fix(articles): 301 junk-tailed article URLs and 404 unknown slugs

middleware.ts

@@ -3,6 +3,8 @@ import { NextRequest, NextResponse } from 'next/server';
 const PUBLIC_FILE = /\.(.*)$/;
 const supportedLocales = ['en', 'ru', 'hy'];
 const defaultLocale = 'en';
+const ARTICLE_PATH = /^(?:\/(?:en|ru|hy))?\/articles\/([^/]+)\/?$/;
+const ARTICLE_SLUG = /^[A-Za-z0-9_-]+$/;
 
 export async function middleware(request: NextRequest) {
   const { pathname } = request.nextUrl;
@@ -30,6 +32,19 @@ export async function middleware(request: NextRequest) {
     return NextResponse.redirect(new URL(`/${matchedLocale}`, request.url));
   }
 
+  // Crawlers follow article URLs with a stray encoded tail (`slug%29`,
+  // `slug%252529`, ...) that grows by one encoding per hop. Article slugs are
+  // [A-Za-z0-9_-] only, so anything after that prefix is cut off with a 301.
+  const article = ARTICLE_PATH.exec(pathname);
+  if (article && !ARTICLE_SLUG.test(article[1])) {
+    const slug = /^[A-Za-z0-9_-]+/.exec(article[1])?.[0];
+    if (slug) {
+      const url = request.nextUrl.clone();
+      url.pathname = `/articles/${slug}`;
+      return NextResponse.redirect(url, 301);
+    }
+  }
+
   return NextResponse.next();
 }
 

src/pages/articles/[page].tsx

@@ -163,8 +163,10 @@ export const getStaticProps: GetStaticProps = async ({
   const articles: TArticle[] = await getArticles(locale === 'ru' ? 'ru' : 'en');
   const article = page && getCurrentArticle(articles, page);
 
-  if (!article) {
-    return { notFound: true };
+  // getCurrentArticle returns {} on a miss, so an unknown slug must be caught
+  // by its keys: rendering it as a 200 page let crawlers loop on junk URLs.
+  if (!article || !Object.keys(article).length) {
+    return { notFound: true, revalidate: 10 };
   }
 
   const recommendedArticles = getRecommendedArticles(articles, page);