paperlint 2.0.0

package/dist/pdf-facts.js

@@ -0,0 +1,239 @@
+/**
+ * READ A FINISHED PDF with pdf.js — the shell around `pdf-geometry.ts` and `adapters/banal/xml.ts`. Page
+ * count, the fonts the pages draw text with, the last page's words, and every page's text boxes
+ * (the input banal measures page geometry from, written as pdftohtml XML by `adapters/banal/xml.ts`).
+ *
+ * pdf.js arrives as the npm package `unpdf` (a zero-dependency build of Mozilla's pdf.js), so
+ * reading a PDF needs nothing installed on the system. It replaced three poppler programs
+ * (`pdfinfo`, `pdffonts`, `pdftotext -bbox`), which every consumer had to install by hand; the
+ * equivalence was measured before the switch (issue #61, see `pdf-geometry.ts`).
+ *
+ * ── THE RESULT IS A UNION, AND A FAILED READ CANNOT LOOK LIKE A CLEAN ONE ─────
+ * `readPdf` never throws for a property of the file: an unreadable or encrypted PDF, and a PDF whose
+ * pages draw text while pdf.js resolves no font for it, come back as `ok: false` with a reason. The
+ * last case is the one that matters: on Node 20 pdf.js returns zero fonts and no error, and an empty
+ * font list reads as "no Type 3, nothing unembedded" — a clean verdict over nothing.
+ *
+ * ── LAZY ────────────────────────────────────────────────────────────────────────
+ * `unpdf` is imported on the first read, not when this module loads: `paperlint lint` imports the build
+ * and must not pay pdf.js's start-up for a command that never opens a PDF.
+ */
+// eslint-disable-next-line boundaries/dependencies -- legacy I/O, moves behind a port in #76
+import { readFileSync } from "node:fs";
+import { DEFAULT_ASCENT, DEFAULT_DESCENT, fontsOf, wordsOf, } from "./pdf-geometry.js";
+// eslint-disable-next-line boundaries/dependencies -- legacy layer, moves behind a port in #76
+import { fillsOf, isUpright } from "./adapters/pdfjs/fill.js";
+/**
+ * The options every read uses, exactly as measured in the spike. `fontExtraProperties` is
+ * load-bearing: without it pdf.js exports no `type`/`subtype`, and every font's program is unknown.
+ * `disableFontFace` keeps pdf.js from trying to install fonts into a page that does not exist here.
+ */
+export const PDFJS_OPTIONS = {
+    disableFontFace: true,
+    fontExtraProperties: true,
+    verbosity: 0,
+    useSystemFonts: false,
+    isEvalSupported: false,
+};
+let pdfjs = null;
+function loadPdfJs() {
+    // eslint-disable-next-line boundaries/dependencies -- legacy layer, moves behind a port in #76
+    pdfjs ??= import("unpdf").then((m) => m.getResolvedPDFJS());
+    return pdfjs;
+}
+const fail = (reason, detail) => ({
+    ok: false,
+    reason,
+    detail,
+});
+/** A font object pdf.js left in `commonObjs`, read into a `RawFont`, or null when it is not one. */
+export function rawFontOf(id, obj) {
+    if (typeof obj !== "object" || obj === null)
+        return null;
+    const o = obj;
+    if (!("loadedName" in o) && !("isType3Font" in o))
+        return null;
+    const str = (v) => typeof v === "string" ? v : undefined;
+    return {
+        id,
+        name: str(o["name"]) ?? "",
+        type: str(o["type"]),
+        subtype: str(o["subtype"]),
+        isType3Font: o["isType3Font"] === true,
+        missingFile: o["missingFile"] === true,
+    };
+}
+const isText = (it) => "str" in it && it.str.trim() !== "";
+/** The font pdf.js resolved for `id`, if it did. `get` throws on an unresolved id; `has` does not. */
+function resolved(page, id) {
+    return page.commonObjs.has(id) ? page.commonObjs.get(id) : null;
+}
+/**
+ * The fonts pdf.js has loaded for the document so far, by id. `commonObjs` is shared by every page
+ * and holds fonts alongside other objects; `rawFontOf` keeps only the fonts.
+ */
+function loadedFonts(page) {
+    const out = new Map();
+    for (const [id, obj] of page.commonObjs) {
+        const f = rawFontOf(id, obj);
+        if (f)
+            out.set(id, f);
+    }
+    return out;
+}
+/**
+ * One page: its text items, and the fonts it drew with — the fonts its text items name, plus every
+ * font first loaded while its operator list ran. The second half is not redundant: pdf.js can
+ * render a run in a font that no text item names (measured on an all-Type-3 page: four fonts
+ * loaded, two named by items; poppler lists four). A page that reuses an earlier page's fonts loads
+ * none, and its items name them.
+ */
+async function readPage(page, seen, lib) {
+    // The operator list is what makes pdf.js load a page's fonts into `commonObjs`.
+    const ops = await page.getOperatorList();
+    const items = (await page.getTextContent()).items.filter(isText);
+    const loaded = loadedFonts(page);
+    const named = new Set(items.map((it) => it.fontName));
+    const fonts = [...loaded.values()].filter((f) => named.has(f.id) || !seen.has(f.id));
+    return {
+        raw: { hasText: items.length > 0, fonts },
+        items,
+        layout: layoutOf(page, items, ops, lib),
+    };
+}
+/**
+ * A font's ascent and descent as fractions of its size, or the defaults when pdf.js has none.
+ *
+ * ⚠️ Type 3 fonts get the defaults: pdf.js exports no ascent for them. The obvious substitute,
+ * FontBBox scaled by FontMatrix (what poppler uses), was measured on `fixtures/pdf-facts/t3-all.pdf`
+ * and came out FURTHER from poppler than the defaults: 8.5 pt off on the left column against 7.2.
+ * Both are far below any balance tolerance, so the simpler rule stays.
+ */
+function metricsOf(font) {
+    const f = (font ?? {});
+    const ascent = f["ascent"];
+    if (typeof ascent !== "number" || !(ascent > 0))
+        return { ascent: DEFAULT_ASCENT, descent: DEFAULT_DESCENT };
+    const d = f["descent"];
+    return { ascent, descent: typeof d === "number" ? d : DEFAULT_DESCENT };
+}
+/** The last page's text runs in top-down coordinates, and its size. */
+function pageText(page, items) {
+    const vp = page.getViewport({ scale: 1 });
+    const runs = items.map((it) => {
+        const [a = 0, b = 0, , d = 0, e = 0, f = 0] = it.transform;
+        const [x, baseline] = vp.convertToViewportPoint(e, f);
+        const size = Math.hypot(a, b) || it.height || Math.abs(d);
+        return {
+            text: it.str,
+            x,
+            baseline,
+            width: it.width,
+            size,
+            ...metricsOf(resolved(page, it.fontName)),
+        };
+    });
+    return {
+        widthPt: vp.width,
+        heightPt: vp.height,
+        words: runs.flatMap(wordsOf),
+    };
+}
+/** The font name pdf.js resolved for an item, else its internal id. */
+function fontNameOf(font, id) {
+    const name = font?.name;
+    return typeof name === "string" && name ? name : id;
+}
+/**
+ * One text item as a box, in points, top-down. The size is the item's vertical scale, which is what
+ * pdftohtml reports as the font size; upright is judged in VIEWER space (the page's /Rotate applied),
+ * because that is where pdftohtml draws.
+ */
+function boxOf(page, vp, it, lib) {
+    const [a = 0, b = 0, c = 0, d = 0, e = 0, f = 0] = it.transform;
+    const [x, baseline] = vp.convertToViewportPoint(e, f);
+    const size = Math.hypot(c, d) || Math.hypot(a, b) || it.height;
+    const font = resolved(page, it.fontName);
+    const m = metricsOf(font);
+    return {
+        top: baseline - m.ascent * size,
+        left: x,
+        width: it.width,
+        height: (m.ascent - m.descent) * size,
+        size,
+        font: fontNameOf(font, it.fontName),
+        text: it.str,
+        upright: isUpright(lib.Util.transform(vp.transform, it.transform)),
+    };
+}
+/** Every text item of a page as a box with its fill — the input of `pdf2xml`. */
+function layoutOf(page, items, ops, lib) {
+    const vp = page.getViewport({ scale: 1 });
+    const fills = fillsOf(lib.OPS, ops, items.map((it) => it.str));
+    const boxes = items.map((it, k) => ({
+        ...boxOf(page, vp, it, lib),
+        fill: fills[k] ?? { kind: "unknown" },
+    }));
+    return { widthPt: vp.width, heightPt: vp.height, boxes };
+}
+/** Every page's fonts, and the last page's words. */
+async function factsOf(doc, lib) {
+    const read = [];
+    const seen = new Set();
+    for (let i = 1; i <= doc.numPages; i++) {
+        const page = await doc.getPage(i);
+        const r = await readPage(page, seen, lib);
+        for (const f of r.raw.fonts)
+            seen.add(f.id);
+        read.push({ page, ...r });
+    }
+    const [last] = read.slice(-1);
+    if (!last)
+        return fail("empty", "the PDF has no pages");
+    const fonts = fontsOf(read.map((r) => r.raw));
+    if (fonts.kind === "zero-fonts-on-text")
+        return fail("zero-fonts-on-text", `page(s) ${fonts.pages.join(", ")} draw text, but pdf.js resolved no font for it ` +
+            // eslint-disable-next-line no-restricted-globals -- legacy I/O, moves behind a port in #76
+            `(Node ${process.version}; paperlint needs Node >= 22.13, where pdf.js reports fonts)`);
+    return {
+        ok: true,
+        facts: {
+            pages: doc.numPages,
+            fonts,
+            last: pageText(last.page, last.items),
+            layout: read.map((r) => r.layout),
+        },
+    };
+}
+/** pdf.js's own exception names for the two conditions a caller can act on. */
+function failureOf(e) {
+    const err = e;
+    const detail = `${err.name ?? "Error"}: ${err.message ?? String(e)}`;
+    return fail(err.name === "PasswordException" ? "encrypted" : "unreadable", detail);
+}
+/** Read a PDF from disk. Never throws for a property of the file; see the module header. */
+export const readPdf = async (path) => {
+    let data;
+    try {
+        data = new Uint8Array(readFileSync(path));
+    }
+    catch (e) {
+        return fail("unreadable", e.message);
+    }
+    const lib = await loadPdfJs();
+    const task = lib.getDocument({ data, ...PDFJS_OPTIONS });
+    try {
+        return await factsOf(await task.promise, lib);
+    }
+    catch (e) {
+        return failureOf(e);
+    }
+    finally {
+        await task.destroy();
+    }
+};
+/** The one wording of a failed read, as a build step or a CLI prints it. */
+export function describeFailure(r, pdf) {
+    return `could not read ${pdf} with pdf.js (${r.reason}): ${r.detail}`;
+}
+//# sourceMappingURL=pdf-facts.js.map

