@effortlessmotion/pdf-to-md 0.1.8 → 0.2.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +5 -5
- package/dist/cjs/analyze.d.cts +3 -1
- package/dist/cjs/analyze.d.ts +3 -1
- package/dist/cjs/analyze.js +38 -100
- package/dist/cjs/analyze.js.map +1 -1
- package/dist/cjs/audit.d.cts +6 -5
- package/dist/cjs/audit.d.ts +6 -5
- package/dist/cjs/audit.js +368 -60
- package/dist/cjs/audit.js.map +1 -1
- package/dist/cjs/cli/extract.js +13 -48
- package/dist/cjs/cli/extract.js.map +1 -1
- package/dist/cjs/collect.js +91 -35
- package/dist/cjs/collect.js.map +1 -1
- package/dist/cjs/errors.d.cts +2 -0
- package/dist/cjs/errors.d.ts +2 -0
- package/dist/cjs/errors.js.map +1 -1
- package/dist/cjs/extract.d.cts +20 -9
- package/dist/cjs/extract.d.ts +20 -9
- package/dist/cjs/extract.js +333 -106
- package/dist/cjs/extract.js.map +1 -1
- package/dist/cjs/lines.js +21 -10
- package/dist/cjs/lines.js.map +1 -1
- package/dist/cjs/markdown.js +140 -54
- package/dist/cjs/markdown.js.map +1 -1
- package/dist/cjs/page-marker.d.cts +1 -1
- package/dist/cjs/page-marker.d.ts +1 -1
- package/dist/cjs/page-marker.js +39 -8
- package/dist/cjs/page-marker.js.map +1 -1
- package/dist/cjs/render-analysis.d.cts +4 -2
- package/dist/cjs/render-analysis.d.ts +4 -2
- package/dist/cjs/render-analysis.js +5 -3
- package/dist/cjs/render-analysis.js.map +1 -1
- package/dist/cjs/report.d.cts +2 -2
- package/dist/cjs/report.d.ts +2 -2
- package/dist/cjs/report.js +27 -8
- package/dist/cjs/report.js.map +1 -1
- package/dist/cjs/text-analysis.d.cts +12 -0
- package/dist/cjs/text-analysis.d.ts +12 -0
- package/dist/cjs/text-analysis.js +175 -0
- package/dist/cjs/text-analysis.js.map +1 -0
- package/dist/cjs/types.d.cts +40 -3
- package/dist/cjs/types.d.ts +40 -3
- package/dist/cjs/types.js +2 -3
- package/dist/cjs/types.js.map +1 -1
- package/dist/esm/analyze.d.ts +3 -1
- package/dist/esm/analyze.js +18 -89
- package/dist/esm/analyze.js.map +1 -1
- package/dist/esm/audit.d.ts +6 -5
- package/dist/esm/audit.js +236 -29
- package/dist/esm/audit.js.map +1 -1
- package/dist/esm/cli/extract.js +10 -23
- package/dist/esm/cli/extract.js.map +1 -1
- package/dist/esm/collect.js +64 -15
- package/dist/esm/collect.js.map +1 -1
- package/dist/esm/errors.d.ts +2 -0
- package/dist/esm/errors.js.map +1 -1
- package/dist/esm/extract.d.ts +20 -9
- package/dist/esm/extract.js +167 -76
- package/dist/esm/extract.js.map +1 -1
- package/dist/esm/lines.js +14 -10
- package/dist/esm/lines.js.map +1 -1
- package/dist/esm/markdown.js +110 -27
- package/dist/esm/markdown.js.map +1 -1
- package/dist/esm/page-marker.d.ts +1 -1
- package/dist/esm/page-marker.js +15 -9
- package/dist/esm/page-marker.js.map +1 -1
- package/dist/esm/render-analysis.d.ts +4 -2
- package/dist/esm/render-analysis.js +9 -5
- package/dist/esm/render-analysis.js.map +1 -1
- package/dist/esm/report.d.ts +2 -2
- package/dist/esm/report.js +12 -7
- package/dist/esm/report.js.map +1 -1
- package/dist/esm/text-analysis.d.ts +12 -0
- package/dist/esm/text-analysis.js +81 -0
- package/dist/esm/text-analysis.js.map +1 -0
- package/dist/esm/types.d.ts +40 -3
- package/dist/esm/types.js +4 -4
- package/dist/esm/types.js.map +1 -1
- package/package.json +1 -1
- package/skills/pdf-to-md/SKILL.md +39 -66
package/dist/cjs/extract.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/formats/pdf-to-md/src/extract.ts"],"sourcesContent":["// The library layer for PDF -> markdown conversion: extractText/extractOcr/\n// pdfToMarkdown each take plain args, return plain data, and throw PdfToMdError — no process.exit, no console.*, no file writing.\n//\n// Warnings are DATA: every notice is returned in `warnings` verbatim and also\n// delivered incrementally via `onWarning` — the exact strings are load-bearing (the test suite asserts them).\nimport { buildOcrAnalysisPages } from './analyze.ts';\nimport { collect } from './collect.ts';\nimport { dictAvailable, dictUnavailableWarning } from './dict.ts';\nimport type { OcrProgressEvent } from './engines/tesseract.ts';\nimport type { PageWord, Tuning } from './geometry.ts';\nimport { toLines } from './lines.ts';\nimport { buildMarkdown } from './markdown.ts';\nimport { renderPagesToMarkdown } from './render-analysis.ts';\nimport { AUTO_OCR_CHARS_PER_PAGE, buildOcrReport, buildReport } from './report.ts';\nimport type { PageLines, Report } from './types.ts';\nimport { validatePages } from './validate.ts';\n\nexport type { OcrProgressEvent } from './engines/tesseract.ts';\n\n// A plain-JS caller (or `.ts` with `as any`) gets no compile-time check on\n// `pages`, so it's validated once at runtime — see src/validate.ts's `validatePages`.\n\n/** One recognized word with its page — the OCR replay input contract (same\n * shape as `geometry.ts`'s `PageWord`, re-exported under this module's public name). */\nexport type OcrWordInput = PageWord;\n\n/** A 1-based, inclusive page range — the shape every entry point's `pages`\n * option takes (the CLI's own `--pages N[-M]` parses into this). */\nexport interface PageSelection {\n first: number;\n /** inclusive; defaults to `first` */\n last?: number;\n}\n\n/** extractText/extractOcr/pdfToMarkdown's common return shape. */\nexport interface ExtractResult {\n markdown: string;\n /** per-document and per-page QA metrics — see `Report` (`src/types.ts`) */\n report: Report;\n /** every notice also delivered incrementally via `onWarning`, collected\n * here in full (hybrid pages, auto-OCR-fallback, empty OCR pages, sparse-retry) */\n warnings: string[];\n}\n\ninterface CommonOptions {\n /** convert only this page range (1-based, inclusive); defaults to the\n * whole document */\n pages?: PageSelection;\n /** delivered as each warning happens (also returned, in full, as\n * `warnings` on the resolved result) */\n onWarning?: (message: string) => void;\n}\n\n/** Options for {@link extractText}. */\nexport interface ExtractTextOptions extends CommonOptions {\n /** emit a `### pN` marker before each page's content (output formatting\n * only — `report.pageStats` is always populated regardless) */\n pageMarkers?: boolean;\n}\n\n/** Options for {@link extractOcr}. */\nexport interface ExtractOcrOptions extends CommonOptions {\n /** override any subset of the OCR-path geometry defaults — see\n * DEFAULT_TUNING (`src/geometry.ts`) for each knob's measured default */\n tuning?: Partial<Tuning>;\n /** page-render resolution for engine recognition (default 288, measured;\n * >~400 measured WORSE on ordinary pages — not a \"bigger is safer\" knob) */\n dpi?: number;\n /** marks report.ocrFallback — set by pdfToMarkdown's auto-fallback */\n fallback?: boolean;\n /** one event per page as recognition finishes (timing/count only — engine\n * chatter and the sparse-retry notice arrive via `onWarning` instead) */\n onProgress?: (e: OcrProgressEvent) => void;\n /** every recognized word entering the geometry, delivered once after\n * recognition — the severity-1 audit instrument (--debug-words) */\n onWords?: (words: OcrWordInput[]) => void;\n}\n\n/** Options for {@link pdfToMarkdown}. */\nexport interface PdfToMarkdownOptions extends ExtractTextOptions, Omit<ExtractOcrOptions, 'fallback'> {\n /** force the OCR path even when a text layer exists */\n ocr?: boolean;\n /** never fall back to OCR, even for image-only pages */\n noOcr?: boolean;\n}\n\n/**\n * Text-layer extraction. Returns markdown + QA report + warnings (hybrid\n * pages carrying large images whose baked-in text the text layer can never see). Makes no OCR decision — see pdfToMarkdown for the fallback.\n *\n * @param pdfPath path to the PDF file\n * @param opts see {@link ExtractTextOptions}\n */\n// One collector per entry point: collects into `warnings`, streams via\n// onWarning, and emits the dictionary-unavailable notice exactly once up front.\nfunction warnCollector(onWarning?: (message: string) => void): { warnings: string[]; warn: (m: string) => void } {\n const warnings: string[] = [];\n const warn = (m: string) => {\n warnings.push(m);\n onWarning?.(m);\n };\n if (!dictAvailable()) warn(dictUnavailableWarning());\n return { warnings, warn };\n}\n\nexport async function extractText(pdfPath: string, opts: ExtractTextOptions = {}): Promise<ExtractResult> {\n validatePages(opts.pages);\n const { warnings, warn } = warnCollector(opts.onWarning);\n\n const range = opts.pages ? { first: opts.pages.first, last: opts.pages.last ?? opts.pages.first } : {};\n const { numPages, pages, heightChars } = await collect(pdfPath, range);\n const pageLines: PageLines[] = pages.map((p) => ({\n page: p.page,\n lines: toLines(p),\n }));\n const pageMarkers = opts.pageMarkers ?? false;\n const { md, stats, bodyH, headingSizes } = buildMarkdown(pageLines, heightChars, { pageMarkers });\n // pageStats needs \"### pN\" markers to segment content; when the caller\n // didn't request them, build a throwaway marked variant (same heightChars) purely to measure stats from, so pageStats is always populated.\n const statsMd = pageMarkers ? md : buildMarkdown(pageLines, heightChars, { pageMarkers: true }).md;\n\n // Advisory, not prescriptive: `hasLargeImage` fires on any large image,\n // even one with no baked-in text, where `--ocr` recovers nothing and can\n // actively degrade a correct result (measured: \"I\" -> \"|\", stray em-dashes) — so this tells the caller to COMPARE the OCR rerun, never blindly replace.\n const hybridPages = pages.filter((p) => p.largeImage).map((p) => p.page);\n for (const n of hybridPages) {\n warn(`p${n}: text layer used but page carries large image(s); if it has baked-in text this output won't have it — compare (don't blindly replace) against --pages ${n} --ocr, which can itself introduce errors`);\n }\n\n const report = buildReport({\n numPages,\n md,\n statsMd,\n stats,\n bodyH,\n headingSizes,\n hybridPages,\n });\n return { markdown: md, report, warnings };\n}\n\n/**\n * OCR extraction. Input is either a PDF (pages render at `dpi` and the\n * engine recognizes them, imported lazily) or an already-recognized word dump (the offline replay input — see `--words-json`).\n *\n * @param input `{ pdfPath }` to recognize, or `{ words }` to replay a dump\n * @param opts see {@link ExtractOcrOptions}\n */\nexport async function extractOcr(input: { pdfPath: string } | { words: OcrWordInput[] }, opts: ExtractOcrOptions = {}): Promise<ExtractResult> {\n validatePages(opts.pages);\n const { warnings, warn } = warnCollector(opts.onWarning);\n\n // Shared with analyze(pdf, {path:'ocr'}) — see that module's header for\n // why this makes the two byte-identical by construction, not by hand.\n const { pages, words } = await buildOcrAnalysisPages(input, opts, warn);\n const md = renderPagesToMarkdown(pages, { forcePageMarkers: true, pageMarkers: true });\n\n const report = buildOcrReport({\n md,\n fallback: opts.fallback,\n // Confidence is on `words` — see report.ts's low-confidence detector for\n // how it handles a confidence-less replay honestly.\n words,\n });\n return { markdown: md, report, warnings };\n}\n\n/** pdfToMarkdown's auto-OCR fallback predicate, shared with the CLI's\n * --format=raw|txt path so the two call sites can never drift apart. Module-internal: not re-exported by `src/index.ts`. */\nexport function needsOcrFallback(charsPerPage: number): boolean {\n return charsPerPage < AUTO_OCR_CHARS_PER_PAGE;\n}\n\n/** The exact stderr/warnings notice printed when needsOcrFallback fires —\n * one copy of the string for both call sites (asserted verbatim by the suite). */\nexport function ocrFallbackNotice(charsPerPage: number): string {\n return `text layer yields ~${charsPerPage} chars/page (< ${AUTO_OCR_CHARS_PER_PAGE}) — pages are images of text; falling back to OCR (suppress with --no-ocr)`;\n}\n\n/**\n * The headline API: extract a PDF to markdown, text layer first, falling\n * back to OCR when the text layer is empty enough the pages are certainly images of text (< `AUTO_OCR_CHARS_PER_PAGE` chars/page) — the same decision the CLI makes.\n *\n * @param pdfPath path to the PDF file\n * @param opts see {@link PdfToMarkdownOptions}\n */\nexport async function pdfToMarkdown(pdfPath: string, opts: PdfToMarkdownOptions = {}): Promise<ExtractResult> {\n if (opts.ocr) return extractOcr({ pdfPath }, opts);\n\n // Warning DELIVERY is deferred on the text pass: if the fallback fires,\n // those hybrid warnings describe output being discarded, so only a kept text result replays them.