Change Effect to Sync in FileService #3124
Conversation
| @@ -6,7 +6,7 @@ import cats.effect._ | |||
|
|
|||
| /** Cache strategy that doesn't cache anything, ever. */ | |||
| class NoopCacheStrategy[F[_]] extends CacheStrategy[F] { | |||
| override def cache(uriPath: String, resp: Response[F])(implicit F: Effect[F]): F[Response[F]] = | |||
| override def cache(uriPath: String, resp: Response[F])(implicit F: Sync[F]): F[Response[F]] = | |||
There was a problem hiding this comment.
Isn't this one just Applicative?
There was a problem hiding this comment.
Ah, no, because of the trait. Perhaps we could move this constraint to a constructor for the strategy.
There was a problem hiding this comment.
You mean remove it from CacheStrategy.cache? Should I do it in this PR?
There was a problem hiding this comment.
If it's on CacheStrategy.cache, it has to be on all the implementations. I'm thinking of moving the constraints to the MemoryCache and NoopCacheStrategy constructors. MemoryCache would require Sync, and NoopCacheStrategy would be fine with just Applicative.
... at least that's how it compiles in my head.
There was a problem hiding this comment.
I've started on it but ran into an issue because NoopCacheStrategy is used as a default argument in FileService.Config.apply(https://github.com/http4s/http4s/blob/master/server/src/main/scala/org/http4s/server/staticcontent/FileService.scala#L38) and that doesn't work if I have the constraint is on the constructor.
I could make the cache strategy a non optional argument but I'm not sure if that's the right call.
There was a problem hiding this comment.
I prefer builder patterns for configuration, because we can add more configuration options later without breaking bincompat. But that's still ugly with regard to this problem: we would need to supply the constraint wherever the default value is materialized.
And getting the strategy to Applicative doesn't even do much in practice, because the service that uses it still requires Sync.
I think maybe we accept this as incremental progress and rethink it for 1.0.
The extra features provided by
Effectare not used and relaxing the constraints makes it easier to use it withZIOinstead ofIO.Edit: I'm a bit confused about the failing spec. Seems unrelated to my change but I don't have much experience with the test suite so I could be wrong.