Skip to content

README.md

Latest commit

 

History

History
50 lines (40 loc) · 2.45 KB

README.md

File metadata and controls

50 lines (40 loc) · 2.45 KB
Raw

conda packaging

recipe/meta.yaml is the maintained source of the conda-forge recipe for jupyter-opensysml-kernel, the pip/conda package of the SysML v2 Jupyter kernel. It carries __TAG__, __VERSION__ and __SHA256_*__ placeholders and is not buildable as-is; scripts/render-conda-recipe.sh fills them from a release's SHA256SUMS.txt:

scripts/render-conda-recipe.sh v0.9.2 > meta.yaml          # fetches the manifest
scripts/render-conda-recipe.sh v0.9.2 dist/SHA256SUMS.txt   # from a local manifest

The recipe is per platform rather than noarch, because conda-forge packages may not download anything at install time: each platform's package carries the release's sysml-jupyter-kernel binary for that platform (the same raw asset pip install would download and verify), and the build installs it into the kernelspec under $PREFIX/share/jupyter/kernels/sysml with python -m jupyter_opensysml_kernel install --prefix "$PREFIX" --binary .... The Python source is the sdist PyPI serves for the same version, so the two installs are the same package; both digests are the manifest's.

Submitting and maintaining

conda-forge takes new packages through conda-forge/staged-recipes: the rendered meta.yaml goes into recipes/jupyter-opensysml-kernel/, and once merged conda-forge creates the jupyter-opensysml-kernel-feedstock repository, which is then the place later versions are updated (its bot opens a pull request when it sees a new PyPI version; the binary digests must be updated in that pull request by hand or by re-rendering from this template). A release must exist, with the kernel binaries and the sdist in it, before a version can be submitted.

The package's maintainers listed under extra.recipe-maintainers must accept the role on the staged-recipes pull request.

Trying the recipe locally

conda build (or rattler-build after converting with conda-recipe-manager) renders and builds the recipe for the host platform:

scripts/render-conda-recipe.sh v0.9.2 > /tmp/recipe/meta.yaml
conda build /tmp/recipe
conda create -n sysml-kernel --use-local jupyter-opensysml-kernel jupyterlab
conda run -n sysml-kernel jupyter kernelspec list

Until the package is on conda-forge, the pip package installs the same kernel into a conda environment: pip install jupyter-opensysml-kernel && python -m jupyter_opensysml_kernel install registers it under the active environment's prefix.