Match site redirects whose source has non-ASCII characters (#4689)

.changeset/redirect-non-ascii-source.md

@@ -0,0 +1,5 @@
+---
+"gitbook": patch
+---
+
+Site redirects whose source path has accented or non-Latin characters now redirect visitors instead of returning 404.

packages/gitbook/src/components/SitePage/fetch.test.ts

@@ -97,11 +97,11 @@ describe('fetchPageData', () => {
 
 describe('resolveMissingPagePath', () => {
     function createRedirectContext(options: {
-        siteRedirect?: { target: string; permanent?: boolean };
+        siteRedirect?: { source?: string; target: string; permanent?: boolean };
         spaceRedirectPageId?: string;
     }) {
         const getSiteRedirectBySource = mock(async ({ source }: { source: string }) =>
-            options.siteRedirect && source === '/old'
+            options.siteRedirect && source === (options.siteRedirect.source ?? '/old')
                 ? {
                       data: {
                           target: options.siteRedirect.target,
@@ -143,6 +143,24 @@ describe('resolveMissingPagePath', () => {
         });
     });
 
+    it('resolves a site redirect whose source has non-ASCII characters', async () => {
+        const context = createRedirectContext({
+            siteRedirect: {
+                source: '/de/kategorien-hinzuf%C3%BCgen',
+                target: 'https://docs.example.com/new',
+            },
+        });
+        const pathname = normalizeURL(
+            new URL('https://docs.example.com/de/kategorien-hinzufügen')
+        ).pathname.slice(1);
+
+        expect(await resolveMissingPagePath(context, pathname)).toEqual({
+            type: 'redirect',
+            destination: '/new',
+            permanent: false,
+        });
+    });
+
     it('resolves a space redirect to a page', async () => {
         const context = createRedirectContext({ spaceRedirectPageId: page.id });
 

packages/gitbook/src/components/SitePage/fetch.ts

@@ -186,7 +186,8 @@ export async function resolveMissingPagePath(
 
 /**
  * Transform a pathname into a candidate source for site redirect matching.
- * We also encode each segment to handle special characters in redirects.
+ * The pathname is already percent-encoded, so each segment is decoded before being encoded
+ * again: redirect sources store special characters percent-encoded exactly once.
  */
 function toSiteRedirectSourceCandidate(pathname: string): string {
     const normalized = withLeadingSlash(pathname);
@@ -194,11 +195,19 @@ function toSiteRedirectSourceCandidate(pathname: string): string {
         normalized
             .slice(1)
             .split('/')
-            .map((segment) => encodeURIComponent(segment))
+            .map((segment) => encodeURIComponent(safeDecodeURIComponent(segment)))
             .join('/')
     );
 }
 
+function safeDecodeURIComponent(segment: string): string {
+    try {
+        return decodeURIComponent(segment);
+    } catch {
+        return segment;
+    }
+}
+
 /**
  * Get the page path from the params.
  */