rapier-markdown-kit 1.2.7 → 1.2.9

package/dist/kit/pdf-glyphs.mjs

@@ -1,6 +1,5 @@
 // SPDX-License-Identifier: MIT
 // The licensed static faces retain the shipped variable font's exact glyphs and metrics.
-// Browser ranges own placement; unsupported shaping stays in the original page paint.
 
 const familyName = value => value.trim().replace(/^['"]|['"]$/g, '');
 
@@ -80,7 +79,6 @@ function shape(value, font, preserve, ligatures) {
     if (code === 0x00ad) {
       at += character.length;
       const own = font.glyph(code), hyphen = font.glyph(0x2d), glyph = own || hyphen;
-      // A discretionary hyphen with no glyph stays in the page paint. Do not skip the lookup.
       if (!glyph || glyph >= font.glyphCount) return null;
       result.push({start, end: at, text: own ? character : '-', glyph, soft: true});
       continue;
@@ -101,7 +99,7 @@ function shape(value, font, preserve, ligatures) {
 }
 
 export async function collectPdfGlyphs(main, {width, check, faces = []}) {
-  const fontCache = new Map(); // Parsed byte copies belong to this PDF operation.
+  const fontCache = new Map();
   const document = main.ownerDocument, window = document.defaultView, origin = main.getBoundingClientRect();
   const nodes = new Map(), hidden = [], canvas = document.createElement('canvas'); canvas.width = canvas.height = 1;
   const context = canvas.getContext('2d', {willReadFrequently: true});
@@ -130,11 +128,9 @@ export async function collectPdfGlyphs(main, {width, check, faces = []}) {
     let visited = 0;
     for (const element of main.querySelectorAll('*')) {
       if (++visited % 128 === 0) {await new Promise(resolve => setTimeout(resolve, 0)); check();}
-      // Leaving the existing element in place preserves selectors, counters, line breaking and inline marks.
-      // Keep preformatted blocks, their inline struts and syntax paint together.
-      if (element.namespaceURI !== 'http://www.w3.org/1999/xhtml' || element.children.length || element.closest('svg,pre,.rapier-will-carrier')) continue;
-      const textNodes = Array.from(element.childNodes).filter(node => node.nodeType === 3);
-      if (textNodes.length !== 1 || !textNodes[0].data.trim() || !safe(element)) continue;
+      if (element.namespaceURI !== 'http://www.w3.org/1999/xhtml' || element.closest('svg,pre,.rapier-will-carrier')) continue;
+      const textNodes = Array.from(element.childNodes).filter(node => node.nodeType === 3 && node.data.trim());
+      if (!textNodes.length || !safe(element)) continue;
       const language = (element.closest('[lang]')?.getAttribute('lang') || document.documentElement.lang || '').toLowerCase();
       if (language && !/^en(?:-|$)/.test(language)) continue;
       const style = styleOf(element), family = familyName(style.fontFamily.split(',')[0]), weight = Number(style.fontWeight), size = Number.parseFloat(style.fontSize);
@@ -148,49 +144,47 @@ export async function collectPdfGlyphs(main, {width, check, faces = []}) {
       if (generated || altered || !Array.from(document.fonts).some(face => familyName(face.family) === family && face.status === 'loaded')) continue;
       const asset = faces.find(face => face.family === family && face.weight === weight);
       if (!asset) continue;
-      const font = readFont(asset, fontCache), preserve = /pre|break-spaces/.test(style.whiteSpace), node = textNodes[0];
-      const shaped = shape(node.data, font, preserve, style.fontVariantLigatures === 'normal' && !(Number.parseFloat(style.letterSpacing) || 0));
-      if (!shaped) continue;
+      const font = readFont(asset, fontCache), preserve = /pre|break-spaces/.test(style.whiteSpace);
       context.font = [style.fontStyle, style.fontWeight, style.fontSize, style.fontFamily].join(' ');
       const metrics = context.measureText('Hg'), ascent = metrics.fontBoundingBoxAscent;
       if (!Number.isFinite(ascent)) continue;
       const rgba = color(style.webkitTextFillColor || style.color), range = document.createRange();
-      let valid = true;
-      const placed = [];
-      for (const item of shaped) {
-        range.setStart(node, item.start); range.setEnd(node, item.end);
-        const ink = Array.from(range.getClientRects()).filter(rect => rect.height > 0);
