UHD_Media_Tool/packages/adapters/src/sites/spider/detail-page.ts

/**
 * 详情页扩展内容(简介 / MediaInfo / 截图 / IMDb·豆瓣·TMDB / 文件清单 / 标签 / 上传者)—— 本项目扩展。
 * 以 MovieClaw detail_*_css(description / imdb / douban / file_list)为起点,其余按 NexusPHP details.php 公开结构
 * 做通用规则:字段标签(td.rowhead 或站点配置的 div 标签)的下一个兄弟元素是值。站点特化写在定义 `umt.detail.page`。
 * 文件清单只取页面上已有的内容,不为它额外请求 viewfilelist.php(请求克制)。
 */
import type { CheerioAPI } from "cheerio";
import {
  emptyDetailPage,
  extractTmdb,
  looksLikeMediaInfo,
  MAX_FILES,
  MAX_SCREENSHOTS,
  screenshotUrl,
  tidyText,
} from "../detail-text.ts";
import type { SiteDetailFile, SiteDetailPage } from "../types.ts";
import { extractDoubanId, extractImdbId } from "../util.ts";
import { mcSelect } from "./mc-selector.ts";
import type { IndexerDefinition } from "./schema.ts";
import { parseSizeMp } from "./spider.ts";

type DetailConf = NonNullable<IndexerDefinition["umt"]["detail"]>;
type El = ReturnType<CheerioAPI>;

const LABELS = {
  description: ["简介", "Description"],
  /** 没有简介容器时的次选:豆瓣 / IMDb 信息块(憨憨把豆瓣资料拆成独立一节) */
  descriptionFallback: ["豆瓣", "豆瓣信息", "影片信息", "IMDb信息", "IMDb"],
  screenshots: ["截图", "Screenshots", "Screenshot"],
  mediaInfo: ["MediaInfo", "媒体信息", "BDInfo"],
  tags: ["标签", "Tags"],
  uploader: ["上传者", "发布者", "发布人", "上传人", "Uploader"],
  files: ["文件列表", "文件清单", "种子文件", "File list", "Files"],
};

/** 简介里可能装 MediaInfo 的块 */
const MI_BLOCKS = "fieldset, div.codemain, pre, code, blockquote, .spoiler-content, .hide, div.nexus-media-info-raw";

const normLabel = (t: string) =>
  t
    .replace(/\s+/g, "")
    .replace(/[:：]+$/, "")
    .toLowerCase();

/** 换行占位:先把 <br> / 块级换行记成它,再去掉 HTML 源码里的换行(浏览器按空白处理),最后还原 */
const BR = "\u0000";

/**
 * 保留换行的文本:<br> 与块级元素换行。源码换行不算换行(观众的简介是 `<br />\r\n`,
 * 两者都算会让每行之间多一空行);<pre> 里的换行保留。
 */
function blockText($: CheerioAPI, el: El): string {
  const c = el.clone();
  c.find("script, style").remove();
  c.find("pre, textarea").each((_, n) => {
    $(n).text($(n).text().replace(/\r?\n/g, BR));
  });
  c.find("br").replaceWith(BR);
  c.find("div, p, li, tr, pre, fieldset, blockquote, h1, h2, h3, h4, h5, h6").each((_, n) => {
    $(n).append(BR);
  });
  const text = c
    .text()
    .replace(/\r?\n/g, " ")
    // `<br />\r\n下一行` 的源码换行变成了行首一个空格
    .replace(new RegExp(`${BR} `, "g"), BR)
    .replaceAll(BR, "\n");
  return tidyText(text);
}

/** NexusPHP MediaInfo 控件的外壳:标题栏 div.codetop、摘要表 div.show、原文 div.hide(观众三者相邻) */
const MI_WIDGET = "div.codetop, div.show, div.hide";

const SIZE_TAIL = /^(.*?)\s*[(（[]\s*([\d.,]+\s*[KMGTP]?i?B)\s*[)）\]]\s*$/i;
const HAS_EXT = /\.[A-Za-z0-9]{2,5}$/;

function fileFromText(text: string): SiteDetailFile | null {
  const t = text.replace(/\s+/g, " ").trim();
  if (!t) return null;
  const m = SIZE_TAIL.exec(t);
  if (m?.[1]) return { name: m[1], sizeBytes: parseSizeMp(m[2]!) || null };
  return HAS_EXT.test(t) ? { name: t, sizeBytes: null } : null;
}

/** 文本里的 IMDb ID:imdb.com/title 链接或「IMDB : tt…」一行;不接受孤立的 tt 串(避免误中哈希) */
export function imdbFromText(text: string): string | null {
  const m =
    /imdb\.com\/title\/(tt\d{7,8})\b/i.exec(text) ??
    /\bIMDb\s*(?:ID|链接|編號|编号)?\s*[:：]?\s*(tt\d{7,8})\b/i.exec(text);
  return m?.[1] ?? null;
}

