Release Notes for Deephaven v0.11.0 | Deephaven

Deephaven continues to extend its capabilities with real-time data toward the Python community, in support of data science and app-dev workflows. The v0.11.0 release notes for deephaven-core list the complete set of deliveries. These oftentimes interact with evolving features in the barrage and web-client-ui projects. Together these make working with “Streaming Tables” (i.e., tables that change) natural and easy.

Please direct feedback, questions, or perspective about product development to Deephaven’s Slack or GitHub Discussions.

As previewed in an early March blog post, we released an upgrade to Deephaven’s server-side Python experience. The table API has been improved, delivering a more Pythonic and cleaner interface in support of the ultimate goals related to easy on-boarding; better unit testing, IDE autocompletion, and modularization; minimal exposure to JPy and Java objects; and consistency with the Python client.

As one simple example:

t = left.naturalJoin(right, "A,B", "X,Y")

becomes

t = left.natural_join(right, on=["A", "B"], joins=["X", "Y"])

Beyond testing and documenting the aforementioned upgrade, and ensuring all deephaven-examples and packaging comply with the new binaries, the Deephaven team has focused on longer-term initiatives:

• Using Deephaven as a Python library: a valuable usage pattern and an important stepping stone for some very slick Jupyter-related experiences.
• The delivery of real-time data via the C++ API, which will then be wrapped to serve the Python client API.
• Performance measurement of both static and real-time loads, with equipment to allow users to self-serve in this capacity.
• Infrastructure to support performance and resource measurement as it relates to Deephaven applications that you build.

Worth noting in the v0.11.0 release notes are the features below.

• Nested JSON is now supported in the Kafka.tools module.
• Related to CDC, which was delivered in the 0.10.0 release:
  • Message-rate performance has improved in the Debezium-to-Kafka integration, through the use of multiple partitions.
  • CDC deployment exists as a standalone module in the Python API.
• Nightly builds now run on both Java11 and Java17.

Changes and improvements to the Deephaven user documentation can be found in this blog post.