plyr-parallel-processing

library(plyr)

#testing gist change

#LLPLY
# Create a list with 2 elements
l <- list(a = 1:10, b = 11:20, c=21:30, d=31:40)

# Apply a function to each element of the list
system.time(lapply(l, mean))

system.time(
  lapply(l, FUN=function(x){
    # THIS FUNCTION TAKES A LONG TIME!
    Sys.sleep(2)
    mean(x)
  }
  )
)

# register parallel backend
# MAC VERSION
#library(doMC)
#registerDoMC(2)
#unregister <- function() {
# env <- foreach:::.foreachGlobals
# rm(list=ls(name=env), pos=env)
#}
# WINDOWS VERSION
library(doSNOW)
cl <- makeCluster(4, type = "SOCK")
registerDoSNOW(cl)
system.time(
  suppressWarnings(
    llply(.data=l, .fun=function(x){
      Sys.sleep(2)
      mean(x)
    }, 
    .parallel=TRUE)
  )
)
# WINDOWS VERSION
stopCluster(cl)
# MAC VERSION
#unregister()

#LDPLY
iris_split_datasets <- list('part1'=iris[1:50,],
                            'part2'=iris[51:100,],
                            'part3'=iris[101:150,])
combined_data <- ldply(.data=iris_split_datasets, .fun=function(x){x})
nrow(combined_data)
names(combined_data)
tail(combined_data)

#DLPLY
models <- dlply(.data=iris, .variables=.(Species), .fun=function(x){
  linear_model <- lm(Sepal.Length~Sepal.Width, data=x)
  log_linear_model <- lm(log(Sepal.Length)~Sepal.Width, data=x)
  list('linear_model'=linear_model, 'log_linear_model'=log_linear_model)
})
names(models)

#NESTED LLPLY
llply(names(models), .fun=function(x){
  species <- models[[x]]
  l <- llply(names(species), .fun=function(x){
    summary(species[[x]])$r.squared
  })
  names(l) <- names(species)
  l
})

#DDPLY
datas <- ddply(.data=iris, .variables=.(Species), .fun=function(x){
  x$new_var <- x$Sepal.Length + x$Petal.Width
  x
})
head(datas)