@reportwright/pdf 0.1.0-beta.0 → 0.1.0-beta.1

package/README.md

@@ -38,6 +38,7 @@ writer to be ready. `pdf.end()` ends a Node stream and closes a `WritableStream`
 | option | meaning |
 |---|---|
 | `title`, `author`, `subject`, `keywords` | document info, mirrored in the XMP metadata (every value escaped) |
+| `metadata` | `true` (default): an XMP metadata stream mirrors the info. `false`: the Info dictionary only, no XMP (a one-page invoice goes from 2.5 KB to 1.6 KB). Refused with `pdfa` or `tagged`: PDF/A and PDF/UA require the XMP |
 | `creationDate` | a `Date` (default: now). With the same input and date the output is byte-for-byte identical |
 | `compress` | `true` (default): content, fonts, object streams and the xref stream are Flate-compressed |
 | `pdfa` | `true`: PDF/A-2b (sRGB output intent bundled, XMP identification). Fonts must be embedded. `{ part, conformance, outputIntent }`: a level (see PDF/A levels) and/or your RGB or CMYK output profile instead of sRGB (a CMYK intent allows CMYK colour and refuses RGB) |
@@ -61,6 +62,9 @@ writer to be ready. `pdf.end()` ends a Node stream and closes a `WritableStream`
   CID-keyed CFF gets an identity charset so every reader maps its glyphs alike); `false` embeds the whole font; or a
   function `(font, codePoints, { glyphs }) => Promise<Uint8Array>` (e.g. a HarfBuzz subsetter, which also
   desubroutinises CFF) that must keep glyph IDs. `shaper`: see Shaping.
+  A variable font (an `fvar` table) is drawn at its default instance, which may be its thinnest weight (Noto Sans JP's
+  is Thin), with a warning naming the axes' values: for another weight, embed a static instance of it (fonttools
+  `varLib.instancer`, or `hb-subset --instance`).
 - `font.widthOfText(text, size, { kerning, characterSpacing, wordSpacing, horizontalScaling, direction })` — the width
   text is drawn at (kerned, shaped, spaced); `font.widthOf(text, size)` — plain advances; `font.lineHeight(size)` —
   ascent − descent + line gap (1.2 × size for standard fonts); `font.ascent`, `font.descent` (fractions of the size).
@@ -123,14 +127,31 @@ Text options (`page.text` and `page.textBox`):
 returns the glyphs in visual order, in font units: `[{ g, cl, ax, dx, dy }]` (glyph ID, cluster = UTF-16 index of
 its first character, x advance, offsets: what HarfBuzz's `hb_shape` gives). So Arabic, Indic and other complex
 scripts draw correctly. The glyphs are drawn by ID with their advances and offsets (TJ adjustments, `Ts` for vertical
-offsets); a glyph standing for its cluster's text maps it in ToUnicode (a ligature maps to all its letters); a cluster
-of several glyphs (reordered matras, conjuncts) or a glyph already standing for other text is wrapped in an
-`/ActualText` span, so copied text is the source text. The shaper's output is checked (glyph IDs in the font, finite
-numbers, clusters inside the text, at most 4 glyphs a character). `examples/harfbuzz-shaper.mjs` is an adapter for
-harfbuzzjs (not a dependency):
+offsets); copied and extracted text is the source text, once, in logical order (checked with Acrobat-style
+ActualText readers, poppler and pdf.js): a cluster of several glyphs (a conjunct, a reordered matra, a letter and its
+dots) has its text on one glyph and WJ (U+2060) on the others, which are drawn after the run or inside an
+`/ActualText` span, and a glyph that stands for different text in different places gets a code for each (a
+`CIDToGIDMap`; a CFF font cannot, so there it gets an ActualText span). The shaper's output is checked (glyph IDs in
+the font, finite numbers, clusters inside the text, at most 4 glyphs a character).
+
+An adapter for harfbuzzjs v1 (`npm i harfbuzzjs`; not a dependency of this package). The same file ships in the package
+as `examples/harfbuzz-shaper.mjs`:
 
 ```js
-import { harfbuzzShaper } from './harfbuzz-shaper.mjs';           // copy it from examples/
+import * as hb from 'harfbuzzjs';
+function harfbuzzShaper(bytes) {
+  const font = new hb.Font(new hb.Face(new hb.Blob(bytes)));     // positions in font units
+  return (text, _font, { direction, kerning }) => {
+    const buf = new hb.Buffer();
+    buf.addText(text);                                           // clusters are UTF-16 indices
+    buf.guessSegmentProperties();                                // script and language from the text
+    buf.setDirection(direction === 'rtl' ? hb.Direction.RTL : hb.Direction.LTR);
+    hb.shape(font, buf, kerning ? [] : [new hb.Feature('kern', 0)]);
+    const pos = buf.getGlyphPositions();
+    return buf.getGlyphInfos().map((g, i) => ({ g: g.codepoint, cl: g.cluster, ax: pos[i].xAdvance, dx: pos[i].xOffset, dy: pos[i].yOffset }));
+  };
+}
+
 const bytes = fs.readFileSync('NotoSansDevanagari-Regular.ttf');
 const hindi = await pdf.embedFont(bytes, { shaper: harfbuzzShaper(bytes) });
 page.text('किताब हिन्दी', { x: 40, y: 40, font: hindi, size: 18 });
@@ -657,7 +678,11 @@ const signed = await doc.saveIncremental({ sign: { field: { name: 'approval', pa
   sign: an empty signature field by name, or a new one (`{ name, page, x, y, width, height }`; no size: invisible), with
   group F's signers, the `/ByteRange` over the whole file including the original bytes. An encrypted file's new objects
   use its own AES key; RC4 files cannot be updated (writing RC4 is refused).
-- `mergePdfs(docs, options)`: every page of each, into a new file.
+- `mergePdfs(docs, options)`: every page of each, into a new file. **A merged PDF/UA (or PDF/A-a) file is no longer
+  tagged**: the structure tree is kept only when every page comes from one document (and the form is not flattened),
+  so a merge drops it with `/MarkInfo` and `/Lang`, the XMP's PDF/UA identification (`pdfuaid`) goes too, PDF/A-a is
+  claimed as PDF/A-b, and `warnings` says "the structure tree (tags) was dropped … the result is not tagged (no
+  longer PDF/UA …)". Tag the merged content again with the writer if the result must be PDF/UA.
 - Errors from the input are `PdfReadError` with a `code`: `E_MALFORMED`, `E_BUDGET`, `E_DEPTH`, `E_CYCLE`,
   `E_TIMEOUT`, `E_ABORTED`, `E_PASSWORD`, `E_UNSUPPORTED`, `E_ACTIVE`.
 
@@ -689,7 +714,7 @@ JavaScript, launch and submit actions cannot be written at all (see Forms and Li
 |---|---|
 | Decoded stream bytes, the whole document / one stream (every filter and predictor counts as it produces, so a zip bomb stops at the cap: a 302-byte stream that inflates to 10 GB stops at 128 MB) | `maxDecoded` 512 MB / `maxStream` 128 MB |
 | Parsing steps (tokens, xref rows, tree nodes, operators); time | `maxSteps` 200 M; `timeoutMs`, `signal` |
-| Cross-reference entries and scanned objects held; one page's text pieces | `maxObjects` 5 M; `maxTextItems` 2 M |
+| Cross-reference entries and scanned objects held; one page's text pieces | `maxObjects` 5 M; `maxTextItems` 1 M (and 16 M characters a page, ActualText included) |
 | Nesting of arrays and dictionaries (an explicit stack, never recursion) | 256 |
 | A chain of indirect references | 32 |
 
@@ -738,3 +763,11 @@ widgets may cover at most 40% of a page in all, and a widget with an opaque back
 
 Image masks (stencils) and colour-managed images (embedded ICC profiles of PNG/JPEG files are ignored); the Unicode
 bidi algorithm, hyphenation, kerning of the standard fonts, WOFF fonts; popup annotations; RC4 encryption (refused; read only), public-key (certificate) encryption (neither written nor read), verifying signatures' cryptography (they are listed with their coverage); PDF/A-1a, PDF/X, attachment annotations in PDF/A-3 (use `pdf.attach`), validating a Factur-X invoice's content; layout-perfect text extraction; page edits in an incremental update (use `save()`); reading encrypted files in browsers (WebCrypto has no synchronous AES; Node only). See NOTES.md for the plan. Extending the library: ARCHITECTURE.md.
+
+## Changes
+
+0.1.0-beta.1: shaped text (Indic, Arabic) extracts once and in order in pdf.js, poppler and `extractText`;
+`extractText` honours ActualText, reads unmapped glyphs as U+FFFD, falls back to an embedded TrueType cmap, spaces
+separate text objects (spreadsheet cells), orders a line along its baseline (rotated text too), and stops a
+2 M-operator page at 170 MB (`maxTextItems` 1 M); standard-font text writes 40% faster; variable fonts warn; the
+package ships `examples/`. The list with details: NOTES.md ("Changes") in the repository.