package/dist/pdf-facts.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"pdf-facts.js","sourceRoot":"","sources":["../src/pdf-facts.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;GAmBG;AACH,6FAA6F;AAC7F,OAAO,EAAE,YAAY,EAAE,MAAM,SAAS,CAAC;AACvC,OAAO,EACL,cAAc,EACd,eAAe,EACf,OAAO,EACP,OAAO,GAMR,MAAM,mBAAmB,CAAC;AAE3B,+FAA+F;AAC/F,OAAO,EAAE,OAAO,EAAE,SAAS,EAAE,MAAM,0BAA0B,CAAC;AAiC9D;;;;GAIG;AACH,MAAM,CAAC,MAAM,aAAa,GAAG;IAC3B,eAAe,EAAE,IAAI;IACrB,mBAAmB,EAAE,IAAI;IACzB,SAAS,EAAE,CAAC;IACZ,cAAc,EAAE,KAAK;IACrB,eAAe,EAAE,KAAK;CACd,CAAC;AAEX,IAAI,KAAK,GAA0B,IAAI,CAAC;AACxC,SAAS,SAAS;IAChB,+FAA+F;IAC/F,KAAK,KAAK,MAAM,CAAC,OAAO,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,gBAAgB,EAAE,CAAC,CAAC;IAC5D,OAAO,KAAK,CAAC;AACf,CAAC;AAED,MAAM,IAAI,GAAG,CAAC,MAAsB,EAAE,MAAc,EAAW,EAAE,CAAC,CAAC;IACjE,EAAE,EAAE,KAAK;IACT,MAAM;IACN,MAAM;CACP,CAAC,CAAC;AAEH,oGAAoG;AACpG,MAAM,UAAU,SAAS,CAAC,EAAU,EAAE,GAAY;IAChD,IAAI,OAAO,GAAG,KAAK,QAAQ,IAAI,GAAG,KAAK,IAAI;QAAE,OAAO,IAAI,CAAC;IACzD,MAAM,CAAC,GAAG,GAA8B,CAAC;IACzC,IAAI,CAAC,CAAC,YAAY,IAAI,CAAC,CAAC,IAAI,CAAC,CAAC,aAAa,IAAI,CAAC,CAAC;QAAE,OAAO,IAAI,CAAC;IAC/D,MAAM,GAAG,GAAG,CAAC,CAAU,EAAsB,EAAE,CAC7C,OAAO,CAAC,KAAK,QAAQ,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;IACxC,OAAO;QACL,EAAE;QACF,IAAI,EAAE,GAAG,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,IAAI,EAAE;QAC1B,IAAI,EAAE,GAAG,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC;QACpB,OAAO,EAAE,GAAG,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC;QAC1B,WAAW,EAAE,CAAC,CAAC,aAAa,CAAC,KAAK,IAAI;QACtC,WAAW,EAAE,CAAC,CAAC,aAAa,CAAC,KAAK,IAAI;KACvC,CAAC;AACJ,CAAC;AAED,MAAM,MAAM,GAAG,CAAC,EAAQ,EAAkB,EAAE,CAC1C,KAAK,IAAI,EAAE,IAAI,EAAE,CAAC,GAAG,CAAC,IAAI,EAAE,KAAK,EAAE,CAAC;AAEtC,sGAAsG;AACtG,SAAS,QAAQ,CAAC,IAAU,EAAE,EAAU;IACtC,OAAO,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC;AAClE,CAAC;AAED;;;GAGG;AACH,SAAS,WAAW,CAAC,IAAU;IAC7B,MAAM,GAAG,GAAG,IAAI,GAAG,EAAmB,CAAC;IACvC,KAAK,MAAM,CAAC,EAAE,EAAE,GAAG,CAAC,IAAI,IAAI,CAAC,UAAyC,EAAE,CAAC;QACvE,MAAM,CAAC,GAAG,SAAS,CAAC,EAAE,EAAE,GAAG,CAAC,CAAC;QAC7B,IAAI,CAAC;YAAE,GAAG,CAAC,GAAG,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC;IACxB,CAAC;IACD,OAAO,GAAG,CAAC;AACb,CAAC;AAED;;;;;;GAMG;AACH,KAAK,UAAU,QAAQ,CACrB,IAAU,EACV,IAAyB,EACzB,GAAU;IAEV,gFAAgF;IAChF,MAAM,GAAG,GAAG,MAAM,IAAI,CAAC,eAAe,EAAE,CAAC;IACzC,MAAM,KAAK,GAAG,CAAC,MAAM,IAAI,CAAC,cAAc,EAAE,CAAC,CAAC,KAAK,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;IACjE,MAAM,MAAM,GAAG,WAAW,CAAC,IAAI,CAAC,CAAC;IACjC,MAAM,KAAK,GAAG,IAAI,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,EAAE,EAAE,EAAE,CAAC,EAAE,CAAC,QAAQ,CAAC,CAAC,CAAC;IACtD,MAAM,KAAK,GAAG,CAAC,GAAG,MAAM,CAAC,MAAM,EAAE,CAAC,CAAC,MAAM,CACvC,CAAC,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAC1C,CAAC;IACF,OAAO;QACL,GAAG,EAAE,EAAE,OAAO,EAAE,KAAK,CAAC,MAAM,GAAG,CAAC,EAAE,KAAK,EAAE;QACzC,KAAK;QACL,MAAM,EAAE,QAAQ,CAAC,IAAI,EAAE,KAAK,EAAE,GAAG,EAAE,GAAG,CAAC;KACxC,CAAC;AACJ,CAAC;AAED;;;;;;;GAOG;AACH,SAAS,SAAS,CAAC,IAAa;IAC9B,MAAM,CAAC,GAAG,CAAC,IAAI,IAAI,EAAE,CAA4B,CAAC;IAClD,MAAM,MAAM,GAAG,CAAC,CAAC,QAAQ,CAAC,CAAC;IAC3B,IAAI,OAAO,MAAM,KAAK,QAAQ,IAAI,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC;QAC7C,OAAO,EAAE,MAAM,EAAE,cAAc,EAAE,OAAO,EAAE,eAAe,EAAE,CAAC;IAC9D,MAAM,CAAC,GAAG,CAAC,CAAC,SAAS,CAAC,CAAC;IACvB,OAAO,EAAE,MAAM,EAAE,OAAO,EAAE,OAAO,CAAC,KAAK,QAAQ,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,eAAe,EAAE,CAAC;AAC1E,CAAC;AAED,uEAAuE;AACvE,SAAS,QAAQ,CAAC,IAAU,EAAE,KAA0B;IACtD,MAAM,EAAE,GAAG,IAAI,CAAC,WAAW,CAAC,EAAE,KAAK,EAAE,CAAC,EAAE,CAAC,CAAC;IAC1C,MAAM,IAAI,GAAc,KAAK,CAAC,GAAG,CAAC,CAAC,EAAE,EAAE,EAAE;QACvC,MAAM,CAAC,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,EAAE,AAAD,EAAG,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,CAAC,SAAqB,CAAC;QACvE,MAAM,CAAC,CAAC,EAAE,QAAQ,CAAC,GAAG,EAAE,CAAC,sBAAsB,CAAC,CAAC,EAAE,CAAC,CAAqB,CAAC;QAC1E,MAAM,IAAI,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC,MAAM,IAAI,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QAC1D,OAAO;YACL,IAAI,EAAE,EAAE,CAAC,GAAG;YACZ,CAAC;YACD,QAAQ;YACR,KAAK,EAAE,EAAE,CAAC,KAAK;YACf,IAAI;YACJ,GAAG,SAAS,CAAC,QAAQ,CAAC,IAAI,EAAE,EAAE,CAAC,QAAQ,CAAC,CAAC;SAC1C,CAAC;IACJ,CAAC,CAAC,CAAC;IACH,OAAO;QACL,OAAO,EAAE,EAAE,CAAC,KAAK;QACjB,QAAQ,EAAE,EAAE,CAAC,MAAM;QACnB,KAAK,EAAE,IAAI,CAAC,OAAO,CAAC,OAAO,CAAC;KAC7B,CAAC;AACJ,CAAC;AAED,uEAAuE;AACvE,SAAS,UAAU,CAAC,IAAa,EAAE,EAAU;IAC3C,MAAM,IAAI,GAAI,IAAkC,EAAE,IAAI,CAAC;IACvD,OAAO,OAAO,IAAI,KAAK,QAAQ,IAAI,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC;AACtD,CAAC;AAED;;;;GAIG;AACH,SAAS,KAAK,CACZ,IAAU,EACV,EAAmC,EACnC,EAAY,EACZ,GAAU;IAEV,MAAM,CAAC,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,CAAC,SAAqB,CAAC;IAC5E,MAAM,CAAC,CAAC,EAAE,QAAQ,CAAC,GAAG,EAAE,CAAC,sBAAsB,CAAC,CAAC,EAAE,CAAC,CAAqB,CAAC;IAC1E,MAAM,IAAI,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,IAAI,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC,MAAM,CAAC;IAC/D,MAAM,IAAI,GAAG,QAAQ,CAAC,IAAI,EAAE,EAAE,CAAC,QAAQ,CAAC,CAAC;IACzC,MAAM,CAAC,GAAG,SAAS,CAAC,IAAI,CAAC,CAAC;IAC1B,OAAO;QACL,GAAG,EAAE,QAAQ,GAAG,CAAC,CAAC,MAAM,GAAG,IAAI;QAC/B,IAAI,EAAE,CAAC;QACP,KAAK,EAAE,EAAE,CAAC,KAAK;QACf,MAAM,EAAE,CAAC,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC,OAAO,CAAC,GAAG,IAAI;QACrC,IAAI;QACJ,IAAI,EAAE,UAAU,CAAC,IAAI,EAAE,EAAE,CAAC,QAAQ,CAAC;QACnC,IAAI,EAAE,EAAE,CAAC,GAAG;QACZ,OAAO,EAAE,SAAS,CAChB,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,CAAC,SAAS,EAAE,EAAE,CAAC,SAAS,CAAa,CAC3D;KACF,CAAC;AACJ,CAAC;AAED,iFAAiF;AACjF,SAAS,QAAQ,CACf,IAAU,EACV,KAA0B,EAC1B,GAAiD,EACjD,GAAU;IAEV,MAAM,EAAE,GAAG,IAAI,CAAC,WAAW,CAAC,EAAE,KAAK,EAAE,CAAC,EAAE,CAAC,CAAC;IAC1C,MAAM,KAAK,GAAG,OAAO,CACnB,GAAG,CAAC,GAAG,EACP,GAAG,EACH,KAAK,CAAC,GAAG,CAAC,CAAC,EAAE,EAAE,EAAE,CAAC,EAAE,CAAC,GAAG,CAAC,CAC1B,CAAC;IACF,MAAM,KAAK,GAAG,KAAK,CAAC,GAAG,CAAC,CAAC,EAAE,EAAE,CAAC,EAAW,EAAE,CAAC,CAAC;QAC3C,GAAG,KAAK,CAAC,IAAI,EAAE,EAAE,EAAE,EAAE,EAAE,GAAG,CAAC;QAC3B,IAAI,EAAE,KAAK,CAAC,CAAC,CAAC,IAAI,EAAE,IAAI,EAAE,SAAS,EAAE;KACtC,CAAC,CAAC,CAAC;IACJ,OAAO,EAAE,OAAO,EAAE,EAAE,CAAC,KAAK,EAAE,QAAQ,EAAE,EAAE,CAAC,MAAM,EAAE,KAAK,EAAE,CAAC;AAC3D,CAAC;AAED,qDAAqD;AACrD,KAAK,UAAU,OAAO,CAAC,GAAQ,EAAE,GAAU;IACzC,MAAM,IAAI,GAKJ,EAAE,CAAC;IACT,MAAM,IAAI,GAAG,IAAI,GAAG,EAAU,CAAC;IAC/B,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,IAAI,GAAG,CAAC,QAAQ,EAAE,CAAC,EAAE,EAAE,CAAC;QACvC,MAAM,IAAI,GAAG,MAAM,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC;QAClC,MAAM,CAAC,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,IAAI,EAAE,GAAG,CAAC,CAAC;QAC1C,KAAK,MAAM,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,KAAK;YAAE,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;QAC5C,IAAI,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC;IAC5B,CAAC;IACD,MAAM,CAAC,IAAI,CAAC,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC;IAC9B,IAAI,CAAC,IAAI;QAAE,OAAO,IAAI,CAAC,OAAO,EAAE,sBAAsB,CAAC,CAAC;IACxD,MAAM,KAAK,GAAG,OAAO,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC;IAC9C,IAAI,KAAK,CAAC,IAAI,KAAK,oBAAoB;QACrC,OAAO,IAAI,CACT,oBAAoB,EACpB,WAAW,KAAK,CAAC,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,iDAAiD;YAChF,2FAA2F;YAC3F,SAAS,OAAO,CAAC,OAAO,8DAA8D,CACzF,CAAC;IACJ,OAAO;QACL,EAAE,EAAE,IAAI;QACR,KAAK,EAAE;YACL,KAAK,EAAE,GAAG,CAAC,QAAQ;YACnB,KAAK;YACL,IAAI,EAAE,QAAQ,CAAC,IAAI,CAAC,IAAI,EAAE,IAAI,CAAC,KAAK,CAAC;YACrC,MAAM,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC;SAClC;KACF,CAAC;AACJ,CAAC;AAED,+EAA+E;AAC/E,SAAS,SAAS,CAAC,CAAU;IAC3B,MAAM,GAAG,GAAG,CAAwC,CAAC;IACrD,MAAM,MAAM,GAAG,GAAG,GAAG,CAAC,IAAI,IAAI,OAAO,KAAK,GAAG,CAAC,OAAO,IAAI,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC;IACrE,OAAO,IAAI,CACT,GAAG,CAAC,IAAI,KAAK,mBAAmB,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC,CAAC,YAAY,EAC7D,MAAM,CACP,CAAC;AACJ,CAAC;AAED,4FAA4F;AAC5F,MAAM,CAAC,MAAM,OAAO,GAAc,KAAK,EAAE,IAAI,EAAE,EAAE;IAC/C,IAAI,IAAgB,CAAC;IACrB,IAAI,CAAC;QACH,IAAI,GAAG,IAAI,UAAU,CAAC,YAAY,CAAC,IAAI,CAAC,CAAC,CAAC;IAC5C,CAAC;IAAC,OAAO,CAAC,EAAE,CAAC;QACX,OAAO,IAAI,CAAC,YAAY,EAAG,CAAW,CAAC,OAAO,CAAC,CAAC;IAClD,CAAC;IACD,MAAM,GAAG,GAAG,MAAM,SAAS,EAAE,CAAC;IAC9B,MAAM,IAAI,GAAG,GAAG,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,GAAG,aAAa,EAAE,CAAC,CAAC;IACzD,IAAI,CAAC;QACH,OAAO,MAAM,OAAO,CAAC,MAAM,IAAI,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC;IAChD,CAAC;IAAC,OAAO,CAAC,EAAE,CAAC;QACX,OAAO,SAAS,CAAC,CAAC,CAAC,CAAC;IACtB,CAAC;YAAS,CAAC;QACT,MAAM,IAAI,CAAC,OAAO,EAAE,CAAC;IACvB,CAAC;AACH,CAAC,CAAC;AAEF,4EAA4E;AAC5E,MAAM,UAAU,eAAe,CAC7B,CAAkC,EAClC,GAAW;IAEX,OAAO,kBAAkB,GAAG,iBAAiB,CAAC,CAAC,MAAM,MAAM,CAAC,CAAC,MAAM,EAAE,CAAC;AACxE,CAAC"}