\n const text = await extractText(pdfPath, { ...opts, onWarning: undefined });\n if (needsOcrFallback(text.report.charsPerPage) && !opts.noOcr) {\n const notice = ocrFallbackNotice(text.report.charsPerPage);\n opts.onWarning?.(notice);\n const ocr = await extractOcr({ pdfPath }, { ...opts, fallback: true });\n ocr.warnings.unshift(notice);\n return ocr;\n }\n for (const w of text.warnings) opts.onWarning?.(w);\n return text;\n}\n"],"names":["extractOcr","extractText","needsOcrFallback","ocrFallbackNotice","pdfToMarkdown","warnCollector","onWarning","warnings","warn","m","push","dictAvailable","dictUnavailableWarning","pdfPath","opts","range","numPages","pages","heightChars","pageLines","pageMarkers","buildMarkdown","md","stats","bodyH","headingSizes","statsMd","hybridPages","n","report","validatePages","first","last","collect","map","p","page","lines","toLines","filter","largeImage","buildReport","markdown","input","words","buildOcrAnalysisPages","renderPagesToMarkdown","forcePageMarkers","buildOcrReport","fallback","charsPerPage","AUTO_OCR_CHARS_PER_PAGE","text","notice","ocr","w","undefined","noOcr","unshift"],"mappings":"AAAA,4EAA4E;AAC5E,kIAAkI;AAClI,EAAE;AACF,8EAA8E;AAC9E,8GAA8G;;;;;;;;;;;;QAgJxFA;eAAAA;;QA3CAC;eAAAA;;QAgENC;eAAAA;;QAMAC;eAAAA;;QAWMC;eAAAA;;;yBArLgB;yBACd;sBAC8B;uBAG9B;0BACM;gCACQ;wBAC+B;0BAEvC;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAuE9B;;;;;;CAMC,GACD,uEAAuE;AACvE,gFAAgF;AAChF,SAASC,cAAcC,SAAqC;IAC1D,IAAMC,WAAqB,EAAE;IAC7B,IAAMC,OAAO,cAACC;QACZF,SAASG,IAAI,CAACD;QACdH,sBAAAA,gCAAAA,UAAYG;IACd;IACA,IAAI,CAACE,IAAAA,qBAAa,KAAIH,KAAKI,IAAAA,8BAAsB;IACjD,OAAO;QAAEL,UAAAA;QAAUC,MAAAA;IAAK;AAC1B;AAEO,SAAeP;wCAAYY,OAAe;YAAEC,MAIWA,kBAMxCA,mBAROT,gBAAnBE,UAAUC,MAEZO,OACmC,MAAjCC,UAAUC,OAAOC,aACnBC,WAIAC,aACqCC,gBAAnCC,IAAIC,OAAOC,OAAOC,cAGpBC,SAKAC,aACD,2BAAA,mBAAA,gBAAA,WAAA,OAAMC,GAILC;;;;;oBAxB2Cf,OAAAA,oEAA2B,CAAC;oBAC7EgB,IAAAA,yBAAa,EAAChB,KAAKG,KAAK;oBACGZ,iBAAAA,cAAcS,KAAKR,SAAS,GAA/CC,WAAmBF,eAAnBE,UAAUC,OAASH,eAATG;oBAEZO,QAAQD,KAAKG,KAAK,GAAG;wBAAEc,OAAOjB,KAAKG,KAAK,CAACc,KAAK;wBAAEC,IAAI,GAAElB,mBAAAA,KAAKG,KAAK,CAACe,IAAI,cAAflB,8BAAAA,mBAAmBA,KAAKG,KAAK,CAACc,KAAK;oBAAC,IAAI,CAAC;oBAC5D;;wBAAME,IAAAA,kBAAO,EAACpB,SAASE;;;oBAAvB,OAAA,eAAjCC,WAAiC,KAAjCA,UAAUC,QAAuB,KAAvBA,OAAOC,cAAgB,KAAhBA;oBACnBC,YAAyBF,MAAMiB,GAAG,CAAC,SAACC;+BAAO;4BAC/CC,MAAMD,EAAEC,IAAI;4BACZC,OAAOC,IAAAA,gBAAO,EAACH;wBACjB;;oBACMf,eAAcN,oBAAAA,KAAKM,WAAW,cAAhBN,+BAAAA,oBAAoB;oBACGO,iBAAAA,IAAAA,yBAAa,EAACF,WAAWD,aAAa;wBAAEE,aAAAA;oBAAY,IAAvFE,KAAmCD,eAAnCC,IAAIC,QAA+BF,eAA/BE,OAAOC,QAAwBH,eAAxBG,OAAOC,eAAiBJ,eAAjBI;oBAC1B,uEAAuE;oBACvE,2IAA2I;oBACrIC,UAAUN,cAAcE,KAAKD,IAAAA,yBAAa,EAACF,WAAWD,aAAa;wBAAEE,aAAa;oBAAK,GAAGE,EAAE;oBAElG,wEAAwE;oBACxE,yEAAyE;oBACzE,wJAAwJ;oBAClJK,cAAcV,MAAMsB,MAAM,CAAC,SAACJ;+BAAMA,EAAEK,UAAU;uBAAEN,GAAG,CAAC,SAACC;+BAAMA,EAAEC,IAAI;;oBAClE,kCAAA,2BAAA;;wBAAL,IAAK,YAAWT,kCAAX,6BAAA,QAAA,yBAAA,iCAAwB;4BAAlBC,IAAN;4BACHpB,KAAK,AAAC,IAA8JoB,OAA3JA,GAAE,2JAA2J,OAAFA,GAAE;wBACxK;;wBAFK;wBAAA;;;iCAAA,6BAAA;gCAAA;;;gCAAA;sCAAA;;;;oBAICC,SAASY,IAAAA,qBAAW,EAAC;wBACzBzB,UAAAA;wBACAM,IAAAA;wBACAI,SAAAA;wBACAH,OAAAA;wBACAC,OAAAA;wBACAC,cAAAA;wBACAE,aAAAA;oBACF;oBACA;;wBAAO;4BAAEe,UAAUpB;4BAAIO,QAAAA;4BAAQtB,UAAAA;wBAAS;;;;IAC1C;;AASO,SAAeP;wCAAW2C,KAAsD;YAAE7B,MAE5DT,gBAAnBE,UAAUC,MAIO,MAAjBS,OAAO2B,OACTtB,IAEAO;;;;;oBATiFf,OAAAA,oEAA0B,CAAC;oBAClHgB,IAAAA,yBAAa,EAAChB,KAAKG,KAAK;oBACGZ,iBAAAA,cAAcS,KAAKR,SAAS,GAA/CC,WAAmBF,eAAnBE,UAAUC,OAASH,eAATG;oBAIO;;wBAAMqC,IAAAA,gCAAqB,EAACF,OAAO7B,MAAMN;;;oBAAzC,OAAA,eAAjBS,QAAiB,KAAjBA,OAAO2B,QAAU,KAAVA;oBACTtB,KAAKwB,IAAAA,uCAAqB,EAAC7B,OAAO;wBAAE8B,kBAAkB;wBAAM3B,aAAa;oBAAK;oBAE9ES,SAASmB,IAAAA,wBAAc,EAAC;wBAC5B1B,IAAAA;wBACA2B,UAAUnC,KAAKmC,QAAQ;wBACvB,yEAAyE;wBACzE,oDAAoD;wBACpDL,OAAAA;oBACF;oBACA;;wBAAO;4BAAEF,UAAUpB;4BAAIO,QAAAA;4BAAQtB,UAAAA;wBAAS;;;;IAC1C;;AAIO,SAASL,iBAAiBgD,YAAoB;IACnD,OAAOA,eAAeC,iCAAuB;AAC/C;AAIO,SAAShD,kBAAkB+C,YAAoB;IACpD,OAAO,AAAC,sBAAmDC,OAA9BD,cAAa,mBAAyC,OAAxBC,iCAAuB,EAAC;AACrF;AASO,SAAe/C;wCAAcS,OAAe;YAAEC,MAapBA,iBARzBsC,MAGJtC,kBADMuC,QAEAC,KAIH,2BAAA,mBAAA,gBAAA,WAAA,OAAMC;;;;;oBAbwCzC,OAAAA,oEAA6B,CAAC;oBACjF,IAAIA,KAAKwC,GAAG,EAAE;;wBAAOtD,WAAW;4BAAEa,SAAAA;wBAAQ,GAAGC;;oBAIhC;;wBAAMb,YAAYY,SAAS,wCAAKC;4BAAMR,WAAWkD;;;;oBAAxDJ,OAAO;yBACTlD,CAAAA,iBAAiBkD,KAAKvB,MAAM,CAACqB,YAAY,KAAK,CAACpC,KAAK2C,KAAK,AAAD,GAAxDvD;;;;oBACImD,SAASlD,kBAAkBiD,KAAKvB,MAAM,CAACqB,YAAY;qBACzDpC,mBAAAA,KAAKR,SAAS,cAAdQ,uCAAAA,sBAAAA,MAAiBuC;oBACL;;wBAAMrD,WAAW;4BAAEa,SAAAA;wBAAQ,GAAG,wCAAKC;4BAAMmC,UAAU;;;;oBAAzDK,MAAM;oBACZA,IAAI/C,QAAQ,CAACmD,OAAO,CAACL;oBACrB;;wBAAOC;;;oBAEJ,kCAAA,2BAAA;;wBAAL,IAAK,YAAWF,KAAK7C,QAAQ,uBAAxB,6BAAA,QAAA,yBAAA;4BAAMgD,IAAN;6BAA0BzC,kBAAAA,KAAKR,SAAS,cAAdQ,sCAAAA,qBAAAA,MAAiByC;;;wBAA3C;wBAAA;;;iCAAA,6BAAA;gCAAA;;;gCAAA;sCAAA;;;;oBACL;;wBAAOH;;;;IACT"}
|
|
1
|
+
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/formats/pdf-to-md/src/extract.ts"],"sourcesContent":["// The library layer for PDF -> markdown conversion: extractText/extractOcr/\n// pdfToMarkdown each take plain args, return plain data, and throw PdfToMdError — no process.exit, no console.*, no file writing.\n//\n// Warnings are DATA: every notice is returned in `warnings` verbatim and also\n// delivered incrementally via `onWarning` — the exact strings are load-bearing (the test suite asserts them).\nimport { buildOcrAnalysisPages } from './analyze.ts';\nimport { collect } from './collect.ts';\nimport { dictAvailable, dictUnavailableWarning } from './dict.ts';\nimport type { OcrProgressEvent } from './engines/tesseract.ts';\nimport type { PageWord, Tuning } from './geometry.ts';\nimport { renderPagesToMarkdown } from './render-analysis.ts';\nimport { AUTO_OCR_CHARS_PER_PAGE, buildOcrReport } from './report.ts';\nimport { buildTextAnalysis, selectCollectedPages } from './text-analysis.ts';\nimport type { Analysis, AnalysisPage, CollectResult, Report } from './types.ts';\nimport { validatePages } from './validate.ts';\n\nexport type { OcrProgressEvent } from './engines/tesseract.ts';\n\n// A plain-JS caller (or `.ts` with `as any`) gets no compile-time check on\n// `pages`, so it's validated once at runtime — see src/validate.ts's `validatePages`.\n\n/** One recognized word with its page — the OCR replay input contract (same\n * shape as `geometry.ts`'s `PageWord`, re-exported under this module's public name). */\nexport type OcrWordInput = PageWord;\n\n/** A 1-based, inclusive page range — the shape every entry point's `pages`\n * option takes (the CLI's own `--pages N[-M]` parses into this). */\nexport interface PageSelection {\n first: number;\n /** inclusive; defaults to `first` */\n last?: number;\n}\n\n/** extractText/extractOcr/pdfToMarkdown's common return shape. */\nexport interface ExtractResult {\n markdown: string;\n /** per-document and per-page QA metrics — see `Report` (`src/types.ts`) */\n report: Report;\n /** every notice also delivered incrementally via `onWarning`, collected\n * here in full (hybrid pages, auto-OCR-fallback, empty OCR pages, sparse-retry) */\n warnings: string[];\n}\n\ninterface CommonOptions {\n /** convert only this page range (1-based, inclusive); defaults to the\n * whole document */\n pages?: PageSelection;\n /** delivered as each warning happens (also returned, in full, as\n * `warnings` on the resolved result) */\n onWarning?: (message: string) => void;\n}\n\n/** Options for {@link extractText}. */\nexport interface ExtractTextOptions extends CommonOptions {\n /** emit a `### pN` marker before each page's content (output formatting\n * only — `report.pageStats` is always populated regardless) */\n pageMarkers?: boolean;\n /** source PDF text runs entering layout, delivered once for the selected\n * final result. Runs retain measured bounds and are not fabricated words. */\n onWords?: (words: OcrWordInput[]) => void;\n}\n\n/** Options for {@link extractOcr}. */\nexport interface ExtractOcrOptions extends CommonOptions {\n /** override any subset of the OCR-path geometry defaults — see\n * DEFAULT_TUNING (`src/geometry.ts`) for each knob's measured default */\n tuning?: Partial<Tuning>;\n /** page-render resolution for engine recognition (default 288, measured;\n * >~400 measured WORSE on ordinary pages — not a \"bigger is safer\" knob) */\n dpi?: number;\n /** marks report.ocrFallback — set by pdfToMarkdown's auto-fallback */\n fallback?: boolean;\n /** one event per page as recognition finishes (timing/count only — engine\n * chatter and the sparse-retry notice arrive via `onWarning` instead) */\n onProgress?: (e: OcrProgressEvent) => void;\n /** every recognized word entering the geometry, delivered once after\n * recognition — the severity-1 audit instrument (--debug-words) */\n onWords?: (words: OcrWordInput[]) => void;\n}\n\n/** Options for {@link pdfToMarkdown}. */\nexport interface PdfToMarkdownOptions extends ExtractTextOptions, Omit<ExtractOcrOptions, 'fallback'> {\n /** force the OCR path even when a text layer exists */\n ocr?: boolean;\n /** never fall back to OCR, even for image-only pages */\n noOcr?: boolean;\n}\n\n// One collector per entry point: collects into `warnings`, streams via\n// onWarning, and emits the dictionary-unavailable notice exactly once up front.\nfunction warnCollector(onWarning?: (message: string) => void): { warnings: string[]; warn: (m: string) => void } {\n const warnings: string[] = [];\n const warn = (m: string) => {\n warnings.push(m);\n onWarning?.(m);\n };\n if (!dictAvailable()) warn(dictUnavailableWarning());\n return { warnings, warn };\n}\n\n/** Extract embedded text with a QA report and warnings; never runs OCR.\n * @param pdfPath path to the PDF file\n * @param opts see {@link ExtractTextOptions}\n */\nexport async function extractText(pdfPath: string, opts: ExtractTextOptions = {}): Promise<ExtractResult> {\n validatePages(opts.pages);\n const { warnings, warn } = warnCollector(opts.onWarning);\n const range = opts.pages ? { first: opts.pages.first, last: opts.pages.last ?? opts.pages.first } : {};\n const collected = await collect(pdfPath, range);\n const text = buildTextAnalysis(collected, opts.pageMarkers ?? false);\n textNotices(collected, new Set(), warn);\n opts.onWords?.(text.words);\n return { markdown: text.markdown, report: text.analysis.report, warnings };\n}\n\n/**\n * OCR extraction. Input is either a PDF (pages render at `dpi` and the\n * engine recognizes them, imported lazily) or an already-recognized word dump (the offline replay input — see `--words-json`).\n *\n * @param input `{ pdfPath }` to recognize, or `{ words }` to replay a dump\n * @param opts see {@link ExtractOcrOptions}\n */\nexport async function extractOcr(input: { pdfPath: string } | { words: OcrWordInput[] }, opts: ExtractOcrOptions = {}): Promise<ExtractResult> {\n validatePages(opts.pages);\n const { warnings, warn } = warnCollector(opts.onWarning);\n\n // Shared with analyze(pdf, {path:'ocr'}) — see that module's header for\n // why this makes the two byte-identical by construction, not by hand.\n const { pages, words } = await buildOcrAnalysisPages(input, opts, warn);\n const md = renderPagesToMarkdown(pages, { forcePageMarkers: true, pageMarkers: true });\n\n const report = buildOcrReport({\n md,\n fallback: opts.fallback,\n // Confidence is on `words` — see report.ts's low-confidence detector for\n // how it handles a confidence-less replay honestly.\n words,\n });\n return { markdown: md, report, warnings };\n}\n\n/** A page needs OCR only when its image signal accompanies an unusable text\n * layer. The caller supplies the image part of that decision. */\nfunction needsPageOcr(chars: number): boolean {\n return chars < AUTO_OCR_CHARS_PER_PAGE;\n}\n\nfunction textNotices(collected: CollectResult, ocrPages: Set<number>, warn: (message: string) => void): void {\n for (const page of collected.pages) {\n if (page.textChars !