[gql_websocket_link] TransportWebSocketLink does not reconnect when graphql server disappears

[bverhagen]

I use the graphql-transport-ws as the connection Link for a Ferry-based grqphl subscription, pretty much as is done in Ferry examples, only with the http link replaced by a ws one:

   final link = Link.from([
      TransportWebSocketLink(
        TransportWsClientOptions(
            socketMaker:
                WebSocketMaker.url(() => _uri)),
      ),
    ]);
    final cache = Cache(possibleTypes: possibleTypesMap);
    final client = Client(link: link, cache: cache);

During development, I regularly restarted the graphql-server. As the README for gql_websocket_link reads: This link support autoReconnect and will resubscribe after reconnecting., I expected the gql_websocket_link-based graphql client to automatically reconnect again after such a restart. However, it does not seem to do that.

The graphql server only supports the graphql-transport-ws protocol. Looking further in the API reference, I am only able to find the autoReconnect configuration option for the WebSocketLink class. Does the TransportWebSocketLink already support a similar autoReconnect feature? If so, can you point me to an example or the relevant configuration options?

[knaeckeKami]

There are options related to auto reconnects in the TransportWsClientOptions class.
There is also a test that should verify to reconnect on a server restart ( https://github.com/gql-dart/gql/blob/master/links/gql_websocket_link/test/gql_websocket_link_test.dart#L1097 ).

[catapop84]

I can confirm this is not working. Here is the error thrown:

E/flutter (14348): [ERROR:flutter/runtime/dart_vm_initializer.cc(41)] Unhandled Exception: Null check operator used on a null value
E/flutter (14348): #0      _ConnectionState._startConnecting.<anonymous closure>.<anonymous closure> (package:gql_websocket_link/src/graphql_transport_ws.dart:853:45)
E/flutter (14348): #1      _RootZone.runGuarded (dart:async/zone.dart:1582:10)
E/flutter (14348): #2      _BufferingStreamSubscription._sendDone.sendDone (dart:async/stream_impl.dart:392:13)
E/flutter (14348): #3      _BufferingStreamSubscription._sendDone (dart:async/stream_impl.dart:402:7)
E/flutter (14348): #4      _BufferingStreamSubscription._close (dart:async/stream_impl.dart:291:7)
E/flutter (14348): #5      _ForwardingStream._handleDone (dart:async/stream_pipe.dart:99:10)
E/flutter (14348): #6      _ForwardingStreamSubscription._handleDone (dart:async/stream_pipe.dart:161:13)
E/flutter (14348): #7      _RootZone.runGuarded (dart:async/zone.dart:1582:10)
E/flutter (14348): #8      _BufferingStreamSubscription._sendDone.sendDone (dart:async/stream_impl.dart:392:13)
E/flutter (14348): #9      _BufferingStreamSubscription._sendDone (dart:async/stream_impl.dart:402:7)
E/flutter (14348): #10     _BufferingStreamSubscription._close (dart:async/stream_impl.dart:291:7)
E/flutter (14348): #11     _SyncStreamControllerDispatch._sendDone (dart:async/stream_controller.dart:792:19)
E/flutter (14348): #12     _StreamController._closeUnchecked (dart:async/stream_controller.dart:647:7)
E/flutter (14348): #13     _StreamController.close (dart:async/stream_controller.dart:640:5)
E/flutter (14348): #14     _RootZone.run (dart:async/zone.dart:1655:54)
E/flutter (14348): #15     _FutureListener.handleWhenComplete (dart:async/future_impl.dart:199:18)
E/flutter (14348): #16     Future._propagateToListeners.handleWhenCompleteCallback (dart:async/future_impl.dart:810:39)
E/flutter (14348): #17     Future._propagateToListeners (dart:async/future_impl.dart:866:11)
E/flutter (14348): #18     Future._completeWithValue (dart:async/future_impl.dart:641:5)
E/flutter (14348): #19     Future._asyncCompleteWithValue.<anonymous closure> (dart:async/future_impl.dart:715:7)
E/flutter (14348): #20     _microtaskLoop (dart:async/schedule_microtask.dart:40:21)
E/flutter (14348): #21     _startMicrotaskLoop (dart:async/schedule_microtask.dart:49:5)

[catapop84]

I found the error. is in this code:

        onDone: () => onClose?.call(
          socket.closeCode == null
              ? "DONE"
              : LikeCloseEvent(
                  code: socket.closeCode!,
                  reason: socket.closeReason!,
                ),
        ),

It assumes that if socket.closeCode is not null then both socket.closeCode and socket.closeReason are available but that's not the case. If I stop the server program I receive code:1002 and reason = null

to fix this you should either make reason optional in LikeCloseEvent or put an empty string if closeReason is null.
Something like this

          socket.closeCode == null
              ? "DONE"
              : LikeCloseEvent(
                  code: socket.closeCode!, // socket.closeCode is 1002
                  reason: socket.closeReason??'', // socket.closeReason is null
                ),

[bverhagen]

@knaeckeKami : Will this be taken up?

[catapop84]

@bverhagen I made this PR that fixes the problem: #418 and we already use it in closed beta with some clients. No complains so far.

You'll need to configure it like this

    final link = TransportWebSocketLink(
      TransportWsClientOptions(
        socketMaker: wsMaker,
        lazy: true,
        keepAlive: const Duration(seconds: 50),
        retryAttempts: 20,
        retryWait: (retries) async {
          log("number of retries: $retries");
          await Future.delayed(const Duration(seconds: 5));
        },
        shouldRetry: (errOrCloseEvent) {
          return true;
        },
        isFatalConnectionProblem: (errOrCloseEvent) {
          return false;
        },
      ),
    );

Basically this PR adds shouldRetry that will replace the deprecated isFatalConnectionProblem (If javascript library is a refference)
You can add your login on shouldRetry based on errOrCloseEvent parameter.

[knaeckeKami]

@knaeckeKami : Will this be taken up?

Yes. Sorry, I have a lot going on at the moment, so triage is slow. Will get better soon.

In the mean time, you can use the fork from the PR.