-        // A break after a soft hyphen can leave the next letter with a phantom box on the
-        // hyphen's line. Keep the line that actually holds the ink.
-        const lines = [];
-        for (const rect of ink) {
-          const line = lines.find(row => Math.abs(row[0].top - rect.top) <= .25);
-          if (line) line.push(rect); else lines.push([rect]);
+      for (const node of textNodes) {
+        const shaped = shape(node.data, font, preserve, style.fontVariantLigatures === 'normal' && !(Number.parseFloat(style.letterSpacing) || 0));
+        if (!shaped) continue;
+        let valid = true;
+        const placed = [];
+        for (const item of shaped) {
+          range.setStart(node, item.start); range.setEnd(node, item.end);
+          const ink = Array.from(range.getClientRects()).filter(rect => rect.height > 0);
+          const lines = [];
+          for (const rect of ink) {
+            const line = lines.find(row => Math.abs(row[0].top - rect.top) <= .25);
+            if (line) line.push(rect); else lines.push([rect]);
+          }
+          lines.sort((a, b) => b.reduce((sum, rect) => sum + rect.width, 0) - a.reduce((sum, rect) => sum + rect.width, 0));
+          const previous = shaped[shaped.indexOf(item) - 1];
+          if (lines.length > 1 && !item.soft && !previous?.soft) {valid = false; break;}
+          const rects = lines[0] || [];
+          const first = (item.soft ? rects.find(rect => rect.width > 0) : null) || rects[0];
+          if (item.soft && !(first?.width > 0)) continue;
+          if (!first) {valid = false; break;}
+          const page = Math.max(0, Math.floor((first.left - origin.left + .01) / width));
+          Object.assign(item, {font, size, rgba, x: first.left - origin.left - page * width, baseline: first.top - origin.top + ascent});
+          placed.push(item);
         }
-        lines.sort((a, b) => b.reduce((sum, rect) => sum + rect.width, 0) - a.reduce((sum, rect) => sum + rect.width, 0));
-        const previous = shaped[shaped.indexOf(item) - 1];
-        if (lines.length > 1 && !item.soft && !previous?.soft) {valid = false; break;}
-        const rects = lines[0] || [];
-        const first = (item.soft ? rects.find(rect => rect.width > 0) : null) || rects[0];
-        // An unbroken soft hyphen has no ink. Leave it out of the vector run.
-        if (item.soft && !(first?.width > 0)) continue;
-        if (!first) {valid = false; break;}
-        const page = Math.max(0, Math.floor((first.left - origin.left + .01) / width));
-        Object.assign(item, {font, size, rgba, x: first.left - origin.left - page * width, baseline: first.top - origin.top + ascent});
-        placed.push(item);
-      }
-      if (!valid || !placed.length) continue;
-      nodes.set(node, placed);
-      if (element.localName === 'li') {
-        // A list item's marker inherits its fill and is not vectorized: hide only the words.
-        const wrap = document.createElement('span');
-        node.replaceWith(wrap); wrap.append(node);
-        wrap.style.setProperty('-webkit-text-fill-color', 'transparent', 'important');
-        hidden.push({wrap});
-        continue;
+        if (!valid || !placed.length) continue;
+        nodes.set(node, placed);
+        if (element.localName === 'li' || element.children.length || textNodes.length > 1) {
+          const wrap = document.createElement('span');
+          node.replaceWith(wrap); wrap.append(node);
+          wrap.style.setProperty('-webkit-text-fill-color', 'transparent', 'important');
+          hidden.push({wrap});
+          continue;
+        }
+        hidden.push({node: element, value: element.style.getPropertyValue('-webkit-text-fill-color'), priority: element.style.getPropertyPriority('-webkit-text-fill-color')});
+        element.style.setProperty('-webkit-text-fill-color', 'transparent', 'important');
       }
-      hidden.push({node: element, value: element.style.getPropertyValue('-webkit-text-fill-color'), priority: element.style.getPropertyPriority('-webkit-text-fill-color')});
-      element.style.setProperty('-webkit-text-fill-color', 'transparent', 'important');
     }
     check(); return {nodes, restore};
   } catch (error) {restore(); throw error;}

