feat: transaction batcher module

[samuelAndalon]

📝 Description

transaction-batcher module is an alternative to data-loaders, using reactive stream to apply batching and deduplication of transactions needed by data fetchers, the implementation is completely agnostic of GraphQL and compatible with any asynchronous approach used to resolve data.
it also supports caching to avoid making a request to a previous completed transaction.

🔗 Related Issues

[dariuszkuc]

@samuelAndalon can you add copyright to all new source files?

[dariuszkuc]

why are we suppressing this warning? is this valid?

[samuelAndalon]

it is valid, given that Subscriber is an interface, so it's valid to implement it, However a warning appears in the IDE when trying to do it.

i can see a lot of Subscriber implementations that need to add that suppress as well
https://github.com/search?p=2&q=%22ReactiveStreamsSubscriberImplementation%22&type=Code

[dariuszkuc]

unsure, I am not getting any warning in my local

[samuelAndalon]

[rocketraman]

Why does this use reactive-streams instead of kotlinx-coroutines Flow? Flow can be transformed to reactive-streams by users, if desired.

[dariuszkuc]

We decided to use generic reactive streams Publisher for flexibility to allow users to use various implementations (including coroutines).

[samuelAndalon]

reactive streams and all his implementations like project-reactor have 100% interoperability with other async technologies like Java Flow, coroutines, completableFuture etc.
So the most generic choice was RX, at the beginning we were scoping for java flow, but that requires target JVM to be 9.

[rocketraman]

Ok. Given this is a Kotlin project, I would have thought coroutines Flow would be the obvious choice as first-party implementation. As stated, it is easy to convert from Flow to other reactive streams implementations i.e. flow.asPublisher().

Is there any documentation in the works for this feature? Seems really interesting.

[samuelAndalon]

Once we add the instrumentation that will signal when to dispatch, we will add the documentation.
It will be pretty much similar to the DataLoaderDispatcherInstrumentation from graphql-java but applied to more than 1 operation if client sent a request with batched operations.

for now you can refer to the unit tests