package/dist/index.js

@@ -1663,14 +1663,14 @@ function createDoc(out, o) {
         }
         await writeNodes(level2);
         const I = o.info, when = I.creationDate;
-        const xmp = doc.xmp ?? new TextEncoder().encode(xmpPacket({ title: I.title, author: I.author, subject: I.subject, keywords: I.keywords, stamp: when.toISOString().replace(/\.\d+Z$/, "Z"), pdfa: doc.pdfaLevel, ua: o.tagged, extensions: doc.xmpExtensions }));
-        const meta = objs.reserve();
-        await objs.write(meta, "<< /Type /Metadata /Subtype /XML >>", xmp);
+        const xmp = o.metadata === false ? null : doc.xmp ?? new TextEncoder().encode(xmpPacket({ title: I.title, author: I.author, subject: I.subject, keywords: I.keywords, stamp: when.toISOString().replace(/\.\d+Z$/, "Z"), pdfa: doc.pdfaLevel, ua: o.tagged, extensions: doc.xmpExtensions }));
+        const meta = xmp ? objs.reserve() : 0;
+        if (xmp) await objs.write(meta, "<< /Type /Metadata /Subtype /XML >>", xmp);
         const info = objs.reserve();
         if (doc.info != null) await objs.write(info, doc.info);
         else await objs.write(info, `<< /Producer ${text(PRODUCER)} /Creator ${text(PRODUCER)}${I.title != null ? ` /Title ${text(I.title)}` : ""}${I.author != null ? ` /Author ${text(I.author)}` : ""}${I.subject != null ? ` /Subject ${text(I.subject)}` : ""}${I.keywords != null ? ` /Keywords ${text(I.keywords)}` : ""} /CreationDate ${pdfDate(when)} /ModDate ${pdfDate(when)} >>`);
         const root = objs.reserve();
-        let cat = `<< /Type /Catalog /Pages ${level2[0].ref} 0 R /Metadata ${meta} 0 R`;
+        let cat = `<< /Type /Catalog /Pages ${level2[0].ref} 0 R${meta ? ` /Metadata ${meta} 0 R` : ""}`;
         for (const [k, v] of catalog) cat += ` ${pdfName(k)} ${v}`;
         await objs.write(root, cat + " >>");
         const r = await objs.finish({ root, info, id: o.id, last: doc.last });
@@ -2028,13 +2028,12 @@ function segments(str3, fonts, o) {
 }
 function runWidth(run2, size, sp) {
   let a = 0, n = 0, s = 0;
-  if (Array.isArray(run2)) {
-    n = run2.length;
-    for (const it of run2) {
-      a += it.a;
-      if (it.sp) s++;
-    }
-  } else ({ a, n, sp: s } = run2);
+  if (Array.isArray(run2)) for (const it of run2) {
+    a += it.a;
+    if (it.sp) s++;
+    if (!it.c) n++;
+  }
+  else ({ a, n, sp: s } = run2);
   return (a * size / 1e3 + n * sp.tc + s * sp.tw) * sp.tz / 100;
 }
 function measure(impl, str3, size, o) {
@@ -2042,11 +2041,11 @@ function measure(impl, str3, size, o) {
   const sp = spacing(o);
   return runWidth(impl.items(str3, { kerning: o?.kerning, rtl: o?.direction === "rtl", ws: sp.tw, measure: true }), size, sp);
 }
-function runOps(run2, size, rise, tw, standard) {
+function runOps(run2, size, rise, tw, standard, tc = 0) {
   if (!Array.isArray(run2)) return run2.body ? `(${run2.body}) Tj
 ` : "";
-  const wsT = standard ? 0 : tw * 1e3 / size;
-  let s = "", arr = "", lit = "", cur = 0, pos = 0, r = rise, left = 0, adjusted = false;
+  const wsT = standard ? 0 : tw * 1e3 / size, tcT = tc * 1e3 / size;
+  let s = "", arr = "", lit = "", cur = 0, pos = 0, r = rise, left = 0, adjusted = false, out = false;
   const endLit = () => {
     if (lit) {
       arr += `(${lit})`;
@@ -2063,13 +2062,7 @@ function runOps(run2, size, rise, tw, standard) {
       adjusted = false;
     }
   };
-  for (const it of run2) {
-    if (it.at !== void 0) {
-      flush();
-      s += `/Span << /ActualText ${text(it.at)} >> BDC
-`;
-      left = it.n || 1;
-    }
+  const put = (it) => {
     const want = rise + it.dy * size / 1e3;
     if (Math.abs(want - r) >= 5e-4) {
       flush();
@@ -2087,6 +2080,20 @@ function runOps(run2, size, rise, tw, standard) {
     }
     lit += it.b;
     cur += it.w;
+  };
+  for (const it of run2) {
+    if (it.c) {
+      pos += it.a;
+      out = true;
+      continue;
+    }
+    if (it.at !== void 0) {
+      flush();
+      s += `/Span << /ActualText ${text(it.at)} >> BDC
+`;
+      left = it.n || 1;
+    }
+    put(it);
     pos += it.a + (it.sp ? wsT : 0);
     if (left && --left === 0) {
       flush();
@@ -2094,6 +2101,23 @@ function runOps(run2, size, rise, tw, standard) {
     }
   }
   flush();
+  if (out) {
+    s += `/Span << /ActualText ${text("")} >> BDC
+0 0 Td
+`;
+    cur = pos = 0;
+    for (const it of run2) {
+      if (!it.c) {
+        pos += it.a + (it.sp ? wsT : 0) + tcT;
+        continue;
+      }
+      put(it);
+      cur += tcT;
+      pos += it.a;
+    }
+    flush();
+    s += "EMC\n";
+  }
   if (r !== rise) s += `${num(rise)} Ts
 `;
   return s;
@@ -2148,7 +2172,7 @@ function drawLine(page, fonts, str3, x2, base3, size, ts, o) {
   }
   let at = x2;
   for (const s of segs) {
-    const ops = runOps(s.run, size, ts.rise, ts.sp.tw, s.impl.standard);
+    const ops = runOps(s.run, size, ts.rise, ts.sp.tw, s.impl.standard, ts.sp.tc);
     if (ops) c.run(s.font.name, size, at, page.height - base3, ops, ts === PLAIN ? PLAIN_STATE : { Tc: ts.sp.tc, Tw: s.impl.standard ? ts.sp.tw : 0, Tz: ts.sp.tz, Ts: ts.rise, Tr: ts.Tr });
     if (segs.length > 1) at += runWidth(s.run, size, ts.sp);
   }
@@ -10009,6 +10033,13 @@ function standardFont(doc, key) {
     /** @type {number[]} */
     m.widths
   );
+  const W8 = new Float64Array(256), SAME = new Uint8Array(256);
+  for (let c = 0; c < 256; c++) {
+    const code = codeOf.get(c);
+    W8[c] = code === void 0 ? -1 : widths[code] || 0;
+    SAME[c] = code === c && esc(code) === String.fromCharCode(c) ? 1 : 0;
+  }
+  const SP32 = codeOf.get(32) === 32;
   const ref = doc.reserve();
   const missing = /* @__PURE__ */ new Set();
   const memo2 = /* @__PURE__ */ new Map();
@@ -10068,8 +10099,31 @@ function standardFont(doc, key) {
     lineHeight: 1.2,
     has: (cp) => codeOf.has(cp),
     items(text3, o) {
-      let body = "", a = 0, n = 0, sp = 0;
       const draw = !o.measure;
+      if (!o.rtl) {
+        let a2 = 0, n2 = 0, sp2 = 0, same = true;
+        for (let i2 = 0; i2 < text3.length; i2++) {
+          const c = text3.charCodeAt(i2);
+          if (c > 255) {
+            same = false;
+            break;
+          }
+          const w = W8[c];
+          if (w < 0) {
+            same = false;
+            break;
+          }
+          a2 += w;
+          n2++;
+          if (c === 32 && SP32) sp2++;
+          if (!SAME[c]) {
+            same = false;
+            break;
+          }
+        }
+        if (same) return { body: draw ? text3 : "", a: a2, n: n2, sp: sp2 };
+      }
+      let body = "", a = 0, n = 0, sp = 0;
       for (const ch of o.rtl ? [...text3].reverse() : text3) {
         const b = byte(ch);
         if (!b) {
@@ -10938,9 +10992,15 @@ function embedFont(doc, input, opt = {}) {
   use(0);
   const two = (g) => esc(g >> 8) + esc(g & 255);
   const kern = kerning(f.tables, f.numGlyphs, (m) => doc.warn(`${f.postscriptName.slice(0, 63)}: ${m}`));
+  const axes = variation(f.tables.get("fvar"));
+  if (axes) doc.warn(`${f.postscriptName.slice(0, 63)} is a variable font: it is drawn at its default instance (${axes}); to draw another weight or width, embed a static instance of it (fonttools varLib.instancer, or HarfBuzz's hb-subset --instance)`);
   const k = 1e3 / f.unitsPerEm;
   const ref = doc.reserve();
-  const glyphs2 = /* @__PURE__ */ new Map();
+  const codes = /* @__PURE__ */ new Map();
+  const codeOf = /* @__PURE__ */ new Map();
+  const copies = !f.cff;
+  const EXACT = [];
+  let next = f.numGlyphs;
   const missing = /* @__PURE__ */ new Set();
   const memo2 = /* @__PURE__ */ new Map();
   const strict = doc.pdfa || doc.tagged;
@@ -10957,24 +11017,31 @@ function embedFont(doc, input, opt = {}) {
       );
       if (!g) lack(ch);
       if (!g && missing.size < 100) missing.add(ch);
-      if (!glyphs2.has(g) && (g || !strict)) {
-        use(g);
