// ===== scanners/parliamentLives.js =====
// Version "costaud" avec parseurs spécifiques FR/BE + fallback générique.

const axios = require("axios");
const cheerio = require("cheerio");
const {
  supaExists,
  askDeepSeek,
  extractDeepSeekResponse,
  sendToDiscord,
  logArticle,
} = require("../utils/shared");

// ---------- ENV attendus ----------
// FR_PARL_SCHEDULE_URL : ex. https://www2.assemblee-nationale.fr/agendas/les-agendas
// FR_PARL_WATCH_URL    : ex. https://videos.assemblee-nationale.fr/direct.php
// FR_PARL_ALT_SCHEDULE : (optionnel) Guide TV LCP ex. https://lcp.fr/guide-tv/100
// FR_PARL_ALT_WATCH    : (optionnel) Direct LCP ex. https://lcp.fr/direct-lcp-5434
//
// BE_PARL_SCHEDULE_URL : ex. https://lachambre.be/  (ou page Agenda)
// BE_PARL_WATCH_URL    : ex. https://media.lachambre.be/
//
// Tu peux ajouter/retirer des sources ci-dessous.

const SOURCES = [
  {
    key: "FR",
    label: "🇫🇷 Assemblée nationale — Direct",
    scheduleUrl: process.env.FR_PARL_SCHEDULE_URL || "",
    watchUrl: process.env.FR_PARL_WATCH_URL || "",
    altScheduleUrl: process.env.FR_PARL_ALT_SCHEDULE || "",
    altWatchUrl: process.env.FR_PARL_ALT_WATCH || "",
    leadMinutes: 90,
    tz: "Europe/Paris",
    parser: parseFRAgenda,       // parseur spécifique
    altParser: parseLCPGuideTV,  // parseur de secours LCP
    keywords: [
      "séance",
      "plénière",
      "questions au gouvernement",
      "qag",
      "hémicycle",
      "direct",
      "en direct",
      "live",
    ],
  },
  {
    key: "BE",
    label: "🇧🇪 Chambre — Direct",
    scheduleUrl: process.env.BE_PARL_SCHEDULE_URL || "",
    watchUrl: process.env.BE_PARL_WATCH_URL || "",
    leadMinutes: 60,
    tz: "Europe/Brussels",
    parser: parseBEChambre, // parseur spécifique pour media.lachambre / page d’accueil vers Agenda
    keywords: ["séance", "plénière", "commission", "en direct", "live"],
  },
];

// ---------- Utils heure & texte ----------

function toMinutes(hh, mm) {
  return hh * 60 + mm;
}
function parseHourString(s) {
  // 9h, 09h, 9h00, 09h30, 9:30, 09:30
  const re = /(?:^|\s)(\d{1,2})[h:](\d{2})(?:\b|h|\s|$)/i;
  const m = re.exec(String(s || ""));
  if (!m) return null;
  const hh = parseInt(m[1], 10);
  const mm = parseInt(m[2], 10);
  if (Number.isNaN(hh) || Number.isNaN(mm) || hh > 23 || mm > 59) return null;
  return { hh, mm };
}
function norm(s) {
  return String(s || "")
    .normalize("NFD")
    .replace(/[\u0300-\u036f]/g, "")
    .toLowerCase();
}
function hasKeyword(s, keywords) {
  const t = norm(s);
  return keywords.some((k) => t.includes(norm(k)));
}
function nowMinutesLocal() {
  const d = new Date();
  return d.getHours() * 60 + d.getMinutes();
}

// ---------- Fallback générique (au cas où) ----------
function parseGeneric(html, { keywords }) {
  const $ = cheerio.load(html);
  const pick = "time, h1, h2, h3, h4, h5, a, li, div, p, span";
  const events = [];

  $(pick).each((_, el) => {
    const text = $(el).text().replace(/\s+/g, " ").trim();
    if (!text || text.length < 4) return;

    const tm = parseHourString(text);
    if (!tm) return;
    if (!hasKeyword(text, keywords || [])) return;

    const title = text;
    events.push({ title, hour: tm.hh, min: tm.mm, raw: text });
  });

  // dédoublonnage simple
  const seen = new Set();
  const uniq = [];
  for (const e of events) {
    const k = `${e.hour}:${e.min}::${norm(e.title)}`;
    if (!seen.has(k)) {
      seen.add(k);
      uniq.push(e);
    }
  }
  return uniq;
}

// ---------- Parseur spécifique FR : https://www2.assemblee-nationale.fr/agendas/les-agendas ----------
function parseFRAgenda(html, { keywords }) {
  const $ = cheerio.load(html);
  const out = [];

  // Heuristiques spécifiques (mais robustes) :
  // - nombreuses pages listent des items d'agenda sous <article>, <li>, ou des blocs .agenda, avec <time> ou texte "HHhMM"
  // 1) time[datetime] & time
  $("time").each((_, el) => {
    const t = $(el).text().replace(/\s+/g, " ").trim();
    const tm = parseHourString(t);
    if (!tm) return;

    // On remonte au parent proche pour choper le titre/description
    const parent = $(el).closest("article, li, .agenda, .evenement, .event, .schedule-item, .bloc, div");
    let title = parent.text().replace(/\s+/g, " ").trim() || t;

    // Filtre mot-clé
    if (!hasKeyword(title, keywords)) return;

    out.push({ title, hour: tm.hh, min: tm.mm, raw: title });
  });

  // 2) lignes textuelles contenant heures + keywords
  const pick = "h1,h2,h3,h4,h5,a,li,div,p,span";
  $(pick).each((_, el) => {
    const text = $(el).text().replace(/\s+/g, " ").trim();
    if (!text || text.length < 5) return;
    const tm = parseHourString(text);
    if (!tm) return;
    if (!hasKeyword(text, keywords)) return;
    out.push({ title: text, hour: tm.hh, min: tm.mm, raw: text });
  });

  // dédoublonner
  const seen = new Set();
  const uniq = [];
  for (const e of out) {
    const k = `${e.hour}:${e.min}::${norm(e.title)}`;
    if (!seen.has(k)) {
      seen.add(k);
      uniq.push(e);
    }
  }
  return uniq;
}

// ---------- Parseur de secours FR (LCP Guide TV) : https://lcp.fr/guide-tv/100 ----------
function parseLCPGuideTV(html, { keywords }) {
  const $ = cheerio.load(html);
  const out = [];

  // Sur le guide TV, les horaires sont souvent dans des cartes/programmes
  // On cible : time, .time, .hour, .schedule__time, etc.
  $("time, .time, .hour, .schedule__time, .program__time").each((_, el) => {
    const t = $(el).text().replace(/\s+/g, " ").trim();
    const tm = parseHourString(t);
    if (!tm) return;

    const parent = $(el).closest("article, li, .program, .schedule__item, .program-card, .program_list_item, .block, div");
    let title = parent.text().replace(/\s+/g, " ").trim() || t;

    if (!hasKeyword(title, keywords)) return;
    out.push({ title, hour: tm.hh, min: tm.mm, raw: title });
  });

  // fallback dans la page
  if (out.length === 0) return parseGeneric(html, { keywords });

  // dédoublonner
  const seen = new Set();
  const uniq = [];
  for (const e of out) {
    const k = `${e.hour}:${e.min}::${norm(e.title)}`;
    if (!seen.has(k)) {
      seen.add(k);
      uniq.push(e);
    }
  }
  return uniq;
}

// ---------- Parseur spécifique BE : Chambre (media.lachambre.be / Agenda) ----------
function parseBEChambre(html, { keywords }) {
  const $ = cheerio.load(html);
  const out = [];

  // Sur media.lachambre.be (ou pages Agenda reliées), on trouve souvent :
  // - des listes d'événements avec <time> (HH:MM) et des titres (plénière/commission).
  // On cible d'abord les time + carte parente :
  $("time").each((_, el) => {
    const t = $(el).text().replace(/\s+/g, " ").trim();
    const tm = parseHourString(t);
    if (!tm) return;

    const parent = $(el).closest("article, li, .event, .event-item, .agenda-item, .media-item, .meeting, .schedule-item, div");
    let title = parent.text().replace(/\s+/g, " ").trim() || t;

    if (!hasKeyword(title, keywords)) return;
    out.push({ title, hour: tm.hh, min: tm.mm, raw: title });
  });

  // Renfort : récupérer d'éventuelles lignes textuelles
  if (out.length === 0) {
    const pick = "h1,h2,h3,h4,h5,a,li,div,p,span";
    $(pick).each((_, el) => {
      const text = $(el).text().replace(/\s+/g, " ").trim();
      if (!text || text.length < 5) return;
      const tm = parseHourString(text);
      if (!tm) return;
      if (!hasKeyword(text, keywords)) return;
      out.push({ title: text, hour: tm.hh, min: tm.mm, raw: text });
    });
  }

  // fallback générique si rien
  if (out.length === 0) return parseGeneric(html, { keywords });

