Requests should be cached on a per-datasource level

[kyri-petrou]

There is a rather annoying bug/issue currently with how we dedup/cache queries. The current Cache takes into consideration only the Request, but not the DataSource. This means that if a Request is reused across different DataSources.

For most usecases this shouldn't be a problem, because having multiple DataSources returning the same type is unlikely. However, this is something that has happened to me before and failed to realise until it hit prod:

  case class EntityId(id: Int) extends Request[Nothing, Boolean]

  val userExistsDs: DataSource[Any, EntityId] =
    DataSource.fromFunctionBatchedZIO[Any, Nothing, EntityId, Boolean]("UserExistsDS") { requests =>
      ???
    }

  val organisationExistsDs: DataSource[Any, EntityId] =
    DataSource.fromFunctionBatchedZIO[Any, Nothing, EntityId, Boolean]("OrganizationExistsDs") { requests =>
      ???
    }

In this case, if we execute a query against both datasources where the id exists in both of them, we would be retrieving the result from the cache even if the request was meant for a different datasource.

While the workaround is simple (create another case class for the 2nd case class), I personally believe that the existing behaviour is not correct. If I'm executing a request against a specific DataSource, I expect it to be cached on the datasource that I am making a request against.

Unfortunately, the only way to fix this is by breaking the Cache API by adding DataSource as a parameter to its methods. @paulpdaniels AFAICT you were the one that asked for the Cache to be made abstract, do you see any issues with this? (assuming you ended up using a custom Cache, otherwise ignore me 😅)

[kyri-petrou]

This test is irrelevant with the PR. It's something I've been meaning to add for some time now just as a safeguard when making changes to batching logic