package/dist/pdf-geometry.d.ts

@@ -0,0 +1,170 @@
+/**
+ * WHAT A PDF'S TEXT AND FONTS SAY, once pdf.js has read them — PURE. No disk, no pdf.js.
+ *
+ * `pdf-facts.ts` is the shell: it opens the PDF with pdf.js (the `unpdf` package) and hands this
+ * module plain data — text runs with their metrics, and the fonts each page resolved. Everything
+ * that DECIDES something is here, so every branch is a row in a table test fed with hand-written
+ * words, including pages the fixture corpus does not contain.
+ *
+ * ── WHY pdf.js AND NOT POPPLER (issue #61) ───────────────────────────────────
+ * The facts used to come from three poppler programs (`pdfinfo`, `pdffonts`, `pdftotext -bbox`), a
+ * system install paperlint could not provide and every consumer had to add by hand. A measurement on 25
+ * PDFs (2026-09-24) found pdf.js equal on page count and on the Type 3 count of every PDF, and the
+ * last page's column heights within 0.2 pt except on an all-Type-3 page (7.2 pt; pdf.js has no
+ * ascent for Type 3). No verdict flipped at a 120 pt tolerance.
+ *
+ * ── THE LAST PAGE IS A MEASUREMENT, NOT A VERDICT ────────────────────────────
+ * `classifyLastPage` returns the heights of the two columns, or says why there is nothing to
+ * measure. It carries no threshold: whether a venue requires balance, and how much difference it
+ * allows, belongs to the lint rule `pdf/last-page-balance`, which a consumer turns on per venue.
+ */
+/** One word on a page, in points, top-down: `y0` is the top edge, `y1` the bottom. */
+export interface Word {
+    readonly x0: number;
+    readonly y0: number;
+    readonly x1: number;
+    readonly y1: number;
+    readonly text: string;
+}
+/** A page's size and its words. */
+export interface PageText {
+    readonly widthPt: number;
+    readonly heightPt: number;
+    readonly words: readonly Word[];
+}
+/** The heights of the last page's two columns, in points: [left, right]. */
+export type Columns = readonly [left: number, right: number];
+/**
+ * What the last page is, as far as balance goes.
+ *
+ * - `measured` — the heights of both columns.
+ * - `stub`     — fewer than `STUB_WORDS` words: a tail of a few lines is not a layout.
+ * - `review`   — numbered lines in the outer margins. The numbers run down the whole page, so every
+ *                column would measure full height; a review build is not judged at all.
+ */
+export type LastPage = {
+    readonly kind: "measured";
+    readonly columns: Columns;
+} | {
+    readonly kind: "stub";
+    readonly words: number;
+} | {
+    readonly kind: "review";
+    readonly lineNumbers: number;
+};
+/** Below this many words the last page is a stub. */
+export declare const STUB_WORDS = 60;
+/** This many short integers in the outer margins make a review build. */
+export declare const REVIEW_LINE_NUMBERS = 5;
+/** The outer margins, as a fraction of the page width, where line numbers sit. */
+export declare const MARGIN_FRACTION = 0.08;
+/**
+ * Classify the last page and, when it has a layout, measure its two columns.
+ *
+ * 🔴 THE PAGE IS SPLIT AT ITS OWN MIDDLE, not between the outermost words: on a page where only the
+ * left column is filled, the words' midpoint falls inside that column and cuts it in two ("77 / 731"
+ * for six lines at the top left, measured 2026-08-29). A word belongs to the side its left edge is on.
+ *
+ * Throws on a page with no width — the caller built `PageText` from the PDF's own viewport, so a
+ * non-positive width is a defect in the caller, not a property of a paper.
+ */
+export declare function classifyLastPage(p: PageText): LastPage;
+/**
+ * One text run as pdf.js reports it, already in top-down page coordinates: where its baseline
+ * starts, how wide it is, its font size, and the font's ascent and descent as fractions of the size
+ * (descent negative, as fonts store it).
+ */
+export interface TextRun {
+    readonly text: string;
+    readonly x: number;
+    readonly baseline: number;
+    readonly width: number;
+    readonly size: number;
+    readonly ascent: number;
+    readonly descent: number;
+}
+/** A font's vertical metrics as fractions of its size, when pdf.js supplies none. */
+export declare const DEFAULT_ASCENT = 0.8;
+export declare const DEFAULT_DESCENT = -0.2;
+/**
+ * Split a run into words. pdf.js gives a run's total width, not each glyph's, so a word's horizontal
+ * extent is estimated in proportion to its characters. Only `x0` decides the column a word falls
+ * in, and a run never crosses the gutter, so the estimate cannot move a word to the other side.
+ */
+export declare function wordsOf(run: TextRun): Word[];
+/** The font programs a PDF can carry, in pdf.js's vocabulary. */
+export type FontProgram = "Type1" | "Type1C" | "MMType1" | "TrueType" | "OpenType" | "CIDType0" | "CIDType0C" | "CIDTrueType" | "Type3";
+/** A program pdf.js named but this module does not know — kept verbatim, never guessed. */
+export type UnknownProgram = `unknown:${string}`;
+/**
+ * One font the PDF draws text with.
+ *
+ * - `embedded`     — a usable font program is in the file.
+ * - `type3`        — a Type 3 font: glyphs drawn as PDF procedures, usually bitmaps. pdf.js gives
+ *                    every Type 3 font the same name, `Type3`, so the name carries nothing.
+ * - `not-embedded` — the PDF names the font but carries no usable program for it: either none at
+ *                    all, or one pdf.js could not parse. pdf.js reports both as one flag
+ *                    (`missingFile`), so the two are not told apart here.
+ */
+export type FontFact = {
+    readonly kind: "embedded";
+    readonly name: string;
+    readonly program: Exclude<FontProgram, "Type3"> | UnknownProgram;
+} | {
+    readonly kind: "type3";
+    readonly name: "Type3";
+} | {
+    readonly kind: "not-embedded";
+    readonly name: string;
+    readonly program: Exclude<FontProgram, "Type3"> | UnknownProgram;
+};
+/** A font as pdf.js exports it with `fontExtraProperties`, reduced to what is read here. */
+export interface RawFont {
+    /** pdf.js's id for the font within the document (`loadedName`). */
+    readonly id: string;
+    readonly name: string;
+    readonly type: string | undefined;
+    readonly subtype: string | undefined;
+    readonly isType3Font: boolean;
+    readonly missingFile: boolean;
+}
+/** One page: whether it draws any text, and the fonts its text resolved to. */
+export interface RawPage {
+    readonly hasText: boolean;
+    readonly fonts: readonly RawFont[];
+}
+/**
+ * The fonts of a document. `drawn` is non-empty BY TYPE: a document whose pages draw text has at
+ * least one font, and "no fonts" is spelled only as `no-text`.
+ */
+export type Fonts = {
+    readonly kind: "drawn";
+    readonly list: readonly [FontFact, ...FontFact[]];
+} | {
+    readonly kind: "no-text";
+};
+/**
+ * 🔴 TEXT ON A PAGE WITH NO FONT BEHIND IT IS A FAILED READ, NOT AN EMPTY LIST. Measured on Node
+ * 20.20.2: pdf.js opens the same PDFs and returns ZERO fonts with no error. An empty list there
+ * would read downstream as "no Type 3, nothing unembedded" — a clean verdict over nothing.
+ */
+export interface ZeroFontsOnText {
+    readonly kind: "zero-fonts-on-text";
+    /** 1-based numbers of the pages that draw text with no font resolved. */
+    readonly pages: readonly number[];
+}
+export declare function programOf(type: string | undefined, subtype: string | undefined): Exclude<FontProgram, "Type3"> | UnknownProgram;
+export declare function fontFact(f: RawFont): FontFact;
+/**
+ * The document's fonts, each once, in the order pages first drew them — or the pages that drew
+ * text with no font, which the caller turns into a failed read.
+ */
+export declare function fontsOf(pages: readonly RawPage[]): Fonts | ZeroFontsOnText;
+/**
+ * poppler's spelling of the font's type column (`pdffonts`), which the facts JSON has always used
+ * and consumers match on (`Type 3`).
+ */
+export declare const POPPLER_TYPE: Readonly<Record<FontProgram, string>>;
+/** The type column for any fact, the unknown ones verbatim. */
+export declare function popplerType(f: FontFact): string;
+//# sourceMappingURL=pdf-geometry.d.ts.map

