Merge branch 'det2'

Co-Authored-By: Claude Fable 5.1 <[email protected]>

# Conflicts:
#	packages/adapters/src/sites/definitions/audiences.json
#	scripts/mp-sync/movieclaw-overrides/audiences.json

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

@@ -43,17 +43,35 @@ const normLabel = (t: string) =>
     .replace(/[:：]+$/, "")
     .toLowerCase();
 
-/** 保留换行的文本:<br> 与块级元素换行 */
+/** 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("br").replaceWith("\n");
-  c.find("div, p, li, tr, pre, fieldset, blockquote, h1, h2, h3, h4, h5, h6").each((_, n) => {
-    $(n).append("\n");
+  c.find("pre, textarea").each((_, n) => {
+    $(n).text($(n).text().replace(/\r?\n/g, BR));
   });
-  return tidyText(c.text());
+  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*$/i;
 const HAS_EXT = /\.[A-Za-z0-9]{2,5}$/;
 
@@ -113,8 +131,9 @@ export function extractDetailPage($: CheerioAPI, conf: DetailConf, base: string)
   };
   addLabels(conf.rowLabel, (e) => e.nextAll(conf.rowValue).first());
   for (const sel of p.labelSelectors ?? []) addLabels(sel, (e) => e.next());
-  const byLabel = (names: string[]): El | null =>
-    pairs.find((x) => names.some((n) => x.label.startsWith(normLabel(n))))?.value ?? null;
+  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
@@ -170,10 +189,30 @@ export function extractDetailPage($: CheerioAPI, conf: DetailConf, base: string)
   page.probe.imageCount = imgs.length;
   page.screenshots = imgs.slice(0, MAX_SCREENSHOTS);
 
-  // 简介纯文本:去掉 MediaInfo 块与图片
+  // 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) c.find(MI_BLOCKS).eq(miBlockIndex).remove();
+    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);
@@ -221,17 +260,16 @@ export function extractDetailPage($: CheerioAPI, conf: DetailConf, base: string)
   // 标签
   const tagEl = p.tags ? $(p.tags) : byLabel(LABELS.tags);
   if (tagEl?.length) {
-    const parts = tagEl.find("span, a").length
-      ? tagEl
-          .find("span, a")
-          .map((_, n) => $(n).text().trim())
-          .get()
-      : tagEl.text().split(/[\s,，|/]+/);
+    // 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
-  const upEl = p.uploader ? $(p.uploader).first() : byLabel(LABELS.uploader);
+  // 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();

