tests/testthat/test-multicore.R

In RStoolbox: Tools for Remote Sensing Data Analysis

context("Multicore/Singlecore")
library(parallel)
library(raster)


test_that(".paraRasterFun is equal to predict, calc, overlay, Both single and multicore.", {
            #skip_on_cran() # hadley says its risky to test parallel code on cran :-)
            for (clusterType in c('PSOCK', 'FORK')){
                if(Sys.info()[["sysname"]] != "Linux" && clusterType == "FORK") next
                r <- raster(ncol=10,nrow=10, vals=1:100)
                r <- stack(r, r^2)
                names(r) <- c("red", "nir")  
                m <- lm(red~nir, data = as.data.frame(r))
                f <- function(a,b){a-b}            
                
                beginCluster(2, type=clusterType)
                cluster <- "multicore"
                for(i in 1:2){
                    expect_equal(.paraRasterFun(r, rasterFun = predict, args = list(model = m)), predict(r, m), label = paste("predict:", cluster))
                    expect_equal(.paraRasterFun(r, rasterFun = calc, args = list(fun = sum)), calc(r, fun = sum), label = paste("calc:", cluster))
                    expect_equal(.paraRasterFun(r, rasterFun = overlay, args = list(fun = f)), overlay(r, fun = f), label = paste("overlay:", cluster))
                    endCluster()
                    cluster <- "singlecore"
                }
            }       
        })


test_that(".parXapply family returns identical results to ?pply family. Both single and multicore.", {
            #skip_on_cran()
            for (clusterType in c('PSOCK', 'FORK')){
              if(Sys.info()[["sysname"]] != "Linux" && clusterType == "FORK") next
              lis <- lapply(1:5, rnorm)
                mat <- matrix(1:100,10,10)
                beginCluster(2, type = clusterType)
                cluster <- "multicore"
                for(i in 1:2){
                    expect_equal(.parXapply(X = lis, XFUN ="lapply", FUN=sum, envir = environment()), lapply(lis, sum), label = paste("lapply:", cluster))
                    expect_equal(.parXapply(X = lis, XFUN ="sapply", FUN=sum, envir = environment()), sapply(lis, sum), label = paste("sapply:", cluster))
                    expect_equal(.parXapply(X = mat, XFUN ="apply", 1, FUN=sum, envir = environment()), apply(mat, MAR =1, sum), label = paste("aapply:", cluster))        
                    endCluster()
                    cluster <- "singlecore"
                }
            }
        })