== 0 || ocrPages.has(page.page)) continue;\n if (page.hasImage) warn(`p${page.page}: no extractable text; page has image content and OCR was not selected`);\n else warn(`p${page.page}: no extractable text; the page may be blank or contain only vector content`);\n }\n for (const page of collected.pages) {\n if (!page.largeImage || ocrPages.has(page.page)) continue;\n warn(`p${page.page}: text layer used but page carries large image(s); if it has baked-in text this output won't have it — compare (don't blindly replace) against --pages ${page.page} --ocr, which can itself introduce errors`);\n }\n}\n\nfunction analysisWords(pages: AnalysisPage[]): OcrWordInput[] {\n return pages.flatMap((page) => page.words.map((word) => ({ page: page.page, text: word.text, ...word.box, confidence: word.confidence })));\n}\n\ninterface AutoAnalysisResult {\n analysis: Analysis;\n markdown: string;\n warnings: string[];\n}\n\n/** Shared automatic route for the library's Markdown result and the CLI's\n * raw/plain-text formats. A page changes producer only when it has image\n * evidence and an unusable text layer, so a blank or vector-only page never\n * becomes speculative OCR input. */\nexport async function analyzePdfAuto(pdfPath: string, opts: PdfToMarkdownOptions = {}): Promise<AutoAnalysisResult> {\n validatePages(opts.pages);\n const { warnings, warn } = warnCollector(opts.onWarning);\n const range = opts.pages ? { first: opts.pages.first, last: opts.pages.last ?? opts.pages.first } : {};\n\n if (opts.ocr) {\n const { pages, words } = await buildOcrAnalysisPages({ pdfPath }, { ...opts, onWords: undefined }, warn);\n const markdown = renderPagesToMarkdown(pages, { forcePageMarkers: true, pageMarkers: true });\n const report = buildOcrReport({ md: markdown, words });\n const analysis: Analysis = { outputVersion: report.outputVersion, report, pages };\n opts.onWords?.(words);\n return { analysis, markdown, warnings };\n }\n\n const collected = await collect(pdfPath, range);\n const ocrPageNumbers = new Set(!opts.noOcr ? collected.pages.filter((page) => page.hasImage && needsPageOcr(page.textChars)).map((page) => page.page) : []);\n if (!ocrPageNumbers.size) {\n const text = buildTextAnalysis(collected, opts.pageMarkers ?? false);\n textNotices(collected, ocrPageNumbers, warn);\n opts.onWords?.(text.words);\n return { analysis: text.analysis, markdown: text.markdown, warnings };\n }\n\n const textPageNumbers = new Set(collected.pages.filter((page) => !ocrPageNumbers.has(page.page)).map((page) => page.page));\n const text = textPageNumbers.size ? buildTextAnalysis(selectCollectedPages(collected, textPageNumbers), true) : undefined;\n const textByPage = new Map(text?.analysis.pages.map((page) => [page.page, page]));\n const ocrByPage = new Map<number, AnalysisPage>();\n const ocrWords: OcrWordInput[] = [];\n const ocrRanges: Array<{ first: number; last: number }> = [];\n for (const page of collected.pages) {\n if (!ocrPageNumbers.has(page.page)) continue;\n warn(`p${page.page}: sparse text layer with image content; using OCR for this page`);\n const lastRange = ocrRanges.at(-1);\n if (lastRange && lastRange.last + 1 === page.page) lastRange.last = page.page;\n else ocrRanges.push({ first: page.page, last: page.page });\n }\n for (const range of ocrRanges) {\n const result = await buildOcrAnalysisPages({ pdfPath }, { ...opts, pages: range, onWords: undefined }, warn);\n for (const page of result.pages) ocrByPage.set(page.page, page);\n ocrWords.push(...result.words);\n }\n textNotices(collected, ocrPageNumbers, warn);\n const pages = collected.pages.map((page) => ocrByPage.get(page.page) ?? (textByPage.get(page.page) as AnalysisPage));\n const markdown = renderPagesToMarkdown(pages, { forcePageMarkers: true, pageMarkers: true });\n const pagePaths = new Map(pages.map((page) => [page.page, page.path]));\n const report = buildOcrReport({\n md: markdown,\n fallback: true,\n words: ocrWords,\n pagePaths,\n hybridPages: text?.analysis.report.hybridPages,\n });\n if (text) {\n report.bodyFontHeight = text.analysis.report.bodyFontHeight;\n report.headingSizes = text.analysis.report.headingSizes;\n report.listItems = text.analysis.report.listItems;\n report.joinedPairs = text.analysis.report.joinedPairs;\n report.closedHyphens = text.analysis.report.closedHyphens;\n report.hybridPages = text.analysis.report.hybridPages;\n }\n const analysis: Analysis = { outputVersion: report.outputVersion, report, pages };\n opts.onWords?.(analysisWords(pages));\n return { analysis, markdown, warnings };\n}\n\n/**\n * Extract each page from embedded text, using OCR only when the page has\n * image content and fewer than `AUTO_OCR_CHARS_PER_PAGE` text characters.\n *\n * @param pdfPath path to the PDF file\n * @param opts see {@link PdfToMarkdownOptions}\n */\nexport async function pdfToMarkdown(pdfPath: string, opts: PdfToMarkdownOptions = {}): Promise<ExtractResult> {\n const result = await analyzePdfAuto(pdfPath, opts);\n return { markdown: result.markdown, report: result.analysis.report, warnings: result.warnings };\n}\n"],"names":["analyzePdfAuto","extractOcr","extractText","pdfToMarkdown","warnCollector","onWarning","warnings","warn","m","push","dictAvailable","dictUnavailableWarning","pdfPath","opts","range","collected","text","validatePages","pages","first","last","collect","buildTextAnalysis","pageMarkers","textNotices","Set","onWords","words","markdown","report","analysis","input","md","buildOcrAnalysisPages","renderPagesToMarkdown","forcePageMarkers","buildOcrReport","fallback","needsPageOcr","chars","AUTO_OCR_CHARS_PER_PAGE","ocrPages","page","textChars","has","hasImage","largeImage","analysisWords","flatMap","map","word","box","confidence","ocrPageNumbers","textPageNumbers","textByPage","ocrByPage","ocrWords","ocrRanges","lastRange","result","pagePaths","ocr","undefined","outputVersion","noOcr","filter","size","selectCollectedPages","Map","at","set","get","path","hybridPages","bodyFontHeight","headingSizes","listItems","joinedPairs","closedHyphens"],"mappings":"AAAA,4EAA4E;AAC5E,kIAAkI;AAClI,EAAE;AACF,8EAA8E;AAC9E,8GAA8G;;;;;;;;;;;;QAyKxFA;eAAAA;;QAnDAC;eAAAA;;QAlBAC;eAAAA;;QA6IAC;eAAAA;;;yBAhPgB;yBACd;sBAC8B;gCAGhB;wBACkB;8BACA;0BAE1B;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AA0E9B,uEAAuE;AACvE,gFAAgF;AAChF,SAASC,cAAcC,SAAqC;IAC1D,IAAMC,WAAqB,EAAE;IAC7B,IAAMC,OAAO,cAACC;QACZF,SAASG,IAAI,CAACD;QACdH,sBAAAA,gCAAAA,UAAYG;IACd;IACA,IAAI,CAACE,IAAAA,qBAAa,KAAIH,KAAKI,IAAAA,8BAAsB;IACjD,OAAO;QAAEL,UAAAA;QAAUC,MAAAA;IAAK;AAC1B;AAMO,SAAeL;wCAAYU,OAAe;YAAEC,MAGWA,kBAElBA,mBAE1CA,eAL2BT,gBAAnBE,UAAUC,MACZO,OACAC,WACAC;;;;;oBAL2CH,OAAAA,oEAA2B,CAAC;oBAC7EI,IAAAA,yBAAa,EAACJ,KAAKK,KAAK;oBACGd,iBAAAA,cAAcS,KAAKR,SAAS,GAA/CC,WAAmBF,eAAnBE,UAAUC,OAASH,eAATG;oBACZO,QAAQD,KAAKK,KAAK,GAAG;wBAAEC,OAAON,KAAKK,KAAK,CAACC,KAAK;wBAAEC,IAAI,GAAEP,mBAAAA,KAAKK,KAAK,CAACE,IAAI,cAAfP,8BAAAA,mBAAmBA,KAAKK,KAAK,CAACC,KAAK;oBAAC,IAAI,CAAC;oBACnF;;wBAAME,IAAAA,kBAAO,EAACT,SAASE;;;oBAAnCC,YAAY;oBACZC,OAAOM,IAAAA,iCAAiB,EAACP,YAAWF,oBAAAA,KAAKU,WAAW,cAAhBV,+BAAAA,oBAAoB;oBAC9DW,YAAYT,WAAW,IAAIU,OAAOlB;qBAClCM,gBAAAA,KAAKa,OAAO,cAAZb,oCAAAA,mBAAAA,MAAeG,KAAKW,KAAK;oBACzB;;wBAAO;4BAAEC,UAAUZ,KAAKY,QAAQ;4BAAEC,QAAQb,KAAKc,QAAQ,CAACD,MAAM;4BAAEvB,UAAAA;wBAAS;;;;IAC3E;;AASO,SAAeL;wCAAW8B,KAAsD;YAAElB,MAE5DT,gBAAnBE,UAAUC,MAIO,MAAjBW,OAAOS,OACTK,IAEAH;;;;;oBATiFhB,OAAAA,oEAA0B,CAAC;oBAClHI,IAAAA,yBAAa,EAACJ,KAAKK,KAAK;oBACGd,iBAAAA,cAAcS,KAAKR,SAAS,GAA/CC,WAAmBF,eAAnBE,UAAUC,OAASH,eAATG;oBAIO;;wBAAM0B,IAAAA,gCAAqB,EAACF,OAAOlB,MAAMN;;;oBAAzC,OAAA,eAAjBW,QAAiB,KAAjBA,OAAOS,QAAU,KAAVA;oBACTK,KAAKE,IAAAA,uCAAqB,EAAChB,OAAO;wBAAEiB,kBAAkB;wBAAMZ,aAAa;oBAAK;oBAE9EM,SAASO,IAAAA,wBAAc,EAAC;wBAC5BJ,IAAAA;wBACAK,UAAUxB,KAAKwB,QAAQ;wBACvB,yEAAyE;wBACzE,oDAAoD;wBACpDV,OAAAA;oBACF;oBACA;;wBAAO;4BAAEC,UAAUI;4BAAIH,QAAAA;4BAAQvB,UAAAA;wBAAS;;;;IAC1C;;AAEA;+DAC+D,GAC/D,SAASgC,aAAaC,KAAa;IACjC,OAAOA,QAAQC,iCAAuB;AACxC;AAEA,SAAShB,YAAYT,SAAwB,EAAE0B,QAAqB,EAAElC,IAA+B;QAC9F,kCAAA,2BAAA;;QAAL,QAAK,YAAcQ,UAAUG,KAAK,qBAA7B,SAAA,6BAAA,QAAA,yBAAA,iCAA+B;YAA/B,IAAMwB,OAAN;YACH,IAAIA,KAAKC,SAAS,KAAK,KAAKF,SAASG,GAAG,CAACF,KAAKA,IAAI,GAAG;YACrD,IAAIA,KAAKG,QAAQ,EAAEtC,KAAK,AAAC,IAAa,OAAVmC,KAAKA,IAAI,EAAC;iBACjCnC,KAAK,AAAC,IAAa,OAAVmC,KAAKA,IAAI,EAAC;QAC1B;;QAJK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;QAKA,mCAAA,4BAAA;;QAAL,QAAK,aAAc3B,UAAUG,KAAK,qBAA7B,UAAA,8BAAA,SAAA,0BAAA,kCAA+B;YAA/B,IAAMwB,QAAN;YACH,IAAI,CAACA,MAAKI,UAAU,IAAIL,SAASG,GAAG,CAACF,MAAKA,IAAI,GAAG;YACjDnC,KAAK,AAAC,IAAsKmC,OAAnKA,MAAKA,IAAI,EAAC,2JAAmK,OAAVA,MAAKA,IAAI,EAAC;QACxL;;QAHK;QAAA;;;iBAAA,8BAAA;gBAAA;;;gBAAA;sBAAA;;;;AAIP;AAEA,SAASK,cAAc7B,KAAqB;IAC1C,OAAOA,MAAM8B,OAAO,CAAC,SAACN;eAASA,KAAKf,KAAK,CAACsB,GAAG,CAAC,SAACC;mBAAU;gBAAER,MAAMA,KAAKA,IAAI;gBAAE1B,MAAMkC,KAAKlC,IAAI;eAAKkC,KAAKC,GAAG;gBAAEC,YAAYF,KAAKE,UAAU;;;;AACvI;AAYO,SAAepD;wCAAeY,OAAe;YAAEC,MAGQA,kBA0D5DA,eA3D2BT,gBAAnBE,UAAUC,MACZO,OAOJD,gBAJyB,MAAjBK,OAAOS,OACTC,UACAC,QACAC,UAKFf,WACAsC,gBAEsCxC,mBAE1CA,gBAFMG,MAMFsC,iBACAtC,OACAuC,YACAC,WACAC,UACAC,WACD,2BAAA,mBAAA,gBAAA,WAAA,OAAMhB,MAGHiB,WAIH,4BAAA,oBAAA,iBAAA,YAAA,QAAM7C,QAGT2C,WAFMG,QACD,4BAAA,oBAAA,iBAAA,YAAA,QAAMlB,YAIPxB,QACAU,WACAiC,WACAhC,SAeAC;;;;;oBA5D8CjB,OAAAA,oEAA6B,CAAC;oBAClFI,IAAAA,yBAAa,EAACJ,KAAKK,KAAK;oBACGd,iBAAAA,cAAcS,KAAKR,SAAS,GAA/CC,WAAmBF,eAAnBE,UAAUC,OAASH,eAATG;oBACZO,QAAQD,KAAKK,KAAK,GAAG;wBAAEC,OAAON,KAAKK,KAAK,CAACC,KAAK;wBAAEC,IAAI,GAAEP,mBAAAA,KAAKK,KAAK,CAACE,IAAI,cAAfP,8BAAAA,mBAAmBA,KAAKK,KAAK,CAACC,KAAK;oBAAC,IAAI,CAAC;yBAEjGN,KAAKiD,GAAG,EAARjD;;;;oBACuB;;wBAAMoB,IAAAA,gCAAqB,EAAC;4BAAErB,SAAAA;wBAAQ,GAAG,wCAAKC;4BAAMa,SAASqC;4BAAaxD;;;oBAA1E,OAAA,eAAjBW,QAAiB,KAAjBA,OAAOS,QAAU,KAAVA;oBACTC,WAAWM,IAAAA,uCAAqB,EAAChB,OAAO;wBAAEiB,kBAAkB;wBAAMZ,aAAa;oBAAK;oBACpFM,SAASO,IAAAA,wBAAc,EAAC;wBAAEJ,IAAIJ;wBAAUD,OAAAA;oBAAM;oBAC9CG,WAAqB;wBAAEkC,eAAenC,OAAOmC,aAAa;wBAAEnC,QAAAA;wBAAQX,OAAAA;oBAAM;qBAChFL,iBAAAA,KAAKa,OAAO,cAAZb,qCAAAA,oBAAAA,MAAec;oBACf;;wBAAO;4BAAEG,UAAAA;4BAAUF,UAAAA;4BAAUtB,UAAAA;wBAAS;;;oBAGtB;;wBAAMe,IAAAA,kBAAO,EAACT,SAASE;;;oBAAnCC,YAAY;oBACZsC,iBAAiB,IAAI5B,IAAI,CAACZ,KAAKoD,KAAK,GAAGlD,UAAUG,KAAK,CAACgD,MAAM,CAAC,SAACxB;+BAASA,KAAKG,QAAQ,IAAIP,aAAaI,KAAKC,SAAS;uBAAGM,GAAG,CAAC,SAACP;+BAASA,KAAKA,IAAI;;oBACpJ,IAAI,CAACW,eAAec,IAAI,EAAE;;;wBAClBnD,OAAOM,IAAAA,iCAAiB,EAACP,YAAWF,oBAAAA,KAAKU,WAAW,cAAhBV,+BAAAA,oBAAoB;wBAC9DW,YAAYT,WAAWsC,gBAAgB9C;yBACvCM,iBAAAA,KAAKa,OAAO,cAAZb,qCAAAA,oBAAAA,MAAeG,KAAKW,KAAK;wBACzB;;4BAAO;gCAAEG,UAAUd,KAAKc,QAAQ;gCAAEF,UAAUZ,KAAKY,QAAQ;gCAAEtB,UAAAA;4BAAS;;oBACtE;oBAEMgD,kBAAkB,IAAI7B,IAAIV,UAAUG,KAAK,CAACgD,MAAM,CAAC,SAACxB;+BAAS,CAACW,eAAeT,GAAG,CAACF,KAAKA,IAAI;uBAAGO,GAAG,CAAC,SAACP;+BAASA,KAAKA,IAAI;;oBAClH1B,QAAOsC,gBAAgBa,IAAI,GAAG7C,IAAAA,iCAAiB,EAAC8C,IAAAA,oCAAoB,EAACrD,WAAWuC,kBAAkB,QAAQS;oBAC1GR,aAAa,IAAIc,IAAIrD,kBAAAA,4BAAAA,MAAMc,QAAQ,CAACZ,KAAK,CAAC+B,GAAG,CAAC,SAACP;+BAAS;4BAACA,KAAKA,IAAI;4BAAEA;yBAAK;;oBACzEc,YAAY,IAAIa;oBAChBZ;oBACAC;oBACD,kCAAA,2BAAA;;wBAAL,IAAK,YAAc3C,UAAUG,KAAK,uBAA7B,6BAAA,QAAA,yBAAA,iCAA+B;4BAAzBwB,OAAN;4BACH,IAAI,CAACW,eAAeT,GAAG,CAACF,KAAKA,IAAI,GAAG;4BACpCnC,KAAK,AAAC,IAAa,OAAVmC,KAAKA,IAAI,EAAC;4BACbiB,YAAYD,UAAUY,EAAE,CAAC,CAAC;4BAChC,IAAIX,aAAaA,UAAUvC,IAAI,GAAG,MAAMsB,KAAKA,IAAI,EAAEiB,UAAUvC,IAAI,GAAGsB,KAAKA,IAAI;iCACxEgB,UAAUjD,IAAI,CAAC;gCAAEU,OAAOuB,KAAKA,IAAI;gCAAEtB,MAAMsB,KAAKA,IAAI;4BAAC;wBAC1D;;wBANK;wBAAA;;;iCAAA,6BAAA;gCAAA;;;gCAAA;sCAAA;;;;oBAOA,mCAAA,4BAAA;;;;;;;;;oBAAA,aAAegB;;;2BAAf,8BAAA,SAAA;;;;oBAAM5C,SAAN;oBACY;;wBAAMmB,IAAAA,gCAAqB,EAAC;4BAAErB,SAAAA;wBAAQ,GAAG,wCAAKC;4BAAMK,OAAOJ;4BAAOY,SAASqC;4BAAaxD;;;oBAAjGqD,SAAS;oBACV,mCAAA,4BAAA;;wBAAL,IAAK,aAAcA,OAAO1C,KAAK,uBAA1B,8BAAA,SAAA,0BAAA;4BAAMwB,QAAN;4BAA4Bc,UAAUe,GAAG,CAAC7B,MAAKA,IAAI,EAAEA;;;wBAArD;wBAAA;;;iCAAA,8BAAA;gCAAA;;;gCAAA;sCAAA;;;;oBACLe,CAAAA,YAAAA,UAAShD,IAAI,OAAbgD,WAAc,qBAAGG,OAAOjC,KAAK;;;oBAH1B;;;;;;;;;;;;oBAAA;oBAAA;;;;;;;6BAAA,8BAAA;4BAAA;;;4BAAA;kCAAA;;;;;;;oBAKLH,YAAYT,WAAWsC,gBAAgB9C;oBACjCW,SAAQH,UAAUG,KAAK,CAAC+B,GAAG,CAAC,SAACP;4BAASc;gCAAAA,iBAAAA,UAAUgB,GAAG,CAAC9B,KAAKA,IAAI,eAAvBc,4BAAAA,iBAA6BD,WAAWiB,GAAG,CAAC9B,KAAKA,IAAI;;oBAC3Fd,YAAWM,IAAAA,uCAAqB,EAAChB,QAAO;wBAAEiB,kBAAkB;wBAAMZ,aAAa;oBAAK;oBACpFsC,YAAY,IAAIQ,IAAInD,OAAM+B,GAAG,CAAC,SAACP;+BAAS;4BAACA,KAAKA,IAAI;4BAAEA,KAAK+B,IAAI;yBAAC;;oBAC9D5C,UAASO,IAAAA,wBAAc,EAAC;wBAC5BJ,IAAIJ;wBACJS,UAAU;wBACVV,OAAO8B;wBACPI,WAAAA;wBACAa,WAAW,EAAE1D,kBAAAA,4BAAAA,MAAMc,QAAQ,CAACD,MAAM,CAAC6C,WAAW;oBAChD;oBACA,IAAI1D,OAAM;wBACRa,QAAO8C,cAAc,GAAG3D,MAAKc,QAAQ,CAACD,MAAM,CAAC8C,cAAc;wBAC3D9C,QAAO+C,YAAY,GAAG5D,MAAKc,QAAQ,CAACD,MAAM,CAAC+C,YAAY;wBACvD/C,QAAOgD,SAAS,GAAG7D,MAAKc,QAAQ,CAACD,MAAM,CAACgD,SAAS;wBACjDhD,QAAOiD,WAAW,GAAG9D,MAAKc,QAAQ,CAACD,MAAM,CAACiD,WAAW;wBACrDjD,QAAOkD,aAAa,GAAG/D,MAAKc,QAAQ,CAACD,MAAM,CAACkD,aAAa;wBACzDlD,QAAO6C,WAAW,GAAG1D,MAAKc,QAAQ,CAACD,MAAM,CAAC6C,WAAW;oBACvD;oBACM5C,YAAqB;wBAAEkC,eAAenC,QAAOmC,aAAa;wBAAEnC,QAAAA;wBAAQX,OAAAA;oBAAM;qBAChFL,gBAAAA,KAAKa,OAAO,cAAZb,oCAAAA,mBAAAA,MAAekC,cAAc7B;oBAC7B;;wBAAO;4BAAEY,UAAAA;4BAAUF,UAAAA;4BAAUtB,UAAAA;wBAAS;;;;IACxC;;AASO,SAAeH;wCAAcS,OAAe;YAAEC,MAC7C+C;;;;;oBAD6C/C,OAAAA,oEAA6B,CAAC;oBAClE;;wBAAMb,eAAeY,SAASC;;;oBAAvC+C,SAAS;oBACf;;wBAAO;4BAAEhC,UAAUgC,OAAOhC,QAAQ;4BAAEC,QAAQ+B,OAAO9B,QAAQ,CAACD,MAAM;4BAAEvB,UAAUsD,OAAOtD,QAAQ;wBAAC;;;;IAChG"}
|
package/dist/cjs/lines.js
CHANGED
|
@@ -74,25 +74,32 @@ function toLines(param) {
|
|
|
74
74
|
}).length < glyphs.length * 0.15;
|
|
75
75
|
// Group on y with a tolerance that scales with font size.