packages/adapters/src/sites/spider/detail.ts

@@ -51,8 +51,15 @@ export function parseDetailPage(
           .text()
   )
     .replace(/\s+/g, " ")
+    // 促销到期的「剩余时间：」是 h1 的直接文本节点(观众),不属于标题
+    .replace(/\s*(?:剩余时间|剩餘時間|Remaining(?: time)?)\s*[:：]?\s*$/i, "")
     .trim();
-  const subtitle = f.subtitle ? (mcSelect($, null, f.subtitle) ?? undefined) : row(conf.labels.subtitle)?.text().trim();
+  const subtitleEl = f.subtitle ? null : row(conf.labels.subtitle)?.clone();
+  // 副标题行右侧的豆瓣 / IMDb / 烂番茄等评分徽标(观众)不算副标题
+  subtitleEl?.find("[title*='评分'], [title*='Rating'], img").remove();
+  const subtitle = f.subtitle
+    ? (mcSelect($, null, f.subtitle) ?? undefined)
+    : subtitleEl?.text().replace(/\s+/g, " ").trim();
   const basicText = row(conf.labels.basicInfo)?.text().replace(/\s+/g, " ") ?? "";
   const sizeText = f.size
     ? (/([\d.,]+\s*[KMGTP]?i?B)/i.exec(mcSelect($, null, f.size) ?? "")?.[1] ?? null)
@@ -64,6 +71,9 @@ export function parseDetailPage(
   const peersText = row(conf.labels.peers)?.text().replace(/\s+/g, " ") ?? "";
   const seeders = /(\d[\d,]*)\s*个?做种者|(\d[\d,]*)\s*seeder/i.exec(peersText);
   const leechers = /(\d[\d,]*)\s*个?下载者|(\d[\d,]*)\s*leecher/i.exec(peersText);
+  // 站点把做种 / 下载 / 完成数放在其它位置时(观众:基本信息行右侧的 #seeders / #leechers / 完成链接)按选择器取
+  const countOf = (sel: string | undefined) => (sel ? parseCount(mcSelect($, null, sel) ?? undefined) : null);
+  const infoHash = mcSelect($, null, f.infoHash)?.toLowerCase();
   const descEl = $(conf.description).first();
   const descText = descEl.text().trim();
   const descHtml = descEl.html() ?? "";
@@ -76,17 +86,22 @@ export function parseDetailPage(
       list: def.torrents.list,
       fields: Object.fromEntries(
         (["downloadvolumefactor", "uploadvolumefactor", "freedate", "hr"] as const)
-          .filter((k) => fields[k])
-          .map((k) => [k, fields[k]]),
+          .map((k) => [k, conf.promo?.[k] ?? fields[k]] as const)
+          .filter(([, v]) => v),
       ),
     },
     browse: undefined,
   };
   const promo = new SiteSpider(promoDef).infoFrom($, $(conf.headerScope).first());
   // 发布时间:带完整时间的 span[title] → 「发布时间 / 上传时间 / 添加时间」字段的文本(憨憨是 div 标签 + 值)
+  // 标题区(headerScope)里的 span[title] 是促销到期时间,不是发布时间
   const timeTitle =
     $("span[title]")
-      .filter((_, el) => /\d{4}-\d{2}-\d{2} \d{2}:\d{2}/.test($(el).attr("title") ?? ""))
+      .filter(
+        (_, el) =>
+          /\d{4}-\d{2}-\d{2} \d{2}:\d{2}/.test($(el).attr("title") ?? "") &&
+          $(el).closest(conf.headerScope).length === 0,
+      )
       .first()
       .attr("title") ??
     /(?:发布时间|上传时间|添加时间|Added)\s*[:：]?\s*(\d{4}-\d{2}-\d{2} \d{2}:\d{2}(?::\d{2})?)/.exec(
@@ -99,9 +114,9 @@ export function parseDetailPage(
     title,
     category: { id: catId, name: catText ?? categoryName(def, catId ?? undefined) },
     sizeBytes: sizeText ? parseSizeMp(sizeText) : null,
-    seeders: seeders ? parseCount(seeders[1] ?? seeders[2]) : null,
-    leechers: leechers ? parseCount(leechers[1] ?? leechers[2]) : null,
-    snatched: null,
+    seeders: seeders ? parseCount(seeders[1] ?? seeders[2]) : countOf(f.seeders),
+    leechers: leechers ? parseCount(leechers[1] ?? leechers[2]) : countOf(f.leechers),
+    snatched: countOf(f.snatched),
     promotion: promotionFromFactors(
       promo.downloadvolumefactor,
       promo.uploadvolumefactor,
@@ -116,6 +131,7 @@ export function parseDetailPage(
     observedAt,
   };
   if (subtitle) detail.subtitle = subtitle;
+  if (infoHash && /^[0-9a-f]{40}$/.test(infoHash)) detail.infoHash = infoHash;
   if (descText) detail.descriptionText = descText.slice(0, 20_000);
   const imdb = extractImdbId(mcSelect($, null, f.imdb)) ?? extractImdbId(descHtml) ?? extractImdbId(descText);
   if (imdb) detail.imdbId = imdb;

packages/adapters/src/sites/spider/schema.ts

@@ -196,8 +196,27 @@ export interface UmtSiteExtension {
      * 用于 HHClub 这类 div 详情页。
      */
     fields?: Partial<
-      Record<"title" | "subtitle" | "size" | "description" | "download" | "imdb" | "douban" | "fileList", string>
+      Record<
+        | "title"
+        | "subtitle"
+        | "size"
+        | "description"
+        | "download"
+        | "imdb"
+        | "douban"
+        | "fileList"
+        | "seeders"
+        | "leechers"
+        | "snatched"
+        | "infoHash",
+        string
+      >
     >;
+    /**
+     * 详情页促销 / H&R 字段(downloadvolumefactor / uploadvolumefactor / freedate / hr),覆盖同名列表字段。
+     * 详情页头部的促销标记常与列表不同(列表 img.pro_free,详情 h1 里 font.free + 「剩余时间」span)。
+     */
+    promo?: Record<string, SpiderFieldSelector>;
     /**
      * 详情页扩展内容(简介 / MediaInfo / 截图 / 文件 / 标签 / 上传者)的站点特化,均可省略(走通用规则,见 detail-page.ts)。
      * 选择器为普通 CSS。
@@ -216,6 +235,8 @@ export interface UmtSiteExtension {
        * 每次详情刷新最多 1 次,失败不影响详情。
        */
       fileListPath?: string;
+      /** 从简介里去掉的站点固定内容(如站点声明),普通 CSS */
+      descriptionStrip?: string;
     };
   };
   /** 下载路径(按资源 ID 拼接) */