-        glyphs2.set(g, g ? ch : "");
-      }
-      memo2.set(ch, b = !g && strict ? "" : two(g));
+      memo2.set(ch, b = !g && strict ? "" : two(code(g, ch)[0]));
     }
     return b;
   };
-  const record = (g, t) => {
+  const code = (g, t) => {
     const m = t.length <= 32 ? t : String.fromCodePoint(
       /** @type {number} */
       t.codePointAt(0)
     );
-    if (!glyphs2.has(g)) {
-      use(g);
-      glyphs2.set(g, g ? m : "");
+    if (codes.get(g)?.t === m) return m === t ? EXACT[g] ??= [g, true] : [g, false];
+    let c = codeOf.get(`${g}:${m}`);
+    if (c === void 0) {
+      if (!codes.has(g)) {
+        use(g);
+        c = g;
+      } else if (copies && next <= 65535) c = next++;
+      else return [
+        g,
+        /** @type {any} */
+        codes.get(g).t === t
+      ];
+      codes.set(c, { g, t: m });
+      codeOf.set(`${g}:${m}`, c);
     }
-    return glyphs2.get(g) === t;
+    return [c, m === t];
   };
   let handle;
   const items = (text3, o) => {
@@ -11025,9 +11092,9 @@ function embedFont(doc, input, opt = {}) {
         if (missing.size < 100 && !o.measure) missing.add(ch);
         if (strict) continue;
       }
-      const ok = o.measure || record(g, ch), w = W(g);
+      const [c, ok] = o.measure ? [g, true] : code(g, ch), w = W(g);
       if (kerned && prev >= 0) out[out.length - 1].a += kern(prev, g) * k;
-      out.push({ b: two(g), w, a: w, dx: 0, dy: 0, sp: ch === " ", at: ok ? void 0 : ch, n: 1 });
+      out.push({ b: two(c), w, a: w, dx: 0, dy: 0, sp: ch === " ", at: ok ? void 0 : ch, n: 1 });
       prev = g;
     }
     return out;
@@ -11051,33 +11118,61 @@ function embedFont(doc, input, opt = {}) {
     });
     const starts = [...new Set(gl.map((x2) => x2.cl))].sort((a, b) => a - b);
     const end = new Map(starts.map((c, i2) => [c, starts[i2 + 1] ?? text3.length]));
-    const out = [];
+    const out = [], said = /* @__PURE__ */ new Set();
+    const at = (a) => {
+      let b = a;
+      while (b < gl.length && gl[b].cl === gl[a].cl) b++;
+      return [b, said.has(gl[a].cl) ? "" : text3.slice(gl[a].cl, end.get(gl[a].cl))];
+    };
+    let glued = null;
     for (let i2 = 0; i2 < gl.length; ) {
-      let j = i2;
-      while (j < gl.length && gl[j].cl === gl[i2].cl) j++;
-      const t = text3.slice(gl[i2].cl, end.get(gl[i2].cl)), first = out.length;
-      let ok = j - i2 === 1;
+      const [j, t0] = at(i2), cl = gl[i2].cl, first = out.length;
+      if (glued && t0 === " " && j - i2 === 1) {
+        const { g, ax, dx, dy } = gl[i2];
+        out.push({ b: two(o.measure ? g : code(g, "⁠")[0]), w: W(g), a: ax * k, dx: dx * k, dy: dy * k, sp: true });
+        if (glued.n) glued.n++;
+        glued = null;
+        said.add(cl);
+        i2 = j;
+        continue;
+      }
+      glued = null;
+      const glue = !o.rtl && !!t0 && t0 !== " " && j < gl.length && at(j)[1] === " " && at(j)[0] - j === 1 && at(j)[0] < gl.length && MN.test(at(at(j)[0])[1]);
+      const t = glue ? `${t0} ` : t0;
+      let lead = -1, span = glue;
+      if (t0) {
+        for (let q = i2; q < j; q++) if ((gl[q].g || !strict) && (lead < 0 || gl[q].ax > gl[lead].ax)) lead = q;
+      }
       for (let q = i2; q < j; q++) {
         const { g, ax, dx, dy } = gl[q];
         if (!g) {
           if (!o.measure) lack(t);
-          if (missing.size < 100 && !o.measure) missing.add(t);
+          if (missing.size < 100 && !o.measure && t) missing.add(t);
           if (strict) {
             if (out.length > first) out[out.length - 1].a += ax * k;
             continue;
           }
         }
-        ok = (o.measure || record(g, t)) && ok;
-        out.push({ b: two(g), w: W(g), a: ax * k, dx: dx * k, dy: dy * k, sp: t === " " });
-      }
-      if (out.length > first && !ok) {
-        out[first].at = t;
-        out[first].n = out.length - first;
+        const [c, exact] = o.measure ? [g, true] : code(g, q !== lead ? "⁠" : o.rtl && RTL3.test(t) ? [...t].reverse().join("") : t);
+        if (!exact) span = true;
+        if (q !== lead && Math.abs(ax * k) >= 1) span ||= !o.rtl;
+        out.push({ b: two(c), w: W(g), a: ax * k, dx: dx * k, dy: dy * k, sp: q === lead && t === " ", c: q !== lead && Math.abs(ax * k) < 1 || void 0 });
       }
+      if (out.length > first) said.add(cl);
+      if (span && !o.measure) {
+        const flow = out.slice(first).filter((x2) => !x2.c);
+        if (flow.length) {
+          flow[0].at = t;
+          flow[0].n = flow.length;
+          if (glue) glued = flow[0];
+        }
+      } else if (glue) glued = /** @type {any} */
+      {};
       i2 = j;
     }
     return out;
   };
+  const MN = /\p{Mn}/u, RTL3 = /[\u0590-\u08ff\ufb1d-\ufdff\ufe70-\ufeff]/;
   const impl = {
     standard: false,
     items,
@@ -11114,12 +11209,12 @@ function embedFont(doc, input, opt = {}) {
       return s;
     },
     async write() {
-      if (!glyphs2.size) glyphs2.set(0, "");
-      const used = [...glyphs2.keys()].sort((a, b) => a - b);
-      const cps = [...glyphs2.values()].filter(Boolean).flatMap((t) => [...t].map((c) => (
+      const cids = [...codes.keys()].sort((a, b) => a - b);
+      const used = [.../* @__PURE__ */ new Set([0, ...[...codes.values()].map((x2) => x2.g)])].sort((a, b) => a - b);
+      const cps = [...new Set([...codes.values()].flatMap((x2) => [...x2.t].map((c) => (
         /** @type {number} */
         c.codePointAt(0)
-      )));
+      ))))];
       let file = bytes;
       if (sub && !(f.fsType & 256)) {
         if (typeof sub === "function") {
@@ -11144,8 +11239,8 @@ function embedFont(doc, input, opt = {}) {
       ) : file);
       let cidSet = "";
       if (subset && doc.pdfaLevel?.part === 1) {
-        const bits2 = new Uint8Array((used[used.length - 1] >> 3) + 1), n = doc.reserve();
-        for (const g of used) bits2[g >> 3] |= 128 >> (g & 7);
+        const all = [.../* @__PURE__ */ new Set([...used, ...cids])], bits2 = new Uint8Array((Math.max(...all) >> 3) + 1), n = doc.reserve();
+        for (const g of all) bits2[g >> 3] |= 128 >> (g & 7);
         await doc.writeStream(n, "<< >>", bits2);
         cidSet = ` /CIDSet ${n} 0 R`;
       }
@@ -11153,16 +11248,27 @@ function embedFont(doc, input, opt = {}) {
       const flags = (f.fixedPitch ? 1 : 0) | 4 | (f.italicAngle ? 64 : 0);
       const [x0, y0, x1, y1] = f.bbox;
       await doc.write(desc, `<< /Type /FontDescriptor /FontName ${base3} /Flags ${flags} /FontBBox [${s(x0)} ${s(y0)} ${s(x1)} ${s(y1)}] /ItalicAngle ${num(f.italicAngle)} /Ascent ${s(f.ascent)} /Descent ${s(f.descent)} /CapHeight ${s(f.capHeight)} /XHeight ${s(f.xHeight)} /StemV 0 /${cff ? "FontFile3" : "FontFile2"} ${fileRef} 0 R${cidSet} >>`);
-      const codes = used.map((g) => [g, g]);
+      const wc = (cids.length ? cids : [0]).map((c) => [c, codes.get(c)?.g ?? 0]);
       let Wa = "";
-      for (let i2 = 0; i2 < codes.length; ) {
+      for (let i2 = 0; i2 < wc.length; ) {
         let j = i2;
-        while (j + 1 < codes.length && codes[j + 1][0] === codes[j][0] + 1) j++;
-        Wa += `${codes[i2][0]} [${codes.slice(i2, j + 1).map(([, g]) => s(f.advance(g))).join(" ")}] `;
+        while (j + 1 < wc.length && wc[j + 1][0] === wc[j][0] + 1) j++;
+        Wa += `${wc[i2][0]} [${wc.slice(i2, j + 1).map(([, g]) => s(f.advance(g))).join(" ")}] `;
         i2 = j + 1;
       }
-      await doc.write(cid, `<< /Type /Font /Subtype /${cff ? "CIDFontType0" : "CIDFontType2"} /BaseFont ${base3} /CIDSystemInfo << /Registry ${ascii("Adobe")} /Ordering ${ascii("Identity")} /Supplement 0 >> /FontDescriptor ${desc} 0 R /W [${Wa}]${cff ? "" : " /CIDToGIDMap /Identity"} >>`);
-      await doc.writeStream(cmap, "<< >>", latin1(toUnicode(glyphs2)));
+      let c2g = " /CIDToGIDMap /Identity";
+      if (!cff && next > f.numGlyphs) {
+        const m = new Uint8Array(2 * next), r = doc.reserve();
