The Flink community intends to release Flink 1.0 after the 0.10 release. The following page lists features we want to include into the release.
Status:
Critical features / mechanisms
APIs
- Annotations that define which classes are public API
- Annotations that define which classes are public experimental / beta (may change)
- Tools that compare public API classes / method signatures across releases to ensure compatibility
- FLINK-1045: Remove Combinable Annotation
- Currently under discussion
- Alternatively change CombineFunction interface of RichGroupReduceFunction
- FLINK-3198: Rename Grouping.getDataSet() method
- FLINK-1903: Remove removal of duplicate expression keys
- FLINK-3233: Add support for partitioning of atomic types (can be done together with FLINK-1666)
- FLINK-3234: Add support for partition sorting with key selectors
- Review the DataSet API
- There are some inconsistencies wrt key definitions and data types that should be resolved.
Project Structure
- Restructure projects as decided on dev-ml
- Proper naming of Scala artifacts
- Introduce code style?
Engine
- Persistent checkpoints (manually created) that allow going back in streams to prior backup points