Naïve question about the semantic of DatasetSubset

[huard]

Hi, in the datasource ontology, DatasetSubset is a subclass of Step, which is a subclass of prov:Derivation. My understanding is that I would write this in prov-n as: wasDerivedFrom(ex:id, generated_entity, used_entity, [prov:type="ds:DatasetSubset"])

The used_entity would be the parent ds:Dataset, but it's not clear what is the generated entity. In metaclipR, a vertex is created with ds:DatasetSubset, and then an edge connects it to ds:Dataset using ds:hadDataset. When I look at graphs exported by metaclip, the parent Dataset looks like this:

<file:///.../pretty.json#CMIP5_ACCESS1.0_historical> a <ipccds:MultiDecadalSimulation> ;
    ...
    ds:hadDatasetSubset <file:///.../pretty.json#DatasetSubset.ilouay>,
        <file:///home/david/Downloads/pretty.json#DatasetSubset.mgzuct> ;
    ...

Where there there are two DatasetSubsets attached to the data through one hadDatasetSubset edge.

I think what I don't understand is why in the PROV-DM a Derivation is an edge, while where it seems to be used as an entity, and why are there two subsets ?

[jbedia]

Hi David, I don't understand very well the issue. Where it reads there is an error, it should read which is the class representing this type of model data belonging to the ds vocabulary. ¿Where does this data come from?, there might be an error somewhere we should fix. If there are two different subsets from the same Dataset, then there should be two different "hadDatasetSubset" edges, and not just one.

[huard]

Hi Joaquin, It's an example you posted here: https://github.com/cf-convention/discuss/issues/33#issuecomment-643187959 that I serialized to prov-n (I believe).

What confuses me is the relation between prov's data model and METACLIP's DatasetSubset. Looking at https://www.w3.org/TR/prov-dm/uml/essentials.png, I would think that subclasses of Derivation would be edges, not vertices. Again, I'm a newbie and this may be a stupid question.