|
|
76
76
|
var lines = [];
|
|
77
|
+
var indexedGlyphs = glyphs.map(function(glyph, index) {
|
|
78
|
+
return {
|
|
79
|
+
glyph: glyph,
|
|
80
|
+
index: index
|
|
81
|
+
};
|
|
82
|
+
});
|
|
77
83
|
var _iteratorNormalCompletion = true, _didIteratorError = false, _iteratorError = undefined;
|
|
78
84
|
try {
|
|
79
|
-
for(var _iterator =
|
|
80
|
-
return a.y - b.y || a.x - b.x;
|
|
85
|
+
for(var _iterator = indexedGlyphs.sort(function(a, b) {
|
|
86
|
+
return a.glyph.y - b.glyph.y || a.glyph.x - b.glyph.x;
|
|
81
87
|
})[Symbol.iterator](), _step; !(_iteratorNormalCompletion = (_step = _iterator.next()).done); _iteratorNormalCompletion = true){
|
|
82
|
-
var
|
|
88
|
+
var part = _step.value;
|
|
89
|
+
var g = part.glyph;
|
|
83
90
|
// NOTE: start-x assignment can misfile a gutter-straddling glyph; an
|
|
84
91
|
// overlap-based alternative was tried and reshuffled stray numerals with
|
|
85
92
|
// no measured win, so start-x stays (src/geometry.ts's OCR path assigns by x-span overlap instead, where the mis-assignment is measured).
|
|
86
93
|
var col = twoCol && g.x > mid ? 1 : 0;
|
|
87
94
|
var last = lines.at(-1);
|
|
88
95
|
if (last && last.col === col && Math.abs(last.y - g.y) < Math.max(2, g.h * 0.5)) {
|
|
89
|
-
last.parts.push(
|
|
96
|
+
last.parts.push(part);
|
|
90
97
|
last.y = (last.y + g.y) / 2;
|
|
91
98
|
} else lines.push({
|
|
92
99
|
col: col,
|
|
93
100
|
y: g.y,
|
|
94
101
|
parts: [
|
|
95
|
-
|
|
102
|
+
part
|
|
96
103
|
]
|
|
97
104
|
});
|
|
98
105
|
}
|
|
@@ -115,14 +122,14 @@ function toLines(param) {
|
|
|
115
122
|
}); // column, then down the page
|
|
116
123
|
return lines.map(function(l) {
|
|
117
124
|
var parts = l.parts.sort(function(a, b) {
|
|
118
|
-
return a.x - b.x;
|
|
125
|
+
return a.glyph.x - b.glyph.x;
|
|
119
126
|
});
|
|
120
127
|
var t = '';
|
|
121
128
|
var prev = null;
|
|
122
129
|
var _iteratorNormalCompletion = true, _didIteratorError = false, _iteratorError = undefined;
|
|
123
130
|
try {
|
|
124
131
|
for(var _iterator = parts[Symbol.iterator](), _step; !(_iteratorNormalCompletion = (_step = _iterator.next()).done); _iteratorNormalCompletion = true){
|
|
125
|
-
var
|
|
132
|
+
var _step_value = _step.value, cur = _step_value.glyph;
|
|
126
133
|
// insert a space when the measured gap is wider than intra-word kerning
|
|
127
134
|
if (prev && cur.x - (prev.x + prev.w) > prev.h * 0.12 && !/\s$/.test(t) && !/^\s/.test(cur.s)) t += ' ';
|
|
128
135
|
t += cur.s;
|
|
@@ -148,7 +155,7 @@ function toLines(param) {
|
|
|
148
155
|
var _iteratorNormalCompletion1 = true, _didIteratorError1 = false, _iteratorError1 = undefined;
|
|
149
156
|
try {
|
|
150
157
|
for(var _iterator1 = parts[Symbol.iterator](), _step1; !(_iteratorNormalCompletion1 = (_step1 = _iterator1.next()).done); _iteratorNormalCompletion1 = true){
|
|
151
|
-
var
|
|
158
|
+
var _step_value1 = _step1.value, p = _step_value1.glyph;
|
|
152
159
|
byH.set(p.h, (byH.get(p.h) || 0) + p.s.length);
|
|
153
160
|
}
|
|
154
161
|
} catch (err) {
|
|
@@ -167,12 +174,16 @@ function toLines(param) {
|
|
|
167
174
|
}
|
|
168
175
|
return {
|
|
169
176
|
text: normalize(t).replace(/\s+/g, ' ').trim(),
|
|
170
|
-
x: parts[0].x,
|
|
177
|
+
x: parts[0].glyph.x,
|
|
171
178
|
y: l.y,
|
|
172
179
|
h: _to_consumable_array(byH.entries()).sort(function(a, b) {
|
|
173
180
|
return b[1] - a[1];
|
|
174
181
|
})[0][0],
|
|
175
|
-
col: l.col
|
|
182
|
+
col: l.col,
|
|
183
|
+
wordIndexes: parts.map(function(param) {
|
|
184
|
+
var index = param.index;
|
|
185
|
+
return index;
|
|
186
|
+
})
|
|
176
187
|
};
|
|
177
188
|
}).filter(function(l) {
|
|
178
189
|
return l.text;
|
package/dist/cjs/lines.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/formats/pdf-to-md/src/lines.ts"],"sourcesContent":["// Pass 2: glyphs -> visual lines — de-braids two-column pages (spanning\n// lines like headings sort with the left column), joins words by measured horizontal gap, and folds typographic codepoints to ASCII so search doesn't miss half the hits.\nimport type { CollectedPage, Glyph, Line } from './types.ts';\n\nconst FOLD: Record<string, string> = {\n ʼ: \"'\",\n '’': \"'\",\n '‘': \"'\",\n '“': '\"',\n '”': '\"',\n '–': '-',\n};\n// Exported for the OCR engine seam (src/engines/tesseract.ts): both paths\n// must fold identically or the same printed apostrophe surfaces as two spellings depending on which path read the page.\nexport const foldTypographic = (s: string): string => s.replace(/[ʼ‘’“”–]/g, (c) => FOLD[c]);\nconst normalize = foldTypographic;\n\n// Intermediate row while grouping glyphs by visual line, before the final\n// text/height reduction below.\ninterface LineDraft {\n col: number;\n y: number;\n parts: Glyph
|
|
1
|
+
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/formats/pdf-to-md/src/lines.ts"],"sourcesContent":["// Pass 2: glyphs -> visual lines — de-braids two-column pages (spanning\n// lines like headings sort with the left column), joins words by measured horizontal gap, and folds typographic codepoints to ASCII so search doesn't miss half the hits.\nimport type { CollectedPage, Glyph, Line } from './types.ts';\n\nconst FOLD: Record<string, string> = {\n ʼ: \"'\",\n '’': \"'\",\n '‘': \"'\",\n '“': '\"',\n '”': '\"',\n '–': '-',\n};\n// Exported for the OCR engine seam (src/engines/tesseract.ts): both paths\n// must fold identically or the same printed apostrophe surfaces as two spellings depending on which path read the page.\nexport const foldTypographic = (s: string): string => s.replace(/[ʼ‘’“”–]/g, (c) => FOLD[c]);\nconst normalize = foldTypographic;\n\n// Intermediate row while grouping glyphs by visual line, before the final\n// text/height reduction below.\ninterface LineDraft {\n col: number;\n y: number;\n parts: Array<{ glyph: Glyph; index: number }>;\n}\n\nexport function toLines({ glyphs, width }: Pick<CollectedPage, 'glyphs' | 'width'>): Line[] {\n if (!glyphs.length) return [];\n\n const mid = width / 2;\n // A real two-column page has a gutter: clusters each side, almost nothing crossing mid.\n // Label/value layouts also cluster but their values cross mid — de-braiding would reorder them.\n const twoCol = glyphs.filter((g) => g.x + g.w < mid).length > glyphs.length * 0.25 && glyphs.filter((g) => g.x > mid).length > glyphs.length * 0.25 && glyphs.filter((g) => g.x < mid && g.x + g.w > mid).length < glyphs.length * 0.15;\n\n // Group on y with a tolerance that scales with font size.\n const lines: LineDraft[] = [];\n const indexedGlyphs = glyphs.map((glyph, index) => ({ glyph, index }));\n for (const part of indexedGlyphs.sort((a, b) => a.glyph.y - b.glyph.y || a.glyph.x - b.glyph.x)) {\n const { glyph: g } = part;\n // NOTE: start-x assignment can misfile a gutter-straddling glyph; an\n // overlap-based alternative was tried and reshuffled stray numerals with\n // no measured win, so start-x stays (src/geometry.ts's OCR path assigns by x-span overlap instead, where the mis-assignment is measured).\n const col = twoCol && g.x > mid ? 1 : 0;\n const last = lines.at(-1);\n if (last && last.col === col && Math.abs(last.y - g.y) < Math.max(2, g.h * 0.5)) {\n last.parts.push(part);\n last.y = (last.y + g.y) / 2;\n } else lines.push({ col, y: g.y, parts: [part] });\n }\n lines.sort((a, b) => a.col - b.col || a.y - b.y); // column, then down the page\n\n return lines\n .map((l): Line => {\n const parts = l.parts.sort((a, b) => a.glyph.x - b.glyph.x);\n let t = '';\n let prev: Glyph | null = null;\n for (const { glyph: cur } of parts) {\n // insert a space when the measured gap is wider than intra-word kerning\n if (prev && cur.x - (prev.x + prev.w) > prev.h * 0.12 && !/\\s$/.test(t) && !/^\\s/.test(cur.s)) t += ' ';\n t += cur.s;\n prev = cur;\n }\n // Line height = the char-weighted dominant height, not the max: a small\n // decorative large-face numeral must not promote a body line to a heading.\n const byH = new Map<number, number>();\n for (const { glyph: p } of parts) byH.set(p.h, (byH.get(p.h) || 0) + p.s.length);\n return {\n text: normalize(t).replace(/\\s+/g, ' ').trim(),\n x: parts[0].glyph.x,\n y: l.y,\n h: [...byH.entries()].sort((a, b) => b[1] - a[1])[0][0],\n col: l.col,\n wordIndexes: parts.map(({ index }) => index),\n };\n })\n .filter((l) => l.text);\n}\n"],"names":["foldTypographic","toLines","FOLD","ʼ","s","replace","c","normalize","glyphs","width","length","mid","twoCol","filter","g","x","w","lines","indexedGlyphs","map","glyph","index","sort","a","b","y","part","col","last","at","Math","abs","max","h","parts","push","l","t","prev","cur","test","byH","Map","p","set","get","text","trim","entries","wordIndexes"],"mappings":"AAAA,wEAAwE;AACxE,0KAA0K;;;;;;;;;;;;QAa7JA;eAAAA;;QAWGC;eAAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AArBhB,IAAMC,OAA+B;IACnCC,GAAG;IACH,KAAK;IACL,KAAK;IACL,KAAK;IACL,KAAK;IACL,KAAK;AACP;AAGO,IAAMH,kBAAkB,yBAACI;WAAsBA,EAAEC,OAAO,CAAC,aAAa,SAACC;eAAMJ,IAAI,CAACI,EAAE;;;AAC3F,IAAMC,YAAYP;AAUX,SAASC,QAAQ,KAA0D;QAAxDO,SAAF,MAAEA,QAAQC,QAAV,MAAUA;IAChC,IAAI,CAACD,OAAOE,MAAM,EAAE,OAAO,EAAE;IAE7B,IAAMC,MAAMF,QAAQ;IACpB,wFAAwF;IACxF,gGAAgG;IAChG,IAAMG,SAASJ,OAAOK,MAAM,CAAC,SAACC;eAAMA,EAAEC,CAAC,GAAGD,EAAEE,CAAC,GAAGL;OAAKD,MAAM,GAAGF,OAAOE,MAAM,GAAG,QAAQF,OAAOK,MAAM,CAAC,SAACC;eAAMA,EAAEC,CAAC,GAAGJ;OAAKD,MAAM,GAAGF,OAAOE,MAAM,GAAG,QAAQF,OAAOK,MAAM,CAAC,SAACC;eAAMA,EAAEC,CAAC,GAAGJ,OAAOG,EAAEC,CAAC,GAAGD,EAAEE,CAAC,GAAGL;OAAKD,MAAM,GAAGF,OAAOE,MAAM,GAAG;IAEnO,0DAA0D;IAC1D,IAAMO,QAAqB,EAAE;IAC7B,IAAMC,gBAAgBV,OAAOW,GAAG,CAAC,SAACC,OAAOC;eAAW;YAAED,OAAAA;YAAOC,OAAAA;QAAM;;QAC9D,kCAAA,2BAAA;;QAAL,QAAK,YAAcH,cAAcI,IAAI,CAAC,SAACC,GAAGC;mBAAMD,EAAEH,KAAK,CAACK,CAAC,GAAGD,EAAEJ,KAAK,CAACK,CAAC,IAAIF,EAAEH,KAAK,CAACL,CAAC,GAAGS,EAAEJ,KAAK,CAACL,CAAC;+BAAzF,SAAA,6BAAA,QAAA,yBAAA,iCAA4F;YAA5F,IAAMW,OAAN;YACH,IAAQN,AAAON,IAAMY,KAAbN;YACR,qEAAqE;YACrE,yEAAyE;YACzE,0IAA0I;YAC1I,IAAMO,MAAMf,UAAUE,EAAEC,CAAC,GAAGJ,MAAM,IAAI;YACtC,IAAMiB,OAAOX,MAAMY,EAAE,CAAC,CAAC;YACvB,IAAID,QAAQA,KAAKD,GAAG,KAAKA,OAAOG,KAAKC,GAAG,CAACH,KAAKH,CAAC,GAAGX,EAAEW,CAAC,IAAIK,KAAKE,GAAG,CAAC,GAAGlB,EAAEmB,CAAC,GAAG,MAAM;gBAC/EL,KAAKM,KAAK,CAACC,IAAI,CAACT;gBAChBE,KAAKH,CAAC,GAAG,AAACG,CAAAA,KAAKH,CAAC,GAAGX,EAAEW,CAAC,AAADA,IAAK;YAC5B,OAAOR,MAAMkB,IAAI,CAAC;gBAAER,KAAAA;gBAAKF,GAAGX,EAAEW,CAAC;gBAAES,OAAO;oBAACR;iBAAK;YAAC;QACjD;;QAXK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAYLT,MAAMK,IAAI,CAAC,SAACC,GAAGC;eAAMD,EAAEI,GAAG,GAAGH,EAAEG,GAAG,IAAIJ,EAAEE,CAAC,GAAGD,EAAEC,CAAC;QAAG,6BAA6B;IAE/E,OAAOR,MACJE,GAAG,CAAC,SAACiB;QACJ,IAAMF,QAAQE,EAAEF,KAAK,CAACZ,IAAI,CAAC,SAACC,GAAGC;mBAAMD,EAAEH,KAAK,CAACL,CAAC,GAAGS,EAAEJ,KAAK,CAACL,CAAC;;QAC1D,IAAIsB,IAAI;QACR,IAAIC,OAAqB;YACpB,kCAAA,2BAAA;;YAAL,QAAK,YAAwBJ,0BAAxB,SAAA,6BAAA,QAAA,yBAAA,iCAA+B;gBAA/B,kBAAA,aAAQd,AAAOmB,kBAAPnB;gBACX,wEAAwE;gBACxE,IAAIkB,QAAQC,IAAIxB,CAAC,GAAIuB,CAAAA,KAAKvB,CAAC,GAAGuB,KAAKtB,CAAC,AAADA,IAAKsB,KAAKL,CAAC,GAAG,QAAQ,CAAC,MAAMO,IAAI,CAACH,MAAM,CAAC,MAAMG,IAAI,CAACD,IAAInC,CAAC,GAAGiC,KAAK;gBACpGA,KAAKE,IAAInC,CAAC;gBACVkC,OAAOC;YACT;;YALK;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;QAML,wEAAwE;QACxE,2EAA2E;QAC3E,IAAME,MAAM,IAAIC;YACX,mCAAA,4BAAA;;YAAL,QAAK,aAAsBR,0BAAtB,UAAA,8BAAA,SAAA,0BAAA;gBAAA,mBAAA,cAAQd,AAAOuB,iBAAPvB;gBAAqBqB,IAAIG,GAAG,CAACD,EAAEV,CAAC,EAAE,AAACQ,CAAAA,IAAII,GAAG,CAACF,EAAEV,CAAC,KAAK,CAAA,IAAKU,EAAEvC,CAAC,CAACM,MAAM;;;YAA1E;YAAA;;;qBAAA,8BAAA;oBAAA;;;oBAAA;0BAAA;;;;QACL,OAAO;YACLoC,MAAMvC,UAAU8B,GAAGhC,OAAO,CAAC,QAAQ,KAAK0C,IAAI;YAC5ChC,GAAGmB,KAAK,CAAC,EAAE,CAACd,KAAK,CAACL,CAAC;YACnBU,GAAGW,EAAEX,CAAC;YACNQ,GAAG,AAAC,qBAAGQ,IAAIO,OAAO,IAAI1B,IAAI,CAAC,SAACC,GAAGC;uBAAMA,CAAC,CAAC,EAAE,GAAGD,CAAC,CAAC,EAAE;cAAC,CAAC,EAAE,CAAC,EAAE;YACvDI,KAAKS,EAAET,GAAG;YACVsB,aAAaf,MAAMf,GAAG,CAAC;oBAAGE,cAAAA;uBAAYA;;QACxC;IACF,GACCR,MAAM,CAAC,SAACuB;eAAMA,EAAEU,IAAI;;AACzB"}
|
package/dist/cjs/markdown.js
CHANGED
|
@@ -65,6 +65,7 @@ function buildMarkdown(pageLines, heightChars) {
|
|
|
65
65
|
];
|
|
66
66
|
}));
|
|
67
67
|
var out = [];
|
|
68
|
+
var blockPages = new Map();
|
|
68
69
|
var stats = {
|
|
69
70
|
headings: 0,
|
|
70
71
|
listItems: 0,
|
|
@@ -74,126 +75,202 @@ function buildMarkdown(pageLines, heightChars) {
|
|
|
74
75
|
callouts: 0,
|
|
75
76
|
closedHyphens: 0
|
|
76
77
|
};
|
|
78
|
+
var currentPage = 0;
|
|
77
79
|
var para = [];
|
|
80
|
+
var addBlock = function addBlock(block) {
|
|
81
|
+
var blocks = blockPages.get(currentPage);
|
|
82
|
+
if (blocks) blocks.push(block);
|
|
83
|
+
else blockPages.set(currentPage, [
|
|
84
|
+
block
|
|
85
|
+
]);
|
|
86
|
+
};
|
|
78
87
|
// a label line ending in ':' whose next line is a bare number belongs on one line
|
|
79
|
-
function attachValue(label, lines,
|
|
80
|
-
var j =
|
|
88
|
+
function attachValue(label, lines, i) {
|
|
89
|
+
var j = i + 1;
|
|
81
90
|
if (/[::]\s*$/.test(label) && j < lines.length && /^[0-9][\d.,]*%?$/.test(lines[j].text)) {
|
|
82
|
-
set(j);
|
|
83
91
|
stats.joinedPairs++;
|
|
84
|
-
return
|
|
92
|
+
return {
|
|
93
|
+
text: "".concat(label, " ").concat(lines[j].text),
|
|
94
|
+
last: j
|
|
95
|
+
};
|
|
85
96
|
}
|
|
86
|
-
return
|
|
97
|
+
return {
|
|
98
|
+
text: label,
|
|
99
|
+
last: i
|
|
100
|
+
};
|
|
87
101
|
}
|
|
88
102
|
var flush = function flush() {
|
|
89
103
|
if (!para.length) return;
|
|
90
|
-
|
|
104
|
+
var text = para.map(function(part) {
|
|
105
|
+
return part.text;
|
|
106
|
+
}).join(' ').replace(/\s+([,.;:%])/g, '$1');
|
|
107
|
+
out.push(text);
|
|
91
108
|
out.push('');
|
|
109
|
+
addBlock({
|
|
110
|
+
type: 'paragraph',
|
|
111
|
+
text: text,
|
|
112
|
+
wordIndexes: para.flatMap(function(part) {
|
|
113
|
+
return part.wordIndexes;
|
|
114
|
+
})
|
|
115
|
+
});
|
|
92
116
|
stats.paragraphs++;
|
|
93
117
|
para = [];
|
|
94
118
|
};
|
|
95
119
|
// append a wrapped line to the open paragraph, resolving a line-final hyphen
|
|
96
|
-
var appendToPara = function appendToPara(text) {
|
|
120
|
+
var appendToPara = function appendToPara(line, text) {
|
|
97
121
|
var prev = para.at(-1);
|
|
98
|
-
if (prev && /\w-$/.test(prev)) {
|
|
122
|
+
if (prev && /\w-$/.test(prev.text)) {
|
|
123
|
+
var _para__wordIndexes;
|
|
99
124
|
var _text_split_;
|
|
100
125
|
// a bare " -" is punctuation, not a wrapped word
|
|
101
|
-
var head = prev.replace(/-$/, '').split(/\s/).pop();
|
|
126
|
+
var head = prev.text.replace(/-$/, '').split(/\s/).pop();
|
|
102
127
|
var tail = ((_text_split_ = text.split(/\s/)[0]) !== null && _text_split_ !== void 0 ? _text_split_ : '').replace(/[^A-Za-z-]/g, '').split('-')[0];
|
|
103
128
|
var close = (0, _dictts.getDict)().has((head + tail).toLowerCase());
|
|
104
129
|
if (close) stats.closedHyphens++;
|
|
105
|
-
para[para.length - 1] = prev.replace(/-$/, close ? '' : '-') + text;
|
|
106
|
-
|
|
130
|
+
para[para.length - 1].text = prev.text.replace(/-$/, close ? '' : '-') + text;
|
|
131
|
+
(_para__wordIndexes = para[para.length - 1].wordIndexes).push.apply(_para__wordIndexes, _to_consumable_array(line.wordIndexes));
|
|
132
|
+
} else para.push({
|
|
133
|
+
text: text,
|
|
134
|
+
wordIndexes: _to_consumable_array(line.wordIndexes)
|
|
135
|
+
});
|
|
107
136
|
};
|
|
108
137
|
var _iteratorNormalCompletion = true, _didIteratorError = false, _iteratorError = undefined;
|
|
109
138
|
try {
|
|
110
139
|
for(var _iterator = pageLines[Symbol.iterator](), _step; !(_iteratorNormalCompletion = (_step = _iterator.next()).done); _iteratorNormalCompletion = true){
|
|
111
|
-
var
|
|
112
|
-
|
|
113
|
-
|
|
140
|
+
var _step_value = _step.value, page = _step_value.page, lines = _step_value.lines;
|
|
141
|
+
var _Math;
|
|
142
|
+
currentPage = page;
|
|
143
|
+
if (opts.pageMarkers) {
|
|
144
|
+
flush();
|
|
145
|
+
out.push((0, _pagemarkerts.pageMarkerLine)(page));
|
|
146
|
+
out.push('');
|
|
147
|
+
}
|
|
148
|
+
if (!lines.length) continue;
|
|
149
|
+
var indentBase = (_Math = Math).min.apply(_Math, _to_consumable_array(lines.map(function(l) {
|
|
150
|
+
return l.x;
|
|
151
|
+
})));
|
|
152
|
+
for(var i = 0; i < lines.length; i++){
|
|
153
|
+
var _lines_i = lines[i], text = _lines_i.text, h = _lines_i.h, x = _lines_i.x;
|
|
154
|
+
var prevLine = i > 0 ? lines[i - 1] : null;
|
|
114
155
|
// a column change or a vertical jump wider than a blank line ends the paragraph
|
|
115
|
-
if (para.length && prevLine && (lines[
|
|
156
|
+
if (para.length && prevLine && (lines[i].col !== prevLine.col || lines[i].y - prevLine.y > h * 2.2)) flush();
|
|
116
157
|
// A large-font line opens a run: gather the consecutive same-size lines
|
|
117
158
|
// below it (a title wraps across lines). A short merged run is one heading; a long one is a callout/pull-quote in a bigger face — prose, not structure — and becomes a blockquote.
|
|
118
159
|
if (levelOf.has(h) && !BULLETS.test(text)) {
|
|
119
160
|
var run = [
|
|
120
|
-
|
|
161
|
+
lines[i]
|
|
121
162
|
];
|
|
122
|
-
while(
|
|
163
|
+
while(i + 1 < lines.length && lines[i + 1].h === h && lines[i + 1].col === lines[i].col && lines[i + 1].y - lines[i].y < h * 2.2 && !BULLETS.test(lines[i + 1].text))run.push(lines[++i]);
|
|
123
164
|
// join wrapped title lines, resolving line-final hyphens like body text
|
|
124
|
-
var merged = run.
|
|
165
|
+
var merged = run.map(function(line) {
|
|
166
|
+
return line.text;
|
|
167
|
+
}).reduce(function(acc, part) {
|
|
125
168
|
var _part_split_;
|
|
126
169
|
return /\w-$/.test(acc) ? acc.replace(/-$/, (0, _dictts.getDict)().has((acc.replace(/-$/, '').split(/\s/).pop() + ((_part_split_ = part.split(/\s/)[0]) !== null && _part_split_ !== void 0 ? _part_split_ : '').replace(/[^A-Za-z-]/g, '').split('-')[0]).toLowerCase()) ? '' : '-') + part : "".concat(acc, " ").concat(part);
|
|
127
170
|
}).replace(/\s+([,.;:%])/g, '$1');
|
|
171
|
+
var wordIndexes = run.flatMap(function(line) {
|
|
172
|
+
return line.wordIndexes;
|
|
173
|
+
});
|
|
128
174
|
flush();
|
|
129
175
|
if (merged.length > 150) {
|
|
130
176
|
out.push("> ".concat(merged));
|
|
131
177
|
out.push('');
|
|
178
|
+
addBlock({
|
|
179
|
+
type: 'paragraph',
|
|
180
|
+
text: "> ".concat(merged),
|
|
181
|
+
wordIndexes: wordIndexes
|
|
182
|
+
});
|
|
132
183
|
stats.callouts++;
|
|
133
184
|
} else {
|
|
134
|
-
|
|
185
|
+
var level = levelOf.get(h);
|
|
186
|
+
out.push("".concat('#'.repeat(level), " ").concat(merged));
|
|
135
187
|
out.push('');
|
|
188
|
+
addBlock({
|
|
189
|
+
type: 'heading',
|
|
190
|
+
text: merged,
|
|
191
|
+
level: level,
|
|
192
|
+
wordIndexes: wordIndexes
|
|
193
|
+
});
|
|
136
194
|
stats.headings++;
|
|
137
195
|
}
|
|
138
|
-
|
|
196
|
+
continue;
|
|
139
197
|
}
|
|
140
198
|
// Lines can carry several stacked glyphs ("• ▪") from nested source lists.
