TODO_before_CRAN.md

In jfallmann/ConsSeq: Weighted Reconciliation of Segmentations

NEED TO DO

Real BUGS?

• check initialization of vectors in R, is it ok that $min_j \Delta[k]$, and $\delta^*$ are NA until AFTER the first border? See figure test_internals.[R|png].

Documentation: README.md, Roxygen and Vignette

• add README.md with github and cran install instructions, and minimal text + example,
• move plots from current test/ to roxygen doc and/or Vignette.
• sanity- and spell-check of existing roxygen doc.
• reconsider which functions in R and cpp should be public for the user (consensus_r/c, potential functions for comparison)
• Add missing autors

EXTRA

Potential Functions

• add more internal pre-defined potential functions,
• vectorize evaluteEquation and generate matrix over L and n.

Test/Debug Mode

• implement the Dslow/test calculations in Rcpp, test against R implementation and use for testing/debugging,
• use the testhat package to write proper tests, based on the current code in tests/,
• test_internals.R and tests/test_compilation.R:
• fuse to systematically test all three implementations: R/slow, R/incremental, and Rcpp,
• test_potentials.R:
• compare built-in vs. R vs. Rcpp user-supplied potential functions.
• allow to use R version from consensus wrapper for testing and debugging, in wrapper.

Optimize Rcpp

• benchmark for R/slow (consensus_r(...,test=TRUE) vs. R/fast vs. Rcpp, and built-in vs. supplied potential functions,
• use int and long double instead of NumericVector to optimize memory usage, cast properly in potential function and for debug return

ADD FUNCTIONALITY

Alternative Algorithms

• implement symmetrized boundary movers distance,
• implement breakpoint only code, using the conjecture that each consensus breakpoint is also a breakpoint in 1+ of the input segmentations,

Utilities

• convert to and from IRanges/GRanges,
• plot_breaklist: allow colors, eg.
  • color gradient for weights,
  • colors of input segment classes (segmenTier cluster association),
• plot_breaklist: dedicated segmentations vs. consensus plot,
  • indicate origin breakpoints in segmentation.

jfallmann/ConsSeq documentation built on March 9, 2021, 3:09 p.m.