blog/src/utils/description.ts

import type { CollectionEntry } from 'astro:content'
import { defaultLocale } from '@/config'
import MarkdownIt from 'markdown-it'
import sanitizeHtml from 'sanitize-html'

const parser = new MarkdownIt()

type ExcerptScene = 'list' | 'meta' | 'og' | 'rss'

// Excerpt length in different scenarios
const EXCERPT_LENGTHS: Record<ExcerptScene, {
  cjk: number
  other: number
}> = {
  list: {
    cjk: 120,
    other: 240,
  },
  meta: {
    cjk: 120,
    other: 240,
  },
  og: {
    cjk: 75,
    other: 150,
  },
  rss: {
    cjk: 120,
    other: 240,
  },
}

const isCJKLang = (lang: string) => ['zh', 'zh-tw', 'ja'].includes(lang)

// Generate an excerpt from Markdown content
export function generateExcerpt(
  content: string,
  scene: ExcerptScene,
  lang: string,
): string {
  if (!content)
    return ''

  const length = isCJKLang(lang)
    ? EXCERPT_LENGTHS[scene].cjk
    : EXCERPT_LENGTHS[scene].other

  // Convert Markdown to plain text
  const plainText = sanitizeHtml(parser.render(content), {
    allowedTags: [],
    allowedAttributes: {},
  })

  // Replace line breaks with spaces
  const normalizedText = plainText.replace(/\s+/g, ' ')
  const excerpt = normalizedText.slice(0, length).trim()
  // Add ellipsis if text was truncated
  return normalizedText.length > length ? `${excerpt}...` : excerpt
}

// Automatically generate a description for the article
export function generateDescription(
  post: CollectionEntry<'posts'>,
  scene: ExcerptScene,
): string {
  // If the article already has a description, return it directly
  if (post.data.description)
    return post.data.description

  const lang = post.data.lang ?? defaultLocale
  return generateExcerpt(post.body, scene, lang)
}