+        for (let c = 0; c < next; c++) {
+          const g = c < f.numGlyphs ? c : codes.get(c)?.g ?? 0;
+          m[2 * c] = g >> 8;
+          m[2 * c + 1] = g & 255;
+        }
+        await doc.writeStream(r, "<< >>", m);
+        c2g = ` /CIDToGIDMap ${r} 0 R`;
+      }
+      await doc.write(cid, `<< /Type /Font /Subtype /${cff ? "CIDFontType0" : "CIDFontType2"} /BaseFont ${base3} /CIDSystemInfo << /Registry ${ascii("Adobe")} /Ordering ${ascii("Identity")} /Supplement 0 >> /FontDescriptor ${desc} 0 R /W [${Wa}]${cff ? "" : c2g} >>`);
+      await doc.writeStream(cmap, "<< >>", latin1(toUnicode(codes)));
       await doc.write(ref, `<< /Type /Font /Subtype /Type0 /BaseFont ${base3} /Encoding /Identity-H /DescendantFonts [${cid} 0 R] /ToUnicode ${cmap} 0 R >>`);
     }
   };
@@ -11193,8 +11299,11 @@ function subsetOtf(bytes, n, used, seac) {
   out.set("CFF ", cff);
   return buildSfnt(out, [79, 84, 84, 79]);
 }
-function toUnicode(glyphs2) {
-  const entries = [...glyphs2].filter(([, t]) => t).sort((a, b) => a[0] - b[0]);
+function toUnicode(codes) {
+  const entries = [...codes].map(([c, x2]) => (
+    /** @type {[number, string]} */
+    [c, x2.t]
+  )).sort((a, b) => a[0] - b[0]);
   let body = "";
   for (let i2 = 0; i2 < entries.length; i2 += 100) {
     const chunk = entries.slice(i2, i2 + 100);
@@ -11218,6 +11327,16 @@ end
 end
 `;
 }
+function variation(t) {
+  if (!t || t.length < 16) return "";
+  const dv = new DataView(t.buffer, t.byteOffset, t.byteLength), off = dv.getUint16(4), n = dv.getUint16(8), size = dv.getUint16(10);
+  const out = [];
+  for (let i2 = 0; i2 < Math.min(n, 16) && size >= 20 && off + i2 * size + 20 <= t.length; i2++) {
+    const p = off + i2 * size;
+    out.push(`${String.fromCharCode(...t.subarray(p, p + 4)).replace(/[^\x20-\x7e]/g, "?")} ${Math.round(dv.getInt32(p + 8) / 655.36) / 100}`);
+  }
+  return out.join(", ") || "no axes read";
+}
 function tag2(bytes) {
   let h = 2166136261;
   for (let i2 = 0; i2 < bytes.length; i2++) h = Math.imul(h ^ bytes[i2], 16777619);
@@ -12883,7 +13002,7 @@ var init_budget = __esm({
     fail = (code, msg) => {
       throw new PdfReadError(code, msg);
     };
-    DEFAULTS = { maxDecoded: 512 << 20, maxStream: 128 << 20, maxSteps: 2e8, maxObjects: 5e6, maxTextItems: 2e6 };
+    DEFAULTS = { maxDecoded: 512 << 20, maxStream: 128 << 20, maxSteps: 2e8, maxObjects: 5e6, maxTextItems: 1e6 };
   }
 });
 
@@ -13648,9 +13767,45 @@ var init_crypt = __esm({
 async function openSource(bytes, budget, o = {}) {
   const size = bytes.length;
   let xref = /* @__PURE__ */ new Map();
-  const cache = /* @__PURE__ */ new Map();
+  let young = /* @__PURE__ */ new Map(), old = /* @__PURE__ */ new Map();
+  const weak = /* @__PURE__ */ new Map();
+  const cache = {
+    /** @param {number} n @returns {any} undefined: not cached */
+    get(n) {
+      let v = young.get(n);
+      if (v !== void 0) return v;
+      v = old.get(n);
+      if (v === void 0) {
+        v = weak.get(n)?.deref();
+        if (v === void 0) {
+          weak.delete(n);
+          return void 0;
+        }
+        weak.delete(n);
+      }
+      cache.set(n, v);
+      return v;
+    },
+    /**
+     * Held strongly for two turns of CACHE_N (a cheap LRU: no reordering per hit), then weakly
+     * @param {number} n @param {any} v
+     */
+    set(n, v) {
+      young.set(n, v);
+      if (young.size < CACHE_N) return;
+      for (const [k, o2] of old) if (o2 !== null && typeof o2 === "object" && !young.has(k)) weak.set(k, new WeakRef(o2));
+      old = young;
+      young = /* @__PURE__ */ new Map();
+    },
+    clear() {
+      young.clear();
+      old.clear();
+      weak.clear();
+    }
+  };
   const loading = /* @__PURE__ */ new Set();
   const objstms = /* @__PURE__ */ new Map();
+  let objstmBytes = 0;
   let endstreams = null;
   let recovered = false;
   let security = null, encryptRef = -1;
@@ -13768,12 +13923,17 @@ async function openSource(bytes, budget, o = {}) {
         if (lx2.next() !== T.NUM) break;
         offs[k] = lx2.val;
       }
-      os = { data, nums: nums.subarray(0, k), offs: offs.subarray(0, k), first };
-      if (objstms.size >= OBJSTM_CACHE) objstms.delete(
-        /** @type {number} */
-        objstms.keys().next().value
-      );
+      os = { data, nums: nums.slice(0, k), offs: offs.slice(0, k), first };
+      while (objstms.size && (objstms.size >= OBJSTM_CACHE || objstmBytes + data.length > OBJSTM_BYTES)) {
+        const [k0, o0] = (
+          /** @type {[number, any]} */
+          objstms.entries().next().value
+        );
+        objstms.delete(k0);
+        objstmBytes -= o0.data.length;
+      }
       objstms.set(s, os);
+      objstmBytes += data.length;
     }
     let j = os.nums[i2] === n ? i2 : -1;
     if (j < 0) {
@@ -13792,7 +13952,8 @@ async function openSource(bytes, budget, o = {}) {
     return v === void 0 || v instanceof Op ? null : v;
   }
   function get(n) {
-    if (cache.has(n)) return cache.get(n);
+    const c = cache.get(n);
+    if (c !== void 0) return c;
     if (loading.has(n)) return null;
     loading.add(n);
     try {
@@ -14046,6 +14207,7 @@ async function openSource(bytes, budget, o = {}) {
     }
     cache.clear();
     objstms.clear();
+    objstmBytes = 0;
   }
   function recoverDeep() {
     const found = xref, tr = trailer;
@@ -14116,6 +14278,7 @@ async function openSource(bytes, budget, o = {}) {
       const keep = encryptRef >= 0 ? cache.get(encryptRef) : void 0;
       cache.clear();
       objstms.clear();
+      objstmBytes = 0;
       if (keep !== void 0) cache.set(encryptRef, keep);
     }
   }
@@ -14124,7 +14287,13 @@ async function openSource(bytes, budget, o = {}) {
     recover();
     recoverDeep();
   }
-  const root = resolveSafe(trailer.get("Root"));
+  let root = null;
+  try {
+    root = resolve2(trailer.get("Root"));
+  } catch (e) {
+    if (!(e instanceof PdfReadError) || e.code === "E_TIMEOUT" || e.code === "E_ABORTED" || e.code === "E_BUDGET") throw e;
+    if (e.code === "E_DEPTH") fail("E_DEPTH", `The document catalog cannot be read: ${e.message}`);
+  }
   if (!(root instanceof Map)) fail("E_MALFORMED", "The PDF has no document catalog (/Root)");
   return {
     bytes,
@@ -14305,7 +14474,7 @@ function inRegion(regions, budget) {
     return reach > off;
   };
 }
-var MAX_SECTIONS, MAX_RESOLVE, MAX_OBJSTM_N, OBJSTM_CACHE, P48, P16, pack, unpack;
+var MAX_SECTIONS, MAX_RESOLVE, MAX_OBJSTM_N, OBJSTM_CACHE, OBJSTM_BYTES, CACHE_N, P48, P16, pack, unpack;
 var init_source = __esm({
   "src/reader/source.js"() {
     init_lexer();
@@ -14316,6 +14485,8 @@ var init_source = __esm({
     MAX_RESOLVE = 32;
     MAX_OBJSTM_N = 1e6;
     OBJSTM_CACHE = 32;
+    OBJSTM_BYTES = 8 << 20;
+    CACHE_N = 1024;
     P48 = 2 ** 48;
     P16 = 2 ** 16;
     pack = (t, a, b) => t * P48 + a * P16 + b;
@@ -14432,7 +14603,10 @@ function parseCMap(data, budget, held = { n: 0 }) {
   }
   return { map, spaces };
 }
-function fontOf(src, font, held = { n: 0 }) {
+function fontOf(src, font, held = (
+  /** @type {Held} */
+  { n: 0, cmaps: /* @__PURE__ */ new Map() }
+)) {
   const R = src.resolve;
   const sub = R(font.get("Subtype"));
   let toU = null, spaces = (
@@ -14489,6 +14663,26 @@ function fontOf(src, font, held = { n: 0 }) {
       }
     }
     const lens = spaces.length ? [...new Set(spaces.map((s) => s[0]))].sort() : [2];
+    let back;
+    const fromFont = (c) => {
+      if (back === void 0) back = fontCmap(src, cid, held);
+      if (!back) return void 0;
+      const m = cid instanceof Map ? R(cid.get("CIDToGIDMap")) : null;
+      let g = c;
+      if (m instanceof PdfStream) {
+        if (!gidMap) {
+          try {
+            gidMap = src.decodeStream(m).data.slice(0, 131072);
+          } catch (e) {
