fix taxila surf scraper

lib/ingestors/taxila/surf_ingestor.rb

@@ -28,41 +28,35 @@ def read(url)
       private
 
       def process_surf(url)
-        Hash.from_xml(Nokogiri::XML(open_url(url, raise: true)).to_s)['urlset']['url'].each do |event_page|
-          next unless event_page['loc'].include?('/en/agenda/')
+        ical_surf_url = "https://www.surf.nl/ical/surf-agenda.ics"
+        ical_events = Icalendar::Event.parse(open_url(ical_surf_url, raise: true).set_encoding('utf-8'))
+        events = {}
+        ical_events.each do |ical_event|
+          title = ical_event.summary.to_s
+          events[title] ||= OpenStruct.new
+          events[title].title = title
+          events[title].url = "https://www.surf.nl/agenda##{title.parameterize(separator: '_')}"
+          events[title].description ||= ical_event.description.to_s
+          my_start = Time.zone.parse(ical_event.dtstart.strftime('%a, %d %b %Y %H:%M:%S'))
+          my_end = Time.zone.parse(ical_event.dtend.strftime('%a, %d %b %Y %H:%M:%S'))
+          events[title].start ||= my_start
+          events[title].end ||= my_end
+          events[title].set_default_times
+          events[title].venue ||= ical_event.location
+          events[title].source ||= 'SURF'
+          events[title].timezone ||= 'Amsterdam'
 
-          sleep(1) unless Rails.env.test? and File.exist?('test/vcr_cassettes/ingestors/surf.yml')
-          data_json = Nokogiri::HTML5.parse(open_url(event_page['loc'], raise: false))&.css('script[type="application/ld+json"]')
-          next unless data_json.present? && data_json.length > 0
-
-          data = JSON.parse(data_json.first.text)
-          begin
-            # create new event
-            event = OpenStruct.new
-
-            # extract event details from
-            attr = data['@graph'].first
-            event.title = convert_title attr['name']
-            event.url = attr['url']&.strip
-            event.description = convert_description attr['description']
-            event.start = attr['startDate']
-            event.end = attr['endDate']
-            event.set_default_times
-            event.venue = if attr['location'].is_a?(Array)
-                            attr['location'].join(' - ')
-                          else
-                            attr['location']
-                          end
-            event.source = 'SURF'
-            event.online = true
-            event.timezone = 'Amsterdam'
-            event.target_audience = parse_audience(event.description)
-
-            # add event to events array
-            add_event(event)
-          rescue Exception => e
-            @messages << "Extract event fields failed with: #{e.message}"
-          end
+          events[title].start = [my_start, events[title].start].min
+          events[title].end = [my_end, events[title].end].max
+        rescue Exception => e
+          puts e
+          @messages << "Extract event fields failed with: #{e.message}"
+        end
+        events.values.each do |event|
+          add_event(event)
+        rescue Exception => e
+          puts e
+          @messages << "Extract event fields failed with: #{e.message}"
         end
       end
     end

test/unit/ingestors/taxila/surf_ingestor_test.rb

@@ -14,20 +14,20 @@ class SurfIngestorTest < ActiveSupport::TestCase
 
   test 'can ingest events from surf' do
     source = @content_provider.sources.build(
-      url: 'https://www.surf.nl/sitemap.xml',
+      url: 'https://www.surf.nl/agenda',
       method: 'surf',
       enabled: true
     )
 
     ingestor = Ingestors::Taxila::SurfIngestor.new
 
     # check event doesn't
-    new_title = 'Master class Privacy Assessment Framework'
-    new_url = 'https://www.surf.nl/en/agenda/masterclass-review-framework-privacy-apr-1'
+    new_title = 'SURF Onderwijsdagen 2026'
+    new_url = 'https://www.surf.nl/agenda#surf_onderwijsdagen_2026'
     refute Event.where(title: new_title, url: new_url).any?
 
     # run task
-    assert_difference 'Event.count', 52 do
+    assert_difference 'Event.count', 20 do
       freeze_time(2019) do
         VCR.use_cassette('ingestors/surf') do
           ingestor.read(source.url)
@@ -36,9 +36,9 @@ class SurfIngestorTest < ActiveSupport::TestCase
       end
     end
 
-    assert_equal 52, ingestor.events.count
+    assert_equal 20, ingestor.events.count
     assert ingestor.materials.empty?
-    assert_equal 52, ingestor.stats[:events][:added]
+    assert_equal 20, ingestor.stats[:events][:added]
     assert_equal 0, ingestor.stats[:events][:updated]
     assert_equal 0, ingestor.stats[:events][:rejected]
 
@@ -51,8 +51,9 @@ class SurfIngestorTest < ActiveSupport::TestCase
     # check other fields
     assert_equal 'Amsterdam', event.timezone
     assert_equal 'SURF', event.source
-    assert event.online?
-    assert_equal Time.zone.parse('Thu, 25 Apr 2024 12:00:00.000000000 UTC +00:00'), event.start
-    assert_equal Time.zone.parse('Thu, 25 Apr 2024 12:00:00.000000000 UTC +00:00'), event.end
+    refute event.online?
+    assert_equal Time.zone.parse('Wed, 10 Nov 2026 09:00:00.000000000 UTC +00:00'), event.start
+    assert_equal Time.zone.parse('Thu, 11 Nov 2026 17:00:00.000000000 UTC +00:00'), event.end
+    assert_equal "Amare, Spuiplein 150, 2511 DG Den Haag", event.venue
   end
 end