Cache /search/count requests in the browser

These requests are typically loaded by the same user repeatedly as they
switch tabs on the search page. Considering that Elasticsearch indexing
is only updated every 1-2 minutes anyway the 1 minute browser side cache
shouldn't really be noticeable to users except for loading counts
faster.

This is also using a `private` cache (default for `expires_in`) so that
it is not cached by intermediate proxies. This is important since each
user's permissions is considered in a search.

app/controllers/search_controller.rb

@@ -49,6 +49,12 @@ class SearchController < ApplicationController
     scope = search_service.scope
     count = search_service.search_results.formatted_count(scope)
 
+    # Users switching tabs will keep fetching the same tab counts so it's a
+    # good idea to cache in their browser just for a short time. They can still
+    # clear cache if they are seeing an incorrect count but inaccurate count is
+    # not such a bad thing.
+    expires_in 1.minute
+
     render json: { count: count }
   end
 

changelogs/unreleased/300150-cache-search-tab-counters.yml

+---
+title: Cache /search/count requests in the browser
+merge_request: 55036
+author:
+type: performance

spec/controllers/search_controller_spec.rb

@@ -252,6 +252,14 @@ RSpec.describe SearchController do
           get :count, params: { search: 'hello' }
         end.to raise_error(ActionController::ParameterMissing)
       end
+
+      it 'sets private cache control headers' do
+        get :count, params: { search: 'hello', scope: 'projects' }
+
+        expect(response).to have_gitlab_http_status(:ok)
+
+        expect(response.headers['Cache-Control']).to include('max-age=60, private')
+      end
     end
 
     describe 'GET #autocomplete' do