On 13-10-22 10:45 PM, Duncan Murdoch wrote:
On 13-10-22 9:45 PM, Michael Sumner wrote:
Hello, if I interrupt Sweave while it's processing a file it seemingly
leaves an open sink connection that hides printed output.

Can this be changed to reset the sink on exit?  I've been baffled by
this for years.

This is seen in Windows (R Under development (unstable) (2013-10-20
r64082))  and an older Linux (R version 3.0.0 (2013-04-03)).

Run the code below in two parts with a manual interrupt to 1) to see it.

That's a bug in the Rweave driver.  It runs the code in try() so that it
can catch errors and undo the sink, but try() doesn't catch user
interrupts, so it never gets undone.

Shouldn't be too hard to fix...

It should now be fixed in R-devel and R-patched.

Likely the reason this went unfixed for so long is that the more common way to use Sweave is in a separate R session. If you run it in the current R session, the results aren't necessarily reproducible, because they may depend on whatever variables you have in your workspace. It's better to run it on its own. One way is to use the command line version

R CMD Sweave ...

but it can also be done by piping specific commands into an R session. I do it that way, because it lets me run some project management code, and lets me patch the Synctex information so it points to the .Rnw file.

Duncan Murdoch


Duncan Murdoch


Cheers, MIke.

## 1)

## this code creates a temporary file to run Sweave

## interrupt this code before Sweave() finishes

txt <-  c("\\documentclass[a4paper]{article}", "\\title{Sweave bail out}",
"\\author{M. Sumner}", "\\begin{document}", "\\maketitle", "",
"Run a loop and bail out when Sweave()ing.", "", "<<>>=",
"for (i in seq_len(1e6)) {", "    if (i %% 1000 == 0)
print(sprintf(\"%i\", i))",
"    Sys.sleep(0.5)", "}", "@", "", "\\end{document}")

f <- tempfile()
writeLines(txt, f)

Sweave(f)

## 2)
## now no printed output is seen
print(1)
##

sink(NULL)

## now it's back
print(1)
## [1] 1

## tidy up
## unlink(f)







______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

Reply via email to