R: pass variable from R to unix - r

I am running an R script via bash script and want to return the output of the R script to the bash script to keep working with it there.
The bash is sth like this:
#!/bin/bash
Rscript MYRScript.R
a=OUTPUT_FROM_MYRScript.R
do sth with a
and the R script is sth like this:
for(i in 1:5){
i
sink(type="message")
}
I want bash to work with one variable from R at the time, meaning: bash receives i=1 and works with that, when that task is done, receives i=2 and so on.
Any ideas how to do that?

One option is to make your R script executable with #!/usr/bin/env Rscript (setting the executable bit; e.g. chmod 0755 myrscript.r, chmod +x myrscript.r, etc...), and just treat it like any other command, e.g. assigning the results to an array variable below:
myrscript.r
#!/usr/bin/env Rscript
cat(1:5, sep = "\n")
mybashscript.sh
#!/bin/bash
RES=($(./myrscript.r))
for elem in "${RES[#]}"
do
echo elem is "${elem}"
done
nrussell$ ./mybashscript.sh
elem is 1
elem is 2
elem is 3
elem is 4
elem is 5

Here is MYRScript.R:
for(iter in 1:5) {
cat(iter, ' ')
}
and here is your bash script:
#!/bin/bash
r_output=`Rscript ~/MYRscript.R`
for iter in `echo $r_output`
do
echo Here is some output from R: $iter
done
Here is some output from R: 1
Here is some output from R: 2
Here is some output from R: 3
Here is some output from R: 4
Here is some output from R: 5

Related

Data from hive to R to bash

Im trying to execute hive query in R using system command. This R function is called from a Bash script.
R Code
hivedata<-function(query)
{
data <- system(paste0("hive -S -e ", query), wait = TRUE,intern=TRUE)
if (identical(data, character(0))){data=NULL}
message("return value:")
message(data)
message("return value type:")
message(class(data))
return(cat(data))
}
if (length(query)>0 && is.na(query)==FALSE){
data=hivedata(query)
print(data)
}
Bash function
gethivedata(){
set -f #disable aterisk in sql getting expanded as filenames
query=$1
data=`Rscript hivedata.r "'$query'"`
echo $data
}
Calling function in Bash
totalcount=$(gethivedata " select count(*) from hivedb.hivetable ")
The outputs
[usr#host dir]$ totalcount=$(execute_sql " select count(*) from
hivedb.hivetable ")
return value:
0
return value type:
character
-------------------------------
[usr#host dir]$ echo $totalcount
0NULL
When cat is not used, the output value comes as [1]"0". Because R returns the index also with the output. When cat is used then the output becomes 0NULL. I want only the actual value which is "0"
What does [1] mean in the output of any command executed on R command line?

Returning an argument from R to bash

I have the following bash script in which an R script is called
#!/bin/bash
declare -x a=33
declare -x b=1
declare -x c=0
Rscript --vanilla MWE.R $a $b $c
echo $a $b $c
I want to modify the bash variables in the R script and return their modified values in the bash script because I am then passing the modified variables somewhere else. The R script is
#!/usr/bin/env Rscript
args = commandArgs(trailingOnly=TRUE)
Rb = as.numeric(args[2])
Rc = as.numeric(args[3])
Rb = Rb + 1
Rc = Rc + 1
args[2]=Rb
args[3]=Rc
print(c(args[1],args[2],args[3]))
However, the output of the print and echo respectively are:
[1] "33" "2" "1"
33 1 0
which shows that the new values aren't passed from R to bash. What am I doing wrong?
As Rscript does not allow environment variable manipulation you will need to capture the R output from the bash program.
One of the many possibilities is to use an array:
#!/bin/bash
declare a=33
declare b=1
declare c=0
declare -a RESULT
RESULT=($(Rscript --vanilla MWE.R $a $b $c))
a=${RESULT[1]}
b=${RESULT[2]}
c=${RESULT[3]}

