R Reference Implementation for Betwixt

library(betwixt)

The reference implementation of Betwixt is written in R and follows tidy data and tidyverse principles. These principles also inform the design of Betwixt itself: candidate and reviewed semantic states are represented as ordinary tabular data that can be inspected, transformed, combined, and projected using familiar data-management operations.

The R implementation covers the complete Betwixt workflow:

source data
    ↓
candidate generation
    ↓
render_review()
    ↓
human review
    ↓
read_review()
    ↓
reviewed state
    ↓
post-processing and serialisation

The review interface itself is based on standard HTML, JavaScript, and CSS. R prepares the candidate data and renders the standalone review document, but the human review does not require a running R session or server.

The main advantage of the R implementation is that both sides of the review can form part of a reproducible computational workflow. Candidate datasets can be generated programmatically from source data, and reviewed states can subsequently be analysed, transformed, validated, projected, or serialised.

Not every Betwixt use case requires this complete programmatic environment. For users who prepare candidate data interactively in Excel, OpenRefine, or similar tabular tools, we plan a minimal standalone Go implementation concentrating on rendering and reading Betwixt review documents.

The R package therefore serves both as a usable implementation and as the reference implementation against which compatible implementations in other programming languages can be developed.