Support CWL Prov with `cwltool` for OGC API - Processes IPT

[fmigneault]

Description

Given the rising need for IPT (Integrity, Provenance, Trust) through OGC APIs and their workflow processing, the provenance capabilities of CWL should be leveraged to accomplish this goal. This would add metadata references within the CWL Application Packages themselves, allowing better open-science and IPT workflow tracking.

To Do

• GET /jobs/{jobId}/run to return the PROV-JSON produced by cwltool --provenance
  (edit: won't do)

• GET /jobs/{jobId}/prov as alternate endpoint

  • see Part 4: Job Provenance endpoint mismatches GDC API Profile opengeospatial/ogcapi-processes#464
  • see Feature/fix issue 464 opengeospatial/ogcapi-processes#473

• consider additional PROV endpoints vs what cwlprov offers
  https://gitlab.ogc.org/ogc/T20-GDC/-/wikis/GDC-Provenance-demonstration-GeoLabs#usage

  • GET /jobs/{jobId}/prov (contents of various metadata/provenance/primary.cwlprov.{ext})
  • GET /jobs/{jobId}/prov/info (as cwlprov info or metadata/manifest.json)
  • GET /jobs/{jobId}/prov/who (as cwlprov who)
  • GET /jobs/{jobId}/prov/inputs (as cwlprov inputs)
  • GET /jobs/{jobId}/prov/inputs/{id} (as cwlprov inputs [<run-id>])
  • GET /jobs/{jobId}/prov/outputs (as cwlprov outputs)
  • GET /jobs/{jobId}/prov/outputs/{id} (as cwlprov outputs [<run-id>])
  • GET /jobs/{jobId}/prov/run (as cwlprov run --inputs --outputs --labels --duration --steps)
    (use all flags to get all available metadata)
  • GET /jobs/{jobId}/prov/run/{id} (as cwlprov run [<run-id>])

• Alternate PROV-XML/RDF/etc. if Accept requests it
  (all variants should already be generated by cwltool as various manifest representations)

• When generating cwltool --provenance results, avoid duplicating results already found in WPS-outputs to save space (use their URI for cross-reference).

  • if File is saved only as the {"class":"File", "path": "..."} definition, this could be allowed to avoid extra code managing the references
  • assume that strings are sufficiently small (as per cwltool's own content limit)

• Any additional metadata/links pointing at the specific job and process executed that should be embedded in the PROV contents

• Cross-walk with Support POST /jobs for various workflow implementations #716 requirements

• Include ORCID and other relevant PROV metadata #783
  (see cwltool --orcid --enable-user-provenance --enable-host-provenance)

• update CLI to provide a provenance operation

• ensure provenance-related requirements are added to conformance

• ensure provenance links are returned in job status response

• add more W3C PROV details about process I/O #780

References

• https://github.com/common-workflow-language/cwlprov (see cwltool --provenance)
• example run with OSPD algae workflow: https://gitlab.ogc.org/ogc/T20-GDC/-/wikis/GDC-Provenance-demonstration-GeoLabs
• consider new options to avoid data duplication of I/O in provenance folder
  Prov data input output common-workflow-language/cwltool#1989
• PROV: https://www.w3.org/TR/prov-overview/
• PROV-LINKS: https://www.w3.org/TR/2013/NOTE-prov-links-20130430/
• PROV-JSON: https://www.w3.org/submissions/prov-json/
• PROV-RDF: https://www.w3.org/TR/2013/REC-prov-o-20130430/
• PROV-HTML: https://www.w3.org/TR/2013/REC-prov-n-20130430/
• PROV-XML: https://www.w3.org/TR/2013/NOTE-prov-xml-20130430/

Implementation

• add CWL PROV support  #778