You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Recently, I've been considering the concept of layers in bindings. I previously believed that all bindings should have their own unique layer abstractions and implementations. However, I discovered that it's illogical for each binding to implement the same layer individually. This contradicts our OpenDAL's VISON.
Looking back at the Java binding, perhaps the main issue lies in how we've designed the API to expose Rust layers in Java. We could consider exposing them as org.apache.opendal.layers.RetryLayer, and construct a similar API like op.layer(new RetryLayer()).
@Xuanwo the layer implemented in Java API should include a way to export Rust's layers, while it's not conflict with making a pure Java layer to leverage Java ecosystem. The first challenge I meet is how to pipeline the accessor and adapt operator class into this absraction :D
The first challenge I meet is how to pipeline the accessor and adapt operator class into this absraction :D
I strongly oppose exposing Accessor to the bindings realm. It's OpenDAL's raw API, which users should never interact with. Technically, it's challenging for Java to access generic types in Rust through FFI.
The text was updated successfully, but these errors were encountered: