IntegerSort

This package uses RAM based algorithm such as radix sort and counting sort and is designed to manipulate large data. For opposite target, insertion sort algorithm is added to operate small data set. Such algorithms give more performance when it used with uniform data and are considered as low level tools (C/C++ resources ) which leads more vigilance and good practice of user.

Installation

IntegerSort package need compilation (basically written with Rcpp) and can be installed from github using devtools :

require(devtools)
devtools::install_github("EMacherki/IntegerSort")

Some time we have to force installation if we won't to update. In this case we have turne force to TRUE in the 'devtools::install_github' function.

require(devtools)
devtools::install_github("EMacherki/IntegerSort",force=TRUE)

After installation restart R session and the package is running.

Benchmarking

counting sort

> x<-sample(1:10000,10000000,TRUE)
> # x is a large integer vector with supported range 
> system.time(sort(x))
utilisateur     système      écoulé 
       0.58        0.05        0.62 
> system.time(countSort(x))
utilisateur     système      écoulé 
       0.05        0.02        0.07 
> all.equal(countSort(x),sort(x)) # return TRUE
[1] TRUE
> # in-place running
> system.time(countSort(x,inplace= TRUE)) 
utilisateur     système      écoulé 
       0.05        0.00        0.05 
> all.equal(sort(x),x)  #return TRUE
[1] TRUE

lsdSort

> x<-isample(10000000)
> # x is an uniform  sample of intgers
> system.time(sort(x))
utilisateur     système      écoulé 
       1.31        0.00        1.31 
> system.time(lsdSort(x))
utilisateur     système      écoulé 
       0.18        0.03        0.20 
> all.equal(sort(x),lsdSort(x)) # return TRUE
[1] TRUE
> system.time(lsdSort(x,TRUE))  # testing in-place
utilisateur     système      écoulé 
       0.14        0.00        0.16 
> all.equal(sort(x),x)          # return TRUE 
[1] TRUE

insertSort

For this example we use microbenchmark package:

> x<-isample(60)
> microbenchmark::microbenchmark(insertSort(x),lsdSort(x),sort(x))
Unit: microseconds
          expr    min      lq     mean median      uq     max neval
 insertSort(x)  2.994  3.4220  4.50364  3.850  5.5600   7.271   100
    lsdSort(x)  5.559  5.9880  7.21058  6.415  8.1260  25.659   100
       sort(x) 65.428 67.5665 70.04705 67.995 69.4915 160.791   100

imedian

This function calculate median:

x<-isample(1e7)
all.equal(imedian(x),median(x)) # return true
> system.time(median(x))
#utilisateur     système      écoulé 
#       0.19        0.05        0.25 
system.time(imedian(x))
#utilisateur     système      écoulé 
#       0.03        0.00        0.03  

pth

This function find pth element in the sorted vector:

x<-isample(1e6)
all.equal(pth(x,5000),sort(x)[5000]) #return TRUE
 

pth_pair

This function find the mean of pth and pth-1 element in the sorted vector:

x<-isample(1e6)
y<-(as.numeric(sort(x)[5000-1])+as.numeric(sort(x)[5000]))/2
all.equal(pth_pair(x,5000),y) #return TRUE
 

NA_rm

Remove NA values in the vector. It is very useful because this package considered by default NA as the smallest integer value:

x<-as.integer(c(NA,NA,88))
NA_rm(x) #return 88