Changes in version 0.6.0 (2026-06-24)                  

  - Enhance Graphviz (DOT) output for publication-quality figures. The
    DOT engine now renders full flow diagrams for complete feature
    parity with grid (default) outputs.
  - Add factorial (double-split) designs, with sub-arms grouped under
    their parent and the trunk centered over all leaves. Compatible with
    both grid and DOT outputs.
  - Add nested exclusion reasons, with each parent reason carrying its
    own named sub-reasons. Accessed by supplying a nested named vector
    (manual mode) or a reason/sub-reason column pair that is
    cross-tabulated automatically (data mode). Compatible with both grid
    and DOT outputs.
  - Rename the criteria parameter of exclude() and assess() to
    criterion. The singular reflects that each step defines one
    accountable exclusion with a single reason breakdown; compound
    conditions (&, |, !) remain supported.
  - Expand arithmetic checks. exclude() and assess() now warn when a
    reason column has no value for some removed rows---such rows are
    grouped under "Other". Like the manual-flow arithmetic checks, this
    is advisory and toggleable via options(selecta.check_arithmetic).
  - Expand vignettes and gallery.

                        Changes in version 0.5.0                        

  - Remove classify(). STARD diagrams now use stratify() on the
    index-test result with a per-arm endpoint(breakdown = ...) for the
    target-condition breakdown, matching the standard STARD layout.
  - cohorts() snapshot fields renamed: remaining to included and
    n_remaining to n_included, mirroring excluded/n_excluded.
  - Complete CRAN-compliant documentation for all exported functions,
    adding @details and @family sections alongside full @param, @return,
    @seealso, and @examples; non-executed examples now use \donttest
    rather than \dontrun.
  - DOT engine: source box and header colors now match the grid engine;
    fix multi-source header alignment; add orthogonal-edge and
    three-source examples.
  - Arithmetic checks for manual flows now also validate per-arm
    exclusion sub-reason totals; checks remain toggleable via
    options(selecta.check_arithmetic).
  - Performance: recdims() no longer issues drawing primitives when only
    measuring dimensions, and data-mode exclusions avoid materializing
    the excluded subset.
  - Expand vignettes and gallery.

                        Changes in version 0.4.0                        

  - Phase boxes rework. Adaptive formatting for both overly long and
    multi-line phase labels, including toggleable soft-returns for the
    former and user-insertable hard-returns (\n) for the latter.
  - Add regional number formatting: US, EU, SI/ISO 31-0 standard,
    custom.
  - Refine vignette workflows and documentation.
  - Rename suggest_size() function to recdims().
  - Add logo.

                        Changes in version 0.3.0                        

  - Add "split-and-recombine" topology, accessible through use of
    successive stratify() and combine() calls.
  - Add exclusion box formatting edits.
  - Make enhancements to classify() to ensure proper alignment.
  - Reorganize mock datasets with expanded documentation.
  - Update documentation headers.

                        Changes in version 0.2.1                        

  - Add refinements to reason ordering.
  - Fix expr parameter expression handling.

                        Changes in version 0.2.0                        

  - Add support for multi-source diagrams (PRISMA, MOOSE).
  - Improve rendering consistency and add specializations for "split"
    diagrams (CONSORT, STROBE).
  - New functions: sources(), combine(), assess(), classify(),
    stratify()/allocate(), suggest_size(), autoflow().
  - Multiple pipeline enhancements for exclude(), endpoint(), and
    stratify()/allocate().
  - Add font-scaled line spacing.
  - Add count-first display mode.
  - Add comprehensive testthat suite.
  - export_diagram() renamed to autodiagram().
  - exclude() default for show_count changed from TRUE to FALSE.
  - Add vectorization where possible for performance enhancements.

                        Changes in version 0.1.0                        

  - Initial commit.
  - Core functions established.
  - First-draft README.
  - Basic CONSORT functionality, tested on 0-, 2-, and n-arm setups.