migrate-lsw/award_of_pride.js

// หน้าแรก > งานบริการ > รางวัลแห่งความภูมิใจ

const { execSync } = require("child_process");
const cheerio = require("cheerio");
const fs = require("fs");
const path = require("path");

const BASE = "https://ladsawai.go.th";
const OUT = path.join(process.cwd(), "รางวัลแห่งความภูมิใจ");

fs.mkdirSync(OUT, { recursive: true });

function curlHtml(url) {
  return execSync(
    `curl -L -s "${url}" -H "User-Agent: Mozilla/5.0" -H "Accept-Language: th-TH,th;q=0.9"`,
    { encoding: "utf8", maxBuffer: 20 * 1024 * 1024 }
  );
}

function absUrl(src) {
  if (!src) return null;
  if (src.startsWith("http")) return src;
  return BASE + src;
}

function scrapeDetailImagesContent(detailUrl) {
  const html = curlHtml(detailUrl);
  const $ = cheerio.load(html);

  // ---------- images ----------
  const imgSet = new Set();

  $(".maingroup.gallery a[href]").each((_, a) => {
    const href = ($(a).attr("href") || "").trim();
    const full = absUrl(href);
    if (full) imgSet.add(full);
  });

  if (imgSet.size === 0) {
    $("a[href]").each((_, a) => {
      const href = ($(a).attr("href") || "").trim();
      const full = absUrl(href);
      if (full && /\.(jpg|jpeg|png|webp|gif)(\?|$)/i.test(full)) imgSet.add(full);
    });
  }

  // ---------- content ----------
  // ✅ เลือกกล่องที่ไม่ใช่ gallery และ "มีข้อความจริง"
  const candidates = $(".col-12.maingroup").not(".gallery");

  let bestBox = null;
  let bestScore = -1;

  candidates.each((_, el) => {
    const $el = $(el);

    // เอา text โดยตัดของไม่เกี่ยว (emoji img, script, style)
    const text = $el
      .clone()
      .find("img, script, style")
      .remove()
      .end()
      .text()
      .replace(/\s+/g, " ")
      .trim();

    const pCount = $el.find("p").length;
    const score = (text ? text.length : 0) + pCount * 50; // ให้ p มีน้ำหนักเพิ่ม

    if (score > bestScore) {
      bestScore = score;
      bestBox = $el;
    }
  });

  let content = "";
  if (bestBox && bestBox.length) {
    const lines = [];

    bestBox.find("p").each((_, p) => {
      const t = $(p)
        .clone()
        .find("img") // ตัดรูป emoji ใน p
        .remove()
        .end()
        .text()
        .replace(/\s+/g, " ")
        .trim();

      if (t) lines.push(t);
    });

    content = lines.length
      ? lines.join("\n")
      : bestBox
          .clone()
          .find("img, script, style")
          .remove()
          .end()
          .text()
          .replace(/\s+/g, " ")
          .trim();
  }

  let mainImageUrl = ''
  try{
    const mainImageDiv = $(".imagestopic img[src]");
    const src = ($(mainImageDiv).attr("src") || "").trim();
    const full = absUrl(src);
    if (full) mainImageUrl = full;
  }
  catch(error){
    mainImageUrl = '' 
  }

  return { imgs: [...imgSet], text: content, mainImage: mainImageUrl };
}

function scrapeOnePage(menuId, page, saveHtml = false) {
  const url = `${BASE}/public/list/data/index/menu/${menuId}/page/${page}`;
  const html = curlHtml(url);

  if (saveHtml) {
    fs.writeFileSync(path.join(OUT, `page-menu-${menuId}-page-${page}.html`), html, "utf8");
  }

  const $ = cheerio.load(html);
  const items = [];

  $(".row.data-row").each((_, row) => {
    const el = $(row);

    const title = el
      .find(".col-12.col-sm-10")
      .text()
      .replace(/\s+/g, " ")
      .trim();

    const detailRef = el
      .find("a.listdataconfig_link ") // a.listdataconfig_link 
      .attr("href")
      .trim();

    const date = el.find(".col-sm-2").last().text().trim();
    const imgSrc = el.find("img").attr("src");

    if (!title) return;

    const linkD = `https://ladsawai.go.th` + detailRef
    const { text, imgs, mainImage } = linkD ? scrapeDetailImagesContent(linkD) : [];

    items.push({
      title,
      detailRef: linkD,
      detail:{
        img: imgs,
        content: text,
        mainImage: mainImage
      },
      date: date || null,
      image: absUrl(imgSrc),
      sourcePage: page,
      sourceUrl: url,
    });
  });
    //   "sourceUrl": "https://ladsawai.go.th/public/list/data/index/menu/1559/page/1"
    // detailRef: https://ladsawai.go.th/public/list/data/detail/id/3826/menu/1559/page/1
    // /public/list/data/detail/id/3826/menu/1559/page/1

  const output = {
    source: url,
    scrapedAt: new Date().toISOString(),
    menuId,
    page,
    count: items.length,
    items,
  };

  const outJson = path.join(OUT, `list-menu-${menuId}-page-${page}.json`);
  fs.writeFileSync(outJson, JSON.stringify(output, null, 2), "utf8");

  console.log(`✅ page ${page} -> items ${items.length}`);
  return items;
}

(function main() {
  const menuId = 1402; // กิจกรรม
  const totalPages = 1;

  const all = [];
  const seen = new Set();

  // ถ้าไม่อยากให้มี HTML 53 ไฟล์ ให้เป็น false
  const saveHtml = true;

  for (let page = 1; page <= totalPages; page++) {
    const items = scrapeOnePage(menuId, page, saveHtml);

    // รวม + กันซ้ำ
    for (const it of items) {
      const key = `${it.title}|${it.date || ""}|${it.image || ""}`;
      if (seen.has(key)) continue;
      seen.add(key);
      all.push(it);
    }
  }

  const merged = {
    menuId,
    totalPages,
    scrapedAt: new Date().toISOString(),
    totalItems: all.length,
    items: all,
  };

  const outAll = path.join(OUT, `list-menu-${menuId}-all.json`);
  fs.writeFileSync(outAll, JSON.stringify(merged, null, 2), "utf8");

  console.log("✅ Saved merged JSON:", outAll);
  console.log("✅ Total unique items:", all.length);
})();