Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
66 lines (42 loc) · 3.3 KB

CONTRIBUTING.md

File metadata and controls

66 lines (42 loc) · 3.3 KB
Raw

Contributing

Setup and Checks

Install the project in editable mode, in a Python environment where DataLab 1.3 or later is available:

python -m pip install -e ".[test]"

Before submitting a change, run:

python -m ruff check .
python -m pytest

Trying the Plugin in DataLab

With DataLab 1.4 or later, add this project folder in Plugins > Configure plugins... > Plugin settings, then reload the plugins: DataLab loads the plugin from src/datalab_camera_characterization/adapters/desktop.py, without installing the project (see DataLab-Platform/DataLab#412). The packages the plugin depends on must be available in DataLab's environment.

Architecture

Dependencies point inward:

adapters -> workflow -> core
  • core holds the scientific code, in NumPy only: simulator, statistics, relative characterization, photon transfer and dark current. It does not import DataLab.
  • workflow turns the core into headless DataLab recipes, registered in CAMERA_RECIPES. It checks inputs, suggests the dark and flat bindings, and converts results to Sigima objects. It never changes a workspace and never imports GUI modules.
  • adapters/desktop.py and adapters/web.py declare the actions, examples, tools and welcome tiles. The run actions call DataLab's generic recipe launcher; DataLab's RecipeRunner owns validation, transactional commit and provenance.
  • simulator.py exposes the core simulator as a DataLab instrument. demo.py generates the photon transfer and dark-ramp examples in memory; the quickstart is a packaged HDF5 file.

tests/unit/test_architecture.py checks these boundaries.

Mean and variance belong to core/aggregation.py. A block accumulator (Chan/Welford) processes at most aggregation_block_size frames at a time, one by default. Memory therefore grows with the block size and the frame size, not with the number of frames. Each exposure level is reduced to scalars before the next one.

Tests

  • tests/unit: core, workflow and adapters.
  • tests/validation: results against the simulator truth.
  • tests/integration: installed wheel, hot reload and DataLab workspace round trip.

benchmarks holds explicit scripts that pytest does not collect. See Qualification for the Alpha and Stable gates.

Maintenance

  • Quickstart resource: regenerate src/datalab_camera_characterization/examples/camera_quickstart.h5 from fixed simulator parameters:

    $env:PYTHONPATH="src;../DataLab;../Sigima;../guidata;../PlotPy;../PythonQwt"
python scripts/generate_quickstart.py

  • Screenshots: regenerate doc/images/*.png with the plugin installed in DataLab's environment:

    python -m scripts.take_screenshots

  • DataLab-Web: the wheel embeds this README. DataLab-Web bundles the wheel and checks its size and SHA-256 (src/runtime/bundledPlugins.ts): update them when bumping the plugin version.

Branches

Day-to-day work lands on develop, the default branch: open pull requests against it. main is the release branch. It will be created at the first release, once the DataLab version that provides the plugin SDK is published; after that, develop is merged into main only to cut a release. This is the same model as DataLab, Sigima and DataLab-Web.