R/mat2dt.R

In philipdelff/NMdata: Preparation, Checking and Post-Processing Data for PK/PD Modeling

##' upper or lower triangle or all values of a matrix as long-format
##' @param x A matrix
##' @param triangle Either `"lower"` (default) or `"upper"`, or
##'     `"all"` for which triangle to return. `"lower"` and `"upper"`
##'     are equivalent for covariance or correlation matrices but the
##'     returned indexes will differ. `"all"` will return the full
##'     matrix which mostly makes sense if matrix is not a covariance
##'     or correlation matrix.
##' @param as.fun See `?NMdataConf`
##' @return A `data.frame`-like object with indexes `i` and `j` for
##'     position and matrix element value in `value` column.
##' @details The matrix is assumed ordered and the index numbers for
##'     rows and columns will be returned in `i` and `j` columns. Row
##'     names and column names will be returned in columns
##'     `parameter.i` and `parameter.j`.
##' @import data.table
##' @export
##' @seealso dt2mat

mat2dt <- function(x,triangle="lower",as.fun){
    

#### Section start: Dummy variables, only not to get NOTE's in pacakge checks ####

    i <- NULL
    j <- NULL

### Section end: Dummy variables, only not to get NOTE's in pacakge checks


    if(missing(as.fun)) as.fun <- NULL
    as.fun <- NMdataDecideOption("as.fun",as.fun)

    names.i <- rownames(x)
    if(is.null(names.i)) names.i <- as.character(1:nrow(x))
    names.j <- colnames(x)
    if(is.null(names.j)) names.j <- as.character(1:ncol(x))
    
    dt <- copy(as.data.table(x))
    
    colnames(dt) <- as.character(1:ncol(dt))
    dt[,i:=.I]
    
    dt <- melt(dt,id.vars="i",variable.name="j")
    dt[,j:=as.numeric(j)]

    dt <- mergeCheck(dt,
                     data.table(i=1:length(names.i),parameter.i=names.i) ,
                     by="i",quiet=TRUE)
    dt <- mergeCheck(dt,data.table(j=1:length(names.j),parameter.j=names.j),by="j",quiet=TRUE)

    dt <- switch(triangle,
                 lower=dt[j<=i]
                ,upper=dt[i<=j]
                ,all=dt
                 )

    as.fun(dt)
}