Author: Frank Raulf<\/p>\n
Normally, it is better to avoid loops in R. But for highly individual tasks a vectorization is not always possible. Hence, a loop is needed \u2013 if the problem is decomposable.<\/p>\n
Which different kinds of loops exist in R and which one to use in which situation?<\/p>\n
In each programming language, for- and while-loops (sometimes until-loops) exist. These loops are sequential and not that fast \u2013 in R.<\/p>\n<\/p>\n
for(i in x)<\/em><\/p>\n {task}<\/em><\/p>\n \u00a0<\/p>\n i=y<\/em><\/p>\n while(i<=x)<\/em><\/p>\n {task<\/em><\/p>\n i=i+1}<\/em><\/p>\n \u00a0<\/p>\n Even for prototyping sometimes too slow.<\/p>\n But how to improve speed?<\/p>\n There are three options in R:<\/p>\n apply loops:<\/b><\/p>\n Normally, you can use apply for calculating some standard statistics of the columns, the rows, or both. But you can use a trick to adjust the apply order for a loop. The syntax is:<\/p>\n<\/p>\n F <- function(i, x, y, z,\u2026)<\/em><\/p>\n {task}<\/em><\/p>\n apply(as.data.frame(1:length(vector)), margin = 1, FUN = F)<\/em><\/p>\n<\/p>\n In this case you use the vector not for direct calculation but as an index \u201ci\u201d instead.<\/p>\n The sapply order is even faster.<\/p>\n<\/p>\n F <- function(i, x, y, z,\u2026)<\/em><\/p>\n {task}<\/em><\/p>\n sapply(1:length(vector), FUN = F)<\/em><\/p>\n<\/p>\n Parallelization:<\/strong><\/p>\n You can use loops and apply orders also in parallel. You need:<\/p>\n<\/p>\n library(“doParallel”)<\/em><\/p>\n library(“parallel”)<\/em><\/p>\n library(“foreach”)<\/em><\/p>\n \u00a0<\/p>\n Firstly defining the number of cores. Leave out at least one:<\/p>\n \u00a0<\/p>\n NumOfCores <- detectCores() - 1<\/em><\/p>\n registerDoParallel(NumOfCores)<\/em><\/p>\n \u00a0<\/p>\n Either using a loop:<\/p>\n \u00a0<\/p>\n foreach::foreach(x = 1:length(vector), .combine = rbind, .inorder = T, .multicombine = F) %dopar%<\/em><\/p>\n {task}<\/em><\/p>\n \u00a0<\/p>\n This loop creates a vector of results.<\/p>\n If the order is not important you can increase performance by .inorder = F<\/em>. This means that a free processor takes the next iteration independent from the sequence of the iterations.<\/p>\n \u00a0<\/p>\n Or using a parSapply order:<\/p>\n \u00a0<\/p>\n clusters <- makeCluster(NumOfCores)<\/em><\/p>\n parSapply(cl = clusters, X = 1:length(vector), FUN = F, x = x, y = y, z = z,\u2026 )<\/em><\/p>\n \u00a0<\/em><\/p>\n In this case it is important to integrate the data in reference within the parentheses \u2013 you cannot directly connect to the workspace like in the ordinary sapply order.<\/p>\n \u00a0<\/p>\n RCPP:<\/p>\n \u00a0<\/p>\n Firstly you need to install RTools.<\/p>\n \u00a0<\/p>\n library(“Rcpp”)<\/em><\/p>\n \u00a0<\/p>\n define a function in C++, create a shared library and compile the code.<\/p>\n \u00a0<\/p>\n #include using namespace Rcpp;<\/em><\/p>\n \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0<\/em><\/p>\n \/\/ [[Rcpp::export]]<\/em><\/p>\n double NameOfFunction (NumericVector Vector)<\/em><\/p>\n {task}<\/em><\/p>\n \u00a0<\/p>\n Then you can call it in R:<\/p>\n \u00a0<\/p>\n sapply(X = 1:length(testVec), FUN = NameOfFunction, y = Vector)<\/em><\/p>\n \u00a0<\/em><\/p>\n But when to use which kind of loop?<\/p>\n \u00a0<\/p>\n Judging from the experience, I recommend to make the decision dependent from the number of iterations and the costs of each iteration.<\/p>\n \u00a0<\/p>\n \u00a0<\/em><\/p>\n<\/td>\n Not costly<\/em><\/strong><\/p>\n<\/td>\n costly<\/em><\/strong><\/p>\n<\/td>\n<\/tr>\n Low number of iterations<\/em><\/strong><\/p>\n<\/td>\n for-loop, while-loop<\/p>\n<\/td>\n RCPP, foreach<\/p>\n<\/td>\n<\/tr>\n Large number of iterations<\/em><\/strong><\/p>\n<\/td>\n RCPP, sapply, apply, lapply<\/p>\n<\/td>\n parSapply, RCPP<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n \u00a0<\/p>\n \u00a0<\/p>\n<\/div>\n Go to Source<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"\n
\n\n
\n \n \n \n \n \n \n \n \n \n \n \n