R/uniTab.R

In BushmanLab/cellTypeCompositions: Cell Type Compositions Gibbs Sampler

##' Combine duplicate rows of a table of counts
##'
##' When a table of counts has many duplicate rows, computations are
##' sometimes faster if duplicate rows are combined and a vector of
##' case weights is created.
##' @title uniTab Combine Duplicate Rows
##' @param tab matrix or table of counts.  Rows correspond to lineages
##'     and columns to cell types.
##' @param omitNullRows logical, if \code{TRUE} omit rows with only
##'     zero cells
##' @return \code{list} with elements \code{tab} - the rows of
##'     \code{unique(tab)},
##'     \code{n} - a vector of replicate row
##'     counts,
##'     \code{data.index} - a mapping of the rows to the
##'     \code{tab} argument to those of \code{uniTab(tab)[["tab"]]},
##'     and
##'     \code{omitted} - \code{NULL} if no rows were omitted, or
##'     an object such that \code{inverse.rle(result$omitted)} is a
##'     logical vector indicating which table rows were all zeros.
##' @export
##' @author Charles Berry
uniTab <- function(tab, omitNullRows=TRUE){
    tab <- unclass(as.matrix(tab))
    rownames(tab) <- NULL
    if (omitNullRows && any(rz <- rowSums(tab)==0)) {
        tab <- tab[!rz,]
        omitted <- rle(rz)
    } else {
        omitted <- NULL
    }
    
    utab <- unique(tab)
    tab.index <- match(
        do.call(paste,as.data.frame(tab)),
        do.call(paste,as.data.frame(utab)))
    list(tab=utab,n=as.vector(table(tab.index)),
         data.index = tab.index,
         omitted = omitted)}