Writing a speedy (wrt factor treatment) replacement for data.frame in R. Boy have I ever felt like I'm reinventing the wheel. Learning loads though.
Immutability is not workable for me. The main idea is a form of imputation (every row in the original dfr gets a given number of 'representatives' in a weighted augmented dataset). For now, it's dubbed NumDfr (see https://r-forge.r-project.org/projects/addendum/ for my progress), as it is a (numeric) matrix (+ a list of levels) acting like a data.frame. Works pretty well, actually.
Add a comment...