server/md.js

// Markdown <-> ProseMirror JSON helpers. Deliberately minimal: paragraphs,
// headings 1-3, bullet/ordered lists (flat), code blocks, blockquotes,
// horizontal rules; inline bold/italic/code/strike/links.

// --- ProseMirror JSON -> markdown ---

export function pmToMarkdown(pmDoc) {
  return (pmDoc.content || []).map(blockToMarkdown).filter(s => s !== null).join('\n\n')
}

export function blockToMarkdown(node) {
  switch (node.type) {
    case 'paragraph':
      return inlineToMarkdown(node.content)
    case 'heading':
      return '#'.repeat(node.attrs?.level || 1) + ' ' + inlineToMarkdown(node.content)
    case 'bulletList':
      return (node.content || []).map(li => '- ' + listItemText(li)).join('\n')
    case 'orderedList':
      return (node.content || []).map((li, i) => `${i + 1}. ` + listItemText(li)).join('\n')
    case 'taskList':
      return (node.content || []).map(li => `- [${li.attrs?.checked ? 'x' : ' '}] ` + listItemText(li)).join('\n')
    case 'table':
      return tableToMarkdown(node)
    case 'codeBlock':
      return '```' + (node.attrs?.language || '') + '\n' + textOf(node) + '\n```'
    case 'blockquote':
      return (node.content || []).map(blockToMarkdown).join('\n\n').split('\n').map(l => '> ' + l).join('\n')
    case 'horizontalRule':
      return '---'
    case 'image':
      return `![${node.attrs?.alt || ''}](${node.attrs?.src || ''})`
    default:
      return textOf(node) || null
  }
}

function listItemText(li) {
  return (li.content || []).map(blockToMarkdown).join(' ')
}

function cellText(cell) {
  return (cell.content || [])
    .map(b => inlineToMarkdown(b.content))
    .join(' ')
    .replace(/\|/g, '\\|')
    .trim()
}

function tableToMarkdown(node) {
  const rows = (node.content || []).map(r => (r.content || []).map(cellText))
  if (!rows.length) return ''
  const cols = Math.max(...rows.map(r => r.length))
  const pad = r => {
    const c = r.slice()
    while (c.length < cols) c.push('')
    return c
  }
  const line = cells => '| ' + pad(cells).join(' | ') + ' |'
  // first row is the header (GFM requires one); a delimiter row follows it
  const out = [line(rows[0]), '| ' + Array(cols).fill('---').join(' | ') + ' |']
  for (const r of rows.slice(1)) out.push(line(r))
  return out.join('\n')
}

function textOf(node) {
  if (node.text) return node.text
  return (node.content || []).map(textOf).join('')
}

function inlineToMarkdown(content) {
  if (!content) return ''
  return content.map(n => {
    if (n.type === 'hardBreak') return '  \n'
    let text = n.text || ''
    for (const mark of n.marks || []) {
      if (mark.type === 'bold') text = `**${text}**`
      else if (mark.type === 'italic') text = `*${text}*`
      else if (mark.type === 'code') text = '`' + text + '`'
      else if (mark.type === 'strike') text = `~~${text}~~`
      else if (mark.type === 'link') text = `[${text}](${mark.attrs?.href || ''})`
    }
    return text
  }).join('')
}

// --- markdown -> block descriptors ---
// Each descriptor: { type, attrs?, inline? (segments), items? (arrays of segments), text? }

export function markdownToBlocks(md) {
  const blocks = []
  const lines = (md || '').replace(/\r\n/g, '\n').split('\n')
  let i = 0
  while (i < lines.length) {
    const line = lines[i]
    if (!line.trim()) { i++; continue }

    const fence = line.match(/^```(\w*)\s*$/)
    if (fence) {
      const body = []
      i++
      while (i < lines.length && !/^```\s*$/.test(lines[i])) { body.push(lines[i]); i++ }
      i++
      blocks.push({ type: 'codeBlock', attrs: fence[1] ? { language: fence[1] } : {}, text: body.join('\n') })
      continue
    }
    if (/^(---|\*\*\*)\s*$/.test(line.trim())) { blocks.push({ type: 'horizontalRule' }); i++; continue }

    const image = line.match(/^!\[([^\]]*)\]\(([^)\s]+)\)\s*$/)
    if (image) {
      blocks.push({ type: 'image', attrs: { alt: image[1] || null, src: image[2] } })
      i++
      continue
    }

    const heading = line.match(/^(#{1,6})\s+(.*)$/)
    if (heading) {
      blocks.push({ type: 'heading', attrs: { level: Math.min(heading[1].length, 3) }, inline: tokenizeInline(heading[2]) })
      i++
      continue
    }

