trialdiff 0.2.0
- Added an analysis/output registry:
td_output() and
output_registry() declare analyses and outputs and their
inputs, and add_outputs(), add_edges() and
remove_edges() graft, extend or override a lineage
graph.
- Added
lineage_from_metadata(), which derives a lineage
graph from a object (or equivalent metadata tables), with provenance and
a review list for ambiguous or unparsed references. aliases
resolve tokens that would otherwise be ambiguous or unknown, and
overrides merge an analysis/output registry into the
generated graph.
- Added a
"diffdf" comparison backend to
compare_cut(), which delegates value comparison to the
package and translates its output into a tdiff object
(schema comparison remains built in).
- Added a case-study vignette using public
pharmaverseadam data.
as_register() now coerces key columns to character,
fixing a failure when key columns were not character.
- New records in visit-based datasets are classified as
new_visit rather than new_assessment.
- Added edge-case tests; package coverage now exceeds 90%.
- Bumped GitHub Actions and added Codecov configuration.
trialdiff 0.1.0
- First tagged release. Five composable layers: compare, classify,
trace, assess and report.
compare_cut() compares two data cuts of a clinical
dataset and reports added, removed and modified observations plus schema
changes.
classify_changes() applies transparent, prioritised
rules to produce a clinical change taxonomy.
define_lineage() and trace_dependencies()
build and query a lineage graph.
assess_impact() grades downstream impact as
definitely/potentially/unlikely.
report_diff() produces HTML, Quarto, JSON and list
reports.
- Synthetic
ADSL/ADLB data cuts and an
example lineage are included.
trialdiff 0.0.0.9000
- Initial development version.