Most prominently, this release features **full Dask array integration**. That means HyperSpy lazy signals and LiberTEM can now be combined seamlessly. Many thanks to Matthew Bryan who implemented major parts of this non-trivial feature!
This enables the following applications:
* Use HyperSpy file readers and other readers that create Dask arrays for LiberTEM.
* Create an ad-hoc file reader for LiberTEM by just building a Dask array from your input data.
* Use LiberTEM file readers for HyperSpy and other software that works with Dask arrays.
* Use the same implementation of an algorithm for live processing with LiberTEM, offline processing with LiberTEM, and offline processing with HyperSpy.
* Simplify implementation of complex processing routines on Dask arrays.
* Chain processing steps together using Dask arrays for intermediate results.
More details about Dask integration can be found at https://libertem.github.io/LiberTEM/dask.html.
Please note that these features are still experimental and cover a large space of possible uses and parameters. Expect the unexpected! Tests, feedback and improvements are highly appreciated.
Other improvements
* Support for TVIPS binary files
* Support for MIB Quad integer and some raw formats
* Parameter guessing for center of mass analysis
* I/O back-end selection and direct I/O, beneficial for very large data on fast SSDs
* Docker image available: https://hub.docker.com/r/libertem/libertem/tags
* Improved parameter handling and the new threaded `ConcurrentJobExecutor` speed up UDFs with large parameters, for example single side band ptychography.
* Various other performance and quality improvements
Please [see our changelog](https://libertem.github.io/LiberTEM/changelog.html#v0-9-0) for the full release notes with details on the changes and new features.