anwar-pi-website/web/src/lib/lexical.ts

/**
 * Helpers over serialized Lexical richText JSON (Payload's editor state).
 * A serialized tree is `{ root: { children: [...] } }`; text lives on
 * `{ text }` nodes, nesting happens via `{ children }`.
 */

export type Node = {
  type?: string
  text?: string
  children?: Node[]
  tag?: string
  url?: string
  listType?: string
  code?: string
  headerState?: number
}

/** Root `children` of a serialized tree, narrowed at runtime. */
function rootChildren(data: unknown): Node[] | null {
  if (data === null || typeof data !== 'object' || !('root' in data)) return null
  const root: unknown = data.root // unknown via `in` narrowing — no shape assumed
  if (root === null || typeof root !== 'object' || !('children' in root)) return null
  const children: unknown = root.children
  if (!Array.isArray(children)) return null
  // Boundary: the array is the editor's serialized node list; every access
  // below goes through Node's all-optional contract, so nothing unchecked.
  const nodes: Node[] = children
  return nodes
}

/** Concatenated text of every text node (secret scanning, llms extraction). */
export function lexicalToText(content: unknown): string {
  let out = ''
  const walk = (node: unknown): void => {
    if (node === null || typeof node !== 'object') return
    if ('text' in node && typeof node.text === 'string') out += node.text + '\n'
    if ('children' in node && Array.isArray(node.children)) {
      for (const child of node.children) walk(child)
    }
  }
  // Serialized trees wrap nodes in `root` — walk it, not the wrapper.
  if (content !== null && typeof content === 'object' && 'root' in content) {
    walk((content as { root: unknown }).root)
  } else {
    walk(content)
  }
  return out
}

/** Count whitespace-delimited words across all text nodes of a Lexical tree. */
export function countLexicalWords(content: unknown): number {
  let words = 0
  const walk = (node: unknown): void => {
    if (node === null || typeof node !== 'object') return
    if ('text' in node && typeof node.text === 'string') {
      words += node.text.trim().split(/\s+/).filter(Boolean).length
    }
    if ('children' in node && Array.isArray(node.children)) {
      for (const child of node.children) walk(child)
    }
  }
  // Same root unwrap as lexicalToText.
  if (content !== null && typeof content === 'object' && 'root' in content) {
    walk((content as { root: unknown }).root)
  } else {
    walk(content)
  }
  return words
}

export function innerText(nodes?: Node[]): string {
  return (nodes || []).map(n => (n.type === 'text' ? n.text || '' : innerText(n.children))).join('')
}

/** H2 headings — TOC and the paid-mode paywall preview. */
export function extractH2s(data: unknown): string[] {
  const children = rootChildren(data)
  if (!children) return []
  return children
    .filter(n => n.type === 'heading' && (n.tag || 'h2') === 'h2')
    .map(n => innerText(n.children))
}

function tagLevel(tag?: string): number {
  const n = Number.parseInt((tag ?? 'h2').replace('h', ''), 10)
  return Number.isFinite(n) && n >= 1 && n <= 6 ? n : 2
}

/**
 * Plain-text rendering for llms-full.txt: headings → `## `, paragraphs and
 * list items → lines, code → fenced blocks, table rows → `| a | b |`.
 */
export function lexicalToPlain(data: unknown): string {
  const children = rootChildren(data)
  if (!children) return ''
  const lines: string[] = []
  for (const n of children) {
    switch (n.type) {
      case 'heading':
        lines.push(`${'#'.repeat(tagLevel(n.tag))} ${innerText(n.children)}`.trim())
        break
      case 'paragraph':
        lines.push(innerText(n.children).trim())
        break
      case 'list':
        for (const item of n.children || []) lines.push(`- ${innerText(item.children).trim()}`)
        break
      case 'listitem':
        lines.push(`- ${innerText(n.children).trim()}`)
        break
      case 'code':
        lines.push('```\n' + (n.code ?? innerText(n.children)) + '\n```')
        break
      case 'table':
        for (const row of n.children || []) {
          const cells = (row.children || []).map(cell => innerText(cell.children).trim().replace(/\|/g, '\\|'))
          lines.push(`| ${cells.join(' | ')} |`)
        }
        break
      case 'quote':
        lines.push(`> ${innerText(n.children).trim()}`)
        break
      case 'horizontalrule':
        lines.push('---')
        break
      default: {
        const text = innerText(n.children).trim()
        if (text) lines.push(text)
      }
    }
  }
  return lines.filter(line => line !== '').join('\n\n')
}