[Rd] suppressing stderr output from system() calls

[Mark Cowley]

Thanks Martin, Simon and Brian,
They're all good suggestions for code that I write, however how can I suppress stderr from pre-existing code. For instance, install.packages has 3 system calls.
Perhaps a better worded question would be: how can I redirect the output of this command:

install.packages("MASS", lib="/tmp", type="source")

As I mentioned before, sink doesn't do the trick
con <- file("stderr.txt", "w")
sink(con, type="message")
install.packages("MASS", lib="/tmp", type="source")
sink(NULL, type="message")
close(con)

Martin will be familiar with the Bioconductor package oligo, which likes to download pd info package on the fly and install them - the embedded call to download.file is noisy, as is the resulting installation of said package.

Short of rewriting install.package & all code which has embedded install.packages calls, are there any other ways?

@Brian, thanks for the pointer to the news article on connections - I did learn something there
cheers,

Mark

On 06/10/2011, at 3:59 AM, Martin Morgan wrote:

> On 10/05/2011 02:36 AM, Mark Cowley wrote:
>> Dear list,
> 
>> I'm trying to suppress/redirect/squash the output from commands like
>> install.packages, or download.file. The problem is that none of:
>> sink(..., type="message"), sink(..., type="output"), capture.output,
>> suppressMessages are quite doing the trick. Output gets written to
>> the stderr stream, despite any combination of the above suppression
>> commands.
> 
> Hi Mark --
> 
> For download.file, the argument quiet=TRUE suppresses output messages; this option can be used in ... for install.packages, too (according to the documentation). This is good enough to quieten the regular chatter on a successful operation. To catch errors and suppress warnings also, maybe a construct like
> 
> url = "http://r-project.org/doesnotexist"
> tryCatch(suppressWarnings(
>    capture.output(download.file(url, tempfile(), quiet=TRUE))),
>    error=function(...) {})
> 
> Martin
> 
>> According to ?sink:
>> Messages sent to ‘stderr()’ (including those from ‘message’, ‘warning’ and ‘stop’) can be diverted by ‘sink(type = "message")’ (see below).
>> 
>> I'm pretty sure it's the system(), or .Internal() calls which are the culprit, which currently write the majority (all?) of their output to the stderr stream.
>> 
>> Simple example:
>> con<- file("stderr.txt", "w")
>> sink(con, type="message")
>> system("ls")
>> sink(NULL, type="message")
>> close(con)
>> # instead of the output going to stderr.txt, it gets printed to the console.
>> 
>> # no good either
>> capture.output(system("ls"))
>> character(0)
>> 
>> This is an issue, since i'm writing GenePattern modules to run R code, and if anything is written to stderr, then the job gets hit with a 'job failed' status, when all that might have happened is an R package got installed, or a file got downloaded via FTP.
>> 
>> Any ideas? Can system() and .Internal() output be redirected to stdout?
>> 
>> cheers,
>> Mark
>> 
>> 
>> sessionInfo()
>> R version 2.13.1 (2011-07-08)
>> Platform: x86_64-apple-darwin9.8.0/x86_64 (64-bit)
>> 
>> locale:
>> [1] en_AU.UTF-8/en_AU.UTF-8/C/C/en_AU.UTF-8/en_AU.UTF-8
>> 
>> attached base packages:
>> [1] stats     graphics  grDevices utils     datasets  methods   base
>>> 
>> 
>> 
>> 
>> -----------------------------------------------------
>> Mark Cowley, PhD
>> 
>> Pancreatic Cancer Program | Peter Wills Bioinformatics Centre
>> Garvan Institute of Medical Research, Sydney, Australia
>> -----------------------------------------------------
>> 
>> 
>> 	[[alternative HTML version deleted]]
>> 
>> 
>> 
>> 
>> ______________________________________________
>> R-devel at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-devel
> 
> 
> -- 
> Computational Biology
> Fred Hutchinson Cancer Research Center
> 1100 Fairview Ave. N. PO Box 19024 Seattle, WA 98109
> 
> Location: M1-B861
> Telephone: 206 667-2793