UHF-8837: remove pages with X-Robots-Tag: noindex from sitemap

helfi_proxy.module

@@ -82,3 +82,38 @@ function helfi_proxy_page_attachments_alter(array &$attachments) {
     $attachments['#attached']['html_head'][] = [$helfi_content_type, $tag_name];
   }
 }
+
+/**
+ * Implements hook_page_attachments_alter().
+ */
+function helfi_proxy_simple_sitemap_links_alter(array &$links, $sitemap_variant) {
+  /** @var \Drupal\Core\Config\ImmutableConfig $config */
+  $config = \Drupal::service('config.factory')->get('helfi_proxy.settings');
+
+  if (!$paths = implode("\n", $config->get(ProxyManagerInterface::ROBOTS_PATHS) ?? [])) {
+    return;
+  }
+
+  /** @var \Drupal\helfi_api_base\Environment\Environment $environment */
+  $environment = \Drupal::service('helfi_api_base.environment_resolver')->getActiveEnvironment();
+
+  /** @var \Drupal\Core\Path\PathMatcherInterface $pathMatcher */
+  $pathMatcher = \Drupal::service('path.matcher');
+
+  // helfi_proxy module sets "X-Robots-Tag: noindex" header for configured
+  // paths. These url should not be included in the sitemap.xml file.
+  foreach ($links as $key => $link) {
+    $baseUrl = $environment->getUrl($link['langcode']);
+    $url = $link['url'];
+
+    if (str_starts_with($url, $baseUrl)) {
+      $path = substr($url, strlen($baseUrl));
+
+      // Remove matched paths from sitemap.xml file.
+      if ($pathMatcher->matchPath($path, $paths)) {
+        unset($links[$key]);
+      }
+    }
+  }
+
+}