/** NexusPHP viewfilelist.php 返回的文件表(按需 AJAX 取到的 HTML 片段) */
export function filesFromFileListHtml($: CheerioAPI): SiteDetailFile[] {
  return filesFrom($, $("tr"));
}

function filesFrom($: CheerioAPI, rows: El): SiteDetailFile[] {
  const out: SiteDetailFile[] = [];
  rows.each((_, n) => {
    if (out.length >= MAX_FILES) return;
    const row = $(n);
    const cells = row.children("td");
    if (cells.length >= 2) {
      const name = cells.first().text().replace(/\s+/g, " ").trim();
      const size = cells.eq(1).text().replace(/\s+/g, " ").trim();
      if (name && HAS_EXT.test(name) && /\d/.test(size)) out.push({ name, sizeBytes: parseSizeMp(size) || null });
      return;
    }
    const f = fileFromText(row.text());
    if (f) out.push(f);
  });
  return out;
}

export function extractDetailPage($: CheerioAPI, conf: DetailConf, base: string): SiteDetailPage {
  const page = emptyDetailPage();
  const p = conf.page ?? {};

  // 字段标签 → 值元素
  const pairs: { label: string; value: El }[] = [];
  const addLabels = (sel: string, valueFor: (e: El) => El) => {
    $(sel).each((_, n) => {
      const e = $(n);
      const label = normLabel(e.text());
      if (!label || label.length > 16) return;
      const value = valueFor(e);
      if (value.length) pairs.push({ label, value });
    });
  };
  addLabels(conf.rowLabel, (e) => e.nextAll(conf.rowValue).first());
  for (const sel of p.labelSelectors ?? []) addLabels(sel, (e) => e.next());
  const byLabel = (names: string[], exact = false): El | null =>
    pairs.find((x) => names.some((n) => (exact ? x.label === normLabel(n) : x.label.startsWith(normLabel(n)))))
      ?.value ?? null;
  page.probe.labels = [...new Set(pairs.map((x) => x.label))].slice(0, 40);

  // 简介容器
  const descSel = conf.fields?.description ?? conf.description;
  let desc: El | null = descSel ? $(descSel).first() : null;
  if (!desc?.length) desc = byLabel(LABELS.description);
  page.probe.hasDescription = !!desc?.length;
  const infoBlock = desc?.length ? null : byLabel(LABELS.descriptionFallback);

  // MediaInfo:站点选择器 → 字段标签 → 简介里的引用 / 代码块(取最内层文本,整块从简介里去掉)
  let mediaInfoEl: El | null = null;
  let miBlockIndex = -1;
  if (p.mediaInfo) {
    const el = $(p.mediaInfo).first();
    if (el.length) mediaInfoEl = el;
  }
  mediaInfoEl ??= byLabel(LABELS.mediaInfo);
  const blocks = desc?.length ? desc.find(MI_BLOCKS) : null;
  if (!mediaInfoEl && blocks) {
    blocks.each((i, n) => {
      if (miBlockIndex >= 0) return;
      if (looksLikeMediaInfo(blockText($, $(n)))) miBlockIndex = i;
    });
    if (miBlockIndex >= 0) {
      const outer = blocks.eq(miBlockIndex);
      const inner = outer
        .find(MI_BLOCKS)
        .filter((_, n) => looksLikeMediaInfo(blockText($, $(n))))
        .last();
      mediaInfoEl = inner.length ? inner : outer;
    }
  }
  if (mediaInfoEl) {
    const t = blockText($, mediaInfoEl);
    if (t) page.mediaInfo = t.slice(0, 50_000);
  }

  // 截图
  const imgs: string[] = [];
  const pushImgs = (els: El | null) =>
    els?.each((_, n) => {
      const e = $(n);
      const u = screenshotUrl(e.attr("data-src") ?? e.attr("data-original") ?? e.attr("src"), base);
      if (u && !imgs.includes(u)) imgs.push(u);
    });
  if (p.screenshots) pushImgs($(p.screenshots));
  else {
    // 独立的「截图」一节(憨憨)优先;没有再取简介里的图
    const shotBlock = byLabel(LABELS.screenshots);
    if (shotBlock) pushImgs(shotBlock.is("img") ? shotBlock : shotBlock.find("img"));
    if (!imgs.length && desc?.length) pushImgs(desc.find("img"));
  }
  page.probe.imageCount = imgs.length;
  page.screenshots = imgs.slice(0, MAX_SCREENSHOTS);

