Momento/memento-note/lib/clip/extract-article.ts

import { Readability } from '@mozilla/readability'
import { JSDOM } from 'jsdom'
import DOMPurify from 'isomorphic-dompurify'
import {
  applyRtlToHtmlBlocks,
  readPageLocaleFromHtml,
  resolveClipLocale,
  wrapClipArticleHtml,
  type ClipLocaleHint,
} from '@/lib/clip/rtl-content'

export interface ExtractedArticle {
  title: string
  content: string
  textContent: string
  excerpt: string
  locale: ClipLocaleHint
}

// Lignes de métadonnées qu'on veut retirer du texte extrait.
// On teste ligne par ligne pour ne pas tronquer du vrai contenu en plein milieu.
const META_LINE_PATTERNS = [
  // "mis à jour", "updated at", "actualizado", "aktualisiert", "aggiornato"
  /\b(mis à jour|updated at|updated|actualizado|actualizada|aktualisiert|aktualisiert am|aggiornato alle|aggiornato)\b/i,
  // "Temps de lecture", "Reading time", etc.
  /\b(temps de lecture|reading time|lesedauer|tiempo de lectura|tempo di lettura|czas czytania|время чтения|阅读时间|読了時間)\b/i,
  // "Aujourd'hui à 09:15", "Today at 09:15", etc.
  /\b(aujourd['’]?hui|today|hoy|heute|oggi|vandaag|dzisiaj|сегодня|今日)\b.*\d{1,2}:\d{2}/i,
  // "Le Progrès - Aujourd'hui à 09:15 | mis à jour..."
  /^.*\s[-–|]\s.*\d{1,2}:\d{2}/i,
  // Heure seule "09:15" ou "09:15 | ..."
  /^\s*\d{1,2}:\d{2}\s*([-|].*)?$/i,
  // "Publié le ...", "Published on ..."
  /\b(publié le|publié|published on|published|veröffentlicht am|publicado el|publicado)\b/i,
  // "Source : ...", "Par ..." en début de ligne
  /^\s*(source\s*:|par\s+|by\s+|auteur\s*:|author\s*:)/i,
]

function isMetaLine(line: string): boolean {
  return META_LINE_PATTERNS.some((pattern) => pattern.test(line))
}

export function cleanClipText(text: string): string {
  return text
    .replace(/\r\n/g, '\n')
    .split('\n')
    .map((line) => line.trim())
    .filter((line) => line.length > 0 && !isMetaLine(line))
    .join('\n')
    .replace(/\n{3,}/g, '\n\n')
    .trim()
}

export function makeExcerpt(text: string, maxLen = 500): string {
  if (text.length <= maxLen) return text
  // Préférer couper à la fin d'une phrase.
  const sentenceEnd = text.lastIndexOf('.', maxLen)
  if (sentenceEnd > maxLen * 0.6) return text.slice(0, sentenceEnd + 1).trim()
  // Sinon couper au dernier espace.
  const space = text.lastIndexOf(' ', maxLen)
  if (space > maxLen * 0.6) return text.slice(0, space).trim() + '…'
  return text.slice(0, maxLen).trim() + '…'
}

export function extractArticleFromHtml(html: string, pageUrl: string): ExtractedArticle | null {
  const dom = new JSDOM(html, { url: pageUrl })
  const reader = new Readability(dom.window.document)
  const article = reader.parse()
  if (!article) return null

  const pageLocale = readPageLocaleFromHtml(html)
  const readabilityDir = article.dir?.toLowerCase() === 'rtl' ? 'rtl' : 'ltr'
  const readabilityLang = article.lang?.split('-')[0]?.toLowerCase()
  const locale = resolveClipLocale(
    pageUrl,
    article.title || '',
    article.textContent || '',
  )
  const mergedLocale: ClipLocaleHint = {
    direction:
      readabilityDir === 'rtl' || pageLocale.direction === 'rtl' || locale.direction === 'rtl'
        ? 'rtl'
        : 'ltr',
    lang:
      (readabilityLang === 'fa' || readabilityLang === 'ar' || readabilityLang === 'he'
        ? readabilityLang
        : undefined) ||
      locale.lang ||
      pageLocale.lang,
  }

  const sanitized = DOMPurify.sanitize(article.content || '')
  const rtlBlocks = applyRtlToHtmlBlocks(sanitized, mergedLocale)
  const content = wrapClipArticleHtml(rtlBlocks, mergedLocale)

  const cleanText = cleanClipText(article.textContent || '')

  return {
    title: (article.title || '').trim(),
    content,
    textContent: cleanText,
    excerpt: makeExcerpt(cleanText, 500),
    locale: mergedLocale,
  }
}

export function clipFooterLocaleTag(lang?: string): string {
  if (lang === 'fa') return 'fa-IR'
  if (lang === 'ar') return 'ar'
  if (lang === 'he') return 'he-IL'
  return 'fr-FR'
}

export function buildClipSourceFooter(domain: string, date: Date, localeTag = 'fr-FR'): string {
  const formatted = date.toLocaleDateString(localeTag, { day: 'numeric', month: 'long', year: 'numeric' })
  const isRtl = localeTag.startsWith('fa') || localeTag.startsWith('ar') || localeTag.startsWith('he')
  const label =
    localeTag.startsWith('fa')
      ? `برگرفته از ${domain} — ${formatted}`
      : localeTag.startsWith('ar')
        ? `مقتبس من ${domain} — ${formatted}`
        : `Extrait de ${domain} le ${formatted}`
  const dirAttr = isRtl ? ' dir="rtl"' : ''
  return `<hr/><p${dirAttr}><small>${DOMPurify.sanitize(label)}</small></p>`
}