    // task list: "- [ ] text" / "- [x] text" (checked before plain bullets)
    if (/^\s*[-*]\s+\[[ xX]\]\s+/.test(line)) {
      const items = []
      while (i < lines.length && /^\s*[-*]\s+\[[ xX]\]\s+/.test(lines[i])) {
        const m = lines[i].match(/^\s*[-*]\s+\[([ xX])\]\s+(.*)$/)
        items.push({ checked: m[1].toLowerCase() === 'x', inline: tokenizeInline(m[2]) })
        i++
      }
      blocks.push({ type: 'taskList', items })
      continue
    }

    if (/^\s*[-*]\s+/.test(line)) {
      const items = []
      while (i < lines.length && /^\s*[-*]\s+/.test(lines[i]) && !/^\s*[-*]\s+\[[ xX]\]\s+/.test(lines[i])) {
        items.push(tokenizeInline(lines[i].replace(/^\s*[-*]\s+/, '')))
        i++
      }
      blocks.push({ type: 'bulletList', items })
      continue
    }

    // GFM table: a "| ... |" row immediately followed by a delimiter row
    if (line.includes('|') && i + 1 < lines.length && /^\s*\|?[\s:|-]*-[\s:|-]*\|?\s*$/.test(lines[i + 1]) && lines[i + 1].includes('-')) {
      const splitCells = l =>
        l
          .trim()
          .replace(/^\||\|$/g, '')
          .split(/(?<!\\)\|/)
          .map(c => c.replace(/\\\|/g, '|').trim())
      const rows = [splitCells(line)]
      i += 2 // header + delimiter
      while (i < lines.length && lines[i].includes('|') && lines[i].trim()) {
        rows.push(splitCells(lines[i]))
        i++
      }
      blocks.push({ type: 'table', rows: rows.map(cells => cells.map(tokenizeInline)) })
      continue
    }

    if (/^\s*\d+[.)]\s+/.test(line)) {
      const items = []
      while (i < lines.length && /^\s*\d+[.)]\s+/.test(lines[i])) {
        items.push(tokenizeInline(lines[i].replace(/^\s*\d+[.)]\s+/, '')))
        i++
      }
      blocks.push({ type: 'orderedList', items })
      continue
    }

    if (/^>\s?/.test(line)) {
      const body = []
      while (i < lines.length && /^>\s?/.test(lines[i])) {
        body.push(lines[i].replace(/^>\s?/, ''))
        i++
      }
      blocks.push({ type: 'blockquote', inline: tokenizeInline(body.join(' ')) })
      continue
    }

    // paragraph: consume until blank line or block marker
    const body = []
    while (i < lines.length && lines[i].trim() && !/^(#{1,6}\s|```|\s*[-*]\s|\s*\d+[.)]\s|>\s?|---\s*$)/.test(lines[i])) {
      body.push(lines[i])
      i++
    }
    blocks.push({ type: 'paragraph', inline: tokenizeInline(body.join(' ')) })
  }
  return blocks
}

// --- inline tokenizer: returns [{ text, attrs }] where attrs maps mark name -> attrs object ---

const INLINE_RE = /(\*\*([^*]+)\*\*)|(\*([^*]+)\*)|(~~([^~]+)~~)|(`([^`]+)`)|(\[([^\]]+)\]\(([^)\s]+)\))/

export function tokenizeInline(text) {
  const segments = []
  let rest = text
  while (rest) {
    const m = rest.match(INLINE_RE)
    if (!m) { segments.push({ text: rest, attrs: {} }); break }
    if (m.index > 0) segments.push({ text: rest.slice(0, m.index), attrs: {} })
    if (m[1]) segments.push({ text: m[2], attrs: { bold: {} } })
    else if (m[3]) segments.push({ text: m[4], attrs: { italic: {} } })
    else if (m[5]) segments.push({ text: m[6], attrs: { strike: {} } })
    else if (m[7]) segments.push({ text: m[8], attrs: { code: {} } })
    else if (m[9]) segments.push({ text: m[10], attrs: { link: { href: m[11] } } })
    rest = rest.slice(m.index + m[0].length)
  }
  return segments.filter(s => s.text)
}