+            if (stops(e)) throw e;
+            gidMap = new Uint8Array(0);
+          }
+        }
+        g = 2 * c + 1 < gidMap.length ? gidMap[2 * c] << 8 | gidMap[2 * c + 1] : 0;
+      }
+      return back.get(g);
+    };
+    let gidMap;
     return {
       /** each glyph to emit, none held (a 16 MB string is not 16 M objects) @param {Uint8Array} b @param {Emit} emit */
       decode(b, emit) {
@@ -14504,7 +14698,7 @@ function fontOf(src, font, held = { n: 0 }) {
           }
           for (let k = 0; k < L && i2 + k < b.length; k++) c = c * 256 + b[i2 + k];
           i2 += L;
-          emit(toU?.get(c) ?? "", W.get(c) ?? dw, false);
+          emit(toU?.get(c) ?? fromFont(c) ?? NONE, W.get(c) ?? dw, false);
         }
       }
     };
@@ -14527,7 +14721,7 @@ function fontOf(src, font, held = { n: 0 }) {
     for (const x0 of d.slice(0, 4096)) {
       const x2 = R(x0);
       if (typeof x2 === "number") c = x2;
-      else if (typeof x2 === "string") diff.set(c++ & 255, glyphText(x2));
+      else if (typeof x2 === "string") diff.set(c++ & 255, x2 === ".notdef" ? "" : glyphText(x2) || NONE);
     }
   }
   const symbolic = base3 === "Symbol" || base3 === "ZapfDingbats";
@@ -14547,7 +14741,7 @@ function fontOf(src, font, held = { n: 0 }) {
         if (t === void 0) t = diff.get(c);
         if (t === void 0) {
           const u = symMap ? symMap.get(c) : baseFn(c);
-          t = u === void 0 ? "" : String.fromCodePoint(u);
+          t = u === void 0 ? NONE : String.fromCodePoint(u);
         }
         let w = Array.isArray(widths) && typeof fc === "number" && c >= fc && c - fc < widths.length ? R(widths[c - fc]) : std ? std.widths[c] : 500;
         if (typeof w !== "number" || !Number.isFinite(w)) w = 0;
@@ -14556,6 +14750,34 @@ function fontOf(src, font, held = { n: 0 }) {
     }
   };
 }
+function fontCmap(src, cid, held) {
+  const R = src.resolve;
+  if (!(cid instanceof Map) || R(cid.get("Subtype")) !== "CIDFontType2") return null;
+  const fd2 = R(cid.get("FontDescriptor")), ff = fd2 instanceof Map ? R(fd2.get("FontFile2")) : null;
+  if (!(ff instanceof PdfStream)) return null;
+  const was = held.cmaps.get(ff);
+  if (was !== void 0) return was;
+  if (held.cmaps.size >= MAX_CMAP_FONTS) fail("E_BUDGET", `A page's text reads the character maps of more than ${MAX_CMAP_FONTS} embedded fonts (fonts with no ToUnicode)`);
+  held.cmaps.set(ff, null);
+  let f;
+  try {
+    f = parseFont(src.decodeStream(ff).data);
+  } catch (e) {
+    if (stops(e)) throw e;
+    return null;
+  }
+  src.budget.tick(65536);
+  const back = /* @__PURE__ */ new Map();
+  for (let cp = 32; cp <= 65535; cp++) {
+    if (cp === 55296) cp = 63744;
+    const g = f.glyphFor(cp);
+    if (g && !back.has(g)) back.set(g, String.fromCharCode(cp));
+  }
+  held.n += 9 * back.size;
+  if (held.n > MAX_HELD) fail("E_BUDGET", `A page's fonts hold more than ${MAX_HELD.toLocaleString("en")} units of code mappings (an entry 8, a character 1)`);
+  held.cmaps.set(ff, back);
+  return back;
+}
 function contentBytes(src, contents) {
   const c = src.resolve(contents);
   const list = Array.isArray(c) ? c.slice(0, 1e4).map(src.resolve) : [c];
@@ -14582,6 +14804,36 @@ function contentBytes(src, contents) {
   }
   return out;
 }
+function logical(parts) {
+  if (!parts.some((t) => RTL2.test(t))) return parts;
+  const kind = (t) => RTL2.test(t) ? "R" : NUM.test(t) ? "N" : STRONG_L.test(t) ? "L" : "N";
+  const k = parts.map(kind), res = parts.slice();
+  for (let i2 = 0; i2 < parts.length; ) {
+    if (k[i2] !== "R") {
+      i2++;
+      continue;
+    }
+    let j = i2, e = i2;
+    while (j < parts.length && k[j] !== "L") {
+      if (k[j] === "R") e = j;
+      j++;
+    }
+    const run2 = parts.slice(i2, e + 1).reverse().map(rev2);
+    for (let a = 0; a < run2.length; ) {
+      if (!NUM.test(run2[a])) {
+        a++;
+        continue;
+      }
+      let b = a;
+      while (b < run2.length && NUM.test(run2[b])) b++;
+      run2.splice(a, b - a, ...run2.slice(a, b).reverse().map(rev2));
+      a = b;
+    }
+    res.splice(i2, run2.length, ...run2);
+    i2 = e + 1;
+  }
+  return res;
+}
 function skipInline(lx) {
   const b = lx.buf;
   let i2 = lx.pos + 1;
@@ -14611,9 +14863,31 @@ function textPieces(src, pageDict, resources, o = {}) {
 }
 function pageText(src, pageDict, resources, raw = false) {
   const R = src.resolve, budget = src.budget;
-  const pieces = [];
+  const PX = [], PY = [], PE = [], PS = [], PG = [];
+  const PA = [], PAE = [], PC = [], PD = [];
+  const PT = [];
+  let chars2 = 0;
+  let moved = true;
   const fonts = /* @__PURE__ */ new Map();
-  const held = { n: 0 };
+  const add2 = (x2, y, end, size, text3, dir) => {
+    if (PT.length >= budget.maxTextItems) throw new PdfReadError("E_BUDGET", `A page has more than ${budget.maxTextItems.toLocaleString("en")} pieces of text (budget.maxTextItems)`);
+    if ((chars2 += text3.length) > MAX_TEXT_CHARS) throw new PdfReadError("E_BUDGET", `A page makes more than ${MAX_TEXT_CHARS.toLocaleString("en")} characters of text`);
+    PX.push(x2);
+    PY.push(y);
+    PE.push(end);
+    PS.push(size);
+    PT.push(text3);
+    PG.push(moved ? 2 : 0);
+    moved = false;
+    PA.push(dir[0]);
+    PAE.push(dir[1]);
+    PC.push(dir[2]);
+    PD.push(dir[3]);
+  };
+  const held = (
+    /** @type {Held} */
+    { n: 0, cmaps: /* @__PURE__ */ new Map() }
+  );
   const fontFor = (res, name) => {
     const fd2 = R(res instanceof Map ? R(res.get("Font")) instanceof Map ? R(res.get("Font")).get(name) : null : null);
     const key = fd2 instanceof Map ? fd2 : null;
@@ -14632,6 +14906,11 @@ function pageText(src, pageDict, resources, raw = false) {
   ), size: 0, Tc: 0, Tw: 0, Th: 1, TL: 0, Ts: 0 };
   const tsStack = [];
   let Tm = ID, Tlm = ID, dos = 0;
+  const marks = [];
+  let deep = 0, actual = (
+    /** @type {any} */
+    null
+  );
   let args = [];
   const show = (b, ctm) => {
     if (!ts.font) return;
@@ -14642,10 +14921,20 @@ function pageText(src, pageDict, resources, raw = false) {
       const x2 = M0[4], y = M0[5], size = Math.hypot(M0[2], M0[3]) || ts.size;
       const tx = (w / 1e3 * ts.size + ts.Tc + (space ? ts.Tw : 0)) * ts.Th;
       Tm = mul2([1, 0, 0, 1, tx, 0], Tm);
-      const end = mul2([ts.size * ts.Th, 0, 0, ts.size, 0, ts.Ts], mul2(Tm, ctm))[4];
-      if (text3) {
-        if (pieces.length >= budget.maxTextItems) throw new PdfReadError("E_BUDGET", `A page has more than ${budget.maxTextItems.toLocaleString("en")} pieces of text (budget.maxTextItems)`);
-        pieces.push({ x: x2, y, end, size, text: text3, gap: false });
+      const M1 = mul2([ts.size * ts.Th, 0, 0, ts.size, 0, ts.Ts], mul2(Tm, ctm)), end = M1[4];
+      const L = Math.hypot(M0[0], M0[1]), ux = L ? M0[0] / L : 1, uy = L ? M0[1] / L : 0;
+      const dir = [x2 * ux + y * uy, M1[4] * ux + M1[5] * uy, y * ux - x2 * uy, Math.round(Math.atan2(uy, ux) * 180 / Math.PI)];
+      if (actual) {
+        if (!actual.seen) Object.assign(actual, { x: x2, y, size, dir, seen: true });
+        actual.end = end;
+        actual.dir[1] = dir[1];
+        return;
+      }
+      const n = PT.length - 1;
+      if (text3) add2(x2, y, end, size, text3, dir);
+      else if (n >= 0 && PAE[n] < dir[1] && PD[n] === dir[3] && Math.abs(PC[n] - dir[2]) < 0.01) {
+        PE[n] = end;
+        PAE[n] = dir[1];
       }
     });
     budget.tick(1 + (k & 1023));
@@ -14686,6 +14975,7 @@ function pageText(src, pageDict, resources, raw = false) {
       }
       case "BT":
         Tm = Tlm = ID;
+        moved = true;
         break;
       case "Tf":
         ts = { ...ts, font: typeof a[0] === "string" ? fontFor(f.res, a[0]) : null, size: typeof a[1] === "number" ? a[1] : 0 };
@@ -14710,19 +15000,23 @@ function pageText(src, pageDict, resources, raw = false) {
       // falls through
