Search code examples
remacssweave

Emacs, R, Sweave: UTF-8 encoding is not recognized when launching Sweave through Emacs

I work with Emacs 24 on Mac OS X 10.7.3 with R 2.14.0. I have a file foo.Rnw containing 

\documentclass[
paper=a4,% 210mm × 297mm
pagesize% write page size to dvi
]{article}

\usepackage[american]{babel}
\usepackage{fancyvrb}

\usepackage[utf8]{inputenc}

\SweaveOpts{engine=R}
\SweaveOpts{pdf=TRUE}

\begin{document}
Some text
<<>>=
sessionInfo()
@
\end{document}

My .emacs for working with Sweave looks like this:

(add-hook 'Rnw-mode-hook
  (lambda ()
    (add-to-list 'TeX-command-list
         '("Sweave" "R CMD Sweave %s"
           TeX-run-command nil t :help "Run Sweave") t)
    (add-to-list 'TeX-command-list
         '("Stangle" "R CMD Stangle %s"
           TeX-run-command nil t :help "Run Stangle") t)
    (setq TeX-command-default "Sweave")))

If I use C-c C-c and then choose Sweave, the .tex file looks like this:

\documentclass[
NA
pagesize% write page size to dvi
]{article}

\usepackage[american]{babel}
\usepackage{fancyvrb}

\usepackage[utf8]{inputenc}




\usepackage{Sweave}
\begin{document}
Some text
\begin{Schunk}
\begin{Sinput}
> sessionInfo()
\end{Sinput}
\begin{Soutput}
R version 2.14.0 (2011-10-31)
Platform: x86_64-apple-darwin11.2.0/x86_64 (64-bit)

locale:
[1] C

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

loaded via a namespace (and not attached):
[1] tools_2.14.0
\end{Soutput}
\end{Schunk}
\end{document}

As you can see, the second line of foo.Rnw containing the non-ASCII character × is replaced by NA. I could figure out that this has to do with the localization in which R runs (see also the output of locale: [1] C. I could include statements of the form export LC_MESSAGES="en_US.UTF-8" in my .bashrc so that R gets the correct locale. I can also perfectly use R CMD Sweave foo.Rnw from the terminal (which of course loads .bashrc and thus sets locale correctly). The question is: How can I make it work through Emacs, so that I can run Sweave via C-c C-c Sweave? I know that Emacs is not aware of variables set in .bashrc , but how can I guarantee that R CMD Sweave issued from Emacs runs with the same localization as when I execute it from the terminal?

I found several posts to this problem, but no solution for Macs and from within Emacs. Again, note that I can perfectly run R CMD Sweave foo.Rnw from the terminal (due to my settings in .bashrc) -- most of the posts address this issue -- it's just not working from within Emacs (which is quite annoying).

UPDATE:

My ~/.bashrc contains:

export LC_COLLATE="en_US.UTF-8"
export LC_CTYPE="en_US.UTF-8"
export LC_MONETARY="en_US.UTF-8"
export LC_TIME="en_US.UTF-8"
export LC_MESSAGES="en_US.UTF-8"

Following the advice of Yihui, I put in the following in my ~/.Rprofile:

invisible(Sys.setlocale("LC_COLLATE", "en_US.UTF-8"))
invisible(Sys.setlocale("LC_CTYPE", "en_US.UTF-8"))
invisible(Sys.setlocale("LC_MONETARY", "en_US.UTF-8"))
invisible(Sys.setlocale("LC_TIME", "en_US.UTF-8"))
invisible(Sys.setlocale("LC_MESSAGES", "en_US.UTF-8"))

This solved the problem.

Solution

  • The final answer can be found under UPDATE.

    @cbeleites: I do not run Mac OS X anymore (neither R 2-14) so I can't tell.