[ESS] ESS in Windows: Unable to run editor "gnuclient.exe"

[Paul Johnson]

Hello, Rich:

I'm very happy to get such a quick answer.

On Tue, Mar 1, 2011 at 10:45 PM, Richard M. Heiberger <rmh at temple.edu> wrote:
> Paul,
>
> you hit two issues.
>
> 1. ESS by default finds R automatically when R is installed in the standard
> places
> defined by R-Core which are (in English locales, it also does the right
> thing in other languages)
> c:/Program Files/R/R-*/
> When you place R in a non-standard place such as c:/Program Files/R/, a
> workaround
> like the one that you used is necessary.  It would be better to put that
> statement
>   (setq-default inferior-R-program-name "c:/Program
> Files/R/bin/i386/rterm.exe")
> in your .emacs, rather than make changes to ess-site.el.  This type of
> workaround is
> illustrated in ess-site.el.
>

OK, well at least I know I inflicted it on myself.  On this one, I
think I'm right and the world is wrong.

R Core doesn't ask for version specific installation folders on Unix
or Mac.   Why Windows? With the most elementary users being clustered
in Windows systems, the accumulation of multiple editions of R is a
support hassle. Only experts want multiple versions, and they could
get that for themselves at install time if they wanted it.

But I've lost that argument a lot of times.
>
> 2. The gnuclient issue.  We have gnuclient hardwired in the code for
> Windows.
> The newer emacs distributes with emacsclient.
>
> You can manually issue the command to R,
>
>    options(STERM='iESS', editor='emacsclientw.exe')
>
> I don't like emacsclient as much as I liked gnuclient.  If you still have an
> old gnuclient
> on your machine you can use it by starting it in your .emacs file.
>

I never heard of "gnuclient" before, don't know what its for, but I'm
sorry I don't have it :)

> ESS intercepts R page() commands at the command line and open them in emacs.
> That is somewhat easier since the pager doesn't have a return value.
>
> It is not obvious to me why you would want to use the R fix() or edit()
> command when you
> are working in the emacs environment.
>


Insert "browser()"!  You already knew, I'm sure, but I just learned
this.  To me it is quite fantastic. In R Extensions manual, it
describes ability to insert "browser()" into code for existing
functions.  To test, I did it to famous ones

> fix (lm)

and, sure enough, I can edit the *actual code* of the lm function, put
browser() in, close the editor, run the function, and then *really
learn* something. But now, it seems quite awesome to me. Its like gdb
on super steriods.  Its about the most obvious benefit of "R is an
interpreted language" that I've ever seen.  To stop a function from a
famous package, and then interact with its variables?  Too cool.

I need this to "just work" in the Windows lab.  It does in Linux. But
in Windows, well, frustrating. I don't mind having to use windows
notepad to do the editing, I just need this to work without users
having to run

> options (editor="whatever")

That's a deal killer.

Couldn't you change from 'gnuclient' to 'notepad'?  we are sure that
exists everywhere.

Couldn't you set it to "runemacs.exe". I don't mind if another
instance gets spawned.  I don't mind starting the server, if
necessary, as long as I can configure it so users never see it.

Anyway,

I figured, "Why not edit Rprofile.site?".  Users will never have to
worry about it.   But Emacs/ESS ignores Rprofile.site?   If I set the
editor "notepad" in there, that will fix it.  But no! When I'm in
ordinary R (not via ESS), then options("editor") shows notepad and it
uses it for functions.  But when I start Emacs/Ess, options("editor")
still says "gnuclient.exe".


Until now, I couldn't understand why there's so much push for Windows
people to use "Tinn-R".

Check out the R for beginners by Zuur from Springer publications, or
Maindonald's notes "Installation of R, of R packages, and Editor
Environments"  http://maths.anu.edu.au/~johnm/r-book/xtras/setup.pdf.

I don't want Windows people to get "stuck" to their OS by Windows-only
solutions, that's why I make them all try to use Emacs, even though
the've never edited a file before :)

> Question for other ESS-core members:  would it make sense to intercept the
> edit()
> and fix() commands?  See defun inferior-R-input-sender in ess-inf.el for
> details.
>
>
> Rich
>


-- 
Paul E. Johnson
Professor, Political Science
1541 Lilac Lane, Room 504
University of Kansas