package/dist/kit/pdf-source.d.mts

@@ -0,0 +1,5 @@
+// SPDX-License-Identifier: MIT
+// Attach the exact source to a PDF Rapier wrote, without rewriting its pages (rapier-markdown-kit/pdf).
+export function readPdfPageCount(input: Uint8Array | ArrayBuffer): number;
+/** The PDF with `source` attached as an associated file; throws `pdf_source_invalid` for a PDF it cannot extend. */
+export function attachPdfSource(input: Uint8Array | ArrayBuffer, options: {source: string; filename: string; maxBytes?: number; check?: () => void}): Promise<Uint8Array>;

package/dist/kit/pdf-source.mjs

@@ -0,0 +1,125 @@
+// SPDX-License-Identifier: MIT
+function pdfStructure(input) {
+  const bytes = input instanceof Uint8Array ? input : new Uint8Array(input);
+  const fail = () => { throw Object.assign(new Error('The PDF source could not be attached.'), {code: 'pdf_source_invalid'}); };
+  let text = '';
+  for (let at = 0; at < bytes.length; at += 32768) text += String.fromCharCode(...bytes.subarray(at, at + 32768));
+  if (!text.startsWith('%PDF-')) fail();
+  const end = /startxref\s+(\d+)\s+%%EOF\s*$/.exec(text), xref = Number(end?.[1]);
+  if (!end || !Number.isSafeInteger(xref) || text.slice(xref, xref + 4) !== 'xref') fail();
+  const offsets = new Map();
+  let at = xref + 4;
+  const space = () => { while (at < text.length && /[\s\0]/.test(text[at])) at++; };
+  for (;;) {
+    space(); if (text.startsWith('trailer', at)) { at += 7; break; }
+    const section = /^(\d+)\s+(\d+)\s+/.exec(text.slice(at)); if (!section) fail();
+    at += section[0].length;
+    const from = Number(section[1]), count = Number(section[2]);
+    if (!Number.isSafeInteger(from + count) || count > bytes.length / 18) fail();
+    for (let i = 0; i < count; i++) {
+      const entry = /^(\d{10})\s+(\d{5})\s+([nf])\s+/.exec(text.slice(at)); if (!entry) fail();
+      at += entry[0].length;
+      if (entry[3] === 'n') offsets.set(from + i, {at: Number(entry[1]), generation: Number(entry[2])});
+    }
+  }
+  const skip = pos => { for (;;) { while (pos < text.length && /[\s\0]/.test(text[pos])) pos++; if (text[pos] !== '%') return pos; while (pos < text.length && !/[\r\n]/.test(text[pos])) pos++; } };
+  const tokenEnd = pos => { while (pos < text.length && !/[\s\0()[\]<>/%]/.test(text[pos])) pos++; return pos; };
+  const valueEnd = (pos, depth = 0) => {
+    if (depth > 32) fail(); pos = skip(pos);
+    if (text.startsWith('<<', pos)) {
+      pos += 2;
+      for (;;) { pos = skip(pos); if (text.startsWith('>>', pos)) return pos + 2; if (text[pos] !== '/') fail(); pos = valueEnd(tokenEnd(pos + 1), depth + 1); }
+    }
+    if (text[pos] === '[') { pos++; for (;;) { pos = skip(pos); if (text[pos] === ']') return pos + 1; pos = valueEnd(pos, depth + 1); } }
+    if (text[pos] === '(') {
+      let nesting = 1; pos++;
+      while (pos < text.length && nesting) { const c = text[pos++]; if (c === '\\') pos++; else if (c === '(') nesting++; else if (c === ')') nesting--; }
+      if (nesting) fail(); return pos;
+    }
+    if (text[pos] === '<') { const close = text.indexOf('>', pos + 1); if (close < 0) fail(); return close + 1; }
+    if (text[pos] === '/') return tokenEnd(pos + 1);
+    const first = tokenEnd(pos); if (first === pos) fail();
+    const reference = /^\d+\s+\d+\s+R\b/.exec(text.slice(pos)); return reference ? pos + reference[0].length : first;
+  };
+  const dictionary = pos => {
+    pos = skip(pos); if (!text.startsWith('<<', pos)) fail();
+    const start = pos, fields = new Map(); pos += 2;
+    for (;;) {
+      pos = skip(pos); if (text.startsWith('>>', pos)) return {start, end: pos + 2, fields};
+      if (text[pos] !== '/') fail();
+      const keyStart = pos, keyEnd = tokenEnd(pos + 1), name = text.slice(pos + 1, keyEnd);
+      if (fields.has(name)) fail();
+      const from = skip(keyEnd), end = valueEnd(from);
+      fields.set(name, {start: keyStart, from, end, value: text.slice(from, end)}); pos = end;
+    }
+  };
+  const trailer = dictionary(at), rootRef = /^(\d+)\s+(\d+)\s+R$/.exec(trailer.fields.get('Root')?.value || '');
+  const size = Number(trailer.fields.get('Size')?.value), rootId = Number(rootRef?.[1]), generation = Number(rootRef?.[2]);
+  let highest = 0; for (const id of offsets.keys()) highest = Math.max(highest, id);
+  if (!rootRef || !Number.isSafeInteger(size) || size <= highest || trailer.fields.has('Encrypt') || trailer.fields.has('Prev')) fail();
+  const readDictionary = (id, gen) => {
+    const place = offsets.get(id); if (!place || place.generation !== gen) fail();
+    const header = new RegExp('^' + id + '\\s+' + gen + '\\s+obj\\b').exec(text.slice(place.at)); if (!header) fail();
+    return dictionary(place.at + header[0].length);
