R/loadEvals.r

In adamlilith/enmSdmPredImport: Testing measures of predictor importance in species distribution models

#' Load evaluation files and compile into a single data frame
#'
#' This is a helper function that loads multiple evaluation files and compiles them into a single data frame. The data frame is then sorted by the order of the algorithms in the argument \code{algo}.
#' @param evalDir Character, name of the \emph{scenario} directory containing evaluation files.
#' @param algos Character, name of algorithm(s) for which to collate evaluation files. Examples: \code{c('omniscient', 'brt', 'gam', 'maxent')}. Algorithms will appear in this order in the data frame output.
#' @param save Logical, if \code{TRUE} then save collated data frame in the evaluation folder. The name of the file will be "!Collated Evaluations.RData".
#' @param redo Logical, if \code{TRUE} and a file named "!Collated Evaluations.RData" is already in the evaluations folder, then this function will load that file instead of re-collating individual evaluation files. If \code{FALSE}, then re-collate evaluation files.
#' @param verbose Logical, if \code{TRUE} then display progress.
#' @return Data frame.
#' @export

loadEvals <- function(
	evalDir,
	algos=c('omniscient', 'bioclim', 'brt', 'gam', 'glm', 'maxent', 'rf'),
	save=TRUE,
	redo=FALSE,
	verbose=TRUE
) {

	# loads evaluation files and compiles them into a single data frame
	if (!redo & file.exists(paste0(evalDir, '/!Collated Evaluations.RData'))) {
	
		load(paste0(evalDir, '/!Collated Evaluations.RData'))
		
	} else {
		
		# load by algorithm
		for (algo in algos) {

			omnibus::say(algo)
			files <- omnibus::listFiles(evalDir, pattern=toupper(algo))

			# load each file
			for (file in files) {

				if (verbose) omnibus::say(file)
				
				# load
				load(file)
				
				# errors
				if (!exists('perform', inherits=FALSE)) stop('File does not contain a variable named "perform."')
				if (any(!(perform$algo %in% algo))) stop(paste('Algorithm listed in perform data frame is not the intended algorithm (data frame is for', perform$algo[1], ' (intended algorithm is', algo, ').'))
								
				# collate with all data
				master <- if (!exists('master', inherits=FALSE)) {
					master <- perform
				} else {
					merge(master, perform, all=TRUE)
				}
				rm(perform)

			} # next file

		} # next algorithm

		master$algo <- as.character(master$algo)

		# sort
		ranks <- rep(NA, nrow(master))
		for (i in seq_along(algos)) {
			ranks[master$algo == algos[i]] <- i
		}
		master <- master[order(ranks), ]

		if (save) save(master, file=paste0(evalDir, '/!Collated Evaluations.RData'))
		
	} # if loading files from scratch
		
	master

}