Skip to content

一分钟搭建影视站,支持Docker等部署方式

fix(search): 拒绝默认目录兜底,补季数别名与流式结果过滤

bestZweifd58b1994989
diff --git a/src/app/api/search/route.test.ts b/src/app/api/search/route.test.ts
index 5b48b40..52b4221 100644
--- a/src/app/api/search/route.test.ts
+++ b/src/app/api/search/route.test.ts
@@ -1,6 +1,7 @@
import { afterEach, beforeAll, describe, expect, it, vi } from 'vitest';
import { POST } from './route';
import { SESSION_COOKIE, signSession } from '@/lib/auth';
+import type { SearchResponse, SearchStreamEvent, SourceConfig } from '@/lib/types';
/**
* 聚合搜索接口单测:跨源聚合、同源去重、精确命中置顶、成人内容过滤、
@@ -106,8 +107,9 @@ describe('POST /api/search', () => {
it('stream=1 以 NDJSON 逐源推送并以 done 事件收尾', async () => {
mockUpstream();
- // 用不同关键词避开同键短缓存(缓存命中时只推 done 事件,见 route.ts 缓存分支)
- const res = await POST(makeRequest({ wd: '流式检测关键词', sources: SOURCES, filterAdult: true }, true));
+ // 用不同源地址避开同键短缓存,保留有实际匹配结果的关键词。
+ const sourcesForStream = SOURCES.map((source) => ({ ...source, url: source.url + '/stream' }));
+ const res = await POST(makeRequest({ wd: '测试剧', sources: sourcesForStream, filterAdult: true }, true));
expect(res.status).toBe(200);
const text = await res.text();
const lines = text
@@ -123,6 +125,74 @@ describe('POST /api/search', () => {
expect(lines[lines.length - 1].type).toBe('done');
});
+ it.each([true, false])('逐源、最终与缓存结果使用相同过滤规则(成人过滤:%s)', async (filterAdult) => {
+ const sources: SourceConfig[] = [
+ { key: 'good', name: '有效源', url: `https://cms.example.com/consistent-${filterAdult}` },
+ { key: 'bad', name: '失败源', url: `https://cms.example.com/failed-${filterAdult}` },
+ ];
+ const fetchSpy = vi.fn(async (input: RequestInfo | URL) => {
+ if (String(input).includes('/failed-')) return new Response('unavailable', { status: 500 });
+ return new Response(cmsList([
+ { vod_id: 1, vod_name: '赛博朋克:边缘行者第二季', type_name: '动漫' },
+ { vod_id: 1, vod_name: '赛博朋克:边缘行者第二季', type_name: '动漫' },
+ { vod_id: 2, vod_name: '10月17日德甲联赛', type_name: '足球' },
+ { vod_id: 3, vod_name: '', type_name: '动漫' },
+ { vod_id: 4, vod_name: '赛博朋克:边缘行者第二季', type_name: '福利视频' },
+ ]), { headers: { 'content-type': 'application/json' } });
+ });
+ vi.stubGlobal('fetch', fetchSpy);
+ const body = { wd: '赛博朋克:边缘行者 2', sources, filterAdult };
+ const stream = await POST(makeRequest(body, true));
+ const events = (await stream.text()).trim().split('\n').map((line) => JSON.parse(line) as SearchStreamEvent);
+ const good = events.find((event) => event.type === 'source' && event.sourceKey === 'good');
+ const failed = events.find((event) => event.type === 'source' && event.sourceKey === 'bad');
+ const done = events.find((event) => event.type === 'done');
+ const ids = filterAdult ? ['1'] : ['1', '4'];
+ expect(good?.list.map((item) => item.vodId)).toEqual(ids);
+ expect(good).toMatchObject({ type: 'source', sourceKey: 'good', ok: true, ms: expect.any(Number) });
+ expect(failed).toMatchObject({ type: 'source', sourceKey: 'bad', ok: false, list: [], error: 'HTTP 500' });
+ expect(done).toMatchObject({ type: 'done', list: good?.list, failures: [{ sourceKey: 'bad', error: 'HTTP 500' }] });
+
