md2docx

npm install docx   # in the working directory
node md2docx.js    # run the converter

MD Source → Parse (line-by-line) → Element Array → docx-js Document → Packer.toBuffer → .docx

// ✅ CORRECT — detect table lines and build proper Table objects
if (line.trim().startsWith('|') && line.trim().endsWith('|') && line.indexOf('|', 1) > 0) {
  // Collect all consecutive | lines, parse header + body, build Table
}

// ❌ WRONG — treating table lines as regular paragraphs
elements.push(new Paragraph({ children: [new TextRun(line)] }));

import zipfile, re
z = zipfile.ZipFile('output.docx')
xml = z.read('word/document.xml').decode('utf-8')
pipe_count = len(re.findall(r'<w:t[^>]*>[^<]*\|[^<]*</w:t>', xml))
# pipe_count should be < 10 (only legitimate in-text pipe usage)

// ✅ CORRECT — strip outer pipes first, then split
function parseTableRow(line) {
  return line.replace(/^\||\|$/g, '').split('|').map(c => c.trim());
}

// ❌ WRONG — creates ghost empty column at end
function parseTableRow(line) {
  return line.split('|').map(c => c.trim()).filter((c, i, arr) => i > 0 && i < arr.length);
}
// Bug: `i < arr.length` is ALWAYS true, so trailing empty '' from `|A|B|` split is kept

const tableWidth = 9026; // A4 content width (11906 - 2*1440 margins)
const colWidths = computeProportional(headerCells, tableWidth);
// colWidths MUST sum exactly to tableWidth — fix rounding on last column
colWidths[colWidths.length - 1] += tableWidth - colWidths.reduce((a, b) => a + b, 0);

new Table({
  width: { size: tableWidth, type: WidthType.DXA },  // NEVER use PERCENTAGE
  columnWidths: colWidths,
  rows: [...]
});

// ❌ FATAL — destroys the Paragraph object, produces empty/broken output
paragraphs[0] = new Paragraph({ ...paragraphs[0], children: undefined });

// ✅ CORRECT — just use the paragraph as-is, or create a new one from scratch

function getPngDimensions(buffer) {
  if (buffer.length < 24) return null;
  if (buffer[0] !== 0x89 || buffer[1] !== 0x50) return null; // Not PNG
  return { width: buffer.readUInt32BE(16), height: buffer.readUInt32BE(20) };
}

function scaleImage(w, h, maxW = 560, maxH = 700) {
  if (w > maxW) { h = Math.round(h * maxW / w); w = maxW; }
  if (h > maxH) { w = Math.round(w * maxH / h); h = maxH; }
  return { width: w, height: h };
}

// Use a manifest.json that maps diagram_NNN.png → [part01.png, part02.png, ...]
for (const partFile of entry.parts) {
  const imgData = fs.readFileSync(path.join(SPLIT_DIR, partFile));
  // ... scale and embed each part as a separate centered Paragraph
}

# Extract ALL mermaid blocks from current MD and re-render them ALL
mmdc -i diagram_NNN.mmd -o diagram_NNN.png -w 2400 -b transparent --scale 2
# Then rebuild manifest.json and split tall images

const S = {
  font: 'Microsoft YaHei',  // or 'SimSun', 'PingFang SC'
  monoFont: 'Consolas',
  bodySize: 24,   // 12pt (half-points)
};

paragraphStyles: [
  { id: 'Heading1', ..., paragraph: { outlineLevel: 0 } },
  { id: 'Heading2', ..., paragraph: { outlineLevel: 1 } },
  // ...
]

// ❌ WRONG — raw markdown syntax leaks into output as plain text
// e.g., "### 8.6.1 表 `cluster`" outputs literal backticks in Word
new Paragraph({
  heading: HeadingLevel.HEADING_3,
  children: [new TextRun({ text: h3Match[1], bold: true, font: S.font, size: S.h3Size })]
});

// ✅ CORRECT — inline formatting parsed, `cluster` renders as mono-font code
new Paragraph({
  heading: HeadingLevel.HEADING_3,
  children: parseInlineFormatting(h3Match[1], { size: S.h3Size, color: S.secondary, bold: true })
});

// ✅ CORRECT — clean page number footer
footers: {
  default: new Footer({
    children: [new Paragraph({
      alignment: AlignmentType.CENTER,
      children: [
        new TextRun({ text: '第 ', font: S.font, size: 18, color: S.footerColor }),
        new TextRun({ children: [PageNumber.CURRENT], font: S.font, size: 18, color: S.footerColor }),
        new TextRun({ text: ' 页 / 共 ', font: S.font, size: 18, color: S.footerColor }),
        new TextRun({ children: [PageNumber.TOTAL_PAGES], font: S.font, size: 18, color: S.footerColor }),
        new TextRun({ text: ' 页', font: S.font, size: 18, color: S.footerColor }),
      ]
    })]
  })
}

// ❌ WRONG — adding "机密 · 内部受控" without user request
new TextRun({ text: '机密 · 内部受控  ·  第 ', ... })

// Skip "第 X 卷说明" and everything until the next H1
if (/^#\s+第\s*\d+\s*卷说明/.test(line)) {
  skipVolumeDescription = true;
  i++; continue;
}
if (skipVolumeDescription) {
  if (/^#\s+/.test(line) && !/^#\s+第\s*\d+\s*卷说明/.test(line)) {
    skipVolumeDescription = false;
    // Fall through to process this heading
  } else { i++; continue; }
}

// Skip "交付自检清单" section (from H2 to the next H2 or higher)
if (/^##\s+.*交付自检清单/.test(line)) {
  skipSelfCheck = true;
  i++; continue;
}
if (skipSelfCheck) {
  if (/^#{1,2}\s+/.test(line) && !/交付自检清单/.test(line)) {
    skipSelfCheck = false;
    // Fall through to process this heading
  } else { i++; continue; }
}

import zipfile, re
z = zipfile.ZipFile('output.docx')
xml = z.read('word/document.xml').decode('utf-8')
print(f"Tables: {xml.count('<w:tbl>')}")
print(f"Pipe remnants: {len(re.findall(r'<w:t[^>]*>[^<]*[|][^<]*</w:t>', xml))}")
print(f"Media: {len([n for n in z.namelist() if n.startswith('word/media/')])}")

for (let vi = 0; vi < volumes.length; vi++) {
  if (vi > 0) allElements.push(new Paragraph({ children: [new PageBreak()] }));
  const { elements } = convertMarkdown(content, vol.diagramStart);
  allElements.push(...elements);
}

properties: {
  page: {
    size: { width: 11906, height: 16838 },
    margin: { top: 1440, right: 1440, bottom: 1440, left: 1440 }
  }
}