v2.2.1

• Improve performance of TipInstability() (and thence ColByStability(), QuickRogue()) through full C++ processing, removing Rfast dependency.

v2.2.0

Performance

• TipInstability() operates on the lower triangle of distance matrices only, halving the work for row statistics (rowMads, rowVars, rowMedians).

• New batch C function LOG_GRAPH_GEODESIC_MULTI computes log-geodesic distances for all trees in a single .Call(), reusing one interim buffer and returning lower-triangle entries directly.

• Cache-friendly extraction loop in graph_geodesic.c (stride-1 access instead of stride-all_nodes).

• Auto-enable OpenMP parallelism in Rfast operations (rowMads, rowVars) when the distance matrix exceeds 1 000 rows.

• Use Rfast::rowMedians() in place of matrixStats::rowMedians().

• RogueTaxa() now calls .PrepareTrees() once and passes .prepared = TRUE to QuickRogue() / Roguehalla(), avoiding redundant tree preparation.

• QuickRogue() precomputes information upper bounds for all iterations rather than recomputing each step.

v2.1.7

• Improve tip instability calculation in identical tree sets (#29).
• Improve variable protection.

v2.1.6

• Legend annotations in documentation.
• Disable parallel evaluation by default in TipInstability(), adding parallel parameter to allow user to override.
• Use format string in REprintf().

v2.1.5

• Call C functions using symbols, not strings.

v2.1.4

• C2X compliant function prototypes.
• Remove unused sprintf() calls.

v2.1.3

• ColByStability() gains pal argument to allow specification of custom palettes.

v2.1.2

• Faster rogue detection when edge lengths provided, per report by Joseph Keating.

• Don't list neverDrop in QuickRogue(fullSeq = TRUE).

v2.1.1

• Handle ColByStability(trees = NULL).

v2.1.0

• Early termination of QuickRogue() when no further improvement possible.

• Cophenetic() renamed to the more accurate GraphGeodesic().

• Calculate information content of consensus trees with p > 0.5 (#15).

• Improve support for multiPhylo objects.

• New vignette detailing rogue detection with Bayesian tree samples.