NEWS

tidyaudit 0.3.0

Audited execution — capture data-frame lineage across a whole script or session without per-step taps:
- audit_source() runs an .R file like source() but returns an audit trail; it owns the evaluation loop, so capture works interactively, under source(), and under Rscript.
- audit_record({ ... }) audits an inline block.
- audit_start() / audit_stop() provide interactive-session convenience via a top-level task callback. (Use audit_source() for scripts: source() is a single top-level task.)
Capture granularity is top-level statement lineage. Snapshots carry versioned-lineage fields (binding-stream object_id, version, event, source, srcref, parent_snapshot_ids, level); parents resolve from the pre-evaluation state and only to data.frames. Lifecycle events (create/update/delete/retire/unchanged_assignment) are tracked.
Capture is metadata-only by default; level (metadata, sample_hash, column_hash, full_hash) opts into per-run-salted content hashing to detect value-only changes.
audit_export() renders audited trails as a tabular report (run summary, step timeline, objects & versions, warnings & errors) plus a per-object lineage graph. Classic tap trails still render the linear flow.

audit_transform() for factor inputs no longer errors under R-devel (r89994+). The internal frequency-table builder previously routed counts through as.data.frame.table(), which now rejects NA in row names and failed for any factor (since the useNA = "always" bucket carries an NA name). The table is now constructed directly. No user-visible output change.

audit_export() — self-contained HTML trail visualization (interactive pipeline flow diagram, light/dark theme, clickable nodes/edges). No server, no Shiny required.
Serialization functions: trail_to_list(), trail_to_df(), write_trail(), read_trail() for machine-readable trail export (RDS and JSON formats).
tab() / tab_tap() — frequency tables and crosstabulations within pipelines. One-way and two-way tables with .sort, .cutoff, .wt, .na, .display options.
Snapshot controls: .numeric_summary, .cols_include / .cols_exclude parameters on all taps to limit snapshot scope on wide datasets.
print.audit_trail() improvements: tabular "Changes" block with From/To/Rows/Cols/NAs columns and +N/-N/= deltas; inline custom diagnostic annotations (.fns results) below each snapshot row with show_custom parameter.

audit_tap(): parameter label renamed to .label for API consistency with other taps (no deprecation — early-stage project).
Join taps: .stat now accepts bare column names (NSE via enquo()/as_label()), matching filter tap behavior. Quoted strings still work.
Internal snapshot field col_info renamed to schema (internal only, no user-facing API change).

audit_transform() is now generic across all vector types (numeric, Date/POSIXct, factor, logical, character) with type-appropriate diagnostics.
audit_report(): removed unimplemented format = "rmd" option (deferred to a future version).