Changes in version 1.2.0 (2026-03-27)                  

Bug fixes

  - Fixed dice not rendering with ggplot2 >= 4.0 — The
    drawDetails.DiceGrob S3 method was not registered in the package
    NAMESPACE, so grid never dispatched to the custom drawing code.
    Neither tiles nor pips were drawn. Fixed by adding
    S3method(grid::drawDetails, DiceGrob) to NAMESPACE via
    @exportS3Method.

  - Fixed invisible pips when fill is not mapped — When the fill
    aesthetic was not mapped (default NA), pip colour was also set to
    NA, making pips invisible. Pips now default to black when fill is
    unmapped.

                 Changes in version 1.1.0 (2026-03-23)                  

New features

  - pip_scale parameter — New argument to geom_dice() that controls pip
    diameter as a fraction (0–1) of the maximum available space inside
    each die face. When size is constant (not mapped), all pips are
    drawn at pip_scale × max_diameter. When size is mapped to a
    variable, pips scale between 25% and pip_scale of the maximum pip
    diameter. Set pip_scale = NULL to disable auto-scaling and use the
    raw size aesthetic (legacy behaviour). Default: 0.75.

  - Deferred rendering via DiceGrob — Pip sizing is now calculated at
    grid draw time rather than ggplot build time, so pip diameters adapt
    correctly to the final figure dimensions.

  - Scale expansion for edge tiles — setup_data now exposes
    xmin/xmax/ymin/ymax so ggplot2 trains axis scales to include the
    full tile area, preventing clipping of border tiles.

  - New datasets — sample_dice_large (480 rows, 60 taxa) and
    sample_dice_miRNA (~90 rows, miRNA dysregulation) are included for
    demonstrating high-density and categorical-fill use cases.

Breaking changes

Visual output changes (existing plots will look different)

  - Default pip sizing changed — Pips are now auto-scaled to fill the
    die face via the new pip_scale parameter (default 0.75). In v1.0.0
    pips used the raw size aesthetic (default 3), which often left
    excessive whitespace or caused pips to overflow tile borders at
    certain figure sizes. Migration: add pip_scale = NULL to your
    geom_dice() call to restore the exact v1.0.0 appearance.

  - Legend key for unmapped fill — When fill is not mapped, legend keys
    now draw a solid black dot instead of an empty circle, improving
    readability. This is a cosmetic change with no opt-out.

  - Axis scale expansion — Edge tiles are no longer clipped because
    setup_data now reports tile extents to ggplot2. Plots may show
    slightly more padding around the border compared to v1.0.0.

API changes

  - draw_panel signature changed — The method now takes explicit named
    arguments (na.rm, ndots, x_length, y_length, pip_scale) instead of
    params, .... Code that subclassed GeomDice or called draw_panel
    directly will need updating.

Data / dependency changes

  - Sample datasets restructured:
    
      - sample_dice_data1: 48 rows → 160 rows (8 taxa × 4 diseases × 5
        specimens). Columns are the same but lfc and q may now contain
        NA.
      - sample_dice_data2: 48 rows → 160 rows; the replicate column has
        been removed. Column count changed from 6 to 5. Scripts that
        reference sample_dice_data2$replicate or hard-code row counts
        from either dataset will break.

  - tibble is now a hard dependency — added to Imports for
    column_to_rownames() / remove_rownames().

  - Minimum R version raised to 4.1.0 (was 4.0.0). R 4.1 is required for
    the native pipe operator (|>) used internally.

Minor changes

  - LICENSE switched to CRAN-standard two-line format (YEAR / COPYRIGHT
    HOLDER).
  - .Rbuildignore added to exclude pixi, demo output, test scripts, and
    other non-package files from the R CMD build tarball.
  - data-raw/ directory created; dataset-generation scripts moved there
    from data/.
  - pixi.toml / pixi.lock added for reproducible environment management.
  - Demo scripts (create_demo_plots.R, test_simple_dice.R) rewritten to
    exercise the new pip_scale feature and run from the project root.
  - README updated with new examples, a parameter table, and the
    published Bioinformatics citation.
  - RoxygenNote bumped to 7.3.3.