|
|
141
199
|
var stripped = text.replace(BULLETS, '').trim();
|
|
142
200
|
var wasBullet = stripped !== text;
|
|
143
201
|
if (wasBullet) text = stripped;
|
|
144
202
|
// bullet with nothing left on the line: attach the following line's text
|
|
145
|
-
if ((!text || GLYPH.test(text)) &&
|
|
203
|
+
if ((!text || GLYPH.test(text)) && i + 1 < lines.length) {
|
|
146
204
|
flush();
|
|
147
205
|
var depth = Math.max(0, Math.round((x - indentBase) / 18));
|
|
148
|
-
var
|
|
149
|
-
|
|
150
|
-
|
|
151
|
-
|
|
152
|
-
|
|
153
|
-
|
|
154
|
-
|
|
206
|
+
var itemLines = [
|
|
207
|
+
lines[i],
|
|
208
|
+
lines[++i]
|
|
209
|
+
];
|
|
210
|
+
var label = lines[i].text.replace(BULLETS, '').trim();
|
|
211
|
+
while(!label && i + 1 < lines.length){
|
|
212
|
+
// stacked glyphs
|
|
213
|
+
itemLines.push(lines[++i]);
|
|
214
|
+
label = lines[i].text.replace(BULLETS, '').trim();
|
|
215
|
+
}
|
|
216
|
+
var attached = attachValue(label, lines, i);
|
|
217
|
+
if (attached.last !== i) itemLines.push(lines[attached.last]);
|
|
218
|
+
i = attached.last;
|
|
219
|
+
var itemText = "".concat(' '.repeat(depth), "- ").concat(attached.text);
|
|
220
|
+
out.push(itemText);
|
|
221
|
+
addBlock({
|
|
222
|
+
type: 'paragraph',
|
|
223
|
+
text: itemText,
|
|
224
|
+
tight: true,
|
|
225
|
+
wordIndexes: itemLines.flatMap(function(line) {
|
|
226
|
+
return line.wordIndexes;
|
|
227
|
+
})
|
|
155
228
|
});
|
|
156
|
-
out.push("".concat(' '.repeat(depth), "- ").concat(label));
|
|
157
229
|
stats.listItems++;
|
|
158
|
-
|
|
230
|
+
continue;
|
|
159
231
|
}
|
|
160
232
|
// glyph(s) followed by text on the same line
|
|
161
233
|
if (wasBullet) {
|
|
162
234
|
flush();
|
|
163
235
|
var depth1 = Math.max(0, Math.round((x - indentBase) / 18));
|
|
164
|
-
|
|
165
|
-
|
|
166
|
-
|
|
167
|
-
|
|
168
|
-
|
|
236
|
+
var itemLines1 = [
|
|
237
|
+
lines[i]
|
|
238
|
+
];
|
|
239
|
+
var attached1 = attachValue(text, lines, i);
|
|
240
|
+
if (attached1.last !== i) itemLines1.push(lines[attached1.last]);
|
|
241
|
+
i = attached1.last;
|
|
242
|
+
var itemText1 = "".concat(' '.repeat(depth1), "- ").concat(attached1.text);
|
|
243
|
+
out.push(itemText1);
|
|
244
|
+
addBlock({
|
|
245
|
+
type: 'paragraph',
|
|
246
|
+
text: itemText1,
|
|
247
|
+
tight: true,
|
|
248
|
+
wordIndexes: itemLines1.flatMap(function(line) {
|
|
249
|
+
return line.wordIndexes;
|
|
250
|
+
})
|
|
251
|
+
});
|
|
169
252
|
stats.listItems++;
|
|
170
|
-
|
|
253
|
+
continue;
|
|
171
254
|
}
|
|
172
255
|
// `label:` on its own line followed by a bare value -> keep together
|
|
173
|
-
if (/[::]\s*$/.test(text) &&
|
|
256
|
+
if (/[::]\s*$/.test(text) && i + 1 < lines.length && /^[0-9][\d.,]*%?$/.test(lines[i + 1].text)) {
|
|
174
257
|
flush();
|
|
175
|
-
|
|
258
|
+
var itemText2 = "- ".concat(text, " ").concat(lines[++i].text);
|
|
259
|
+
out.push(itemText2);
|
|
260
|
+
addBlock({
|
|
261
|
+
type: 'paragraph',
|
|
262
|
+
text: itemText2,
|
|
263
|
+
tight: true,
|
|
264
|
+
wordIndexes: _to_consumable_array(lines[i - 1].wordIndexes).concat(_to_consumable_array(lines[i].wordIndexes))
|
|
265
|
+
});
|
|
176
266
|
stats.joinedPairs++;
|
|
177
|
-
|
|
267
|
+
continue;
|
|
178
268
|
}
|
|
179
269
|
// otherwise body text: re-flow wrapped lines into a paragraph
|
|
180
|
-
appendToPara(text);
|
|
270
|
+
appendToPara(lines[i], text);
|
|
181
271
|
if (/[.!?:]$/.test(text)) flush();
|
|
182
272
|
else stats.reflowed++;
|
|
183
|
-
i = i1, void 0;
|
|
184
|
-
};
|
|
185
|
-
var _step_value = _step.value, page = _step_value.page, lines = _step_value.lines;
|
|
186
|
-
var _Math;
|
|
187
|
-
if (opts.pageMarkers) {
|
|
188
|
-
flush();
|
|
189
|
-
out.push((0, _pagemarkerts.pageMarkerLine)(page));
|
|
190
|
-
out.push('');
|
|
191
273
|
}
|
|
192
|
-
if (!lines.length) continue;
|
|
193
|
-
var indentBase = (_Math = Math).min.apply(_Math, _to_consumable_array(lines.map(function(l) {
|
|
194
|
-
return l.x;
|
|
195
|
-
})));
|
|
196
|
-
for(var i = 0; i < lines.length; i++)_loop(i);
|
|
197
274
|
flush();
|
|
198
275
|
}
|
|
199
276
|
} catch (err) {
|
|
@@ -212,11 +289,20 @@ function buildMarkdown(pageLines, heightChars) {
|
|
|
212
289
|
}
|
|
213
290
|
flush();
|
|
214
291
|
var md = "".concat(out.join('\n').replace(/\n{3,}/g, '\n\n').trim(), "\n");
|
|
292
|
+
var pages = pageLines.map(function(param) {
|
|
293
|
+
var page = param.page;
|
|
294
|
+
var _blockPages_get;
|
|
295
|
+
return {
|
|
296
|
+
page: page,
|
|
297
|
+
blocks: (_blockPages_get = blockPages.get(page)) !== null && _blockPages_get !== void 0 ? _blockPages_get : []
|
|
298
|
+
};
|
|
299
|
+
});
|
|
215
300
|
return {
|
|
216
301
|
md: md,
|
|
217
302
|
stats: stats,
|
|
218
303
|
bodyH: bodyH,
|
|
219
|
-
headingSizes: headingSizes
|
|
304
|
+
headingSizes: headingSizes,
|
|
305
|
+
pages: pages
|
|
220
306
|
};
|
|
221
307
|
}
|
|
222
308
|
/* CJS INTEROP */ if (exports.__esModule && exports.default) { try { Object.defineProperty(exports.default, '__esModule', { value: true }); for (var key in exports) { exports.default[key] = exports[key]; } } catch (_) {}; module.exports = exports.default; }
|
package/dist/cjs/markdown.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/formats/pdf-to-md/src/markdown.ts"],"sourcesContent":["// Pass 3: lines -> markdown. Structure is measured, not guessed — heading\n// levels come from the font-height histogram, hyphen-closes only against a dictionary (see the inline comments below for each rule).\nimport { getDict } from './dict.ts';\nimport { pageMarkerLine } from './page-marker.ts';\nimport type { BuildMarkdownOptions, Line, MarkdownResult, MarkdownStats, PageLines } from './types.ts';\n\nconst GLYPH = /^[•◦▪▫●○·‣⁃]$/;\nconst BULLETS = /^(?:[•◦▪▫●○·‣⁃]\\s*)+/;\n\nexport function buildMarkdown(pageLines: PageLines[], heightChars: Map<number, number>, opts: BuildMarkdownOptions = {}): MarkdownResult {\n // body height = char-weighted mode; headings are the rarer, larger sizes.\n // The 1.35 multiplier keeps emphasized body text out of the outline.\n const bodyH = [...heightChars.entries()].sort((a, b) => b[1] - a[1])[0]?.[0] ?? 0;\n const headingSizes = [...heightChars.keys()].filter((h) => h > bodyH * 1.35).sort((a, b) => b - a);\n // with page markers on, content headings start below the page ### level;\n // sizes beyond the ###### budget share level 6 rather than being dropped\n const offset = opts.pageMarkers ? 3 : 0;\n const levelOf = new Map(headingSizes.slice(0, 6).map((h, i) => [h, Math.min(6, i + 1 + offset)]));\n\n const out: string[] = [];\n const stats: MarkdownStats = {\n headings: 0,\n listItems: 0,\n paragraphs: 0,\n joinedPairs: 0,\n reflowed: 0,\n callouts: 0,\n closedHyphens: 0,\n };\n let para: string[] = [];\n\n // a label line ending in ':' whose next line is a bare number belongs on one line\n function attachValue(label: string, lines: Line[], get: () => number, set: (n: number) => void): string {\n const j = get() + 1;\n if (/[::]\\s*$/.test(label) && j < lines.length && /^[0-9][\\d.,]*%?$/.test(lines[j].text)) {\n set(j);\n stats.joinedPairs++;\n return `${label} ${lines[j].text}`;\n }\n return label;\n }\n const flush = () => {\n if (!para.length) return;\n out.push(para.join(' ').replace(/\\s+([,.;:%])/g, '$1'));\n out.push('');\n stats.paragraphs++;\n para = [];\n };\n // append a wrapped line to the open paragraph, resolving a line-final hyphen\n const appendToPara = (text: string) => {\n const prev = para.at(-1);\n if (prev && /\\w-$/.test(prev)) {\n // a bare \" -\" is punctuation, not a wrapped word\n const head = prev.replace(/-$/, '').split(/\\s/).pop() as string;\n const tail = (text.split(/\\s/)[0] ?? '').replace(/[^A-Za-z-]/g, '').split('-')[0];\n const close = getDict().has((head + tail).toLowerCase());\n if (close) stats.closedHyphens++;\n para[para.length - 1] = prev.replace(/-$/, close ? '' : '-') + text;\n } else para.push(text);\n };\n\n for (const { page, lines } of pageLines) {\n if (opts.pageMarkers) {\n flush();\n out.push(pageMarkerLine(page));\n out.push('');\n }\n if (!lines.length) continue;\n const indentBase = Math.min(...lines.map((l) => l.x));\n\n for (let i = 0; i < lines.length; i++) {\n let { text, h, x } = lines[i];\n const prevLine = i > 0 ? lines[i - 1] : null;\n // a column change or a vertical jump wider than a blank line ends the paragraph\n if (para.length && prevLine && (lines[i].col !== prevLine.col || lines[i].y - prevLine.y > h * 2.2)) flush();\n\n // A large-font line opens a run: gather the consecutive same-size lines\n // below it (a title wraps across lines). A short merged run is one heading; a long one is a callout/pull-quote in a bigger face — prose, not structure — and becomes a blockquote.\n if (levelOf.has(h) && !BULLETS.test(text)) {\n const run = [text];\n while (i + 1 < lines.length && lines[i + 1].h === h && lines[i + 1].col === lines[i].col && lines[i + 1].y - lines[i].y < h * 2.2 && !BULLETS.test(lines[i + 1].text)) run.push(lines[++i].text);\n // join wrapped title lines, resolving line-final hyphens like body text\n const merged = run.reduce((acc, part) => (/\\w-$/.test(acc) ? acc.replace(/-$/, getDict().has(((acc.replace(/-$/, '').split(/\\s/).pop() as string) + (part.split(/\\s/)[0] ?? '').replace(/[^A-Za-z-]/g, '').split('-')[0]).toLowerCase()) ? '' : '-') + part : `${acc} ${part}`)).replace(/\\s+([,.;:%])/g, '$1');\n flush();\n if (merged.length > 150) {\n out.push(`> ${merged}`);\n out.push('');\n stats.callouts++;\n } else {\n out.push(`${'#'.repeat(levelOf.get(h) as number)} ${merged}`);\n out.push('');\n stats.headings++;\n }\n continue;\n }\n\n // Lines can carry several stacked glyphs (\"• ▪\") from nested source lists.\n const stripped = text.replace(BULLETS, '').trim();\n const wasBullet = stripped !== text;\n if (wasBullet) text = stripped;\n\n // bullet with nothing left on the line: attach the following line's text\n if ((!text || GLYPH.test(text)) && i + 1 < lines.length) {\n flush();\n const depth = Math.max(0, Math.round((x - indentBase) / 18));\n let label = lines[++i].text.replace(BULLETS, '').trim();\n while (!label && i + 1 < lines.length)\n // stacked glyphs\n label = lines[++i].text.replace(BULLETS, '').trim();\n label = attachValue(\n label,\n lines,\n () => i,\n (n) => {\n i = n;\n }\n );\n out.push(`${' '.repeat(depth)}- ${label}`);\n stats.listItems++;\n continue;\n }\n // glyph(s) followed by text on the same line\n if (wasBullet) {\n flush();\n const depth = Math.max(0, Math.round((x - indentBase) / 18));\n out.push(\n `${' '.repeat(depth)}- ${attachValue(\n text,\n lines,\n () => i,\n (n) => {\n i = n;\n }\n )}`\n );\n stats.listItems++;\n continue;\n }\n // `label:` on its own line followed by a bare value -> keep together\n if (/[::]\\s*$/.test(text) && i + 1 < lines.length && /^[0-9][\\d.,]*%?