Only process maven repository once if it's the same #2982
Conversation
| @@ -15,6 +15,9 @@ import java.util.* | |||
|
|
|||
|
|
|||
| class MavenRepositoriesProjectResolver: AbstractProjectResolverExtension() { | |||
|
|
|||
| val processedRepositoryModels = IdentityHashMap<RepositoryModels, Unit>() | |||
There was a problem hiding this comment.
Also storing state here will cause memory leak, as this class is a long-living service. It is not re-created on each sync.
There was a problem hiding this comment.
Oh, I didn't realize that. Thanks! What would be the alternative? I've seen a pattern where this sort of cache was stored inside DataNode, is that what I should do?
There was a problem hiding this comment.
Clearing the map in an import finished callback should be fine, too, right?
| @@ -15,6 +15,9 @@ import java.util.* | |||
|
|
|||
|
|
|||
| class MavenRepositoriesProjectResolver: AbstractProjectResolverExtension() { | |||
|
|
|||
| val processedRepositoryModels = IdentityHashMap<RepositoryModels, Unit>() | |||
There was a problem hiding this comment.
Does the IdentityHashMap work as expected here?
At first glance, different instances of RepositoryModels can contain duplicates - different instances of MavenRepositoryModel that are logically equal, but have different references.
There was a problem hiding this comment.
That was my experience, but I don't fully understand how it ends up the same. I will double check and update here.
There was a problem hiding this comment.
For reference, I ran the debugger with a sample project with multiple modules logging the object ids and contents each time it runs and here it is showing they are all the same.
identity-debugger-output.txt
This should improve import performance for large projects