Developer meeting summary 24/04/2023

[alessandrofelder]

• The immediate priority is for the initial milestone for cellfinder-core to be wrapped up and released in the next month under @dstansby 's lead (tensorflow to be pinned to 2.12.0 for now)
  • David to help document release process as part of this.
  • This will fix Investigate memory usage cellfinder-core#73 and will include the cython to numba port.
  • dask is not a viable option due to numba class, but memory issue fixed in different way
• We discussed testing and tutorial-driven development, to help make what BrainGlobe provides explicit and easy-to-check. In the medium term, ideally we'd like to have
  • a new documentation webpage
  • formalised, testable tutorials (and corresponding notebooks?) on (smallest possible) real-life data that can run on the average student laptop
  • benchmarks for the tutorials (to be run - possibly manually or on internal CI) before a release/big PR merge.
  • fast CI runs on synthetic data
  • version-controlled test data
  • a simple way to install the BG tools for users, likely a conda metapackage or environment
  • some slow tests (again to be run manually or internally before release/big merge) that ensure all tools in the metapackage work well together
• We also discussed error-handling and data validation strategies (both for GUI and libraries)
  • as we tackle this, it might be good to be in close discussion with napari community, and possibly contribute to the best practices for plugins section.