Incorrect transformation of IRIs with a document fragment

[markwilkinson]

Hi Guys!

I'm seeing a transformation problem, and I've dropped the problem down as simple as I can.

csv:

org_id
http://www.bancodegermoplasma.upm.es

yarrrml1:

    - objects:
        type: iri
        value: $(org_id)
      predicate: prov:wasAssociatedWith

yarrrml2:

    - objects:
        type: iri
        value: $(org_id)#Organization
      predicate: prov:wasAssociatedWith

yarrrml1 results in: <http://www.bancodegermoplasma.upm.es> (correct!)

yarrrml2 results in: <nullhttp%3A%2F%2Fwww.bancodegermoplasma.upm.es#Organization> (OH NO!)

This happens also when the entity is in the subject position.

Not sure where in the process things go wrong, but... it's definitely wrong ;-)

Cheers!

[ghsnd]

Hi Mark,

You're right, this is not what one would expect. It is a known issue in the RML language, to which YARRRML is parsed before actually executing the rules.

What happens is this: when you use just $(org_id) as value, it is seen as a reference to the org_id field in your CSV input and it just literally takes the values in that field. However, if you add #Organization, it is considered a template, because there is no field org_id#Organization in the CSV. And in the current version of RML, templates are IRI-encoded when the type is iri.

There is a dirty workaround using a function to assemble the IRI you want:

prefixes:
 idlab-fn: "http://example.com/idlab/function/"
...
        objects:
          - function : idlab-fn:concat
            parameters: 
              - [idlab-fn:str, $(org_id)]
              - [idlab-fn:otherStr, 'Organization']
              - [idlab-fn:delimiter, '#']
            type: iri

Hope this helps...

Best regards,
Gerald

[markwilkinson]

that is very helpful, thank you!