+  };
+  const root = readDictionary(rootId, generation);
+  return {bytes, text, offsets, root, rootId, generation, trailer, size, xref, readDictionary, dictionary, fail};
+}
+
+export function readPdfPageCount(input) {
+  const {root, readDictionary, fail} = pdfStructure(input);
+  const reference = /^(\d+)\s+(\d+)\s+R$/.exec(root.fields.get('Pages')?.value || '');
+  if (!reference) fail();
+  const pages = readDictionary(Number(reference[1]), Number(reference[2])), count = Number(pages.fields.get('Count')?.value);
+  if (pages.fields.get('Type')?.value !== '/Pages' || !Number.isSafeInteger(count) || count < 1) fail();
+  return count;
+}
+
+export async function attachPdfSource(input, {source, filename, maxBytes = Infinity, check = () => {}}) {
+  const {bytes, text, offsets, root, rootId, generation, trailer, size, readDictionary, dictionary, fail} = pdfStructure(input);
+  if (typeof source !== 'string' || typeof filename !== 'string' || !filename) fail();
+  const ascii = value => Uint8Array.from(value, character => character.charCodeAt(0));
+  const encode = new TextEncoder(), hex = value => Array.from({length: value.length}, (_, i) => value.charCodeAt(i).toString(16).padStart(4, '0')).join('');
+  if (root.fields.get('Type')?.value !== '/Catalog' || root.fields.has('AF')) fail();
+  let names = '<<', remove = root.fields.get('Names');
+  if (remove) {
+    const ref = /^(\d+)\s+(\d+)\s+R$/.exec(remove.value);
+    const held = ref ? readDictionary(Number(ref[1]), Number(ref[2])) : dictionary(remove.from);
+    if (held.fields.has('EmbeddedFiles')) fail();
+    names = text.slice(held.start, held.end - 2);
+  }
+  const original = remove ? text.slice(root.start, remove.start) + text.slice(remove.end, root.end - 2) : text.slice(root.start, root.end - 2);
+  const payload = encode.encode(source); check();
+  const compressed = new Uint8Array(await new Response(new Blob([payload]).stream().pipeThrough(new CompressionStream('deflate'))).arrayBuffer()); check();
+  const fileId = size, streamId = size + 1, infoId = size + 2, totalSize = size + 3;
+  let infoBody = '<< /Producer (Rapier) /Creator (Rapier) >>';
+  const existingInfo = trailer.fields.get('Info');
+  if (existingInfo) {
+    const ref = /^(\d+)\s+(\d+)\s+R$/.exec(existingInfo.value);
+    const held = ref ? readDictionary(Number(ref[1]), Number(ref[2])) : dictionary(existingInfo.from);
+    const named = ['Producer', 'Creator'].map(key => held.fields.get(key)).filter(Boolean).sort((a, b) => a.start - b.start);
+    let kept = '', at = held.start;
+    for (const field of named) { kept += text.slice(at, field.start); at = field.end; }
+    infoBody = kept + text.slice(at, held.end - 2) + ' /Producer (Rapier) /Creator (Rapier) >>';
+  }
+  const escaped = '(' + filename.replace(/[^\x20-\x7e]/g, '_').replace(/[\\()]/g, '\\$&') + ')';
+  const parts = [bytes, ascii('\n')], merged = new Map(offsets);
+  let length = bytes.length + 1;
+  const append = (id, gen, chunks) => { merged.set(id, {at: length, generation: gen}); const body = [ascii(id + ' ' + gen + ' obj\n'), ...chunks, ascii('\nendobj\n')]; for (const chunk of body) {parts.push(chunk); length += chunk.length;} };
+  append(rootId, generation, [ascii(original + ' /Names ' + names + ' /EmbeddedFiles << /Names [<feff' + hex(filename) + '> ' + fileId + ' 0 R] >> >> /AF [' + fileId + ' 0 R] >>')]);
+  append(fileId, 0, [ascii('<< /Type /Filespec /F ' + escaped + ' /UF <feff' + hex(filename) + '> /AFRelationship /Source /EF << /F ' + streamId + ' 0 R /UF ' + streamId + ' 0 R >> >>')]);
+  append(streamId, 0, [ascii('<< /Type /EmbeddedFile /Subtype /text#2Fmarkdown /Params << /Size ' + payload.length + ' >> /Filter /FlateDecode /Length ' + compressed.length + ' >>\nstream\n'), compressed, ascii('\nendstream')]);
+  append(infoId, 0, [ascii(infoBody)]);
+  let table = 'xref\n0 ' + totalSize + '\n0000000000 65535 f \n';
+  for (let id = 1; id < totalSize; id++) {
+    const place = merged.get(id);
+    table += place ? String(place.at).padStart(10, '0') + ' ' + String(place.generation).padStart(5, '0') + ' n \n' : '0000000000 00000 f \n';
+  }
+  const identity = trailer.fields.get('ID');
+  table += 'trailer\n<< /Size ' + totalSize + ' /Root ' + rootId + ' ' + generation + ' R /Info ' + infoId + ' 0 R' + (identity ? ' /ID ' + identity.value : '') + ' >>\nstartxref\n' + length + '\n%%EOF\n';
+  parts.push(ascii(table)); length += table.length;
+  if (length > maxBytes) throw Object.assign(new Error('The exported file exceeds the file limit.'), {code: 'export_too_large'});
+  const result = new Uint8Array(length); let at = 0;
+  for (const chunk of parts) { result.set(chunk, at); at += chunk.length; }
+  check(); return result;
+}