$/.test(lines[i + 1].text)) {\n flush();\n out.push(`- ${text} ${lines[++i].text}`);\n stats.joinedPairs++;\n continue;\n }\n // otherwise body text: re-flow wrapped lines into a paragraph\n appendToPara(text);\n if (/[.!?:]$/.test(text)) flush();\n else stats.reflowed++;\n }\n flush();\n }\n flush();\n\n const md = `${out\n .join('\\n')\n .replace(/\\n{3,}/g, '\\n\\n')\n .trim()}\\n`;\n return { md, stats, bodyH, headingSizes };\n}\n"],"names":["buildMarkdown","GLYPH","BULLETS","pageLines","heightChars","opts","bodyH","entries","sort","a","b","headingSizes","keys","filter","h","offset","pageMarkers","levelOf","Map","slice","map","i","Math","min","out","stats","headings","listItems","paragraphs","joinedPairs","reflowed","callouts","closedHyphens","para","attachValue","label","lines","get","set","j","test","length","text","flush","push","join","replace","appendToPara","prev","at","head","split","pop","tail","close","getDict","has","toLowerCase","x","prevLine","col","y","run","merged","reduce","acc","part","repeat","stripped","trim","wasBullet","depth","max","round","indentBase","n","page","pageMarkerLine","l","md"],"mappings":"AAAA,0EAA0E;AAC1E,qIAAqI;;;;;+BAQrHA;;;eAAAA;;;sBAPQ;4BACO;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAG/B,IAAMC,QAAQ;AACd,IAAMC,UAAU;AAET,SAASF,cAAcG,SAAsB,EAAEC,WAAgC;QAAEC,OAAAA,iEAA6B,CAAC;;QAGtG;IAFd,0EAA0E;IAC1E,qEAAqE;IACrE,IAAMC,iBAAQ,SAAA,AAAC,qBAAGF,YAAYG,OAAO,IAAIC,IAAI,CAAC,SAACC,GAAGC;eAAMA,CAAC,CAAC,EAAE,GAAGD,CAAC,CAAC,EAAE;MAAC,CAAC,EAAE,cAAzD,6BAAA,MAA2D,CAAC,EAAE,uCAAI;IAChF,IAAME,eAAe,AAAC,qBAAGP,YAAYQ,IAAI,IAAIC,MAAM,CAAC,SAACC;eAAMA,IAAIR,QAAQ;OAAME,IAAI,CAAC,SAACC,GAAGC;eAAMA,IAAID;;IAChG,yEAAyE;IACzE,yEAAyE;IACzE,IAAMM,SAASV,KAAKW,WAAW,GAAG,IAAI;IACtC,IAAMC,UAAU,IAAIC,IAAIP,aAAaQ,KAAK,CAAC,GAAG,GAAGC,GAAG,CAAC,SAACN,GAAGO;eAAM;YAACP;YAAGQ,KAAKC,GAAG,CAAC,GAAGF,IAAI,IAAIN;SAAQ;;IAE/F,IAAMS,MAAgB,EAAE;IACxB,IAAMC,QAAuB;QAC3BC,UAAU;QACVC,WAAW;QACXC,YAAY;QACZC,aAAa;QACbC,UAAU;QACVC,UAAU;QACVC,eAAe;IACjB;IACA,IAAIC,OAAiB,EAAE;IAEvB,kFAAkF;IAClF,SAASC,YAAYC,KAAa,EAAEC,KAAa,EAAEC,GAAiB,EAAEC,GAAwB;QAC5F,IAAMC,IAAIF,QAAQ;QAClB,IAAI,WAAWG,IAAI,CAACL,UAAUI,IAAIH,MAAMK,MAAM,IAAI,mBAAmBD,IAAI,CAACJ,KAAK,CAACG,EAAE,CAACG,IAAI,GAAG;YACxFJ,IAAIC;YACJd,MAAMI,WAAW;YACjB,OAAO,AAAC,GAAWO,OAATD,OAAM,KAAiB,OAAdC,KAAK,CAACG,EAAE,CAACG,IAAI;QAClC;QACA,OAAOP;IACT;IACA,IAAMQ,QAAQ;QACZ,IAAI,CAACV,KAAKQ,MAAM,EAAE;QAClBjB,IAAIoB,IAAI,CAACX,KAAKY,IAAI,CAAC,KAAKC,OAAO,CAAC,iBAAiB;QACjDtB,IAAIoB,IAAI,CAAC;QACTnB,MAAMG,UAAU;QAChBK,OAAO,EAAE;IACX;IACA,6EAA6E;IAC7E,IAAMc,eAAe,sBAACL;QACpB,IAAMM,OAAOf,KAAKgB,EAAE,CAAC,CAAC;QACtB,IAAID,QAAQ,OAAOR,IAAI,CAACQ,OAAO;gBAGfN;YAFd,iDAAiD;YACjD,IAAMQ,OAAOF,KAAKF,OAAO,CAAC,MAAM,IAAIK,KAAK,CAAC,MAAMC,GAAG;YACnD,IAAMC,OAAO,EAACX,eAAAA,KAAKS,KAAK,CAAC,KAAK,CAAC,EAAE,cAAnBT,0BAAAA,eAAuB,IAAII,OAAO,CAAC,eAAe,IAAIK,KAAK,CAAC,IAAI,CAAC,EAAE;YACjF,IAAMG,QAAQC,IAAAA,eAAO,IAAGC,GAAG,CAAC,AAACN,CAAAA,OAAOG,IAAG,EAAGI,WAAW;YACrD,IAAIH,OAAO7B,MAAMO,aAAa;YAC9BC,IAAI,CAACA,KAAKQ,MAAM,GAAG,EAAE,GAAGO,KAAKF,OAAO,CAAC,MAAMQ,QAAQ,KAAK,OAAOZ;QACjE,OAAOT,KAAKW,IAAI,CAACF;IACnB;QAEK,kCAAA,2BAAA;;QAAL,QAAK,YAAyBvC,8BAAzB,SAAA,6BAAA,QAAA,yBAAA,iCAAoC;;gBAUrC,IAAqBiC,WAAAA,KAAK,CAACf,GAAE,EAAvBqB,OAAeN,SAAfM,MAAM5B,IAASsB,SAATtB,GAAG4C,IAAMtB,SAANsB;gBACf,IAAMC,WAAWtC,KAAI,IAAIe,KAAK,CAACf,KAAI,EAAE,GAAG;gBACxC,gFAAgF;gBAChF,IAAIY,KAAKQ,MAAM,IAAIkB,YAAavB,CAAAA,KAAK,CAACf,GAAE,CAACuC,GAAG,KAAKD,SAASC,GAAG,IAAIxB,KAAK,CAACf,GAAE,CAACwC,CAAC,GAAGF,SAASE,CAAC,GAAG/C,IAAI,GAAE,GAAI6B;gBAErG,wEAAwE;gBACxE,mLAAmL;gBACnL,IAAI1B,QAAQuC,GAAG,CAAC1C,MAAM,CAACZ,QAAQsC,IAAI,CAACE,OAAO;oBACzC,IAAMoB,MAAM;wBAACpB;qBAAK;oBAClB,MAAOrB,KAAI,IAAIe,MAAMK,MAAM,IAAIL,KAAK,CAACf,KAAI,EAAE,CAACP,CAAC,KAAKA,KAAKsB,KAAK,CAACf,KAAI,EAAE,CAACuC,GAAG,KAAKxB,KAAK,CAACf,GAAE,CAACuC,GAAG,IAAIxB,KAAK,CAACf,KAAI,EAAE,CAACwC,CAAC,GAAGzB,KAAK,CAACf,GAAE,CAACwC,CAAC,GAAG/C,IAAI,OAAO,CAACZ,QAAQsC,IAAI,CAACJ,KAAK,CAACf,KAAI,EAAE,CAACqB,IAAI,EAAGoB,IAAIlB,IAAI,CAACR,KAAK,CAAC,EAAEf,GAAE,CAACqB,IAAI;oBAC/L,wEAAwE;oBACxE,IAAMqB,SAASD,IAAIE,MAAM,CAAC,SAACC,KAAKC;4BAAqHA;+BAA3G,OAAO1B,IAAI,CAACyB,OAAOA,IAAInB,OAAO,CAAC,MAAMS,IAAAA,eAAO,IAAGC,GAAG,CAAC,AAAC,CAAA,AAACS,IAAInB,OAAO,CAAC,MAAM,IAAIK,KAAK,CAAC,MAAMC,GAAG,KAAgB,EAACc,eAAAA,KAAKf,KAAK,CAAC,KAAK,CAAC,EAAE,cAAnBe,0BAAAA,eAAuB,IAAIpB,OAAO,CAAC,eAAe,IAAIK,KAAK,CAAC,IAAI,CAAC,EAAE,AAAD,EAAGM,WAAW,MAAM,KAAK,OAAOS,OAAO,AAAC,GAASA,OAAPD,KAAI,KAAQ,OAALC;uBAASpB,OAAO,CAAC,iBAAiB;oBAC1SH;oBACA,IAAIoB,OAAOtB,MAAM,GAAG,KAAK;wBACvBjB,IAAIoB,IAAI,CAAC,AAAC,KAAW,OAAPmB;wBACdvC,IAAIoB,IAAI,CAAC;wBACTnB,MAAMM,QAAQ;oBAChB,OAAO;wBACLP,IAAIoB,IAAI,CAAC,AAAC,GAA0CmB,OAAxC,IAAII,MAAM,CAAClD,QAAQoB,GAAG,CAACvB,KAAc,KAAU,OAAPiD;wBACpDvC,IAAIoB,IAAI,CAAC;wBACTnB,MAAMC,QAAQ;oBAChB;oBACA,eAAA;gBACF;gBAEA,2EAA2E;gBAC3E,IAAM0C,WAAW1B,KAAKI,OAAO,CAAC5C,SAAS,IAAImE,IAAI;gBAC/C,IAAMC,YAAYF,aAAa1B;gBAC/B,IAAI4B,WAAW5B,OAAO0B;gBAEtB,yEAAyE;gBACzE,IAAI,AAAC,CAAA,CAAC1B,QAAQzC,MAAMuC,IAAI,CAACE,KAAI,KAAMrB,KAAI,IAAIe,MAAMK,MAAM,EAAE;oBACvDE;oBACA,IAAM4B,QAAQjD,KAAKkD,GAAG,CAAC,GAAGlD,KAAKmD,KAAK,CAAC,AAACf,CAAAA,IAAIgB,UAAS,IAAK;oBACxD,IAAIvC,QAAQC,KAAK,CAAC,EAAEf,GAAE,CAACqB,IAAI,CAACI,OAAO,CAAC5C,SAAS,IAAImE,IAAI;oBACrD,MAAO,CAAClC,SAASd,KAAI,IAAIe,MAAMK,MAAM,CACnC,iBAAiB;oBACjBN,QAAQC,KAAK,CAAC,EAAEf,GAAE,CAACqB,IAAI,CAACI,OAAO,CAAC5C,SAAS,IAAImE,IAAI;oBACnDlC,QAAQD,YACNC,OACAC,OACA;+BAAMf;uBACN,SAACsD;wBACCtD,KAAIsD;oBACN;oBAEFnD,IAAIoB,IAAI,CAAC,AAAC,GAAyBT,OAAvB,KAAKgC,MAAM,CAACI,QAAO,MAAU,OAANpC;oBACnCV,MAAME,SAAS;oBACf,eAAA;gBACF;gBACA,6CAA6C;gBAC7C,IAAI2C,WAAW;oBACb3B;oBACA,IAAM4B,SAAQjD,KAAKkD,GAAG,CAAC,GAAGlD,KAAKmD,KAAK,CAAC,AAACf,CAAAA,IAAIgB,UAAS,IAAK;oBACxDlD,IAAIoB,IAAI,CACN,AAAC,GAAyBV,OAAvB,KAAKiC,MAAM,CAACI,SAAO,MAOpB,OAPwBrC,YACxBQ,MACAN,OACA;+BAAMf;uBACN,SAACsD;wBACCtD,KAAIsD;oBACN;oBAGJlD,MAAME,SAAS;oBACf,eAAA;gBACF;gBACA,qEAAqE;gBACrE,IAAI,WAAWa,IAAI,CAACE,SAASrB,KAAI,IAAIe,MAAMK,MAAM,IAAI,mBAAmBD,IAAI,CAACJ,KAAK,CAACf,KAAI,EAAE,CAACqB,IAAI,GAAG;oBAC/FC;oBACAnB,IAAIoB,IAAI,CAAC,AAAC,KAAYR,OAARM,MAAK,KAAmB,OAAhBN,KAAK,CAAC,EAAEf,GAAE,CAACqB,IAAI;oBACrCjB,MAAMI,WAAW;oBACjB,eAAA;gBACF;gBACA,8DAA8D;gBAC9DkB,aAAaL;gBACb,IAAI,UAAUF,IAAI,CAACE,OAAOC;qBACrBlB,MAAMK,QAAQ;;YACrB;YAxFG,kBAAA,aAAQ8C,mBAAAA,MAAMxC,oBAAAA;gBAOEd;YANnB,IAAIjB,KAAKW,WAAW,EAAE;gBACpB2B;gBACAnB,IAAIoB,IAAI,CAACiC,IAAAA,4BAAc,EAACD;gBACxBpD,IAAIoB,IAAI,CAAC;YACX;YACA,IAAI,CAACR,MAAMK,MAAM,EAAE;YACnB,IAAMiC,aAAapD,CAAAA,QAAAA,MAAKC,GAAG,OAARD,OAAS,qBAAGc,MAAMhB,GAAG,CAAC,SAAC0D;uBAAMA,EAAEpB,CAAC;;YAEnD,IAAK,IAAIrC,IAAI,GAAGA,IAAIe,MAAMK,MAAM,EAAEpB;YAgFlCsB;QACF;;QA1FK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IA2FLA;IAEA,IAAMoC,KAAK,AAAC,GAGF,OAHIvD,IACXqB,IAAI,CAAC,MACLC,OAAO,CAAC,WAAW,QACnBuB,IAAI,IAAG;IACV,OAAO;QAAEU,IAAAA;QAAItD,OAAAA;QAAOnB,OAAAA;QAAOK,cAAAA;IAAa;AAC1C"}
|
|
1
|
+
{"version":3,"sources":["/Users/kevin/Dev/OpenSource/formats/pdf-to-md/src/markdown.ts"],"sourcesContent":["// Pass 3: lines -> markdown. Structure is measured, not guessed — heading\n// levels come from the font-height histogram, hyphen-closes only against a dictionary (see the inline comments below for each rule).\nimport { getDict } from './dict.ts';\nimport { pageMarkerLine } from './page-marker.ts';\nimport type { AnalysisBlock, BuildMarkdownOptions, Line, MarkdownPage, MarkdownResult, MarkdownStats, PageLines } from './types.ts';\n\nconst GLYPH = /^[•◦▪▫●○·‣⁃]$/;\nconst BULLETS = /^(?:[•◦▪▫●○·‣⁃]\\s*)+/;\n\nexport function buildMarkdown(pageLines: PageLines[], heightChars: Map<number, number>, opts: BuildMarkdownOptions = {}): MarkdownResult {\n // body height = char-weighted mode; headings are the rarer, larger sizes.\n // The 1.35 multiplier keeps emphasized body text out of the outline.\n const bodyH = [...heightChars.entries()].sort((a, b) => b[1] - a[1])[0]?.[0] ?? 0;\n const headingSizes = [...heightChars.keys()].filter((h) => h > bodyH * 1.35).sort((a, b) => b - a);\n // with page markers on, content headings start below the page ### level;\n // sizes beyond the ###### budget share level 6 rather than being dropped\n const offset = opts.pageMarkers ? 3 : 0;\n const levelOf = new Map(headingSizes.slice(0, 6).map((h, i) => [h, Math.min(6, i + 1 + offset)]));\n\n const out: string[] = [];\n const blockPages = new Map<number, AnalysisBlock[]>();\n const stats: MarkdownStats = {\n headings: 0,\n listItems: 0,\n paragraphs: 0,\n joinedPairs: 0,\n reflowed: 0,\n callouts: 0,\n closedHyphens: 0,\n };\n let currentPage = 0;\n let para: { text: string; wordIndexes: number[] }[] = [];\n\n const addBlock = (block: AnalysisBlock) => {\n const blocks = blockPages.get(currentPage);\n if (blocks) blocks.push(block);\n else blockPages.set(currentPage, [block]);\n };\n\n // a label line ending in ':' whose next line is a bare number belongs on one line\n function attachValue(label: string, lines: Line[], i: number): { text: string; last: number } {\n const j = i + 1;\n if (/[::]\\s*$/.test(label) && j < lines.length && /^[0-9][\\d.,]*%?$/.test(lines[j].text)) {\n stats.joinedPairs++;\n return { text: `${label} ${lines[j].text}`, last: j };\n }\n return { text: label, last: i };\n }\n const flush = () => {\n if (!para.length) return;\n const text = para\n .map((part) => part.text)\n .join(' ')\n .replace(/\\s+([,.;:%])/g, '$1');\n out.push(text);\n out.push('');\n addBlock({ type: 'paragraph', text, wordIndexes: para.flatMap((part) => part.wordIndexes) });\n stats.paragraphs++;\n para = [];\n };\n // append a wrapped line to the open paragraph, resolving a line-final hyphen\n const appendToPara = (line: Line, text: string) => {\n const prev = para.at(-1);\n if (prev && /\\w-$/.test(prev.text)) {\n // a bare \" -\" is punctuation, not a wrapped word\n const head = prev.text.replace(/-$/, '').split(/\\s/).pop() as string;\n const tail = (text.split(/\\s/)[0] ?? '').replace(/[^A-Za-z-]/g, '').split('-')[0];\n const close = getDict().has((head + tail).toLowerCase());\n if (close) stats.closedHyphens++;\n para[para.length - 1].text = prev.text.replace(/-$/, close ? '' : '-') + text;\n para[para.length - 1].wordIndexes.push(...line.wordIndexes);\n } else para.push({ text, wordIndexes: [...line.wordIndexes] });\n };\n\n for (const { page, lines } of pageLines) {\n currentPage = page;\n if (opts.pageMarkers) {\n flush();\n out.push(pageMarkerLine(page));\n out.push('');\n }\n if (!lines.length) continue;\n const indentBase = Math.min(...lines.map((l) => l.x));\n\n for (let i = 0; i < lines.length; i++) {\n let { text, h, x } = lines[i];\n const prevLine = i > 0 ? lines[i - 1] : null;\n // a column change or a vertical jump wider than a blank line ends the paragraph\n if (para.length && prevLine && (lines[i].col !== prevLine.col || lines[i].y - prevLine.y > h * 2.2)) flush();\n\n // A large-font line opens a run: gather the consecutive same-size lines\n // below it (a title wraps across lines). A short merged run is one heading; a long one is a callout/pull-quote in a bigger face — prose, not structure — and becomes a blockquote.\n if (levelOf.has(h) && !BULLETS.test(text)) {\n const run = [lines[i]];\n while (i + 1 < lines.length && lines[i + 1].h === h && lines[i + 1].col === lines[i].col && lines[i + 1].y - lines[i].y < h * 2.2 && !BULLETS.test(lines[i + 1].text)) run.push(lines[++i]);\n // join wrapped title lines, resolving line-final hyphens like body text\n const merged = run\n .map((line) => line.text)\n .reduce((acc, part) => (/\\w-$/.test(acc) ? acc.replace(/-$/, getDict().has(((acc.replace(/-$/, '').split(/\\s/).pop() as string) + (part.split(/\\s/)[0] ?? '').replace(/[^A-Za-z-]/g, '').split('-')[0]).toLowerCase()) ? '' : '-') + part : `${acc} ${part}`))\n .replace(/\\s+([,.;:%])/g, '$1');\n const wordIndexes = run.flatMap((line) => line.wordIndexes);\n flush();\n if (merged.length > 150) {\n out.push(`> ${merged}`);\n out.push('');\n addBlock({ type: 'paragraph', text: `> ${merged}`, wordIndexes });\n stats.callouts++;\n } else {\n const level = levelOf.get(h) as number;\n out.push(`${'#'.repeat(level)} ${merged}`);\n out.push('');\n addBlock({ type: 'heading', text: merged, level, wordIndexes });\n stats.headings++;\n }\n continue;\n }\n\n // Lines can carry several stacked glyphs (\"• ▪\") from nested source lists.\n const stripped = text.replace(BULLETS, '').trim();\n const wasBullet = stripped !