  // dédoublonner
  const seen = new Set();
  const uniq = [];
  for (const e of out) {
    const k = `${e.hour}:${e.min}::${norm(e.title)}`;
    if (!seen.has(k)) {
      seen.add(k);
      uniq.push(e);
    }
  }
  return uniq;
}

// ---------- Publication ----------
async function publishLiveAlert(logger, src, ev) {
  const hh = String(ev.hour).padStart(2, "0");
  const mm = String(ev.min).padStart(2, "0");

  const watch =
    src.watchUrl ||
    src.altWatchUrl ||
    src.scheduleUrl ||
    src.altScheduleUrl ||
    "#";

  const originalTitle = `${src.label} — Live prévu à ${hh}h${mm}`;

  const base = `Alerte direct — ${src.label}\n\n` +
    `Début prévu : ${hh}h${mm} (heure locale)\n\n` +
    `Programme : ${ev.title}\n\n` +
    `Regarder : ${watch}\n\n` +
    `Note : alerte générée automatiquement à l’approche d’un direct.`;

  let rewritten;
  try {
    rewritten = await askDeepSeek(
      `Réécris en français une courte alerte “live imminent” (factuelle, 3 phrases max, sans opinion). 
       Titre: ${originalTitle}\n\nTexte:\n${base}`,
      originalTitle
    );
  } catch {
    rewritten = null;
  }

  const { title: aiTitle, content: summary } =
    rewritten ? extractDeepSeekResponse(rewritten) : { title: originalTitle, content: base };

  // Déduplication via une URL synthétique
  const urlKey = `${src.key}::${hh}:${mm}::${norm(ev.title).slice(0, 120)}`;
  const dedupUrl = `about:live-alert#${encodeURIComponent(urlKey)}`;

  if (await supaExists(dedupUrl)) return false;

  if (typeof logger === "function") {
    const ok = typeof logger.reserve === "function" ? logger.reserve() : true;
    if (!ok) return false;
    const allowed = await logger(
      aiTitle || originalTitle,
      summary || base,
      dedupUrl,
      null,
      `${src.label} • Alerte live`
    );
    return !!allowed;
  }

  await sendToDiscord(aiTitle || originalTitle, summary || base, dedupUrl, null, `${src.label} • Alerte live`);
  await logArticle(aiTitle || originalTitle, summary || base, dedupUrl, null, `${src.label}`);
  return true;
}

// ---------- Scanner principal ----------
async function scanParliamentLives(logger, opts = {}) {
  const oneArticle = !!opts.oneArticle;
  let posted = 0;
  const nowMin = nowMinutesLocal();

  for (const src of SOURCES) {
    try {
      const minutesLead = Math.max(15, parseInt(src.leadMinutes || 60, 10));

      // 1) Page principale
      let events = [];
      if (src.scheduleUrl) {
        const { data: html } = await axios.get(src.scheduleUrl, {
          headers: {
            "User-Agent": "Mozilla/5.0",
            "Accept-Language": "fr-FR,fr;q=0.9,en;q=0.8",
          },
          timeout: 20000,
        });
        if (typeof src.parser === "function") {
          events = src.parser(html, src);
        } else {
          events = parseGeneric(html, src);
        }
      }

      // 2) Si rien, page de secours (FR : LCP)
      if ((!events || events.length === 0) && src.altScheduleUrl) {
        try {
          const { data: html2 } = await axios.get(src.altScheduleUrl, {
            headers: {
              "User-Agent": "Mozilla/5.0",
              "Accept-Language": "fr-FR,fr;q=0.9,en;q=0.8",
            },
            timeout: 20000,
          });
          if (typeof src.altParser === "function") {
            events = src.altParser(html2, src);
          } else {
            events = parseGeneric(html2, src);
          }
        } catch (e) {
          // ignore
        }
      }

      if (!events || events.length === 0) continue;

      const soon = events
        .map((e) => ({ ...e, startMin: toMinutes(e.hour, e.min) }))
        .filter((e) => e.startMin >= nowMin && (e.startMin - nowMin) <= minutesLead)
        .sort((a, b) => a.startMin - b.startMin);

      if (soon.length === 0) continue;

      const next = soon[0];
      const ok = await publishLiveAlert(logger, src, next);
      if (ok) {
        posted += 1;
        if (oneArticle) break;
      }
    } catch (e) {
      console.error(`❌ ParliamentLives(${src.key}):`, e.message || e);
    }
  }

  console.log(`✅ ParliamentLives: ${posted} alerte(s) publiée(s).`);
  return { status: "ok", count: posted };
}

module.exports = { scanParliamentLives };