add_marginal() (and mark_density() / mark_histogram()) currently compute the KDE / bins in R via stats::density() and .stat_bin() (see R/compile-stat.R). For very large data this could mirror the existing Rust 2-D binning (rs_aggregate_2d in the vellum backend's src/rust/src/aggregate.rs) with a 1-D fast path.
This is an optional performance enhancement, not a correctness prerequisite — the R implementation is exact and fine for typical sizes. Worth revisiting if profiling shows the 1-D stats becoming a bottleneck on large marginals/histograms.
Context: added in the add_marginal() feature (marginal distributions along the top/right panel edges, the ggMarginal analogue).
add_marginal()(andmark_density()/mark_histogram()) currently compute the KDE / bins in R viastats::density()and.stat_bin()(seeR/compile-stat.R). For very large data this could mirror the existing Rust 2-D binning (rs_aggregate_2din the vellum backend'ssrc/rust/src/aggregate.rs) with a 1-D fast path.This is an optional performance enhancement, not a correctness prerequisite — the R implementation is exact and fine for typical sizes. Worth revisiting if profiling shows the 1-D stats becoming a bottleneck on large marginals/histograms.
Context: added in the
add_marginal()feature (marginal distributions along the top/right panel edges, the ggMarginal analogue).