tlux / graphql_ws_client Goto Github PK

View Code? Open in Web Editor NEW

3.0 3.0 0.0 278 KB

An Elixir client for connecting with GraphQL over Websockets following the graphql-ws conventions

License: MIT License

Elixir 100.00%

graphql graphql-ws websocket

graphql_ws_client's People

Contributors

Stargazers

Watchers

graphql_ws_client's Issues

prevent `Driver.parse_message/2` being called without `Conn`

     18:04:32.606 [error] GenServer #PID<0.830.0> terminating
     ** (FunctionClauseError) no function clause matching in GraphQLWSClient.Driver.parse_message/2
         (graphql_ws_client 0.1.1) lib/graphql_ws_client/driver.ex:82: GraphQLWSClient.Driver.parse_message(nil, {:gun_error, #PID<0.831.0>, #Reference<0.1050648357.2088501250.11223>, {:closed, :normal}})
         (graphql_ws_client 0.1.1) lib/graphql_ws_client.ex:509: GraphQLWSClient.handle_info/2
         (connection 1.1.0) lib/connection.ex:810: Connection.handle_async/3
         (stdlib 4.3) gen_server.erl:1123: :gen_server.try_dispatch/4
         (stdlib 4.3) gen_server.erl:1200: :gen_server.handle_msg/6
         (stdlib 4.3) proc_lib.erl:240: :proc_lib.init_p_do_apply/3
     Last message: {:gun_error, #PID<0.831.0>, #Reference<0.1050648357.2088501250.11223>, {:closed, :normal}}

Halt on GraphQLError in Stream

At the moment, the iterator exits which probably is a bit error prone. Instead we should rather halt and log the error (the latter is something we are already doing)

Add option `:resubscribe_on_reconnect`

When the connection is terminated by the server, all listeners get notified that the socket has been closed. However, in some cases (for example if the client is running as part of the application supervision tree) it may be desirable to automatically restore all subscriptions when the connected is reopened. That means, we have to store all subscription queries and the associated listener processes in the client state and re-send all subscriptions right after the connection is restored.

Hence, I would like to introduce the :resubscribe_on_reconnect config option.

We need to keep in mind, that graphq-ws sockets prohibit to reuse a subscription ID unless the old subscription was removed. So we should make sure that the old subscription is removed before resubscribing with the same ID.

Improve query timeout handling

At the moment a query timeout will lead to reconnection. Instead we should remove the timed out query from the state, reply to the caller that the call timed out (%SocketError{cause: :timeout}) and ignore any messages that arrive after the timeout. We should probably also explicitly unsubscribe after a read timeout by sending a complete message to the socket.

Subscriptions in constrast should never time out.

{:gun_down, #PID<0.831.0>, :ws, :normal, [#Reference<0.2401964237.290979843.123374>]}

Prevent open to be callable multiple times

Resubscribe automatically when connection lost

Add option :resubscribe_on_reconnect
Resubscribe only when disconnect occurred due to an error
Alternatively, notify all subscriptions that the connection was lost so that they can resubscribe manually

tlux / graphql_ws_client Goto Github PK

graphql_ws_client's People

Contributors

Stargazers

Watchers

graphql_ws_client's Issues

prevent `Driver.parse_message/2` being called without `Conn`

Halt on GraphQLError in Stream

Add option `:resubscribe_on_reconnect`

Improve query timeout handling

Add `open` function that allows passing a payload for `connection_init`

Use only one monitor per listener

Allow passing connect options to gun driver

:gun_down message not handled

Prevent open to be callable multiple times

Resubscribe automatically when connection lost

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent