ggvariant 0.2.0

New features

Added plot_oncoprint() / plot_waterfall(): a gene-by-sample mutation matrix (the same visualisation under the ComplexHeatmap/cBioPortal and GenVisR/maftools community names, respectively). Genes are ranked by descending number of altered samples; samples are ordered by the memo-sort/cascade algorithm for the characteristic staircase pattern. Multi-hit cells (a gene mutated more than once in one sample) render as a distinct "Multi_Hit" category. Supports an optional clinical annotation track from a sample-metadata data frame. Suggested by Nour-al-dain Marzouka (#N).
Added plot_tmb(): a per-sample tumour mutational burden bar chart, raw counts or normalised to mutations/Mb.
read_vcf()’s gvf objects now have real print.gvf() and summary.gvf() methods, showing a compact header and consequence/ sample/chromosome breakdowns instead of a raw data.frame dump.

read_vcf() no longer attaches a variant to a homozygous-reference ("0/0") sample. .pivot_samples()’s presence check only excluded literal missing-genotype codes, so every sample was pivoted in as carrying every variant regardless of its actual genotype on any multi-sample VCF. This affected every per-sample dimension of read_vcf()’s output: plot_lollipop(color_by = "sample"), plot_consequence_summary()’s per-sample bars, and both new sample-aware plots above. Predates this release entirely (present since 0.1.0) (#3).
plot_variant_spectrum()’s context/genome arguments and read_vcf()’s info_fields argument now abort with an informative error instead of being silently ignored. None of the three were actually implemented, despite being documented (#1, #2).
A VCF missing its #CHROM header line, or containing a data line with the wrong number of tab-separated fields, now aborts with a clear message instead of failing deep inside on an opaque base-R error.
A nonexistent file path now aborts with a clear message instead of normalizePath()’s base-R error.

read_vcf()’s .parse_ann_csq() (ANN/CSQ INFO field parsing) is now vectorised over the whole INFO column instead of looping row by row. ~8.75x faster in isolation on a synthetic 100,000-record VCF (8.05s -> 0.92s); full read_vcf() on the same file drops from 11.62s to 3.68s.

Added a pkgdown site. The introductory vignette was renamed to ggvariant.Rmd to activate the “Get started” navbar convention, and three new articles (Gallery, interactive plots, customising with ggplot2) live in vignettes/articles/ rather than the shipped vignettes, per R Packages’ guidance for graphics-heavy content.
Removed stray scratch files (Mutation 1.pdf, Rplot*.png/.pdf, etc.) that were shipping inside the package tarball and triggering an R CMD check WARNING and NOTE.
Added R-CMD-check, test-coverage, and pkgdown GitHub Actions workflows, and fixed the Codecov badge, which pointed at a nonexistent main branch.
Test coverage increased from 68.12% to over 92%, including the first tests of malformed/edge-case input and the first vdiffr visual regression snapshots.

Initial CRAN release.
read_vcf() and coerce_variants() read VCF files or plain data frames into a tidy gvf object.
plot_lollipop(), plot_consequence_summary(), and plot_variant_spectrum() provide ggplot2-native variant visualizations.
gv_palette() and theme_ggvariant() provide built-in colour palettes and a shared plot theme.