- Support Apache Spark 2.x codebase including Spark 2.2.0
- Enable Toree to run in Yarn cluster mode
- Create spark context lazily to avoid long startup times for the kernel
- Properly cleanup of temporary files/directories upon kernel shutdown
- %AddJAR now supports HDFS file format
- %AddDEP now defaults to default configuration
- Cell Interrupt now cancel running Spark jobs and works in background process
- Interpreters now have the ability to send results other than text/plain
This is the first release of Toree since it joined Apache as an incubator project on December 2nd, 2015.
As part of moving to Apache, the original codebase from SparkKernel has been renamed, repackaged and improved in a variety of areas. It is also important to note that the version has been reset back to
0.1.0 in favor or the version scheme used in the old project.
- Support for installation as a Jupyter kernel using pip
- New plugin framework for extending Toree. Currently only used for Magics.
- Support for sharing Spark context across different language interpreters.
- Improved AddDeps magic by using Coursier
- Binder support for easy trial on Jupyter Notebook
- Demonstration of building an interactive streaming dashboard