package/dist/kit/pdf-vectors.mjs

@@ -1,5 +1,4 @@
 // SPDX-License-Identifier: MIT
-// SVG keeps its geometry in PDF; only paint operations that need browser compositing are rasterized.
 const SVG = 'http://www.w3.org/2000/svg';
 const number = value => Number(Number(value).toFixed(6));
 const nums = values => values.map(number).join(' ');
@@ -132,7 +131,6 @@ export async function collectPdfVectors(main, {width, height, count, capture, si
         root = document.importNode(parsed.documentElement, true);
         root.style.setProperty('width', originalBox.width + 'px', 'important'); root.style.setProperty('height', originalBox.height + 'px', 'important');
         mount = document.createElement('div'); mount.style.cssText = `position:fixed;left:${originalBox.left}px;top:${originalBox.top}px;width:${originalBox.width}px;height:${originalBox.height}px;pointer-events:none`;
-        // A shadow root keeps the picture's own <style> on the picture, as an <img> does.
         mount.attachShadow({mode: 'open'}).append(root); document.body.append(mount); temporary.push(mount); await document.fonts.ready;
         textRoots.push(root);
       }
@@ -177,7 +175,6 @@ export async function collectPdfVectors(main, {width, height, count, capture, si
         let path, fill, stroke;
         try {
           path = geometry(node); fill = color(style.fill);
-          // SVG zero-width strokes are absent; PDF's zero width instead paints a hairline.
           stroke = Number.parseFloat(style.strokeWidth) === 0 ? null : color(style.stroke);
           if (style.vectorEffect !== 'none' || style.markerStart !== 'none' || style.markerMid !== 'none' || style.markerEnd !== 'none' || style.paintOrder !== 'normal') unsupported();
         } catch (_) { sequence.push({raster: node}); return; }

package/dist/kit/print-capitals.mjs

@@ -9,8 +9,6 @@ export function followingPrintProse(paragraph) {
   return next;
 }
 
-// The print compositor owns pagination. Only a normal-flow picture initial moves into
-// its adjacent prose, so a native float can remain with the lines it introduces.
 export function markPrintCapitals(root) {
   for (const image of root.querySelectorAll('img[data-rapier-image-layout]')) {
     const layout = parseLayoutAttribute(image.getAttribute('data-rapier-image-layout'));
@@ -23,7 +21,6 @@ export function markPrintCapitals(root) {
   }
 }
 
-// This factory is also embedded in the offline page, with the same line metrics owner.
 export function createPrintCapitals(lineMetrics, linesHeight, followingProse) {
   const selector = 'img[data-rapier-print-lines]', states = new WeakMap();
   const restoreStyle = (node, value) => value == null ? node.removeAttribute('style') : node.setAttribute('style', value);
@@ -55,7 +52,6 @@ export function createPrintCapitals(lineMetrics, linesHeight, followingProse) {
       const availableWidth = owner.getBoundingClientRect().width -
         ['paddingLeft', 'paddingRight', 'borderLeftWidth', 'borderRightWidth'].reduce((sum, name) => sum + (parseFloat(style[name]) || 0), 0);
       if (!(availableWidth > 0)) continue;
-      // As imageBox does, cap the width first and retain the picture's natural ratio.
       const width = Math.min(availableWidth, height * naturalWidth / naturalHeight);
       const occurrence = state?.occurrence || (image.parentElement.tagName === 'A' ? image.parentElement : image);
       if (!state) {

package/dist/kit/render-entry.mjs

@@ -1,5 +1,4 @@
 // SPDX-License-Identifier: MIT
-// The public renderer door; each factory binds explicit host ports once per document realm.
 export {createRenderer} from './render.mjs';
 export {createMarkdownRenderer} from './render-markdown.mjs';
 export {createRenderStyles} from './render-styles.mjs';

package/dist/kit/render-ink.mjs

@@ -1,5 +1,4 @@
 // SPDX-License-Identifier: MIT
-// Published typography is measured once; an annotation keeps its words and paths in one inert, scalable group.
 import {check, pause} from './render-work.mjs';
 import {apply as applyTypography} from '../layout/typography.mjs';
 import {documentLanguage} from '../spec/document-settings.mjs';
@@ -75,7 +74,6 @@ export async function freezePublishedInk(root, {document: doc, css, fontCss = ''
       group.style.cssText = 'display:flow-root;position:relative;width:100%;margin:0;padding:0;border:0;overflow:visible';
       main.insertBefore(group, captured.nodes[first]);
       for (let index = first; index <= last; index++) group.appendChild(captured.nodes[index]);
-      // A fixed group cannot grow after disclosure. Show each marked passage in full.
       for (const span of group.querySelectorAll(MARK)) {
         for (let parent = span.parentElement; parent && parent !== group; parent = parent.parentElement)
           if (parent.localName === 'details') parent.open = true;
@@ -88,7 +86,6 @@ export async function freezePublishedInk(root, {document: doc, css, fontCss = ''
       try { await image.decode(); }
       catch (_) { throw failure('an embedded picture could not be decoded.'); }
     })));
-    // The captured CSS supplies the faces; validate them before measuring typography or ink.
     main.getBoundingClientRect();
     await wait(waitForFonts(main, {settings, css, fontCss, install: false}));
     if (paper.fonts?.ready) await wait(paper.fonts.ready);
@@ -114,8 +111,6 @@ export async function freezePublishedInk(root, {document: doc, css, fontCss = ''
       results.push(svg);
       await pause(work);
     }
-    // Keep the measured typography outside annotated groups too. The caller's detached
-    // projection changes only after every resource and group has succeeded.
     for (const [index, group] of prepared.entries()) group.replaceWith(results[index]);
     const projected = doc.createDocumentFragment();
     for (const node of main.childNodes) {