       case "Td":
         if (typeof a[0] === "number" && typeof a[1] === "number") Tm = Tlm = mul2([1, 0, 0, 1, a[0], a[1]], Tlm);
+        moved = true;
         break;
       case "Tm": {
         const m = matrixOf(a);
         if (m) Tm = Tlm = m;
+        moved = true;
         break;
       }
       case "T*":
         Tm = Tlm = mul2([1, 0, 0, 1, 0, -ts.TL], Tlm);
+        moved = true;
         break;
       case "Tj":
         if (a[0] instanceof PdfString) show(a[0].bytes, f.ctm);
         break;
       case "'":
+        moved = true;
         Tm = Tlm = mul2([1, 0, 0, 1, 0, -ts.TL], Tlm);
         if (a[0] instanceof PdfString) show(a[0].bytes, f.ctm);
         break;
@@ -14730,6 +15024,7 @@ function pageText(src, pageDict, resources, raw = false) {
         if (typeof a[0] === "number") ts = { ...ts, Tw: a[0] };
         if (typeof a[1] === "number") ts = { ...ts, Tc: a[1] };
         Tm = Tlm = mul2([1, 0, 0, 1, 0, -ts.TL], Tlm);
+        moved = true;
         if (a[2] instanceof PdfString) show(a[2].bytes, f.ctm);
         break;
       case "TJ":
@@ -14737,10 +15032,38 @@ function pageText(src, pageDict, resources, raw = false) {
           if (x2 instanceof PdfString) show(x2.bytes, f.ctm);
           else if (typeof x2 === "number") {
             Tm = mul2([1, 0, 0, 1, -x2 / 1e3 * ts.size * ts.Th, 0], Tm);
-            if (x2 < -180 && pieces.length) pieces[pieces.length - 1].gap = true;
+            if (x2 < -180 && PT.length) PG[PT.length - 1] |= 1;
           }
         }
         break;
+      case "BMC":
+      case "BDC": {
+        if (marks.length >= MAX_MARKS) {
+          deep++;
+          break;
+        }
+        let at = null;
+        if (op === "BDC" && !actual) {
+          const pr = a[1] instanceof Map ? a[1] : typeof a[1] === "string" && f.res instanceof Map ? R(R(f.res.get("Properties")) instanceof Map ? R(f.res.get("Properties")).get(a[1]) : null) : null;
+          const t = pr instanceof Map ? R(pr.get("ActualText")) : null;
+          if (t instanceof PdfString && t.bytes.length > 2 * MAX_ACTUAL + 2) src.warn(`an /ActualText longer than ${MAX_ACTUAL} characters was ignored`);
+          else if (t instanceof PdfString) at = actual = { at: decodeText(t), x: 0, y: 0, end: 0, size: 0, dir: [0, 0, 0, 0], seen: false };
+        }
+        marks.push(at);
+        break;
+      }
+      case "EMC": {
+        if (deep) {
+          deep--;
+          break;
+        }
+        const m = marks.pop();
+        if (m && m === actual) {
+          actual = null;
+          if (m.seen && m.at) add2(m.x, m.y, m.end, m.size, m.at, m.dir);
+        }
+        break;
+      }
       case "BI": {
         for (let k = 0; k < 4096; k++) {
           const t = parseValue(f.lx);
@@ -14774,30 +15097,44 @@ function pageText(src, pageDict, resources, raw = false) {
       default:
     }
   }
-  if (raw) return pieces;
+  if (raw) return PT.map((text3, i2) => ({ x: PX[i2], y: PY[i2], end: PE[i2], size: PS[i2], text: text3, gap: !!(PG[i2] & 1) }));
   const out = [];
-  let ly = NaN, lend = 0, lsize = 0, lgap = false, last = "";
-  for (const p of pieces) {
-    const tol = Math.max(p.size, lsize, 1) * 0.5;
-    if (out.length && Math.abs(p.y - ly) > tol) out.push("\n");
-    else if (out.length && last !== " " && !p.text.startsWith(" ") && (lgap || p.x - lend > Math.max(p.size, 1) * 0.25 || p.x < lend - Math.max(p.size, 1) * 2)) out.push(" ");
-    out.push(p.text);
-    last = p.text[p.text.length - 1];
-    ly = p.y;
-    lend = p.end;
-    lsize = p.size;
-    lgap = p.gap;
+  for (let i2 = 0; i2 < PT.length; ) {
+    let j = i2 + 1;
+    while (j < PT.length && PD[j] === PD[j - 1] && Math.abs(PC[j] - PC[j - 1]) <= Math.max(PS[j], PS[j - 1], 1) * 0.5) j++;
+    const ix = [];
+    for (let k = i2; k < j; k++) ix.push(k);
+    let sorted = true;
+    for (let k = i2 + 1; k < j && sorted; k++) if (PA[k] < PA[k - 1] - 0.01) sorted = false;
+    if (!sorted) ix.sort((a, b) => PA[a] - PA[b]);
+    const line3 = [];
+    let lend = 0, lgap = false, last = "";
+    for (const k of ix) {
+      const text3 = PT[k], size = Math.max(PS[k], 1);
+      if (line3.length && last !== " " && !text3.startsWith(" ") && (lgap || PA[k] - lend > size * (PG[k] & 2 ? 0.07 : 0.25))) line3.push(" ");
+      line3.push(text3);
+      last = text3[text3.length - 1];
+      lend = Math.max(PAE[k], line3.length > 1 ? lend : -Infinity);
+      lgap = !!(PG[k] & 1);
+    }
+    if (out.length) out.push("\n");
+    out.push(logical(line3).join(""));
+    i2 = j;
   }
   return out.join("");
 }
-var MAX_FORMS, MAX_CMAP, MAX_DST, MAX_CMAP_CHARS, MAX_HELD, MAX_W, MAX_DO, TEXT_TIMEOUT_MS, DOC, WIN, MAC_HIGH, BASE, STD_HIGH, NAMES2, ACCENT, own, STD_ALIAS, stops, mul2, ID, matrixOf;
+var MAX_FORMS, MAX_CMAP, MAX_MARKS, MAX_ACTUAL, MAX_TEXT_CHARS, MAX_DST, MAX_CMAP_CHARS, MAX_HELD, MAX_W, MAX_DO, TEXT_TIMEOUT_MS, DOC, WIN, MAC_HIGH, BASE, STD_HIGH, NAMES2, ACCENT, own, STD_ALIAS, NONE, stops, MAX_CMAP_FONTS, mul2, ID, matrixOf, RTL2, STRONG_L, NUM, rev2;
 var init_text2 = __esm({
   "src/reader/text.js"() {
     init_lexer();
     init_standard_metrics();
     init_budget();
+    init_truetype();
     MAX_FORMS = 16;
     MAX_CMAP = 1e6;
+    MAX_MARKS = 1e3;
+    MAX_ACTUAL = 4096;
+    MAX_TEXT_CHARS = 16e6;
     MAX_DST = 128;
     MAX_CMAP_CHARS = 4e6;
     MAX_HELD = 16e6;
@@ -15002,7 +15339,9 @@ var init_text2 = __esm({
       t[k]
     ) : void 0;
     STD_ALIAS = { Arial: "Helvetica", "Arial-BoldMT": "Helvetica-Bold", ArialMT: "Helvetica", TimesNewRoman: "Times-Roman", TimesNewRomanPSMT: "Times-Roman", CourierNew: "Courier" };
+    NONE = "�";
     stops = (e) => e?.code === "E_TIMEOUT" || e?.code === "E_ABORTED" || e?.code === "E_BUDGET";
+    MAX_CMAP_FONTS = 100;
     mul2 = (A, B) => (
       /** @type {M} */
       [A[0] * B[0] + A[1] * B[2], A[0] * B[1] + A[1] * B[3], A[2] * B[0] + A[3] * B[2], A[2] * B[1] + A[3] * B[3], A[4] * B[0] + A[5] * B[2] + B[4], A[4] * B[1] + A[5] * B[3] + B[5]]
@@ -15013,6 +15352,10 @@ var init_text2 = __esm({
       /** @type {M} */
       a.slice(0, 6)
     ) : null;
+    RTL2 = /[\u0590-\u05ff\u0600-\u065f\u066a-\u06ef\u06fa-\u08ff\ufb1d-\ufdff\ufe70-\ufeff]/;
+    STRONG_L = /[\p{L}]/u;
+    NUM = /^[\p{Nd}.,٫٬]+$/u;
+    rev2 = (t) => [...t].reverse().join("");
   }
 });
 
@@ -15254,7 +15597,7 @@ function createCopier(t) {
     let t0 = trees.get(owner);
     if (t0) return t0;
     const src = owner.r.src, R = src.resolve;
-    t0 = { objs: /* @__PURE__ */ new WeakSet(), nums: /* @__PURE__ */ new Set(), types: new Set(STRUCT_TYPES) };
+    t0 = { objs: /* @__PURE__ */ new Set(), nums: /* @__PURE__ */ new Set(), types: new Set(STRUCT_TYPES) };
     trees.set(owner, t0);
     const root = R(owner.r.root.get("StructTreeRoot"));
     if (!(root instanceof Map)) return t0;
@@ -15476,6 +15819,7 @@ function createCopier(t) {
           continue;
         }
         const rep = t.replace?.(owner, f.n);
+        f.rep = rep;
         f.v = rep !== void 0 ? rep : owner.r.src.get(f.n);
         if (!t.keepActive && f.v instanceof Map) {
           const r = judge(f.v, owner.r.src.resolve);
@@ -15500,7 +15844,7 @@ function createCopier(t) {
         continue;
       }
       stack.pop();
-      const v = f.v;
+      const v = f.v, rep0 = f.rep;
       let open = false;
       const refText = (o, r) => {
         if (mapOf(o).get(r.n) === GRAY) open = true;
@@ -15515,7 +15859,7 @@ function createCopier(t) {
       let key = "";
       const raw = v instanceof PdfStream ? owner.r.src.rawStream(v) : null;
       if (t.dedupe && early === void 0 && !open && shareable(v)) {
-        key = hex(await (t.hash ?? sha256)(raw ? concat2(dict, raw) : new TextEncoder().encode(dict)));
+        key = `${dict.length} ${fnv(dict)} ${raw ? `${raw.length} ${fnv(raw)}` : ""}`;