+ // 同一缓存同时服务 JSON 和流式客户端,不能重新引入原始杂项。
+ const calls = fetchSpy.mock.calls.length;
+ const json = await POST(makeRequest(body));
+ const cached = await json.json() as SearchResponse;
+ expect(cached.list).toEqual(good?.list);
+ expect(cached.failures).toEqual(done?.type === 'done' ? done.failures : undefined);
+ const cachedStream = await POST(makeRequest(body, true));
+ expect((await cachedStream.text()).trim().split('\n').map((line) => JSON.parse(line))).toEqual([done]);
+ expect(fetchSpy).toHaveBeenCalledTimes(calls);
+
+ // 新地址走非缓存 JSON 分支,仍须得到相同的筛选结果。
+ const fresh = await POST(makeRequest({ ...body, sources: sources.map((source) => ({ ...source, url: source.url + '/fresh' })) }));
+ const freshData = await fresh.json() as SearchResponse;
+ expect(freshData.list.map((item) => item.vodId)).toEqual(ids);
+ expect(freshData.failures).toEqual(cached.failures);
+ });
+
+ it('没有相关结果时,各响应模式均保留空列表', async () => {
+ vi.stubGlobal('fetch', vi.fn(async () => new Response(cmsList([
+ { vod_id: 1, vod_name: '10月17日德甲联赛', type_name: '足球' },
+ { vod_id: 2, vod_name: '', type_name: '动漫' },
+ ]))));
+ const body = {
+ wd: '赛博朋克:边缘行者 2',
+ sources: [{ key: 'empty', name: '不相关源', url: 'https://cms.example.com/no-relevant-results' }],
+ };
+ const res = await POST(makeRequest(body, true));
+ const events = (await res.text()).trim().split('\n').map((line) => JSON.parse(line) as SearchStreamEvent);
+ expect(events).toHaveLength(2);
+ for (const event of events) expect(event.list).toEqual([]);
+ expect(events[0]).toMatchObject({ type: 'source', ok: true });
+ expect(events[1]).toMatchObject({ type: 'done', failures: [] });
+ const cached = await POST(makeRequest(body));
+ expect(await cached.json()).toEqual({ list: [], failures: [] });
+ const cachedStream = await POST(makeRequest(body, true));
+ expect(JSON.parse((await cachedStream.text()).trim())).toEqual({ type: 'done', list: [], failures: [] });
+ const fresh = await POST(makeRequest({ ...body, sources: [{ ...body.sources[0], url: body.sources[0].url + '/fresh' }] }));
+ expect(await fresh.json()).toEqual({ list: [], failures: [] });
+ });
+
it('未登录返回 401,未配置密码返回 503', async () => {
vi.stubGlobal('fetch', vi.fn());
const saved = process.env.PASSWORD;
diff --git a/src/app/api/search/route.ts b/src/app/api/search/route.ts
index be3b712..1a050ed 100644
--- a/src/app/api/search/route.ts
+++ b/src/app/api/search/route.ts
@@ -249,7 +249,9 @@ export async function POST(req: Request) {
sources.map(async (source, i) => {
const outcome = await searchSource(source, wd);
outcomes[i] = outcome;
- send({ type: 'source', ...outcome });
+ // 中间结果与最终结果共用去重和过滤规则,避免无关/成人条目短暂出现在页面上。
+ const { list } = aggregateOutcomes([outcome], wd, filterAdult);
+ send({ type: 'source', ...outcome, list });
})
);
diff --git a/src/lib/cms-parser.test.ts b/src/lib/cms-parser.test.ts
index ed04b1f..3da93c7 100644
--- a/src/lib/cms-parser.test.ts
+++ b/src/lib/cms-parser.test.ts
@@ -179,9 +179,41 @@ describe('isRelevant', () => {
expect(isRelevant('钢铁侠 Iron Man 2008', '钢铁侠 2008')).toBe(true);
});
- it('空关键词或空标题不参与过滤', () => {
+ it.each([
