Skip to contents

Implementing Betwixt in Python

Source: vignettes/python.Rmd
python.Rmd

Betwixt is intended as a portable semantic review model rather than an R-specific workflow. The current reference implementation is written in R, but its main objects are ordinary tabular data, standalone HTML review documents, and RDF serialisations.

A Python implementation could therefore begin by reproducing the same transformations:

candidate dataset
       ↓
standalone review
       ↓
reviewed state
       ↓
wide / long projections
       ↓
RDF serialisation

The simplest starting point would be to reproduce the candidate dataset conventions with pandas, including reviewable columns, optional _range and _definition metadata, and display-only context_* columns.

The next step would be to reproduce the standalone HTML review and ensure that a review created by the R implementation can be read by Python, and vice versa. Compatibility should be defined by the exchanged artefacts and their semantics rather than by reproducing the internal structure of the R package.

Finally, a Python implementation could reproduce the wide and long projections and serialise the reviewed assertions using the Betwixt ontology and PROV-O.

The R source code can therefore be treated as a reference for behaviour, while the Betwixt candidate dataset, review artefact, projections, and RDF vocabulary provide the interoperability contract.