The lux-api v0.2.0 release contains major feature additions, refactoring, and changes towards a more usable and minimal API. This release documents the contributions from Spring to Summer 2020, with the goal of making Lux more robust, bug-free, and easier to use. This is the first release supporting a stable API. This release is associated with lux-widget v0.1.0, refer to the release notes for details on frontend changes.

Major Features

• New programmatic objects Vis, VisList
• Ability to set Vis as intent
• Simplified property access for intent, exported, and protected LuxDataFrame properties
• Added plot_config setting to customize plot style
• Default option to fallback to Pandas for errors
• Added support for Row and Column Group based on dataframe index
• Extended semantic type detection, temporal and ID fields
• Pandas Executor added support for colored bar/line chart (#48)
• UI Warning message support for ID fields (#71)
• Improve interestingness scoring (#72, #59)
• Basic history mechanism to support df.head and df.tail (#73)
• standalone option for to_Altair (#75)
• Warning message for temporal column (#80)
• Pandas compatibility and coverage test and related bugfixes (#86, #92)

Scalability

• Improve metadata maintenance and expiration mechanism to avoid costly __init__ calls (#63, #81)
• Pandas Executor supports generic random sampling for above 10k and above 30k rows (capped) with caching (#96)
• Render large scatterplots (above 5k rows) as binned heatmaps. 2D binning support in Pandas Executor with correlation-based lazy evaluation (#95).

Installation, Documentation, Tutorials

• Improved ReadTheDocs documentation
• New demo and multi-part tutorial notebooks

Contributors

We thank the following users who have contributed to this release: @caitlynachen, @cjachekang, @dorisjlee, @jaywoo123, @jerrysong1324, @jinimukh, @jrdzha, @thyneb19, @westernguy2.
Additional thanks to Aditya Parameswaran and Marti Hearst for providing valuable feedback that contributed to this release.