         const hit = seen.get(key);
         if (hit && hit.dict === dict && same(hit.raw(), raw)) {
           m.set(f.n, hit.n);
@@ -15526,8 +15870,11 @@ function createCopier(t) {
       const n = early ?? t.reserve();
       await t.write(n, dict, data, v instanceof PdfStream ? v.dict : v instanceof Map ? v : null);
       if (key) {
-        const o2 = owner, v2 = v;
-        seen.set(key, { n, dict, raw: () => v2 instanceof PdfStream ? o2.r.src.rawStream(v2) : null });
+        const o2 = owner, n2 = f.n, mine = rep0 === void 0 ? null : v;
+        seen.set(key, { n, dict, raw: () => {
+          const v2 = mine ?? o2.r.src.get(n2);
+          return v2 instanceof PdfStream ? o2.r.src.rawStream(v2) : null;
+        } });
       }
       m.set(f.n, n);
     }
@@ -15547,13 +15894,11 @@ function createCopier(t) {
     }
     return typeof v !== "object" || v === null || v instanceof PdfString;
   };
-  const concat2 = (s, b) => {
-    const a = new TextEncoder().encode(s);
-    const o = new Uint8Array(a.length + b.length + 1);
-    o.set(a);
-    o[a.length] = 0;
-    o.set(b, a.length + 1);
-    return o;
+  const fnv = (x2) => {
+    let h = 2166136261;
+    if (typeof x2 === "string") for (let i2 = 0; i2 < x2.length; i2++) h = Math.imul(h ^ x2.charCodeAt(i2), 16777619);
+    else for (let i2 = 0; i2 < x2.length; i2++) h = Math.imul(h ^ x2[i2], 16777619);
+    return (h >>> 0).toString(36);
   };
   function streamDict(owner, s) {
     const R = owner.r.src.resolve;
@@ -15622,16 +15967,14 @@ function createCopier(t) {
     }
   };
 }
-var sha256, SHARE, GRAY, HEXD, Raw, Foreign, OPEN_A, CLOSE_A, OPEN_D, CLOSE_D;
+var SHARE, GRAY, HEXD, Raw, Foreign, OPEN_A, CLOSE_A, OPEN_D, CLOSE_D;
 var init_copy = __esm({
   "src/reader/copy.js"() {
     init_lexer();
     init_pdf();
-    init_crypto();
     init_policy();
     init_deflate();
     init_policy();
-    sha256 = (b) => sha("SHA-256", b);
     SHARE = /* @__PURE__ */ new Set(["Font", "FontDescriptor", "ExtGState", "Encoding", "Pattern", "Shading", "Halftone", "Function", "XObject", "Mask", "CMap", "OutputIntent", "Filespec", "EmbeddedFile"]);
     GRAY = -2;
     HEXD = "0123456789abcdef";
@@ -15926,7 +16269,8 @@ async function run(self, sink, o) {
   const keepStruct = single && o.keepStructure !== false && !self.edits.flatten;
   const info = self.info();
   const edited = !!self.edits.info || level?.conformance !== self.pdfa()?.conformance;
-  const srcXmp = self.edits.xmp ?? (edited || !level && self.pdfa() ? null : self.xmp());
+  const uaLost = !keepStruct && /pdfuaid:part/.test(self.xmp() ?? "");
+  const srcXmp = self.edits.xmp ?? (edited || uaLost || !level && self.pdfa() ? null : self.xmp());
   const created = o.creationDate ?? (info.creationDate instanceof Date ? info.creationDate : /* @__PURE__ */ new Date());
   let intentOpt = null;
   if (level && profile) {
@@ -16006,7 +16350,8 @@ async function run(self, sink, o) {
     }
   }
   for (let k = 0; k < plan.length; k++) {
-    const p = plan[k], owner = p.owner, src = owner.r.src, OR = src.resolve, pg = owner.r.pages[p.index], d = pg.dict;
+    const p = plan[k], owner = p.owner, src = owner.r.src, OR = src.resolve, pg = pageAt(owner.r, p.index), d = pg.dict;
+    if (pg.emptied) warn(pg.emptied);
     const b = owner.boxes(p.index);
     const W = b.mediaBox[2] - b.mediaBox[0], H = b.mediaBox[3] - b.mediaBox[1];
     const page = (
@@ -16130,7 +16475,7 @@ async function run(self, sink, o) {
     };
     for (const k of ["ViewerPreferences", "PageMode", "PageLayout", "OCProperties", "Collection", "URI"]) await cat(k);
     if (keepStruct) for (const k of ["StructTreeRoot", "MarkInfo", "Lang"]) await cat(k);
-    else if (root.has("StructTreeRoot")) warn("the structure tree (tags) was dropped: the pages come from several documents, or the form was flattened");
+    else if (root.has("StructTreeRoot")) warn("the structure tree (tags) was dropped: the pages come from several documents, the form was flattened, or keepStructure is false; the result is not tagged (no longer PDF/UA; PDF/A-a is claimed as PDF/A-b)");
     if (level && level.part === 3) await cat("AF");
     if (single && plan.length === self.r.pages.length && plan.every((p, i2) => p.index === i2)) await cat("PageLabels");
     const oa = R(root.get("OpenAction"));
@@ -16943,6 +17288,26 @@ function flatTree(rootNode, src, kind = "Names") {
   }
   return out;
 }
+function pageAt(r, i2) {
+  const pg = r.pages[i2];
+  if (pg.checked) return pg;
+  pg.checked = true;
+  const R = r.src.resolve;
+  try {
+    const c = R(pg.dict.get("Contents"));
+    if (Array.isArray(c)) {
+      r.src.budget.tick(Math.min(c.length, 1e4));
+      for (const x2 of c.slice(0, 1e4)) R(x2);
+    }
+  } catch (e) {
+    if (!(e instanceof PdfReadError) || e.code !== "E_CYCLE") throw e;
+    pg.emptied = `page ${i2 + 1}'s /Contents has a reference loop; the page is read as empty`;
+    r.src.warn(pg.emptied);
+    pg.dict = new Map(pg.dict);
+    pg.dict.delete("Contents");
+  }
+  return pg;
+}
 function makeDoc(r, plan, edits) {
   const { src, root, pages, pageIndex: pageIndex2 } = r;
   const R = src.resolve;
@@ -17029,7 +17394,7 @@ function makeDoc(r, plan, edits) {
        * document's timeoutMs when sooner) @param {{ timeoutMs?: number }} [o]
        */
       extractText: (o) => {
-        const pg = owner.r.pages[p.index];
+        const pg = pageAt(owner.r, p.index);
         return extractText(owner.r.src, pg.dict, owner.r.src.resolve(pg.inh.Resources), o);
       },
       /** the annotations: subtype, rect, contents, uri, the page index a link goes to */
@@ -17215,7 +17580,7 @@ function makeDoc(r, plan, edits) {
       const contents = R(v.get("Contents"));
       const end = range ? range[2] + range[3] : 0;
       const ok = !!range && range[0] === 0 && range[1] <= range[2] && end <= src.size;
-      const rev2 = ok ? src.revisions.findIndex((e) => e >= end) : -1;
+      const rev3 = ok ? src.revisions.findIndex((e) => e >= end) : -1;
       out.push({
         field: f.name,
         subFilter: typeof R(v.get("SubFilter")) === "string" ? R(v.get("SubFilter")) : null,
@@ -17229,7 +17594,7 @@ function makeDoc(r, plan, edits) {
         /** the signature covers every byte of the file but its /Contents (no change after it) */
         coversWholeFile: ok && end === src.size,
         /** the revision (0 first) whose end it covers, -1 when none */
-        revision: rev2,
+        revision: rev3,
         /** the bytes after it: an incremental update was appended after signing */
         bytesAfter: ok ? src.size - end : null,
         cms: contents instanceof PdfString ? contents.bytes : null,
@@ -17551,6 +17916,8 @@ function createPdf(sink, options2 = {}) {
     /** @type {const} */
     ["title", "author", "subject", "keywords"]
   ) if (o[k] != null && (typeof o[k] !== "string" || o[k].length > 65536)) throw new TypeError(`createPdf: ${k} is a string of at most 65,536 characters`);
+  if (o.metadata != null && typeof o.metadata !== "boolean") throw new TypeError("createPdf: metadata is true (the default: an XMP stream) or false (Info only)");
+  if (o.metadata === false && (o.pdfa || o.tagged)) throw new TypeError("createPdf: metadata: false cannot be combined with pdfa or tagged: PDF/A and PDF/UA require the XMP metadata stream");
   const level = pdfaLevel(o.pdfa, o.tagged);
   o.signal?.throwIfAborted();
   const ctl = new AbortController();
@@ -17575,6 +17942,7 @@ function createPdf(sink, options2 = {}) {
     pdfa: !!o.pdfa,
     pdfaLevel: level,
     tagged: !!o.tagged,
+    metadata: o.metadata !== false,
     signal: ctl.signal,
     encrypt: enc,
     hash: signHash ? hasher(

package/examples/factur-x.mjs

@@ -0,0 +1,37 @@
+// A Factur-X (ZUGFeRD 2.1+) invoice: a tagged PDF/A-3a page with the CII XML embedded as factur-x.xml.
