feat: Classify exchanges when loader changes

Author: route

CHANGELOG.md

@@ -6,9 +6,14 @@
 - `Ferrum::Network::Request#headers` are enhanced and supplemented with `Network.requestWillBeSentExtraInfo` [#506]
 - `Ferrum::Page#off` to unsubscribe from CDP events [#455]
 - `Ferrum::Mouse#scroll_by` to be able to scroll by, as alternative to `scroll_to` [#514]
+- `Ferrum::Network::Exchange#unknown` determines if the exchange is in an unknown state, meaning that browser might not return info about it [#426]
+- `Ferrum::Network::Exchange#loader_id` returns loader id [#426]
 
 ### Changed
 
+- `Ferrum::Network::Exchange#finished?` takes into account that exchange might be in an unknown state after loader changed(next `go_to` call) and considers
+  such as finished [#426]
+
 ### Fixed
 
 - Correctly set mouse events buttons property [#509]

lib/ferrum/network.rb

@@ -86,11 +86,11 @@ def idle?(connections = 0)
     end
 
     def total_connections
-      exchange_connections.count
+      @traffic.size
     end
 
     def finished_connections
-      exchange_connections.count(&:finished?)
+      @traffic.count(&:finished?)
     end
 
     def pending_connections
@@ -100,7 +100,7 @@ def pending_connections
     #
     # Page request of the main frame.
     #
-    # @return [Request]
+    # @return [Request, nil]
     #
     # @example
     #   browser.go_to("https://github.com/")
@@ -391,7 +391,7 @@ def subscribe_request_will_be_sent
 
         if exchange.navigation_request?(@page.main_frame.id)
           @exchange = exchange
-          mark_pending_exchanges_as_unknown(exchange)
+          classify_pending_exchanges(exchange.loader_id)
         end
       end
 
@@ -403,17 +403,6 @@ def subscribe_request_will_be_sent
       end
     end
 
-    # When the main frame navigates Chrome doesn't send `Network.loadingFailed`
-    # for pending async requests. Therefore, we mark pending connections as unknown since
-    # they are not relevant to the current navigation.
-    def mark_pending_exchanges_as_unknown(navigation_exchange)
-      @traffic.each do |exchange|
-        break if exchange.id == navigation_exchange.id
-
-        exchange.unknown = true if exchange.pending?
-      end
-    end
-
     def subscribe_response_received
       @page.on("Network.responseReceived") do |params|
         exchange = select(params["requestId"]).last
@@ -514,15 +503,16 @@ def whitelist?
       Array(@whitelist).any?
     end
 
-    def exchange_connections
-      @traffic.select { |e| exchange_connection?(e) }
-    end
-
-    def exchange_connection?(exchange)
-      return false if !@page.frames.any? { |f| f.id == exchange.request&.frame_id }
-      return false if exchange.request&.frame_id != @exchange.request&.frame_id
+    # When the main frame navigates Chrome doesn't send `Network.loadingFailed`
+    # for pending async requests. Therefore, we mark pending connections as unknown since
+    # they are not relevant to the current navigation.
+    def classify_pending_exchanges(new_loader_id)
+      @traffic.each do |exchange|
+        break if exchange.loader_id == new_loader_id
+        next unless exchange.pending?
 
-      exchange.request&.loader_id == @exchange.request&.loader_id
+        exchange.unknown = true
+      end
     end
   end
 end

lib/ferrum/network/exchange.rb

@@ -65,6 +65,15 @@ def navigation_request?(frame_id)
         request&.type?(:document) && request&.frame_id == frame_id
       end
 
+      #
+      # The loader ID of the request.
+      #
+      # @return [String, nil]
+      #
+      def loader_id
+        request&.loader_id
+      end
+
       #
       # Determines if the network exchange has a request.
       #

lib/ferrum/page/frames.rb

@@ -124,7 +124,7 @@ def subscribe_frame_stopped_loading
         on("Page.frameStoppedLoading") do |params|
           # `DOM.performSearch` doesn't work without getting #document node first.
           # It returns node with nodeId 1 and nodeType 9 from which descend the
-          # tree and we save it in a variable because if we call that again root
+          # tree, and we save it in a variable because if we call that again root
           # node will change the id and all subsequent nodes have to change id too.
           if @main_frame.id == params["frameId"]
             @event.set if idling?

spec/network/exchange_spec.rb

@@ -185,6 +185,22 @@
       )
       expect(last_exchange.pending?).to be true
     end
+
+    it "classifies request if loader changes" do
+      page.go_to("/ferrum/with_slow_ajax_connection")
+      xhr = page.network.traffic.find(&:pending?)
+
+      expect(page.body).to include("Slow AJAX")
+      expect(xhr.url).to include("really_slow")
+      expect(xhr.unknown).to be_falsey
+      expect(xhr).to be_pending
+
+      page.go_to("/")
+
+      expect(page.body).to include("Hello world!")
+      expect(xhr.unknown).to be_truthy
+      expect(xhr).not_to be_pending
+    end
   end
 
   describe "#intercepted?" do