In the ontology shacl is used to define columns, constraints and a mapping to the database tables.
ad:bakShape a sh:NodeShape ;
adshapes:areaaldataversie "4.2" ;
adshapes:begingeldigheid "2019-09-25"^^xsd:date ;
sh:property adprop:bak-AD_ID,
...
adprop:bak-hasGeometry ;
sh:target "av_bak_p" ;
sh:targetClass ad:bak .
Because there is no formal way to map SHACL to database tables we got a little bit creative and decided to use sh:target 'by convention' to map to the database table name.
ad:bak a owl:Class ;
rdfs:label "bak"@nl-NL ;
ad:areaaldataversie "4.2" ;
ad:begingeldigheid "2019-09-25"^^xsd:date ;
ad:herkomstDefinitie "PNH" ;
ad:isBGT_Object "Nee" ;
ad:positioneleNauwkeurigheid "None" ;
dcterms:subject adc:bak ;
rdfs:subClassOf nta8035:PhysicalObject
...
sh:targetClass is used to map to the owl definition, which in turn points to the skos vocabulary via dcterms:subject.
adc:bak
a skos:Concept ;
adc:areaaldataversie "4.2" ;
adc:begingeldigheid "2019-09-25"^^xsd:date ;
dcterms:creator ... ;
dcterms:description "Areaaldata versie: 4.2"@nl ;
dcterms:modified "2019-09-25"^^xsd:date ;
skos:definition "Object met een permanent karakter dat dient om iets in op te bergen of te verzamelen."@nl ;
skos:exactMatch <https://definities.geostandaarden.nl/imgeo/id/begrip/Bak> ;
skos:inScheme adc:Areaaldata ;
skos:prefLabel "bak"@nl ;
skos:topConceptOf adc:Areaaldata ;
.
See the jupyter notebook for the code to perform the mapping between tables and terms in datahub.
result:
issues:
- for some reason mapping doesn't seem to work both ways all the time.
so e.g. a table has a mapped term, but the term does not show it's related enitities
Potential improvements:
-
code is currently plenty 'hacky'
- improve filtering for datasets
- improve handling of term mapping between skos and datahub
- ...
-
might be nice to have a lineage map between entities and terms, or between terms and external mapping?
-
Entities (tables) are related to an external knowledge graph -> would be nice to document that relationship somehow.