R/Transform.R

In zoonproject/modules: Species Distribution Modelling Code Modules for the ZOON Package

#' @name Transform
#'
#' @title Apply Transformations to Covariates
#'
#' @description Apply a tranformation function (e.g. square, log, something else) to one or more covariates, pixelwise. These can either overwrite the named covariates, or be added to the covariate set.
#'
#' @details  This module is essentially a wrapper around the \code{raster} function \code{\link[raster]{calc}}.
#'
#' @param .data \strong{Internal parameter, do not use in the workflow function}. \code{.data} is a list of a data frame and a raster object returned from occurrence modules and covariate modules respectively. \code{.data} is passed automatically in workflow from the occurrence and covariate modules to the process module(s) and should not be passed by the user.#'
#' @param trans a function defining a transformation to apply to each pixel in the covariate(s) raster. This can be a definition (e.g. \code{function(x) x^2} for a square) or a function defined in base R (e.g. \code{log1p} for the log(x + 1) transformation). The default transformation does nothing.#'
#' @param which_cov a character vector naming all of the covariates in the RasterStack to which the new transformation should be applied.#'
#' @param replace a logical, determining whether to overwrite the values of the named layers with the transformed values (if \code{replace = TRUE}), or to add extra layers with the transformed covariates. If additional layers are added, they will be renamed accoring to the \code{raster} package's renaming conventions. This normally involves adding a '.2' at the end of the name.
#'
#' @family process
#'
#' @author ZOON Developers, \email{zoonproject@@gmail.com}
#'
#' @section Data type: presence-only, presence/absence, presence/background, abundance, proportion
#'
#' @section Version: 0.1
#'
#' @section Date submitted:  2016-06-15
Transform <- function (.data, trans = function(x) {
    x
}, which_cov = NULL, replace = TRUE) 
{
    df <- .data$df
    ras <- .data$ras
    # Keep attributes
    Atts <- attributes(df)[!names(attributes(df)) %in% c('names', 'class', 'row.names')]
    
    if (is.null(which_cov)){ 
        which_cov <- names(ras)
    }
    stopifnot(all(which_cov %in% names(ras)))
    if (!replace) {
      # which will be the new ones
      new_idx <- seq_along(which_cov) + nlayers(ras)
      # get good new names
      new_names <- c(names(ras), which_cov)
      new_names <- make.names(new_names, unique = TRUE)
      # extend the raster & give it the good names
      ras <- stack(ras, ras[[which_cov]])
      names(ras) <- new_names
      which_cov <- names(ras)[new_idx]
    }
    if (nlayers(ras) == 1) {
        ras <- calc(ras, trans)
    } else {
        for (cov in which_cov) {
            ras[[cov]] <- calc(ras[[cov]], trans)
        }
    }
    vals <- extract(ras, df[, c("longitude", "latitude")])
    if(is.null(ncol(vals))){
      vals <- data.frame(vals)
      names(vals) <- which_cov
      df <- cbind(df[, !(names(df) %in% attr(df, 'covCols'))], vals)
    } else {
      df <- cbind(df[, !(names(df) %in% attr(df, 'covCols'))], vals)
    }
    
    attributes(df) <- c(attributes(df), Atts)
    attr(df, 'covCols') <- names(vals)
    return(list(df = df, ras = ras))
}