Some feedback on CoW from user perspective

[MiconSchorsij]

In the last months I've been using CoW intensively for the transformation of the ANDB datasets (working together with @wouterbeek ). Some of the issues that we encountered might be of interest to the CoW developers community. I am not sure if everything qualifies as an issue, therefore I would like to start by presenting them as a list. If there's anything in here that draws your attention please to not hesitate to contact me.

• It's not possible to create literals with datatype xsd:anyURI
• To create a language tag it is necessary to use lang:nl and xsd:string together in your CoW script. But actually that combination is not allowed in RDF (not sure if this is a problem, but it is confusing)
• The options to prevent the creation of triples with empty subjects appear insufficient (please check [CoW] Afhandeling empty objects wouterbeek/etl-andb#12 - in Dutch)
• The transformation process of a large dataset can take up a looong time (ANDB dataset >2 hours)
• It is not always clear what is the purpose of the different metadata graphs that are created. Do you always want to publish those? (As happens when uploading to Druid)
• Some triples appear to present properties of the transformation, not of the contents of the dataset
• SHACL validation is not supported (Validation of the complete dataset is not possible therefore it should be performed as part of the ETL)