Jannis Pohlmann

Agreed. The tests are non-exhaustive and I've only tried this with https://github.com/graphprotocol/graph-cli. My feeling is that unless there is a global CLI spec built from all commands and their options,...

May I propose https://thegraph.com / https://github.com/graphprotocol/graph-node? We leave the mutation part of GraphQL up to blockchains etc. but we are generating a GraphQL API (from schemas that only contain types)...

Could the `orderBy` input type be a union, like this? ```graphql orderBy: UserOrderByInput! | [UserOrderByInput!]! ``` Assuming GraphiQL and other tools handle this right, they would still provide auto-completion for...

You're referring to https://github.com/apollographql/subscriptions-transport-ws/blob/master/PROTOCOL.md#gql_data I assume. I agree, this is something we are not doing yet but should do.

The reason I haven't merged the `subscription-listener` branch yet is that there already is way to achieve the same thing: by wrapping the default `subscriptionManager` in your own `SubscriptionManager` interface...

@ccamel That sounds like a good approach. It will force people to think about why they'd return `nil` and we can add a doc string pointing at the spec.

An alternative approach could be to give developers a way to invalidate the user from the outside whenever they want. The reasoning here is that the lifetime of a token...

I'm writing this from a beach chair while wearing swimming trousers and eating ice cream. The docstring approach seems like a bad idea for the downside you mentioned. As long...

Also odd: the timeout passed to `waitForTransaction` is `240000`. This is also the time that `waitForTransaction` claims to have timed out with. _But_ this happens after 10 minutes, not 4...

I did some more experimentation today by setting the `waitForTransaction` timeout to `0`. The resulting behavior (waiting for 3 confirmations): transactions are always detected as being mined and having 3...