package/dist/pdf-geometry.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"pdf-geometry.d.ts","sourceRoot":"","sources":["../src/pdf-geometry.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;GAmBG;AAEH,sFAAsF;AACtF,MAAM,WAAW,IAAI;IACnB,QAAQ,CAAC,EAAE,EAAE,MAAM,CAAC;IACpB,QAAQ,CAAC,EAAE,EAAE,MAAM,CAAC;IACpB,QAAQ,CAAC,EAAE,EAAE,MAAM,CAAC;IACpB,QAAQ,CAAC,EAAE,EAAE,MAAM,CAAC;IACpB,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;CACvB;AAED,mCAAmC;AACnC,MAAM,WAAW,QAAQ;IACvB,QAAQ,CAAC,OAAO,EAAE,MAAM,CAAC;IACzB,QAAQ,CAAC,QAAQ,EAAE,MAAM,CAAC;IAC1B,QAAQ,CAAC,KAAK,EAAE,SAAS,IAAI,EAAE,CAAC;CACjC;AAED,4EAA4E;AAC5E,MAAM,MAAM,OAAO,GAAG,SAAS,CAAC,IAAI,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,CAAC,CAAC;AAE7D;;;;;;;GAOG;AACH,MAAM,MAAM,QAAQ,GAChB;IAAE,QAAQ,CAAC,IAAI,EAAE,UAAU,CAAC;IAAC,QAAQ,CAAC,OAAO,EAAE,OAAO,CAAA;CAAE,GACxD;IAAE,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IAAC,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAA;CAAE,GACjD;IAAE,QAAQ,CAAC,IAAI,EAAE,QAAQ,CAAC;IAAC,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAA;CAAE,CAAC;AAE9D,qDAAqD;AACrD,eAAO,MAAM,UAAU,KAAK,CAAC;AAC7B,yEAAyE;AACzE,eAAO,MAAM,mBAAmB,IAAI,CAAC;AACrC,kFAAkF;AAClF,eAAO,MAAM,eAAe,OAAO,CAAC;AAqBpC;;;;;;;;;GASG;AACH,wBAAgB,gBAAgB,CAAC,CAAC,EAAE,QAAQ,GAAG,QAAQ,CAgBtD;AAID;;;;GAIG;AACH,MAAM,WAAW,OAAO;IACtB,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,CAAC,EAAE,MAAM,CAAC;IACnB,QAAQ,CAAC,QAAQ,EAAE,MAAM,CAAC;IAC1B,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IACvB,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC;IACxB,QAAQ,CAAC,OAAO,EAAE,MAAM,CAAC;CAC1B;AAED,qFAAqF;AACrF,eAAO,MAAM,cAAc,MAAM,CAAC;AAClC,eAAO,MAAM,eAAe,OAAO,CAAC;AAEpC;;;;GAIG;AACH,wBAAgB,OAAO,CAAC,GAAG,EAAE,OAAO,GAAG,IAAI,EAAE,CAQ5C;AAID,iEAAiE;AACjE,MAAM,MAAM,WAAW,GACnB,OAAO,GACP,QAAQ,GACR,SAAS,GACT,UAAU,GACV,UAAU,GACV,UAAU,GACV,WAAW,GACX,aAAa,GACb,OAAO,CAAC;AAEZ,2FAA2F;AAC3F,MAAM,MAAM,cAAc,GAAG,WAAW,MAAM,EAAE,CAAC;AAEjD;;;;;;;;;GASG;AACH,MAAM,MAAM,QAAQ,GAChB;IACE,QAAQ,CAAC,IAAI,EAAE,UAAU,CAAC;IAC1B,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,OAAO,EAAE,OAAO,CAAC,WAAW,EAAE,OAAO,CAAC,GAAG,cAAc,CAAC;CAClE,GACD;IAAE,QAAQ,CAAC,IAAI,EAAE,OAAO,CAAC;IAAC,QAAQ,CAAC,IAAI,EAAE,OAAO,CAAA;CAAE,GAClD;IACE,QAAQ,CAAC,IAAI,EAAE,cAAc,CAAC;IAC9B,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,OAAO,EAAE,OAAO,CAAC,WAAW,EAAE,OAAO,CAAC,GAAG,cAAc,CAAC;CAClE,CAAC;AAEN,4FAA4F;AAC5F,MAAM,WAAW,OAAO;IACtB,mEAAmE;IACnE,QAAQ,CAAC,EAAE,EAAE,MAAM,CAAC;IACpB,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,IAAI,EAAE,MAAM,GAAG,SAAS,CAAC;IAClC,QAAQ,CAAC,OAAO,EAAE,MAAM,GAAG,SAAS,CAAC;IACrC,QAAQ,CAAC,WAAW,EAAE,OAAO,CAAC;IAC9B,QAAQ,CAAC,WAAW,EAAE,OAAO,CAAC;CAC/B;AAED,+EAA+E;AAC/E,MAAM,WAAW,OAAO;IACtB,QAAQ,CAAC,OAAO,EAAE,OAAO,CAAC;IAC1B,QAAQ,CAAC,KAAK,EAAE,SAAS,OAAO,EAAE,CAAC;CACpC;AAED;;;GAGG;AACH,MAAM,MAAM,KAAK,GACb;IACE,QAAQ,CAAC,IAAI,EAAE,OAAO,CAAC;IACvB,QAAQ,CAAC,IAAI,EAAE,SAAS,CAAC,QAAQ,EAAE,GAAG,QAAQ,EAAE,CAAC,CAAC;CACnD,GACD;IAAE,QAAQ,CAAC,IAAI,EAAE,SAAS,CAAA;CAAE,CAAC;AAEjC;;;;GAIG;AACH,MAAM,WAAW,eAAe;IAC9B,QAAQ,CAAC,IAAI,EAAE,oBAAoB,CAAC;IACpC,yEAAyE;IACzE,QAAQ,CAAC,KAAK,EAAE,SAAS,MAAM,EAAE,CAAC;CACnC;AAgBD,wBAAgB,SAAS,CACvB,IAAI,EAAE,MAAM,GAAG,SAAS,EACxB,OAAO,EAAE,MAAM,GAAG,SAAS,GAC1B,OAAO,CAAC,WAAW,EAAE,OAAO,CAAC,GAAG,cAAc,CAKhD;AAMD,wBAAgB,QAAQ,CAAC,CAAC,EAAE,OAAO,GAAG,QAAQ,CAO7C;AAED;;;GAGG;AACH,wBAAgB,OAAO,CAAC,KAAK,EAAE,SAAS,OAAO,EAAE,GAAG,KAAK,GAAG,eAAe,CAY1E;AAED;;;GAGG;AACH,eAAO,MAAM,YAAY,EAAE,QAAQ,CAAC,MAAM,CAAC,WAAW,EAAE,MAAM,CAAC,CAU9D,CAAC;AAEF,+DAA+D;AAC/D,wBAAgB,WAAW,CAAC,CAAC,EAAE,QAAQ,GAAG,MAAM,CAK/C"}

package/dist/pdf-geometry.js

@@ -0,0 +1,158 @@
+/**
+ * WHAT A PDF'S TEXT AND FONTS SAY, once pdf.js has read them — PURE. No disk, no pdf.js.
+ *
+ * `pdf-facts.ts` is the shell: it opens the PDF with pdf.js (the `unpdf` package) and hands this
+ * module plain data — text runs with their metrics, and the fonts each page resolved. Everything
+ * that DECIDES something is here, so every branch is a row in a table test fed with hand-written
+ * words, including pages the fixture corpus does not contain.
+ *
+ * ── WHY pdf.js AND NOT POPPLER (issue #61) ───────────────────────────────────
+ * The facts used to come from three poppler programs (`pdfinfo`, `pdffonts`, `pdftotext -bbox`), a
+ * system install paperlint could not provide and every consumer had to add by hand. A measurement on 25
+ * PDFs (2026-09-24) found pdf.js equal on page count and on the Type 3 count of every PDF, and the
+ * last page's column heights within 0.2 pt except on an all-Type-3 page (7.2 pt; pdf.js has no
+ * ascent for Type 3). No verdict flipped at a 120 pt tolerance.
+ *
+ * ── THE LAST PAGE IS A MEASUREMENT, NOT A VERDICT ────────────────────────────
+ * `classifyLastPage` returns the heights of the two columns, or says why there is nothing to
+ * measure. It carries no threshold: whether a venue requires balance, and how much difference it
+ * allows, belongs to the lint rule `pdf/last-page-balance`, which a consumer turns on per venue.
+ */
+/** Below this many words the last page is a stub. */
+export const STUB_WORDS = 60;
+/** This many short integers in the outer margins make a review build. */
+export const REVIEW_LINE_NUMBERS = 5;
+/** The outer margins, as a fraction of the page width, where line numbers sit. */
+export const MARGIN_FRACTION = 0.08;
+const round1 = (n) => Number(n.toFixed(1));
+/** Top of the highest word to bottom of the lowest one; 0 for no words. */
+function heightOf(words) {
+    if (words.length === 0)
+        return 0;
+    const top = Math.min(...words.map((w) => w.y0));
+    const bottom = Math.max(...words.map((w) => w.y1));
+    return round1(bottom - top);
+}
+/** A short integer standing in the left or right margin: a line number of a review build. */
+function isLineNumber(w, widthPt) {
+    const margin = widthPt * MARGIN_FRACTION;
+    return (/^\d{1,4}$/.test(w.text.trim()) &&
+        (w.x1 < margin || w.x0 > widthPt - margin));
+}
+/**
+ * Classify the last page and, when it has a layout, measure its two columns.
+ *
+ * 🔴 THE PAGE IS SPLIT AT ITS OWN MIDDLE, not between the outermost words: on a page where only the
+ * left column is filled, the words' midpoint falls inside that column and cuts it in two ("77 / 731"
+ * for six lines at the top left, measured 2026-08-29). A word belongs to the side its left edge is on.
+ *
+ * Throws on a page with no width — the caller built `PageText` from the PDF's own viewport, so a
+ * non-positive width is a defect in the caller, not a property of a paper.
+ */
+export function classifyLastPage(p) {
+    if (!(p.widthPt > 0))
+        throw new RangeError(`page width must be positive, got ${p.widthPt}`);
+    if (p.words.length < STUB_WORDS)
+        return { kind: "stub", words: p.words.length };
+    const lineNumbers = p.words.filter((w) => isLineNumber(w, p.widthPt)).length;
+    if (lineNumbers >= REVIEW_LINE_NUMBERS)
+        return { kind: "review", lineNumbers };
+    const mid = p.widthPt / 2;
+    return {
+        kind: "measured",
+        columns: [
+            heightOf(p.words.filter((w) => w.x0 < mid)),
+            heightOf(p.words.filter((w) => w.x0 >= mid)),
+        ],
+    };
+}
+/** A font's vertical metrics as fractions of its size, when pdf.js supplies none. */
+export const DEFAULT_ASCENT = 0.8;
+export const DEFAULT_DESCENT = -0.2;
+/**
+ * Split a run into words. pdf.js gives a run's total width, not each glyph's, so a word's horizontal
+ * extent is estimated in proportion to its characters. Only `x0` decides the column a word falls
+ * in, and a run never crosses the gutter, so the estimate cannot move a word to the other side.
+ */
+export function wordsOf(run) {
+    const perChar = run.width / Math.max(1, run.text.length);
+    const y0 = run.baseline - run.ascent * run.size;
+    const y1 = run.baseline - run.descent * run.size;
+    return [...run.text.matchAll(/\S+/g)].map((m) => {
+        const x0 = run.x + (m.index ?? 0) * perChar;
+        return { x0, y0, x1: x0 + m[0].length * perChar, y1, text: m[0] };
+    });
+}
+/**
+ * pdf.js's `type`/`subtype` pair → the program, keyed `type/subtype`. Composite fonts carry their
+ * descendant's type. A pair not listed is kept verbatim as `unknown:…`, never guessed.
+ */
+const PROGRAMS = {
+    "Type1/undefined": "Type1",
+    "Type1/Type1C": "Type1C",
+    "MMType1/undefined": "MMType1",
+    "TrueType/undefined": "TrueType",
+    "CIDFontType0/undefined": "CIDType0",
+    "CIDFontType0/CIDFontType0C": "CIDType0C",
+    "CIDFontType2/undefined": "CIDTrueType",
+};
+export function programOf(type, subtype) {
+    // pdf.js names an OpenType-wrapped program by the wrapper, on either field, whatever it wraps.
+    if (subtype === "OpenType" || type === "OpenType")
+        return "OpenType";
+    const key = `${String(type)}/${String(subtype)}`;
+    return PROGRAMS[key] ?? `unknown:${key}`;
+}
+/** Subset fonts carry a six-letter tag, `ABCDEF+Name`; the name is what follows it. */
+const withoutSubsetTag = (name) => name.replace(/^[A-Z]{6}\+/, "");
+export function fontFact(f) {
+    if (f.isType3Font)
+        return { kind: "type3", name: "Type3" };
+    const name = withoutSubsetTag(f.name);
+    const program = programOf(f.type, f.subtype);
+    return f.missingFile
+        ? { kind: "not-embedded", name, program }
+        : { kind: "embedded", name, program };
+}
+/**
+ * The document's fonts, each once, in the order pages first drew them — or the pages that drew
+ * text with no font, which the caller turns into a failed read.
+ */
+export function fontsOf(pages) {
+    const bad = pages.flatMap((p, i) => p.hasText && p.fonts.length === 0 ? [i + 1] : []);
+    if (bad.length > 0)
+        return { kind: "zero-fonts-on-text", pages: bad };
+    const seen = new Map();
+    for (const p of pages)
+        for (const f of p.fonts)
+            if (!seen.has(f.id))
+                seen.set(f.id, fontFact(f));
+    const [first, ...rest] = seen.values();
+    return first
+        ? { kind: "drawn", list: [first, ...rest] }
+        : { kind: "no-text" };
+}
+/**
+ * poppler's spelling of the font's type column (`pdffonts`), which the facts JSON has always used
+ * and consumers match on (`Type 3`).
+ */
+export const POPPLER_TYPE = {
+    Type1: "Type 1",
+    Type1C: "Type 1C",
+    MMType1: "Type 1",
+    TrueType: "TrueType",
+    OpenType: "OpenType",
+    CIDType0: "CID Type 0",
+    CIDType0C: "CID Type 0C",
+    CIDTrueType: "CID TrueType",
+    Type3: "Type 3",
+};
+/** The type column for any fact, the unknown ones verbatim. */
+export function popplerType(f) {
+    if (f.kind === "type3")
+        return POPPLER_TYPE.Type3;
+    return f.program.startsWith("unknown:")
+        ? f.program
+        : POPPLER_TYPE[f.program];
+}
+//# sourceMappingURL=pdf-geometry.js.map

package/dist/pdf-geometry.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"pdf-geometry.js","sourceRoot":"","sources":["../src/pdf-geometry.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;GAmBG;AAkCH,qDAAqD;AACrD,MAAM,CAAC,MAAM,UAAU,GAAG,EAAE,CAAC;AAC7B,yEAAyE;AACzE,MAAM,CAAC,MAAM,mBAAmB,GAAG,CAAC,CAAC;AACrC,kFAAkF;AAClF,MAAM,CAAC,MAAM,eAAe,GAAG,IAAI,CAAC;AAEpC,MAAM,MAAM,GAAG,CAAC,CAAS,EAAU,EAAE,CAAC,MAAM,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC;AAE3D,2EAA2E;AAC3E,SAAS,QAAQ,CAAC,KAAsB;IACtC,IAAI,KAAK,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,CAAC,CAAC;IACjC,MAAM,GAAG,GAAG,IAAI,CAAC,GAAG,CAAC,GAAG,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC;IAChD,MAAM,MAAM,GAAG,IAAI,CAAC,GAAG,CAAC,GAAG,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC;IACnD,OAAO,MAAM,CAAC,MAAM,GAAG,GAAG,CAAC,CAAC;AAC9B,CAAC;AAED,6FAA6F;AAC7F,SAAS,YAAY,CAAC,CAAO,EAAE,OAAe;IAC5C,MAAM,MAAM,GAAG,OAAO,GAAG,eAAe,CAAC;IACzC,OAAO,CACL,WAAW,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,EAAE,CAAC;QAC/B,CAAC,CAAC,CAAC,EAAE,GAAG,MAAM,IAAI,CAAC,CAAC,EAAE,GAAG,OAAO,GAAG,MAAM,CAAC,CAC3C,CAAC;AACJ,CAAC;AAED;;;;;;;;;GASG;AACH,MAAM,UAAU,gBAAgB,CAAC,CAAW;IAC1C,IAAI,CAAC,CAAC,CAAC,CAAC,OAAO,GAAG,CAAC,CAAC;QAClB,MAAM,IAAI,UAAU,CAAC,oCAAoC,CAAC,CAAC,OAAO,EAAE,CAAC,CAAC;IACxE,IAAI,CAAC,CAAC,KAAK,CAAC,MAAM,GAAG,UAAU;QAC7B,OAAO,EAAE,IAAI,EAAE,MAAM,EAAE,KAAK,EAAE,CAAC,CAAC,KAAK,CAAC,MAAM,EAAE,CAAC;IACjD,MAAM,WAAW,GAAG,CAAC,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,YAAY,CAAC,CAAC,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC;IAC7E,IAAI,WAAW,IAAI,mBAAmB;QACpC,OAAO,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,CAAC;IACzC,MAAM,GAAG,GAAG,CAAC,CAAC,OAAO,GAAG,CAAC,CAAC;IAC1B,OAAO;QACL,IAAI,EAAE,UAAU;QAChB,OAAO,EAAE;YACP,QAAQ,CAAC,CAAC,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,EAAE,GAAG,GAAG,CAAC,CAAC;YAC3C,QAAQ,CAAC,CAAC,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,EAAE,IAAI,GAAG,CAAC,CAAC;SAC7C;KACF,CAAC;AACJ,CAAC;AAmBD,qFAAqF;AACrF,MAAM,CAAC,MAAM,cAAc,GAAG,GAAG,CAAC;AAClC,MAAM,CAAC,MAAM,eAAe,GAAG,CAAC,GAAG,CAAC;AAEpC;;;;GAIG;AACH,MAAM,UAAU,OAAO,CAAC,GAAY;IAClC,MAAM,OAAO,GAAG,GAAG,CAAC,KAAK,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,GAAG,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;IACzD,MAAM,EAAE,GAAG,GAAG,CAAC,QAAQ,GAAG,GAAG,CAAC,MAAM,GAAG,GAAG,CAAC,IAAI,CAAC;IAChD,MAAM,EAAE,GAAG,GAAG,CAAC,QAAQ,GAAG,GAAG,CAAC,OAAO,GAAG,GAAG,CAAC,IAAI,CAAC;IACjD,OAAO,CAAC,GAAG,GAAG,CAAC,IAAI,CAAC,QAAQ,CAAC,MAAM,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE;QAC9C,MAAM,EAAE,GAAG,GAAG,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI,CAAC,CAAC,GAAG,OAAO,CAAC;QAC5C,OAAO,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,MAAM,GAAG,OAAO,EAAE,EAAE,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;IACpE,CAAC,CAAC,CAAC;AACL,CAAC;AAiFD;;;GAGG;AACH,MAAM,QAAQ,GAA4D;IACxE,iBAAiB,EAAE,OAAO;IAC1B,cAAc,EAAE,QAAQ;IACxB,mBAAmB,EAAE,SAAS;IAC9B,oBAAoB,EAAE,UAAU;IAChC,wBAAwB,EAAE,UAAU;IACpC,4BAA4B,EAAE,WAAW;IACzC,wBAAwB,EAAE,aAAa;CACxC,CAAC;AAEF,MAAM,UAAU,SAAS,CACvB,IAAwB,EACxB,OAA2B;IAE3B,+FAA+F;IAC/F,IAAI,OAAO,KAAK,UAAU,IAAI,IAAI,KAAK,UAAU;QAAE,OAAO,UAAU,CAAC;IACrE,MAAM,GAAG,GAAG,GAAG,MAAM,CAAC,IAAI,CAAC,IAAI,MAAM,CAAC,OAAO,CAAC,EAAE,CAAC;IACjD,OAAO,QAAQ,CAAC,GAAG,CAAC,IAAI,WAAW,GAAG,EAAE,CAAC;AAC3C,CAAC;AAED,uFAAuF;AACvF,MAAM,gBAAgB,GAAG,CAAC,IAAY,EAAU,EAAE,CAChD,IAAI,CAAC,OAAO,CAAC,aAAa,EAAE,EAAE,CAAC,CAAC;AAElC,MAAM,UAAU,QAAQ,CAAC,CAAU;IACjC,IAAI,CAAC,CAAC,WAAW;QAAE,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,CAAC;IAC3D,MAAM,IAAI,GAAG,gBAAgB,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;IACtC,MAAM,OAAO,GAAG,SAAS,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC;IAC7C,OAAO,CAAC,CAAC,WAAW;QAClB,CAAC,CAAC,EAAE,IAAI,EAAE,cAAc,EAAE,IAAI,EAAE,OAAO,EAAE;QACzC,CAAC,CAAC,EAAE,IAAI,EAAE,UAAU,EAAE,IAAI,EAAE,OAAO,EAAE,CAAC;AAC1C,CAAC;AAED;;;GAGG;AACH,MAAM,UAAU,OAAO,CAAC,KAAyB;IAC/C,MAAM,GAAG,GAAG,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CACjC,CAAC,CAAC,OAAO,IAAI,CAAC,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,CACjD,CAAC;IACF,IAAI,GAAG,CAAC,MAAM,GAAG,CAAC;QAAE,OAAO,EAAE,IAAI,EAAE,oBAAoB,EAAE,KAAK,EAAE,GAAG,EAAE,CAAC;IACtE,MAAM,IAAI,GAAG,IAAI,GAAG,EAAoB,CAAC;IACzC,KAAK,MAAM,CAAC,IAAI,KAAK;QACnB,KAAK,MAAM,CAAC,IAAI,CAAC,CAAC,KAAK;YAAE,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC;gBAAE,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,QAAQ,CAAC,CAAC,CAAC,CAAC,CAAC;IAC5E,MAAM,CAAC,KAAK,EAAE,GAAG,IAAI,CAAC,GAAG,IAAI,CAAC,MAAM,EAAE,CAAC;IACvC,OAAO,KAAK;QACV,CAAC,CAAC,EAAE,IAAI,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,KAAK,EAAE,GAAG,IAAI,CAAC,EAAE;QAC3C,CAAC,CAAC,EAAE,IAAI,EAAE,SAAS,EAAE,CAAC;AAC1B,CAAC;AAED;;;GAGG;AACH,MAAM,CAAC,MAAM,YAAY,GAA0C;IACjE,KAAK,EAAE,QAAQ;IACf,MAAM,EAAE,SAAS;IACjB,OAAO,EAAE,QAAQ;IACjB,QAAQ,EAAE,UAAU;IACpB,QAAQ,EAAE,UAAU;IACpB,QAAQ,EAAE,YAAY;IACtB,SAAS,EAAE,aAAa;IACxB,WAAW,EAAE,cAAc;IAC3B,KAAK,EAAE,QAAQ;CAChB,CAAC;AAEF,+DAA+D;AAC/D,MAAM,UAAU,WAAW,CAAC,CAAW;IACrC,IAAI,CAAC,CAAC,IAAI,KAAK,OAAO;QAAE,OAAO,YAAY,CAAC,KAAK,CAAC;IAClD,OAAO,CAAC,CAAC,OAAO,CAAC,UAAU,CAAC,UAAU,CAAC;QACrC,CAAC,CAAC,CAAC,CAAC,OAAO;QACX,CAAC,CAAC,YAAY,CAAC,CAAC,CAAC,OAAsB,CAAC,CAAC;AAC7C,CAAC"}

