Haskell on the JVM with Eta • Brian McKenna • YOW! 2017

Learn how Eta compiles Haskell to JVM bytecode, enabling cross-language interop and deployment to JVM platforms while maintaining key Haskell features like laziness and type classes.

Key takeaways

Eta is a fork of GHC that compiles Haskell to JVM bytecode instead of native code
Performance is currently 4-10x slower than GHC, with significant room for optimization
Uses the STG (Spineless Tagless Graph) intermediate representation before generating JVM bytecode
Provides Java FFI capabilities to interact with Java code and libraries
Successfully tested with 221 Haskell packages that don’t depend on C code
Useful for running Haskell on platforms where JVM is required (e.g., Android, enterprise environments)
Can expose Haskell functions to be called from Java/Scala code
Supports key Haskell features like laziness and type classes
Requires patching some GHC libraries to remove C dependencies and replace with Java code
Enables integration with Java ecosystem tools like Spark, Kafka, and JavaFX
Project originated from GHCVM during Haskell Summer of Code, now maintained by TypeLead company
Supports inheritance and subtyping when interfacing with Java code
Uses explicit memory allocation and evaluation model similar to GHC’s compilation pipeline

Haskell on the JVM with Eta • Brian McKenna • YOW! 2017

More talks