crs4 / repo2rocrate Goto Github PK

Snakemake workflows are often split into multiple files across a number of directories (this is the best practice). At the moment, except for the main Snakefile, repo2rocrate seems to label the other elements with type Dataset, rather than using a more specific type.

There is probably enough information to label these files with a more appropriate type.

E.g., author from nextflow.config.

Add an option to turn the input repository itself into an RO-Crate, by only adding ro-crate-metadata.json. Note that this is technically already possible, by setting the output directory to the input one, but that needlessly overwrites all files.

This is a bit tricky. We certainly want to reuse code from the IWC tool, which will then be able to use this one as a dependency. Ideally, this tool could be made to cover all the functionality currently implemented by the IWC tool, so that it could simply be installed and used by the IWC CI workflow. However, some features of the IWC tool are quite specific to the way the IWC repo works as a whole (e.g., its workflows/<CATEGORY><WORKFLOW> layout, its automatic deployment to iwc-workflows, etc.), rather than to the layout of individual workflow repositories.

E.g.:

{
    "@id": ".tests",
    "@type": ["Dataset", "TestDefinition"],
}

This is useful information even if it's not actionable. We need to relax the "A test definition MUST refer to the test engine it is written for via conformsTo and to the engine’s version via engineVersion" requirement in the spec though.

The current default for the workflow language is snakemake, but that's arbitrary. We should try to detect the workflow language, e.g. by running the various workflow-finding functions in sequence.