Better string interpolation in R

I need to build up long command lines in R and pass them to system(). I find it is very inconvenient to use paste0/paste function, or even sprintf function to build each command line. Is there a simpler way to do like this:
Instead of this hard-to-read-and-too-many-quotes:
cmd <- paste("command", "-a", line$elem1, "-b", line$elem3, "-f", df$Colum5[4])
or:
cmd <- sprintf("command -a %s -b %s -f %s", line$elem1, line$elem3, df$Colum5[4])
Can I have this:
cmd <- buildcommand("command -a %line$elem1 -b %line$elem3 -f %df$Colum5[4]")
For a tidyverse solution see https://github.com/tidyverse/glue. Example
name="Foo Bar"
glue::glue("How do you do, {name}?")
With version 1.1.0 (CRAN release on 2016-08-19), the stringr package has gained a string interpolation function str_interp() which is an alternative to the gsubfn package.
# sample data
line <- list(elem1 = 10, elem3 = 30)
df <- data.frame(Colum5 = 1:4)
# do the string interpolation
stringr::str_interp("command -a ${line$elem1} -b ${line$elem3} -f ${df$Colum5[4]}")
#[1] "command -a 10 -b 30 -f 4"
This comes pretty close to what you are asking for. When any function f is prefaced with fn$, i.e. fn$f, character interpolation will be performed replacing ... with the result of running ... as an R expression.
library(gsubfn)
cmd <- fn$identity("command -a `line$elem1` -b `line$elem3` -f `df$Colum5[4]`")
Here is a self contained reproducible example:
library(gsubfn)
# test inputs
line <- list(elem1 = 10, elem3 = 30)
df <- data.frame(Colum5 = 1:4)
fn$identity("command -a `line$elem1` -b `line$elem3` -f `df$Colum5[4]`")
## [1] "command -a 10 -b 30 -f 4"
system
Since any function can be used we could operate directly on the system call like this. We have used echo here to make it executable but any command could be used.
exitcode <- fn$system("echo -a `line$elem1` -b `line$elem3` -f `df$Colum5[4]`")
## -a 10 -b 30 -f 4
Variation
This variation would also work. fn$f also performs substitution of $whatever with the value of variable whatever. See ?fn for details.
with(line, fn$identity("command -a $elem1 -b $elem3 -f `df$Colum5[4]`"))
## [1] "command -a 10 -b 30 -f 4"
Another option would be to use whisker.render from https://github.com/edwindj/whisker which is a {{Mustache}} implementation in R. Usage example:
require(dplyr); require(whisker)
bedFile="test.bed"
whisker.render("processing {{bedFile}}") %>% print
Not really a string interpolation solution, but still a very good option for the problem is to use the processx package instead of system() and then you don't need to quote anything.
library(GetoptLong)
str = qq("region = (#{region[1]}, #{region[2]}), value = #{value}, name = '#{name}'")
cat(str)
qqcat("region = (#{region[1]}, #{region[2]}), value = #{value}, name = '#{name}'")
https://cran.r-project.org/web/packages/GetoptLong/vignettes/variable_interpolation.html

two layers of quotes around a bash variable

I've written a bash script that makes a series of R scripts. However, Ive run into difficulty quoting a bash variable to echo to the R script as a file to be read into R. I have
echo "loadings_file <- $loadings ; calls_file <- $file" | cat - template.R > temp && mv temp $scriptname
$loadings and $file are files I want R to read in. But when I run it as is they end up in the R script with no quotes aroudn them for R to treat as a string. How do I make sure they're quoted in R but still expanded in bash first?
echo "loadings_file <- '$loadings' ; calls_file <- '$file'"
If you specifically need double quoting:
echo "loadings_file <- \"$loadings\" ; calls_file <- \"$file\""
You have to escape your quotes (\") around the variables:
echo "loadings_file <- \"$loadings\" ; calls_file <- \"$file\"" | cat - template.R > temp && mv temp $scriptname

Combining R + awk + bash commands

I want to combine awk and R language. The thing is that I have a set of *.txt files in a specified directory and that I don't know the length of the header from the files. In some cases I have to skip 25 lines while in others I have to skip 27 and etc. So I want to type some awk commands to get the number of lines to skip. Once I have this value, I can begin processing the data with R.
Furthermore, in the R file I combine R an bash so my code looks like this :
!/usr/bin/env Rscript
...
argv <- commandArgs(T)
**error checking...**
import_file <- argv[1]
export_file <- argv[2]
**# your function call**
format_windpro(import_file, export_file)
Where and how can i type my awk command. Thanks!
I tried to do what you told me about awk commands and I still get an error. The program doesn't recognize my command and so I can not enter the number of lines to skip to my function. Here is my code:
**nline <- paste('$(grep -n 'm/s' import_file |awk -F":" '{print $1}')')
nline <- scan(pipe(nline),quiet=T)**
I look for the pattern m/s in the first column in order to know where I have my header text. I use R under w7.
Besides Vincent's hint of using system("awk ...", intern=TRUE), you can also use the pipe() function that is part of the usual text connections:
R> sizes <- read.table(pipe("ls -l /tmp | awk '!/^total/ {print $5}'"))
R> summary(sizes)
V1
Min. : 0
1st Qu.: 482
Median : 4096
Mean : 98746
3rd Qu.: 13952
Max. :27662342
R>
Here I am piping a command into awk and then read all the output from awk, that could also be a single line:
R> cmd <- "ls -l /tmp | awk '!/^total/ {sum = sum + $5} END {print sum}'"
R> totalsize <- scan(pipe(cmd), quiet=TRUE)
R> totalsize
[1] 116027050
R>
You can use system to run an external program from R.
system("gawk --version", intern=TRUE)

Resources