Add invalidation for the schema-list cache 2.2.0 branch (close #215) #218
Conversation
| @@ -111,7 +111,8 @@ final case class Resolver[F[_]](repos: List[Registry], cache: Option[ResolverCac | |||
| def listSchemasResult( | |||
| vendor: Vendor, | |||
| name: Name, | |||
| model: Model | |||
| model: Model, | |||
| mustIncludeKey: Option[SchemaKey] = None | |||
There was a problem hiding this comment.
As this is public method, such change would mean breaking compatibility? So to avoid that, we could either:
- don't touch this method, then basically copy all that content but with
mustIncludeKeyto the newlistSchemasLikemethod. - extract this new method body to private method,
listSchemasResultwould call it withNone,listSchemasLikewithSome
There was a problem hiding this comment.
I have a suspicion that this method should not be public. Cached/NotCached should be transparent for the consumers. Could we just make it private?
However, since it is already public, I've added a 3 argument version of it. I'm pretty sure that the default value argument would compile to the same code. But just for the good measure.
| F: Monad[F], | ||
| L: RegistryLookup[F], | ||
| C: Clock[F] | ||
| ): F[Either[ResolutionError, SchemaList]] = |
There was a problem hiding this comment.
Is it possible to return here full SchemaListLookupResult? I suspect it would be useful for caching in transformer?
There was a problem hiding this comment.
This functions in the same the the listSchemas does. The same argument as before as consumers shouldn't need to know if it came from the cache or not.
The transformer's result would be cached by the iglu-client, so I won't be building another caching wrapper.
There was a problem hiding this comment.
In general yes, caching details should be hidden. But we needed that to build related cache in the transfromer -> see : #207 and usage in transformer snowplow/snowplow-rdb-loader@756b4a7
That's what I meant by 'caching in transformer' - on top of the cache in iglu-client build another cache for ddl Properties (in case of shredded type) and cache for ddl Field (in case of parquet type)
There was a problem hiding this comment.
What about:
Is it possible to return here full SchemaListLookupResult
? When we add mentioned Field cache to transfromer, we'd need that. So maybe it's better to add full SchemaListLookupResult now and map on a consumer side. Otherwise, we would have to add this method in the next release or use full listSchemasResult(schemaKey.vendor, schemaKey.name, schemaKey.version.model, Some(schemaKey)) which is a bit verbose?
| val resultTwo = resolver.listSchemasLike(schema101) | ||
|
|
||
| resultOne must beRight(SchemaList(List(schema100))) | ||
| resultTwo must beRight(SchemaList(List(schema100, schema101))) |
There was a problem hiding this comment.
If there is 1-0-2 added to the iglucentral this test would fail?
CHANGELOG
Outdated
| @@ -1,3 +1,7 @@ | |||
| Version 2.2.0 (2022-11-18) | |||
There was a problem hiding this comment.
Aaand I think this commit shouldn't be here? ;)
There was a problem hiding this comment.
Why not? It was created during the release in master, it should be picked to develop.
There was a problem hiding this comment.
But...2.2.0 hasn't been released yet? I think it will be when all those pending PRs are merged :)
voropaevp
force-pushed
the
feature/list-schema-list-2.2.0
branch
from
162c1ee
to
e5a9666
Compare
voropaevp
force-pushed
the
feature/list-schema-list-2.2.0
branch
from
e5a9666
to
6c30d7f
Compare
No description provided.