = _iterator1.next()).done); _iteratorNormalCompletion1 = true){
|
|
674
|
+
column = _step1.value;
|
|
675
|
+
idx = rowIndexOf.get(column.name);
|
|
676
|
+
value = idx === undefined ? undefined : row[idx];
|
|
677
|
+
if (idx === undefined) appender.appendDefault();
|
|
678
|
+
else if (value === null || value === undefined) appender.appendNull();
|
|
679
|
+
else if (column.type === 'VARIANT') appender.appendVariant(variantValue(value));
|
|
680
|
+
else appender.appendValue(value);
|
|
681
|
+
}
|
|
682
|
+
} catch (err) {
|
|
683
|
+
_didIteratorError1 = true;
|
|
684
|
+
_iteratorError1 = err;
|
|
685
|
+
} finally{
|
|
686
|
+
try {
|
|
687
|
+
if (!_iteratorNormalCompletion1 && _iterator1.return != null) {
|
|
688
|
+
_iterator1.return();
|
|
689
|
+
}
|
|
690
|
+
} finally{
|
|
691
|
+
if (_didIteratorError1) {
|
|
692
|
+
throw _iteratorError1;
|
|
693
|
+
}
|
|
694
|
+
}
|
|
695
|
+
}
|
|
696
|
+
appender.endRow();
|
|
697
|
+
}
|
|
698
|
+
} catch (err) {
|
|
699
|
+
_didIteratorError = true;
|
|
700
|
+
_iteratorError = err;
|
|
701
|
+
} finally{
|
|
702
|
+
try {
|
|
703
|
+
if (!_iteratorNormalCompletion && _iterator.return != null) {
|
|
704
|
+
_iterator.return();
|
|
705
|
+
}
|
|
706
|
+
} finally{
|
|
707
|
+
if (_didIteratorError) {
|
|
708
|
+
throw _iteratorError;
|
|
709
|
+
}
|
|
710
|
+
}
|
|
711
|
+
}
|
|
712
|
+
appender.flushSync();
|
|
713
|
+
} finally{
|
|
714
|
+
appender.closeSync();
|
|
715
|
+
}
|
|
716
|
+
return [
|
|
717
|
+
2
|
|
718
|
+
];
|
|
719
|
+
}
|
|
720
|
+
});
|
|
721
|
+
})();
|
|
722
|
+
})
|
|
723
|
+
];
|
|
724
|
+
case 4:
|
|
725
|
+
_state.sent();
|
|
726
|
+
return [
|
|
727
|
+
2
|
|
728
|
+
];
|
|
729
|
+
}
|
|
730
|
+
});
|
|
731
|
+
})();
|
|
732
|
+
},
|
|
613
733
|
runBatch: // One crossing regardless of row count: rewriteBatch folds recognized shapes into a multi-row statement (batch.ts),
|
|
614
734
|
// else falls back to a bind-and-run loop, one call either way. Joins the caller's transaction when there is one, else opens its own.
|
|
615
735
|
function runBatch(sql, paramRows) {
|