package/dist/ports/download.d.ts

@@ -0,0 +1,9 @@
+/** Fetching a URL — the generic port for the network. */
+import type { Result } from "../domain/result.ts";
+/** Fetch a URL's bytes. */
+export interface Download {
+    fetch(url: string, timeoutMs: number): Result<Uint8Array, {
+        readonly detail: string;
+    }>;
+}
+//# sourceMappingURL=download.d.ts.map

package/dist/ports/download.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"download.d.ts","sourceRoot":"","sources":["../../src/ports/download.ts"],"names":[],"mappings":"AAAA,yDAAyD;AACzD,OAAO,KAAK,EAAE,MAAM,EAAE,MAAM,qBAAqB,CAAC;AAElD,2BAA2B;AAC3B,MAAM,WAAW,QAAQ;IACvB,KAAK,CACH,GAAG,EAAE,MAAM,EACX,SAAS,EAAE,MAAM,GAChB,MAAM,CAAC,UAAU,EAAE;QAAE,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAA;KAAE,CAAC,CAAC;CACpD"}

package/dist/ports/download.js

@@ -0,0 +1,2 @@
+export {};
+//# sourceMappingURL=download.js.map

package/dist/ports/download.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"download.js","sourceRoot":"","sources":["../../src/ports/download.ts"],"names":[],"mappings":""}

package/dist/ports/files.d.ts

@@ -0,0 +1,11 @@
+/** Files by path — the generic port for the disk. */
+import type { AbsolutePath } from "../domain/paths.ts";
+/** Files by path. A read that fails for any reason but absence throws: that is an adapter bug. */
+export interface Files {
+    isFile(p: AbsolutePath): boolean;
+    /** The file's bytes, or null when there is no such file. */
+    readBytes(p: AbsolutePath): Uint8Array | null;
+    /** Write through `<p>.part` and a rename, creating the directory: a reader never sees half a file. */
+    writeAtomic(p: AbsolutePath, bytes: Uint8Array): void;
+}
+//# sourceMappingURL=files.d.ts.map

package/dist/ports/files.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"files.d.ts","sourceRoot":"","sources":["../../src/ports/files.ts"],"names":[],"mappings":"AAAA,qDAAqD;AACrD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAEvD,kGAAkG;AAClG,MAAM,WAAW,KAAK;IACpB,MAAM,CAAC,CAAC,EAAE,YAAY,GAAG,OAAO,CAAC;IACjC,4DAA4D;IAC5D,SAAS,CAAC,CAAC,EAAE,YAAY,GAAG,UAAU,GAAG,IAAI,CAAC;IAC9C,sGAAsG;IACtG,WAAW,CAAC,CAAC,EAAE,YAAY,EAAE,KAAK,EAAE,UAAU,GAAG,IAAI,CAAC;CACvD"}

package/dist/ports/files.js

@@ -0,0 +1,2 @@
+export {};
+//# sourceMappingURL=files.js.map

package/dist/ports/files.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"files.js","sourceRoot":"","sources":["../../src/ports/files.ts"],"names":[],"mappings":""}