+//
+//   node examples/factur-x.mjs invoice.xml font.ttf out.pdf [profile]
+//
+// The XML must be a UN/CEFACT CrossIndustryInvoice whose GuidelineSpecifiedDocumentContextParameter ID names the
+// profile (EN 16931 by default: urn:cen.eu:en16931:2017). pdf.facturX checks that, refuses any DOCTYPE or entity, and
+// writes the fx XMP schema; it does not validate the invoice itself (run the Factur-X XSD and Schematron for that).
+// The page here only shows a few fields; a real one draws the whole invoice from the same data as the XML.
+import fs from 'node:fs';
+import { createPdf } from '../dist/index.js';
+
+const [xmlFile, fontFile, outFile, profile = 'EN 16931'] = process.argv.slice(2);
+if (!outFile) { console.error('node examples/factur-x.mjs invoice.xml font.ttf out.pdf [profile]'); process.exit(2); }
+const xml = fs.readFileSync(xmlFile);
+const text = xml.toString('utf8');
+/** the first <ram:tag> after anchor (display only: the page shows a few fields of the XML) */
+const field = (tag, anchor = '') => new RegExp(`${anchor}[\\s\\S]*?<ram:${tag}[^>]*>([^<]*)<`).exec(text)?.[1] ?? '';
+const number = field('ID', '<rsm:ExchangedDocument>'), seller = field('Name', '<ram:SellerTradeParty>');
+
+const pdf = createPdf(fs.createWriteStream(outFile), {
+  title: `Invoice ${number}`, author: seller, subject: 'Factur-X invoice', keywords: `invoice, Factur-X, ${profile}`,
+  pdfa: { part: 3, conformance: 'a' }, tagged: { lang: 'en' },
+});
+const font = await pdf.embedFont(fs.readFileSync(fontFile));
+const page = pdf.addPage({ trimBox: [0, 0, 595.28, 841.89] });
+const line = (s, y, size = 11) => page.text(s, { x: 56, y, font, size });
+page.artifact({ type: 'Pagination', subtype: 'Header' }, () => line('Factur-X invoice: the XML inside this PDF is the legal invoice', 24, 8));
+page.tag('H1', () => line(`Invoice ${number}`, 60, 22));
+page.tag('P', () => line(`Seller: ${seller}`, 110));
+page.tag('Table', () => {
+  page.tag('TR', () => ['Net', 'VAT', 'Total'].forEach((h, i) => page.tag('TH', () => page.text(h, { x: 56 + i * 140, y: 150, font }), { scope: 'Column', id: `h${i}` })));
+  page.tag('TR', () => [field('TaxBasisTotalAmount'), field('TaxTotalAmount'), field('GrandTotalAmount')].forEach((v, i) =>
+    page.tag('TD', () => page.text(`${v} ${field('InvoiceCurrencyCode')}`, { x: 56 + i * 140, y: 170, font }), { headers: [`h${i}`] })));
+});
+await pdf.facturX(xml, { profile });
+await pdf.end();
+console.log(`wrote ${outFile}: PDF/A-3a with factur-x.xml (${profile})`);

package/examples/harfbuzz-shaper.mjs

@@ -0,0 +1,27 @@
+// A shaper for @reportwright/pdf's embedFont(bytes, { shaper }) built on harfbuzzjs (npm i harfbuzzjs). harfbuzzjs is
+// not a dependency of @reportwright/pdf: copy this file, or import it where harfbuzzjs is installed.
+//
+//   import { harfbuzzShaper } from './harfbuzz-shaper.mjs';
+//   const bytes = fs.readFileSync('NotoSansArabic-Regular.ttf');
+//   const arabic = await pdf.embedFont(bytes, { shaper: harfbuzzShaper(bytes) });
+//   page.text('مرحبا بالعالم', { x: 40, y: 40, font: arabic, size: 18, direction: 'rtl' });
+//
+// The shaper gets one run (one font, one direction) and returns its glyphs in visual order, in font units:
+// { g: glyph ID, cl: cluster (UTF-16 index of its first character), ax: x advance, dx, dy: offsets }.
+import * as hb from 'harfbuzzjs';
+
+const NO_KERN = [new hb.Feature('kern', 0)];
+
+/** @param {Uint8Array} fontBytes the same bytes passed to embedFont */
+export function harfbuzzShaper(fontBytes) {
+  const font = new hb.Font(new hb.Face(new hb.Blob(fontBytes))); // scale: units per em, so positions are font units
+  return (text, _font, { direction, kerning }) => {
+    const b = new hb.Buffer();
+    b.addText(text); // clusters are UTF-16 indices, as JavaScript strings count
+    b.guessSegmentProperties(); // script and language from the text
+    b.setDirection(direction === 'rtl' ? hb.Direction.RTL : hb.Direction.LTR);
+    hb.shape(font, b, kerning ? [] : NO_KERN);
+    const pos = b.getGlyphPositions();
+    return b.getGlyphInfos().map((g, i) => ({ g: g.codepoint, cl: g.cluster, ax: pos[i].xAdvance, dx: pos[i].xOffset, dy: pos[i].yOffset }));
+  };
+}

package/index.d.ts

@@ -74,6 +74,11 @@ export interface PdfOptions {
   pdfa?: boolean | { part?: 1 | 2 | 3; conformance?: 'a' | 'b' | 'u'; outputIntent?: Uint8Array | ArrayBuffer };
   /** PDF/UA-1: a structure tree from page.tag(). Needs title; fonts must be embedded. */
   tagged?: boolean | { lang?: string };
+  /**
+   * false: no XMP metadata stream, the Info dictionary only (about 1 KB smaller: a one-page invoice). Default true.
+   * Refused with pdfa or tagged: PDF/A and PDF/UA require the XMP.
+   */
+  metadata?: boolean;
   /** stop the export: pending and later calls reject with signal.reason */
   signal?: AbortSignal;
   /** stop the export after this many milliseconds (a TimeoutError, code 'ETIMEDOUT') */
@@ -703,7 +708,7 @@ export interface ReadBudget {
   maxSteps?: number;
   /** cross-reference entries and scanned objects held (default 5,000,000) */
   maxObjects?: number;
-  /** pieces of text one page's extraction holds (default 2,000,000) */
+  /** pieces of text one page's extraction holds (default 1,000,000; a page also stops at 16,000,000 characters) */
   maxTextItems?: number;
 }
 export interface LoadOptions {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@reportwright/pdf",
-  "version": "0.1.0-beta.0",
+  "version": "0.1.0-beta.1",
   "description": "A streaming PDF writer for Node and browsers with no runtime dependencies: flat memory, PDF/A-1b to 3a, PDF/UA-1, Factur-X, embedded TrueType with Unicode text.",
   "type": "module",
   "license": "MIT",
@@ -15,7 +15,7 @@
     "./package.json": "./package.json"
   },
   "types": "./index.d.ts",
-  "files": ["dist", "index.d.ts", "README.md", "LICENSE", "THIRD-PARTY-NOTICES.md"],
+  "files": ["dist", "index.d.ts", "examples", "README.md", "LICENSE", "THIRD-PARTY-NOTICES.md"],
   "engines": { "node": ">=20.16" },
   "sideEffects": false
 }