  // 简介纯文本:去掉站点固定内容、MediaInfo 块(连同控件标题栏 / 摘要表)与图片
  if (desc?.length) {
    const c = desc.clone();
    if (miBlockIndex >= 0) {
      const isMi = (n: Parameters<typeof $>[0]) => looksLikeMediaInfo(blockText($, $(n)));
      // 同一份 MediaInfo 常出现两次(摘要 div.show 与原文 div.hide):另外去掉最内层的 MediaInfo 块,
      // 在控件外壳(div.codetop / div.show / div.hide)里时连同相邻外壳一起去掉
      const extra = c
        .find(MI_BLOCKS)
        .filter((_, n) => isMi(n) && !$(n).find(MI_BLOCKS).toArray().some(isMi))
        .map((_, n) => {
          const e = $(n);
          const shell = e.parent().is(MI_WIDGET) ? e.parent() : e;
          const out = [shell.get(0)!];
          if (!shell.is(MI_WIDGET)) return out;
          for (let x = shell.prev(); x.is(MI_WIDGET) && out.length < 3; x = x.prev()) out.push(x.get(0)!);
          for (let x = shell.next(); x.is(MI_WIDGET) && out.length < 4; x = x.next()) out.push(x.get(0)!);
          return out;
        })
        .get();
      c.find(MI_BLOCKS).eq(miBlockIndex).remove();
      for (const w of extra) $(w).remove();
    }
    if (p.descriptionStrip) c.find(p.descriptionStrip).remove();
    c.find("img").remove();
    const t = blockText($, c);
    if (t) page.descriptionText = t.slice(0, 20_000);
  } else if (infoBlock) {
    const c = infoBlock.clone();
    c.find("img").remove();
    // 块里的「标签」span 与值之间补上冒号 / 换行,读起来是「片名:xxx」一行一项
    c.find(p.labelSelectors?.join(", ") || "span.font-bold").each((_, n) => {
      const e = $(n);
      const t = e.text().trim();
      if (t && t.length <= 16 && !/[:：]$/.test(t)) e.text(`${t}：`);
      e.before("\n");
    });
    const t = blockText($, c)
      .replace(/[ \t]*\n[ \t]*/g, "\n")
      .replace(/：\s*\n(?=\S)/g, "：");
    if (t) page.descriptionText = t.slice(0, 20_000);
  }

  // 外部 ID:页面上的链接优先,其次简介 / 信息块正文,再次 MediaInfo 里的 IMDB / TMDB 行(憨憨的 MediaInfo 带这两行)
  const descHtml = `${desc?.html() ?? ""}\n${infoBlock?.html() ?? ""}`;
  const idText = `${descHtml}\n${page.mediaInfo ?? ""}`;
  const f = conf.fields ?? {};
  page.imdbId =
    extractImdbId(mcSelect($, null, f.imdb ?? "a[href*='imdb.com']::attr(href)")) ??
    imdbFromText(idText) ??
    extractImdbId(descHtml) ??
    null;
  page.doubanId =
    extractDoubanId(mcSelect($, null, f.douban ?? "a[href*='douban.com']::attr(href)")) ??
    extractDoubanId(idText) ??
    null;
  page.tmdb = extractTmdb($("a[href*='themoviedb.org']").first().attr("href")) ?? extractTmdb(idText);

  // 文件清单:站点选择器 → MovieClaw file_list(去伪元素)→ 字段标签 → #filelist 表格
  const fileSel = p.files ?? f.fileList?.replace(/::text$/, "");
  let files = fileSel ? filesFrom($, $(fileSel)) : [];
  if (!files.length) {
    const v = byLabel(LABELS.files);
    if (v) files = filesFrom($, v.find("li, tr"));
  }
  if (!files.length) files = filesFrom($, $("#filelist tr"));
  page.files = files;

  // 标签
  const tagEl = p.tags ? $(p.tags) : byLabel(LABELS.tags);
  if (tagEl?.length) {
    // 只取最内层的 span / a(观众外面还包了一层 span.detail-tag-list,取外层会把全部标签连成一个)
    const leaves = tagEl.find("span, a").filter((_, n) => $(n).find("span, a").length === 0);
    const parts = leaves.length ? leaves.map((_, n) => $(n).text().trim()).get() : tagEl.text().split(/[\s,，|/]+/);
    page.tags = [...new Set(parts.map((t) => t.trim()).filter((t) => t && t.length <= 20))].slice(0, 20);
  }

  // 上传者
  // 上传者标签精确匹配:观众另有「发布者带宽」一行
  const upSel = p.uploader ? $(p.uploader).first() : null;
  const upEl = upSel?.length ? upSel : byLabel(LABELS.uploader, true);
  if (upEl?.length) {
    const a = upEl.is("a") ? upEl : upEl.find("a[href*='userdetails.php']").first();
    const name = (a.length ? a.text() : upEl.text()).replace(/\s+/g, " ").trim();
    if (name) page.uploader = name.slice(0, 64);
  }
  if (!page.uploader) {
    for (const x of pairs) {
      const m = /(?:上传者|发布者|发布人|Uploader)\s*[:：]\s*(\S+)/i.exec(x.value.text());
      if (m?.[1]) {
        page.uploader = m[1].slice(0, 64);
        break;
      }
    }
  }
  return page;
}