== text;\n if (wasBullet) text = stripped;\n\n // bullet with nothing left on the line: attach the following line's text\n if ((!text || GLYPH.test(text)) && i + 1 < lines.length) {\n flush();\n const depth = Math.max(0, Math.round((x - indentBase) / 18));\n const itemLines = [lines[i], lines[++i]];\n let label = lines[i].text.replace(BULLETS, '').trim();\n while (!label && i + 1 < lines.length) {\n // stacked glyphs\n itemLines.push(lines[++i]);\n label = lines[i].text.replace(BULLETS, '').trim();\n }\n const attached = attachValue(label, lines, i);\n if (attached.last !== i) itemLines.push(lines[attached.last]);\n i = attached.last;\n const itemText = `${' '.repeat(depth)}- ${attached.text}`;\n out.push(itemText);\n addBlock({ type: 'paragraph', text: itemText, tight: true, wordIndexes: itemLines.flatMap((line) => line.wordIndexes) });\n stats.listItems++;\n continue;\n }\n // glyph(s) followed by text on the same line\n if (wasBullet) {\n flush();\n const depth = Math.max(0, Math.round((x - indentBase) / 18));\n const itemLines = [lines[i]];\n const attached = attachValue(text, lines, i);\n if (attached.last !== i) itemLines.push(lines[attached.last]);\n i = attached.last;\n const itemText = `${' '.repeat(depth)}- ${attached.text}`;\n out.push(itemText);\n addBlock({ type: 'paragraph', text: itemText, tight: true, wordIndexes: itemLines.flatMap((line) => line.wordIndexes) });\n stats.listItems++;\n continue;\n }\n // `label:` on its own line followed by a bare value -> keep together\n if (/[::]\\s*$/.test(text) && i + 1 < lines.length && /^[0-9][\\d.,]*%?$/.test(lines[i + 1].text)) {\n flush();\n const itemText = `- ${text} ${lines[++i].text}`;\n out.push(itemText);\n addBlock({ type: 'paragraph', text: itemText, tight: true, wordIndexes: [...lines[i - 1].wordIndexes, ...lines[i].wordIndexes] });\n stats.joinedPairs++;\n continue;\n }\n // otherwise body text: re-flow wrapped lines into a paragraph\n appendToPara(lines[i], text);\n if (/[.!?:]$/.test(text)) flush();\n else stats.reflowed++;\n }\n flush();\n }\n flush();\n\n const md = `${out\n .join('\\n')\n .replace(/\\n{3,}/g, '\\n\\n')\n .trim()}\\n`;\n const pages: MarkdownPage[] = pageLines.map(({ page }) => ({ page, blocks: blockPages.get(page) ?? [] }));\n return { md, stats, bodyH, headingSizes, pages };\n}\n"],"names":["buildMarkdown","GLYPH","BULLETS","pageLines","heightChars","opts","bodyH","entries","sort","a","b","headingSizes","keys","filter","h","offset","pageMarkers","levelOf","Map","slice","map","i","Math","min","out","blockPages","stats","headings","listItems","paragraphs","joinedPairs","reflowed","callouts","closedHyphens","currentPage","para","addBlock","block","blocks","get","push","set","attachValue","label","lines","j","test","length","text","last","flush","part","join","replace","type","wordIndexes","flatMap","appendToPara","line","prev","at","head","split","pop","tail","close","getDict","has","toLowerCase","page","pageMarkerLine","indentBase","l","x","prevLine","col","y","run","merged","reduce","acc","level","repeat","stripped","trim","wasBullet","depth","max","round","itemLines","attached","itemText","tight","md","pages"],"mappings":"AAAA,0EAA0E;AAC1E,qIAAqI;;;;;+BAQrHA;;;eAAAA;;;sBAPQ;4BACO;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAG/B,IAAMC,QAAQ;AACd,IAAMC,UAAU;AAET,SAASF,cAAcG,SAAsB,EAAEC,WAAgC;QAAEC,OAAAA,iEAA6B,CAAC;;QAGtG;IAFd,0EAA0E;IAC1E,qEAAqE;IACrE,IAAMC,iBAAQ,SAAA,AAAC,qBAAGF,YAAYG,OAAO,IAAIC,IAAI,CAAC,SAACC,GAAGC;eAAMA,CAAC,CAAC,EAAE,GAAGD,CAAC,CAAC,EAAE;MAAC,CAAC,EAAE,cAAzD,6BAAA,MAA2D,CAAC,EAAE,uCAAI;IAChF,IAAME,eAAe,AAAC,qBAAGP,YAAYQ,IAAI,IAAIC,MAAM,CAAC,SAACC;eAAMA,IAAIR,QAAQ;OAAME,IAAI,CAAC,SAACC,GAAGC;eAAMA,IAAID;;IAChG,yEAAyE;IACzE,yEAAyE;IACzE,IAAMM,SAASV,KAAKW,WAAW,GAAG,IAAI;IACtC,IAAMC,UAAU,IAAIC,IAAIP,aAAaQ,KAAK,CAAC,GAAG,GAAGC,GAAG,CAAC,SAACN,GAAGO;eAAM;YAACP;YAAGQ,KAAKC,GAAG,CAAC,GAAGF,IAAI,IAAIN;SAAQ;;IAE/F,IAAMS,MAAgB,EAAE;IACxB,IAAMC,aAAa,IAAIP;IACvB,IAAMQ,QAAuB;QAC3BC,UAAU;QACVC,WAAW;QACXC,YAAY;QACZC,aAAa;QACbC,UAAU;QACVC,UAAU;QACVC,eAAe;IACjB;IACA,IAAIC,cAAc;IAClB,IAAIC,OAAkD,EAAE;IAExD,IAAMC,WAAW,kBAACC;QAChB,IAAMC,SAASb,WAAWc,GAAG,CAACL;QAC9B,IAAII,QAAQA,OAAOE,IAAI,CAACH;aACnBZ,WAAWgB,GAAG,CAACP,aAAa;YAACG;SAAM;IAC1C;IAEA,kFAAkF;IAClF,SAASK,YAAYC,KAAa,EAAEC,KAAa,EAAEvB,CAAS;QAC1D,IAAMwB,IAAIxB,IAAI;QACd,IAAI,WAAWyB,IAAI,CAACH,UAAUE,IAAID,MAAMG,MAAM,IAAI,mBAAmBD,IAAI,CAACF,KAAK,CAACC,EAAE,CAACG,IAAI,GAAG;YACxFtB,MAAMI,WAAW;YACjB,OAAO;gBAAEkB,MAAM,AAAC,GAAWJ,OAATD,OAAM,KAAiB,OAAdC,KAAK,CAACC,EAAE,CAACG,IAAI;gBAAIC,MAAMJ;YAAE;QACtD;QACA,OAAO;YAAEG,MAAML;YAAOM,MAAM5B;QAAE;IAChC;IACA,IAAM6B,QAAQ;QACZ,IAAI,CAACf,KAAKY,MAAM,EAAE;QAClB,IAAMC,OAAOb,KACVf,GAAG,CAAC,SAAC+B;mBAASA,KAAKH,IAAI;WACvBI,IAAI,CAAC,KACLC,OAAO,CAAC,iBAAiB;QAC5B7B,IAAIgB,IAAI,CAACQ;QACTxB,IAAIgB,IAAI,CAAC;QACTJ,SAAS;YAAEkB,MAAM;YAAaN,MAAAA;YAAMO,aAAapB,KAAKqB,OAAO,CAAC,SAACL;uBAASA,KAAKI,WAAW;;QAAE;QAC1F7B,MAAMG,UAAU;QAChBM,OAAO,EAAE;IACX;IACA,6EAA6E;IAC7E,IAAMsB,eAAe,sBAACC,MAAYV;QAChC,IAAMW,OAAOxB,KAAKyB,EAAE,CAAC,CAAC;QACtB,IAAID,QAAQ,OAAOb,IAAI,CAACa,KAAKX,IAAI,GAAG;gBAOlCb;gBAJca;YAFd,iDAAiD;YACjD,IAAMa,OAAOF,KAAKX,IAAI,CAACK,OAAO,CAAC,MAAM,IAAIS,KAAK,CAAC,MAAMC,GAAG;YACxD,IAAMC,OAAO,EAAChB,eAAAA,KAAKc,KAAK,CAAC,KAAK,CAAC,EAAE,cAAnBd,0BAAAA,eAAuB,IAAIK,OAAO,CAAC,eAAe,IAAIS,KAAK,CAAC,IAAI,CAAC,EAAE;YACjF,IAAMG,QAAQC,IAAAA,eAAO,IAAGC,GAAG,CAAC,AAACN,CAAAA,OAAOG,IAAG,EAAGI,WAAW;YACrD,IAAIH,OAAOvC,MAAMO,aAAa;YAC9BE,IAAI,CAACA,KAAKY,MAAM,GAAG,EAAE,CAACC,IAAI,GAAGW,KAAKX,IAAI,CAACK,OAAO,CAAC,MAAMY,QAAQ,KAAK,OAAOjB;YACzEb,CAAAA,qBAAAA,IAAI,CAACA,KAAKY,MAAM,GAAG,EAAE,CAACQ,WAAW,EAACf,IAAI,OAAtCL,oBAAuC,qBAAGuB,KAAKH,WAAW;QAC5D,OAAOpB,KAAKK,IAAI,CAAC;YAAEQ,MAAAA;YAAMO,aAAc,qBAAGG,KAAKH,WAAW;QAAE;IAC9D;QAEK,kCAAA,2BAAA;;QAAL,QAAK,YAAyBpD,8BAAzB,SAAA,6BAAA,QAAA,yBAAA,iCAAoC;YAApC,kBAAA,aAAQkE,mBAAAA,MAAMzB,oBAAAA;gBAQEtB;YAPnBY,cAAcmC;YACd,IAAIhE,KAAKW,WAAW,EAAE;gBACpBkC;gBACA1B,IAAIgB,IAAI,CAAC8B,IAAAA,4BAAc,EAACD;gBACxB7C,IAAIgB,IAAI,CAAC;YACX;YACA,IAAI,CAACI,MAAMG,MAAM,EAAE;YACnB,IAAMwB,aAAajD,CAAAA,QAAAA,MAAKC,GAAG,OAARD,OAAS,qBAAGsB,MAAMxB,GAAG,CAAC,SAACoD;uBAAMA,EAAEC,CAAC;;YAEnD,IAAK,IAAIpD,IAAI,GAAGA,IAAIuB,MAAMG,MAAM,EAAE1B,IAAK;gBACrC,IAAqBuB,WAAAA,KAAK,CAACvB,EAAE,EAAvB2B,OAAeJ,SAAfI,MAAMlC,IAAS8B,SAAT9B,GAAG2D,IAAM7B,SAAN6B;gBACf,IAAMC,WAAWrD,IAAI,IAAIuB,KAAK,CAACvB,IAAI,EAAE,GAAG;gBACxC,gFAAgF;gBAChF,IAAIc,KAAKY,MAAM,IAAI2B,YAAa9B,CAAAA,KAAK,CAACvB,EAAE,CAACsD,GAAG,KAAKD,SAASC,GAAG,IAAI/B,KAAK,CAACvB,EAAE,CAACuD,CAAC,GAAGF,SAASE,CAAC,GAAG9D,IAAI,GAAE,GAAIoC;gBAErG,wEAAwE;gBACxE,mLAAmL;gBACnL,IAAIjC,QAAQkD,GAAG,CAACrD,MAAM,CAACZ,QAAQ4C,IAAI,CAACE,OAAO;oBACzC,IAAM6B,MAAM;wBAACjC,KAAK,CAACvB,EAAE;qBAAC;oBACtB,MAAOA,IAAI,IAAIuB,MAAMG,MAAM,IAAIH,KAAK,CAACvB,IAAI,EAAE,CAACP,CAAC,KAAKA,KAAK8B,KAAK,CAACvB,IAAI,EAAE,CAACsD,GAAG,KAAK/B,KAAK,CAACvB,EAAE,CAACsD,GAAG,IAAI/B,KAAK,CAACvB,IAAI,EAAE,CAACuD,CAAC,GAAGhC,KAAK,CAACvB,EAAE,CAACuD,CAAC,GAAG9D,IAAI,OAAO,CAACZ,QAAQ4C,IAAI,CAACF,KAAK,CAACvB,IAAI,EAAE,CAAC2B,IAAI,EAAG6B,IAAIrC,IAAI,CAACI,KAAK,CAAC,EAAEvB,EAAE;oBAC1L,wEAAwE;oBACxE,IAAMyD,SAASD,IACZzD,GAAG,CAAC,SAACsC;+BAASA,KAAKV,IAAI;uBACvB+B,MAAM,CAAC,SAACC,KAAK7B;4BAAqHA;+BAA3G,OAAOL,IAAI,CAACkC,OAAOA,IAAI3B,OAAO,CAAC,MAAMa,IAAAA,eAAO,IAAGC,GAAG,CAAC,AAAC,CAAA,AAACa,IAAI3B,OAAO,CAAC,MAAM,IAAIS,KAAK,CAAC,MAAMC,GAAG,KAAgB,EAACZ,eAAAA,KAAKW,KAAK,CAAC,KAAK,CAAC,EAAE,cAAnBX,0BAAAA,eAAuB,IAAIE,OAAO,CAAC,eAAe,IAAIS,KAAK,CAAC,IAAI,CAAC,EAAE,AAAD,EAAGM,WAAW,MAAM,KAAK,OAAOjB,OAAO,AAAC,GAASA,OAAP6B,KAAI,KAAQ,OAAL7B;uBACrPE,OAAO,CAAC,iBAAiB;oBAC5B,IAAME,cAAcsB,IAAIrB,OAAO,CAAC,SAACE;+BAASA,KAAKH,WAAW;;oBAC1DL;oBACA,IAAI4B,OAAO/B,MAAM,GAAG,KAAK;wBACvBvB,IAAIgB,IAAI,CAAC,AAAC,KAAW,OAAPsC;wBACdtD,IAAIgB,IAAI,CAAC;wBACTJ,SAAS;4BAAEkB,MAAM;4BAAaN,MAAM,AAAC,KAAW,OAAP8B;4BAAUvB,aAAAA;wBAAY;wBAC/D7B,MAAMM,QAAQ;oBAChB,OAAO;wBACL,IAAMiD,QAAQhE,QAAQsB,GAAG,CAACzB;wBAC1BU,IAAIgB,IAAI,CAAC,AAAC,GAAuBsC,OAArB,IAAII,MAAM,CAACD,QAAO,KAAU,OAAPH;wBACjCtD,IAAIgB,IAAI,CAAC;wBACTJ,SAAS;4BAAEkB,MAAM;4BAAWN,MAAM8B;4BAAQG,OAAAA;4BAAO1B,aAAAA;wBAAY;wBAC7D7B,MAAMC,QAAQ;oBAChB;oBACA;gBACF;gBAEA,2EAA2E;gBAC3E,IAAMwD,WAAWnC,KAAKK,OAAO,CAACnD,SAAS,IAAIkF,IAAI;gBAC/C,IAAMC,YAAYF,aAAanC;gBAC/B,IAAIqC,WAAWrC,OAAOmC;gBAEtB,yEAAyE;gBACzE,IAAI,AAAC,CAAA,CAACnC,QAAQ/C,MAAM6C,IAAI,CAACE,KAAI,KAAM3B,IAAI,IAAIuB,MAAMG,MAAM,EAAE;oBACvDG;oBACA,IAAMoC,QAAQhE,KAAKiE,GAAG,CAAC,GAAGjE,KAAKkE,KAAK,CAAC,AAACf,CAAAA,IAAIF,UAAS,IAAK;oBACxD,IAAMkB,YAAY;wBAAC7C,KAAK,CAACvB,EAAE;wBAAEuB,KAAK,CAAC,EAAEvB,EAAE;qBAAC;oBACxC,IAAIsB,QAAQC,KAAK,CAACvB,EAAE,CAAC2B,IAAI,CAACK,OAAO,CAACnD,SAAS,IAAIkF,IAAI;oBACnD,MAAO,CAACzC,SAAStB,IAAI,IAAIuB,MAAMG,MAAM,CAAE;wBACrC,iBAAiB;wBACjB0C,UAAUjD,IAAI,CAACI,KAAK,CAAC,EAAEvB,EAAE;wBACzBsB,QAAQC,KAAK,CAACvB,EAAE,CAAC2B,IAAI,CAACK,OAAO,CAACnD,SAAS,IAAIkF,IAAI;oBACjD;oBACA,IAAMM,WAAWhD,YAAYC,OAAOC,OAAOvB;oBAC3C,IAAIqE,SAASzC,IAAI,KAAK5B,GAAGoE,UAAUjD,IAAI,CAACI,KAAK,CAAC8C,SAASzC,IAAI,CAAC;oBAC5D5B,IAAIqE,SAASzC,IAAI;oBACjB,IAAM0C,WAAW,AAAC,GAAyBD,OAAvB,KAAKR,MAAM,CAACI,QAAO,MAAkB,OAAdI,SAAS1C,IAAI;oBACxDxB,IAAIgB,IAAI,CAACmD;oBACTvD,SAAS;wBAAEkB,MAAM;wBAAaN,MAAM2C;wBAAUC,OAAO;wBAAMrC,aAAakC,UAAUjC,OAAO,CAAC,SAACE;mCAASA,KAAKH,WAAW;;oBAAE;oBACtH7B,MAAME,SAAS;oBACf;gBACF;gBACA,6CAA6C;gBAC7C,IAAIyD,WAAW;oBACbnC;oBACA,IAAMoC,SAAQhE,KAAKiE,GAAG,CAAC,GAAGjE,KAAKkE,KAAK,CAAC,AAACf,CAAAA,IAAIF,UAAS,IAAK;oBACxD,IAAMkB,aAAY;wBAAC7C,KAAK,CAACvB,EAAE;qBAAC;oBAC5B,IAAMqE,YAAWhD,YAAYM,MAAMJ,OAAOvB;oBAC1C,IAAIqE,UAASzC,IAAI,KAAK5B,GAAGoE,WAAUjD,IAAI,CAACI,KAAK,CAAC8C,UAASzC,IAAI,CAAC;oBAC5D5B,IAAIqE,UAASzC,IAAI;oBACjB,IAAM0C,YAAW,AAAC,GAAyBD,OAAvB,KAAKR,MAAM,CAACI,SAAO,MAAkB,OAAdI,UAAS1C,IAAI;oBACxDxB,IAAIgB,IAAI,CAACmD;oBACTvD,SAAS;wBAAEkB,MAAM;wBAAaN,MAAM2C;wBAAUC,OAAO;wBAAMrC,aAAakC,WAAUjC,OAAO,CAAC,SAACE;mCAASA,KAAKH,WAAW;;oBAAE;oBACtH7B,MAAME,SAAS;oBACf;gBACF;gBACA,qEAAqE;gBACrE,IAAI,WAAWkB,IAAI,CAACE,SAAS3B,IAAI,IAAIuB,MAAMG,MAAM,IAAI,mBAAmBD,IAAI,CAACF,KAAK,CAACvB,IAAI,EAAE,CAAC2B,IAAI,GAAG;oBAC/FE;oBACA,IAAMyC,YAAW,AAAC,KAAY/C,OAARI,MAAK,KAAmB,OAAhBJ,KAAK,CAAC,EAAEvB,EAAE,CAAC2B,IAAI;oBAC7CxB,IAAIgB,IAAI,CAACmD;oBACTvD,SAAS;wBAAEkB,MAAM;wBAAaN,MAAM2C;wBAAUC,OAAO;wBAAMrC,aAAa,AAAC,qBAAGX,KAAK,CAACvB,IAAI,EAAE,CAACkC,WAAW,SAAE,qBAAGX,KAAK,CAACvB,EAAE,CAACkC,WAAW;oBAAE;oBAC/H7B,MAAMI,WAAW;oBACjB;gBACF;gBACA,8DAA8D;gBAC9D2B,aAAab,KAAK,CAACvB,EAAE,EAAE2B;gBACvB,IAAI,UAAUF,IAAI,CAACE,OAAOE;qBACrBxB,MAAMK,QAAQ;YACrB;YACAmB;QACF;;QAjGK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAkGLA;IAEA,IAAM2C,KAAK,AAAC,GAGF,OAHIrE,IACX4B,IAAI,CAAC,MACLC,OAAO,CAAC,WAAW,QACnB+B,IAAI,IAAG;IACV,IAAMU,QAAwB3F,UAAUiB,GAAG,CAAC;YAAGiD,aAAAA;YAA4B5C;eAAhB;YAAE4C,MAAAA;YAAM/B,MAAM,GAAEb,kBAAAA,WAAWc,GAAG,CAAC8B,mBAAf5C,6BAAAA,kBAAwB,EAAE;QAAC;;IACtG,OAAO;QAAEoE,IAAAA;QAAInE,OAAAA;QAAOpB,OAAAA;QAAOK,cAAAA;QAAcmF,OAAAA;IAAM;AACjD"}
|
|
@@ -12,5 +12,5 @@ export declare function splitByPageMarker(md: string): {
|
|
|
12
12
|
/** Per-page body text, duplicate-marker slices CONCATENATED — a word anywhere
|
|
13
13
|
* on page N counts as on page N (the auditor's membership semantics). */
|
|
14
14
|
export declare function bodyByPage(md: string): Map<number, string>;
|
|
15
|
-
/** Count of
|
|
15
|
+
/** Count of visible or invisible page-provenance marker lines in `md`. */
|
|
16
16
|
export declare function countPageMarkers(md: string): number;
|
|
@@ -12,5 +12,5 @@ export declare function splitByPageMarker(md: string): {
|
|
|
12
12
|
/** Per-page body text, duplicate-marker slices CONCATENATED — a word anywhere
|
|
13
13
|
* on page N counts as on page N (the auditor's membership semantics). */
|
|
14
14
|
export declare function bodyByPage(md: string): Map<number, string>;
|
|
15
|
-
/** Count of
|
|
15
|
+
/** Count of visible or invisible page-provenance marker lines in `md`. */
|
|
16
16
|
export declare function countPageMarkers(md: string): number;
|