+ ['赛博朋克:边缘行者第二季', '赛博朋克:边缘行者 2'],
+ ['赛博朋克:边缘行者 2', '赛博朋克:边缘行者第二季'],
+ ['赛博朋克:边缘行者第2季', '赛博朋克:边缘行者第二季'],
+ ['示例剧 第十二季', '示例剧 12'],
+ ['示例剧 第二十季', '示例剧 第20季'],
+ ['赛博朋克:边缘行者第二季 花絮', '赛博朋克:边缘行者 2'],
+ ['赛博朋克:边缘行者第二季【影视解说】', '赛博朋克:边缘行者 2'],
+ ])('匹配明确的季数别名:%s / %s', (name, query) => {
+ expect(isRelevant(name, query)).toBe(true);
+ });
+
+ it.each([
+ ['赛博朋克:边缘行者第三季', '赛博朋克:边缘行者 2'],
+ ['赛博朋克:边缘行者第十二季', '赛博朋克:边缘行者 2'],
+ ['赛博朋克:边缘行者第20季', '赛博朋克:边缘行者 2'],
+ ['赛博朋克:边缘行者第十二季 花絮', '赛博朋克:边缘行者 2'],
+ ['赛博朋克:边缘行者第20季【影视解说】', '赛博朋克:边缘行者 2'],
+ ['赛博朋克:边缘行者第二季', '赛博朋克:边缘行者第二季 花絮'],
+ ['赛博朋克:边缘行者 20', '赛博朋克:边缘行者 2'],
+ ['赛博朋克:边缘行者12', '赛博朋克:边缘行者 2'],
+ ['电影第二季', '电影2'],
+ ['电影2', '电影第二季'],
+ ['钢铁侠 第二季', '钢铁侠 2008'],
+ ['钢铁侠 12008', '钢铁侠 2008'],
+ ['别的剧 第二季', '赛博朋克:边缘行者 2'],
+ ['10月17日 足球联赛', '赛博朋克:边缘行者 2'],
+ ])('拒绝无关标题、不同季数或数字的部分命中:%s / %s', (name, query) => {
+ expect(isRelevant(name, query)).toBe(false);
+ });
+
+ it('空关键词不参与过滤,但空标题不能命中非空关键词', () => {
expect(isRelevant('任意标题', '')).toBe(true);
- expect(isRelevant('', '任意')).toBe(true);
+ expect(isRelevant('', '任意')).toBe(false);
+ expect(isRelevant(' ! ', '任意')).toBe(false);
});
});
@@ -199,8 +231,15 @@ describe('filterRelevantResults', () => {
expect(out.map((i) => i.name)).toEqual(['摔跤吧!爸爸', '摔跤吧!爸爸【影视解说】']);
});
- it('全部不相关时回退为原列表', () => {
+ it('全部不相关时返回空列表,不能把源站的默认目录当成搜索结果', () => {
const junk = [{ name: '爸爸当家' }, { name: '爸爸别走' }];
- expect(filterRelevantResults(junk, wd)).toEqual(junk);
+ expect(filterRelevantResults(junk, wd)).toEqual([]);
+ });
+
+ it('过滤缺失标题,同时保留匹配的第二季而不保留无关体育目录', () => {
+ const relevant = { name: '赛博朋克:边缘行者第二季' };
+ expect(filterRelevantResults([
+ { name: '10月17日 足球联赛' }, {}, { name: '' }, relevant,
+ ], '赛博朋克:边缘行者 2')).toEqual([relevant]);
});
});
diff --git a/src/lib/cms-parser.ts b/src/lib/cms-parser.ts
index a49e753..afd8e1e 100644
--- a/src/lib/cms-parser.ts
+++ b/src/lib/cms-parser.ts
@@ -175,34 +175,100 @@ export function normalizeTitle(s: string): string {
return s.toLowerCase().replace(/[\s\p{P}\p{S}]+/gu, '');
}
+/** 仅将独立数字或明确的「第 N 季」后缀作为季数别名。 */
+function parseSeasonSuffix(title: string, allowAnnotations = false): { base: string; season: number; explicit: boolean } | undefined {
+ // 标题侧允许「第二季」后跟任意附注:粤语、预告片、【影视解说】、无分隔符的「第二季粤语」都算
+ const explicitPattern = allowAnnotations
+ ? /^(.*?)\s*第\s*([1-9]\d?|[一二两三四五六七八九十]{1,3})\s*季.*$/u
+ : /^(.*?)\s*第\s*([1-9]\d?|[一二两三四五六七八九十]{1,3})\s*季$/u;
+ const explicit = title.trim().match(explicitPattern);
+ const bare = title.trim().match(/^(.*?)\s+([1-9]\d?)$/u);
+ const match = explicit ?? bare;
+ if (!match || !normalizeTitle(match[1])) return undefined;
+ const value = match[2];
+ const digits = '零一二三四五六七八九';
+ let season = Number(value);
+ if (!Number.isFinite(season)) {
+ const chinese = value.replace('两', '二');
+ if (/^[一二三四五六七八九]$/u.test(chinese)) {
+ season = digits.indexOf(chinese);
+ } else if (/^[一二三四五六七八九]?十[一二三四五六七八九]?$/u.test(chinese)) {
+ const [tens, units] = chinese.split('十');
+ season = (tens ? digits.indexOf(tens) : 1) * 10 + (units ? digits.indexOf(units) : 0);
+ }
+ }
+ if (!Number.isFinite(season)) return undefined;
