Generate a split sitemap (also fix robots.txt) (#4639)

news/4638.bugfix

@@ -0,0 +1 @@
+Generate a split sitemap @reebalazs

src/express-middleware/sitemap.js

@@ -1,11 +1,34 @@
 import express from 'express';
-import { generateSitemap } from '@plone/volto/helpers';
+import {
+  generateSitemap,
+  generateSitemapIndex,
+  SITEMAP_BATCH_SIZE,
+} from '@plone/volto/helpers/Sitemap/Sitemap';
 
 export const sitemap = function (req, res, next) {
-  generateSitemap(req).then((sitemap) => {
+  let start = 0;
+  let size = undefined;
+  const { batch: batchStr } = req.params;
+  if (batchStr !== undefined) {
+    const batch = parseInt(batchStr);
+    if (isNaN(batch) || batch === 0 || '' + batch !== batchStr) {
+      res.status(404);
+      // Some data, such as the internal API address, may be sensitive to be published
+      res.send(
+        `Invalid sitemap name, use sitemap.xml.gz, or batched sitemapN.xml.gz where N is a positive integer.`,
+      );
+      return;
+    }
+    start = SITEMAP_BATCH_SIZE * (batch - 1);
+    size = SITEMAP_BATCH_SIZE;
+  }
+  generateSitemap(req, start, size).then((sitemap) => {
     if (Buffer.isBuffer(sitemap)) {
       res.set('Content-Type', 'application/x-gzip');
-      res.set('Content-Disposition', 'attachment; filename="sitemap.xml.gz"');
+      res.set(
+        'Content-Disposition',
+        `attachment; filename="sitemap${batchStr || ''}.xml.gz"`,
+      );
       res.send(sitemap);
     } else {
       // {"errno":-111, "code":"ECONNREFUSED", "host": ...}
@@ -16,10 +39,20 @@ export const sitemap = function (req, res, next) {
   });
 };
 
+export const sitemapIndex = function (req, res, next) {
+  generateSitemapIndex(req).then((sitemapIndex) => {
+    res.set('Content-Type', 'application/xml');
+    res.set('Content-Disposition', 'attachment; filename="sitemap-index.xml"');
+    res.send(sitemapIndex);
+  });
+};
+
 export default function () {
   const middleware = express.Router();
 
   middleware.all('**/sitemap.xml.gz', sitemap);
+  middleware.all('**/sitemap:batch.xml.gz', sitemap);
+  middleware.all('**/sitemap-index.xml', sitemapIndex);
   middleware.id = 'sitemap.xml.gz';
   return middleware;
 }

src/helpers/Robots/Robots.js

@@ -15,12 +15,9 @@ import { addHeadersFactory } from '@plone/volto/helpers/Proxy/Proxy';
  */
 export const generateRobots = (req) =>
   new Promise((resolve) => {
-    //const url = `${req.protocol}://${req.get('Host')}`;
-    const request = superagent.get(
-      `${
-        config.settings.internalApiPath ?? config.settings.apiPath
-      }/robots.txt`,
-    );
+    const internalUrl =
+      config.settings.internalApiPath ?? config.settings.apiPath;
+    const request = superagent.get(`${internalUrl}/robots.txt`);
     request.set('Accept', 'text/plain');
     const authToken = req.universalCookies.get('auth_token');
     if (authToken) {
@@ -31,6 +28,12 @@ export const generateRobots = (req) =>
       if (error) {
         resolve(text || error);
       } else {
+        // Plone has probably returned the sitemap link with the internal url.
+        // If so, let's replace it with the current one.
+        const url = `${req.protocol}://${req.get('Host')}`;
+        text = text.replace(internalUrl, url);
+        // Replace the sitemap with the sitemap index.
+        text = text.replace('sitemap.xml.gz', 'sitemap-index.xml');
         resolve(text);
       }
     });

src/helpers/Sitemap/Sitemap.js

@@ -11,19 +11,23 @@ import { addHeadersFactory } from '@plone/volto/helpers/Proxy/Proxy';
 
 import config from '@plone/volto/registry';
 
+export const SITEMAP_BATCH_SIZE = 5000;
+
 /**
  * Generate sitemap
  * @function generateSitemap
  * @param {Object} _req Request object
  * @return {string} Generated sitemap
  */
-export const generateSitemap = (_req) =>
+export const generateSitemap = (_req, start = 0, size = undefined) =>
   new Promise((resolve) => {
     const { settings } = config;
     const APISUFIX = settings.legacyTraverse ? '' : '/++api++';
     const apiPath = settings.internalApiPath ?? settings.apiPath;
     const request = superagent.get(
-      `${apiPath}${APISUFIX}/@search?metadata_fields=modified&b_size=100000000&use_site_search_settings=1`,
+      `${apiPath}${APISUFIX}/@search?metadata_fields=modified&b_start=${start}&b_size=${
+        size !== undefined ? size : 100000000
+      }&use_site_search_settings=1`,
     );
     request.set('Accept', 'application/json');
     request.use(addHeadersFactory(_req));
@@ -50,3 +54,41 @@ export const generateSitemap = (_req) =>
       }
     });
   });
+
+/**
+ * Generate sitemap
+ * @function generateSitemapIndex
+ * @param {Object} _req Request object
+ * @return {string} Generated sitemap index
+ */
+export const generateSitemapIndex = (_req) =>
+  new Promise((resolve) => {
+    const { settings } = config;
+    const APISUFIX = settings.legacyTraverse ? '' : '/++api++';
+    const apiPath = settings.internalApiPath ?? settings.apiPath;
+    const request = superagent.get(
+      `${apiPath}${APISUFIX}/@search?metadata_fields=modified&b_size=0&use_site_search_settings=1`,
+    );
+    request.set('Accept', 'application/json');
+    const authToken = _req.universalCookies.get('auth_token');
+    if (authToken) {
+      request.set('Authorization', `Bearer ${authToken}`);
+    }
+    request.end((error, { body } = {}) => {
+      if (error) {
+        resolve(body || error);
+      } else {
+        const items = Array.from(
+          { length: Math.ceil(body.items_total / SITEMAP_BATCH_SIZE) },
+          (_, i) =>
+            `  <sitemap>
+    <loc>${toPublicURL('/sitemap' + (i + 1) + '.xml.gz')}</loc>
+  </sitemap>`,
+        );
+        const result = `<?xml version="1.0" encoding="UTF-8"?>
+<sitemapindex xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
+${items.join('\n')}\n</sitemapindex>`;
+        resolve(result);
+      }
+    });
+  });