search

linkedpipes / etl

LinkedPipes ETL is an RDF based, lightweight ETL tool
https://etl.linkedpipes.com
Other
147 stars 30 forks source link

Enhancement: support blank nodes in configuration #400

Open jindrichmynarz opened 7 years ago

jindrichmynarz commented 7 years ago

Blank nodes are not supported in configuration for LP-ETL components. This is already documented. However, if there is not a strong reason not to support them, it would be convenient for users if they were supported. First, users could use full syntax of RDF without worrying about LP-ETL-specific restrictions. Second, in some cases blank nodes are easier to generate than unique IRIs (in the absence of unique keys, unique IRIs would require to use uuid()).

If blank nodes will not be supported, then can we get a more precise error? Currently, if blank nodes are used in configuration, the exception Not a valid (absolute) IRI is thrown:

Caused by: java.lang.IllegalArgumentException: Not a valid (absolute) IRI: node1bhjqgl77x119679
    at org.eclipse.rdf4j.model.impl.SimpleIRI.setIRIString(SimpleIRI.java:76)
    at org.eclipse.rdf4j.model.impl.SimpleIRI.<init>(SimpleIRI.java:65)
    at org.eclipse.rdf4j.model.impl.AbstractValueFactory.createIRI(AbstractValueFactory.java:88)
    at com.linkedpipes.etl.rdf.utils.rdf4j.BufferedTripleWriter.typed(BufferedTripleWriter.java:37)
    at com.linkedpipes.etl.dataunit.core.rdf.DefaultSingleGraphDataUnit.writeStatement(DefaultSingleGraphDataUnit.java:130)
    at com.linkedpipes.etl.dataunit.core.rdf.DefaultSingleGraphDataUnit.lambda$write$0(DefaultSingleGraphDataUnit.java:83)
    at com.linkedpipes.etl.dataunit.core.rdf.ActionExecutor.execute(ActionExecutor.java:18)
    ... 13 common frames omitted

While it is true that blank nodes are not valid (absolute) IRIs, the error could be more specific. Upon seeing this error, the users may suspect they have syntax errors in the IRIs. The error can report explicitly that blank nodes are unsupported.

Tested using the develop version fd194f2aa4f49eb44e79b45d2b9e2057e3745c14.

jakubklimek commented 7 years ago

The support is not planned as it is always possible to use uuid() - I do it this way - and blank nodes only cause trouble.

However, the error could be more desriptive.

jindrichmynarz commented 7 years ago

So perhaps the component can simply raise an error that the configuration contains blank nodes if the following query evaluates to true:

ASK {
  ?s ?p ?o .
  FILTER (isBlank(?s) || isBlank(?o))
}