Dan Allan, Kira Evans, & Sanket Verma - SciPy 2024 Tools Plenary | SciPy 2024

Join Dan Allan, Kira Evans & Sanket Verma for updates on AnyWidget, X-ray, Dask, SciPy, Ibis & JOSS. Learn about performance upgrades & new features in scientific tools.

Key takeaways
  • AnyWidget is a new tool for creating custom Jupyter widgets, offering a specification and cross-platform compatibility for notebook environments including Jupyter Lab, Google Colab, and VS Code

  • X-ray has seen significant improvements in performance and functionality, including:

    • Named array support for working with labeled dimensions
    • Data tree feature for efficient handling of hierarchical datasets
    • Better integration with array libraries through experimental support
    • Enhanced CPU/GPU backend compatibility
  • Dask received major performance upgrades:

    • Up to 20x faster performance for certain operations
    • Improved task queuing for data loading
    • Better shuffling and merge operations
    • Enhanced query optimizer
    • Cloud deployment capabilities through Coil
  • SciPy developments include:

    • NumPy 2.0 support
    • Code modernization and transition from Fortran 77
    • New transform class for handling distortions
    • Implementation of array API support
    • Enhanced sparse array functionality
  • Ibis updates:

    • Ground-up rewrite with improved query optimization
    • Integration with multiple databases including Oracle
    • Pyodide support for browser-based operation
    • New ML preprocessing capabilities
    • Streaming database support including Flink and rising wave
  • Journal of Open Source Software (JOSS) growth:

    • Expanded from 69 to 80+ editors
    • Implementation of editorial rotations
    • New validation tools for submissions
    • Launch of JOSScast podcast
    • Enhanced automation for reviewer selection