With R-4, make Dotchart1() just a dotchart() wrapper which converts non-numeric objects first (and signals it), think how to keep the reverse compatibility

With R-4, make Ex.cols() use new palettes and allow it to change palettes

With R-4, make Str(..., as.factor=TRUE) use df <- data.frame(as.list(df), stringsAsFactors=TRUE) and also invisibly output new df

Make Booster() in a way analogous to Tapkee(); modify BootA() to use it

Extend Bclust() to accumulate and accept the list of 'hclust' objects and probably to employ 'booster'

Extend data with Hansen & Rahn (1969) matrix (from colby.edu plant_id.c)

Biokey(): test more, fill gaps in conversion pairs, keep node labels (descriptions, taxon names) everywhere, add object type checks, modularize

Experimental Misclass() code needs more testing, especially with table()-specific functions and NAs (but "ignore" seems to work)

Think how to employ J. Oksanens's "ordihulltest" (proportion of cases when randomized classes give better overlap), vegan also has simple polyarea() and polycenter() within summary.ordihull()

Think if "manhattan" and "ward.D" are needed as default in Bclust() and Jclust(); and if "manhattan" needed in examples at all

Think if prcomp(iris..., scale=TRUE) is needed in examples (from experiments, scaling of _iris_ makes results worse)

Think how to standardize MrBayes() writing Nexus (now there are two ways, "old ips" and "direct")

In Ellipses() and Hulls(), it is better to calculate "coords" first, then plot

In bivariate plots, it is better to use xy.coords()

Gap.code(): try to optimize more; possible alternative: rle() each sequence, then extract pos/len of gaps and match (and check inclusions) sequence gaps in the union of all gaps

BestOverlap(): try to optimize with parallel::mclapply()

Idea: Hhomonyms() to work with API or directly with hemihomonyms database; maybe, it deserves the separate package

Idea: Coarse(phylo) on the base of ape::di2multi() which allows for given number of hierarchies (maybe, by iteration), collapsing and keeping nodes using e.g., bootstrap support

Idea: Normalize.ranks(classif) to (1) fit ranks into given list and (2) propagate main (whole numbers) ranks (based on code from Biokey())

Idea: create dichotomous key from binary table using rpart::rpart() as helper; look on DichotomousKey Github package

Idea: using Hcoords() and Tcoords(), it is in principle possible to plot 'hclust' object horizontally, without conversion into dendrogram, like series of segments

Idea: extract 'clipper' and adjacent polygon functions from PBSmapping:: sources and make them work from within shipunov::; however, in that case Windows distribution will become more complicated

Note: while Fligner-Policello tests are numerous in R packages, pairwise variant seemengly absents so it is better to keep Rro.test() and pairwise.Rro.test()

Note: there is a package MRH for Multi-Resolution Estimation of the Hazard Rate, and MRH::MRH() so probably rename shipunov::MRH()