+ return { base: match[1], season, explicit: Boolean(explicit) };
+}
+
+/** 数字必须完整命中,避免第 2 季误匹配 12 或 20。 */
+function includesKeyword(title: string, keyword: string): boolean {
+ if (!title || !keyword) return false;
+ let index = title.indexOf(keyword);
+ while (index >= 0) {
+ const before = title[index - 1] ?? '';
+ const after = title[index + keyword.length] ?? '';
+ if ((!/^\d/u.test(keyword) || !/\d/u.test(before)) &&
+ (!/\d$/u.test(keyword) || !/\d/u.test(after))) return true;
+ index = title.indexOf(keyword, index + 1);
+ }
+ return false;
+}
+
/**
- * 相关性判断:结果标题是否真的命中了搜索词。
- * 部分采集站会做分词/OR 模糊搜索(搜「摔跤吧!爸爸」返回一堆「爸爸XXX」),
- * 逐条校验后只保留真正相关的结果。
- *
- * 规则:归一化后的标题包含归一化后的完整关键词;
- * 多词查询(空格分隔)时放宽为「每个词都命中」即可(如「钢铁侠 2008」)。
+ * 关键词只解析一次。
+ * filterRelevantResults 会对同一个查询过滤几百上千条结果,
+ * 关键词的归一化、季数解析、分词结果对每条都相同,没必要重复算。
*/
-export function isRelevant(name: string, wd: string): boolean {
+interface PreparedQuery {
+ normalized: string;
+ season?: { base: string; season: number; explicit: boolean };
+ tokens: string[];
+}
+
+function prepareQuery(wd: string): PreparedQuery | undefined {
+ const normalized = normalizeTitle(wd);
+ if (!normalized) return undefined;
+ return {
+ normalized,
+ season: parseSeasonSuffix(wd),
+ tokens: wd.trim().split(/\s+/).map(normalizeTitle).filter(Boolean),
+ };
+}
+
+function matchesPrepared(nName: string, query: PreparedQuery): boolean {
+ if (includesKeyword(nName, query.normalized)) return true;
+ return query.tokens.length > 1 && query.tokens.every((token) => includesKeyword(nName, token));
+}
+
+function isRelevantPrepared(name: string, query: PreparedQuery): boolean {
const nName = normalizeTitle(name);
- const nWd = normalizeTitle(wd);
- if (!nName || !nWd) return true; // 空关键词不参与过滤
- if (nName.includes(nWd)) return true;
- const tokens = wd.trim().split(/\s+/);
- if (tokens.length > 1) {
- return tokens.every((t) => {
- const n = normalizeTitle(t);
- return n && nName.includes(n);
- });
+ if (!nName) return false;
+ const titleSeason = parseSeasonSuffix(name, true);
+ if (query.season && titleSeason) {
+ if (query.season.season !== titleSeason.season) return false;
+ if (query.season.explicit || titleSeason.explicit) {
+ const baseQuery = prepareQuery(query.season.base);
+ if (baseQuery) return matchesPrepared(normalizeTitle(titleSeason.base), baseQuery);
+ }
}
- return false;
+ return matchesPrepared(nName, query);
}
/**
- * 结果相关性过滤:丢弃与关键词无关的条目。
- * 过滤后为空时回退到原列表——宁可多展示也不把源站唯一可用的结果清空。
+ * 忽略标点和大小写,多词查询要求每个词都命中。
+ * 独立的末尾数字也可匹配明确的中文季数后缀;
+ * 标题内部的数字与四位年份不作为季数别名。
*/
+export function isRelevant(name: string, wd: string): boolean {
+ const query = prepareQuery(wd);
+ if (!query) return true;
+ return isRelevantPrepared(name, query);
+}
+
+/** 丢弃无关结果;即使全部不相关,也不能返回源站的原始目录。 */
export function filterRelevantResults<T extends { name?: string }>(items: T[], wd: string): T[] {
- const filtered = items.filter((item) => isRelevant(item.name || '', wd));
- return filtered.length > 0 ? filtered : items;
+ const query = prepareQuery(wd);
+ if (!query) return items;
+ return items.filter((item) => isRelevantPrepared(item.name || '', query));
}