[R-gui] basic-statistics GUI for R

[John Fox]

Dear Martin,

At 10:02 AM 5/13/2003 +0200, you wrote:
>Thanks a lot! {finally something I can try on Linux! ;-)}
>
>I think your general concept --- work with one *active dataset*
>makes very much sense when a "simple easy GUI" is the goal.
>
>A few notes from playing (on Linux):
>- I miss the [Quit] or [Close] option in the [File] menu.

This was just laziness. I'll add a [Close] item and prompt for saving the log.


>- I'm running with R-patched and seem unable to get a dataset
>   from an attached package via the 1st entry of  [File].
>
>   The error message is a bit embarassing (I don't know for whom):
>
>   >> Warning: The use of _ is soon to be removed: you will be warned 
> repeatedly
>   >> Warning: The use of _ is soon to be removed: you will be warned 
> repeatedly
>   >> Error in parse(file, n, text, prompt) : parse error

Needless to say, none of this happens under Windows. Since I want the 
package to be portable across platforms, this is distressing.

I never use underscores, and so the warnings shouldn't originate from my 
code!  The only line in the function called by [File] -> [Read data set 
from an attached package] that involves a parse is

         check <- try(eval(parse(text=logger(paste("data(", dsnameValue, 
")", sep="")))),
             silent=TRUE)

and an error here should have been intercepted. It's hard to debug this by 
long distance, but I wonder if I could impose on you to try two things?

(1) get a traceback

(2) try to read data via [File] -> [Read data set from text file], which 
invokes read.table().

>- Hence I couldn't try much else but the [Distribution] menu
>   which was quite nice.

Not terribly exciting in the absence of data, I'm afraid.

>- I wonder if the  [Data] and [File] menus shouldn't be merged
>   or organized differently.
>   To play with Rcmdr, I'd want to load a standard dataset, i.e.
>   [File] -> [Read data set from attached package].
>   But then, the  *list* of available data sets is not available
>   there, but from the  [Data] menu only.

The general rationale was to put input and output operations under the File 
menu and other operations on data sets in the Data menu. I wouldn't want to 
merge the menus, but organizing them differently might make sense, even if 
it's only to duplicate one or two menu items.

>- "R CMD check Rcmdr"  (using R-patched on Linux) was ok apart
>   from the fact that many files had "wrong permissions", namely
>   they had the "x" (execution) flag set -- typical for files
>   created in windows.
>
>   To fix permissions for R packages, recently, Tony (Rossini)
>   posted a small sh-script to one of the Bioconductor lists --
>   which I have slightly improved and attach here.
>
>#!/bin/sh
>## Fix R permissions (unix shell script)
>find . -type d -print | xargs chmod 755
>find . -type f -a \! -name configure -print | xargs chmod 644
>
>---

I guess that I should add this information to my web page for Linux/Unix 
users (until the package is ready for contribution to CRAN).

>BTW, I started  "library(Rcmdr)"  from ESS  which is a bit
>awkward, but I'm sure nobody in your target audience will have
>this problem! ;-)
>But in any case, for a non-Windows environment (maybe even
>unconditionally for consistency ?),
>I think you (Rcmdr) should set
>    options(pager = tkpager)
>such that clicking the [Help] button will use a nice tcltk pager
>window.
>[Note that you should keep the previous pager setting and reset the
>  pager to that, once Rcmdr is left]

I like compiled html help in Windows, so perhaps I should add a switch 
determining whether one of the Windows help options is set, and use the 
tkpager otherwise.

Thank you very much for your comments,

  John



-----------------------------------------------------
John Fox
Department of Sociology
McMaster University
Hamilton, Ontario, Canada L8S 4M4
email: jfox at mcmaster.ca
phone: 905-525-9140x23604
web: www.socsci.mcmaster.ca/jfox