Implementing the DDI4 Model in the R Language

Abstract

Model based DDI4 lends itself to a representation in a language with object oriented structures. This presentation describes an ongoing project implementing the complete DDI4 model in R. We currently have the complete structure represented as R6 objects. Future work will focus on a graphical user interface for entering metadata and on capturing metadata about transformations. Import and export of DDI and a human readable codebook are also planned.

Each DDI4 class is modeled as an (object oriented) R6 class. This approach yields several useful capabilities. Properties and relationships of objects are validated as to correct datatype and cardinality as they are instantiated. Existence of DDI URNs is validated against an internal registry and via an XML catalog. The latter allows for validating the existence of an external object. DDI metadata are attached to existing R objects as attributes having DDI URNs as their values.

The DDI4R package, including documentation, is generated programmatically directly from the DDI4 model XMI file available on the DDI4 website.

Having metadata objects in R may also offer interesting future possibilities for operations directly on metadata. Sum and difference operators on Collections, for example might be useful in harmonization.