Provenance Run Crate validation issues

[elichad]

I'm having a look at RO-Crates exported from Galaxy and testing them against Provenance Run Crate. I have a couple of questions:

• provenance-run-crate-0.5.4.3 "If specified, position must be a string representing an integer" - where does this come from? https://schema.org/position says it can be either Integer or String, and the spec doesn't mention this
• provenance-run-crate-0.5.1 "ComputationalWorkflow MUST refer to orchestrated tools via hasPart" - currently this is applied to any ComputationalWorkflow entity in the crate. Should it only apply to the mainEntity? This is not explicit in the spec, but it does refer to "the workflow" (as in "The workflow MUST refer to the orchestrated tools via hasPart") which I interpret as meaning just the main workflow.

In case it's useful, I've attached an RO-Crate exported from Galaxy which fails these criteria (and some others). Notably, Galaxy exports the same workflow in multiple formats, but currently only links the tools to one of them (the one that is used as mainEntity).

Invocation-of-sort-and-change-case-imported-from-uploaded-file-at-2025-03-19T112541788214.rocrate.zip

[simleo]

I'm having a look at RO-Crates exported from Galaxy and testing them against Provenance Run Crate

RO-Crates generated by Galaxy for workflow runs conform to Workflow Run Crate, not Provenance Run Crate. In practice they add some elements from the latter, but the root data entity's conformsTo does not reference it.

"If specified, position must be a string representing an integer" - where does this come from?

In the requirements table in https://w3id.org/ro/wfrun/provenance/0.5, the description of "position" says "An integer indicating the step's position in the execution order".

provenance-run-crate-0.5.1 "ComputationalWorkflow MUST refer to orchestrated tools via hasPart" - currently this is applied to any ComputationalWorkflow entity in the crate. Should it only apply to the mainEntity? This is not explicit in the spec, but it does refer to "the workflow"

In the main text it's a bit of a grey zone, but in the requirements table the constraint is specified to apply to ComputationalWorkflow, not just the main workflow.

[elichad]

RO-Crates generated by Galaxy for workflow runs conform to Workflow Run Crate, not Provenance Run Crate. In practice they add some elements from the latter, but the root data entity's conformsTo does not reference it.

I am hacking on this with the Galaxy team this week :) we may not get to full compliance with Provenance Run Crate but we're making some improvements, guided by the validator.

In the requirements table in https://w3id.org/ro/wfrun/provenance/0.5, the description of "position" says "An integer indicating the step's position in the execution order".

I saw this, but I don't understand why having "position": 1 in the metadata is then considered invalid. Is the 1 (not a string) not interpreted as an integer?

In the main text it's a bit of a grey zone, but in the requirements table the constraint is specified to apply to ComputationalWorkflow, not just the main workflow.

Ah yes, thanks. I wonder if this might cause issues with referencing workflows that are described elsewhere (e.g. an external WROC), but that's a separate issue really.

[simleo]

I saw this, but I don't understand why having "position": 1 in the metadata is then considered invalid. Is the 1 (not a string) not interpreted as an integer?

I've never been sure about JSON typing for property values, see ResearchObject/ro-crate#269. What I usually see in Schema.org examples is that numbers are represented as JSON strings, see for instance the example at https://schema.org/HowToStep (which includes usage of position). In fact, I'm looking for an example that uses JSON numbers right now and I cannot find it! Anyway, the validator in this case has just a misleading message, since it actually allows both an integer and a string representing an integer:

    sh:property [
        a sh:PropertyShape ;
        sh:name "HowToStep position type" ;
        sh:description "If specified, position must be a string representing an integer" ;
        sh:path schema:position ;
        sh:or (
            [ sh:datatype xsd:string ; ]
            [ sh:datatype xsd:integer ; ]
        ) ;
        sh:pattern "\\d+" ;
        sh:message "If specified, position must be a string representing an integer" ;
    ] ;

I'm not sure whether it should allow strings not representing an integer, e.g. "first". I could not find such an example on Schema.org. I think it should not allow that for better machine actionability.