[R-gui] R tcltk/tcltk2 vs. R GTk2

[Dan Putler]

Hi Phillipe,

An excellent and well argued post. One thing I'm curious about is where
does the RwxWidgets package (http://www.omegahat.org/RwxWidgets) fit
into the mix? This seems to potentially deal with some of the OS X
issues. In any event, I completely agree that the gWidgets approach of
creating an interface that is both "R like" and separates R from the
underlying widget tool kit is the correct one.

Dan

On Wed, 2007-01-08 at 12:09 +0200, Philippe Grosjean wrote:
> Just my 0.02$ added to this discussion: tcl/tk could be automatically 
> installed under Windows (if the option is checked), and tile is readily 
> usable once the tcltk2 package is installed. Also, using tcltk2, you 
> have more Windows look&feel than with Gtk2. There are also a couple of 
> other advantages of tcltk on specific tasks, mainly thanks to nice 
> aspects in the underlying tcl language readily available. So, under 
> Windows, tcltk (+ tcltk2) solution is much more easily in operation on 
> any users computer than a solution based on RGtk2.
> 
> For Linux, Gtk2 is certainly at home, although there is special tile 
> version in development to support Qt widgets, I think.
> 
> For MacOS X neither provide the appropriate look&feel. There a special 
> Tcl/Tk distro installed on all MacOS X (at least on Tiger), but the 
> regular R distribution with the AQUA GUI cannot use it. It is possible 
> to compile R to use it and then, you got an excellent tcltk + tcltk2 
> solution with MacOS X look&feel that is almost perfect. However, this is 
> something only for experienced users, while most GUIs target beginners.
> 
> For the rest, I pretty much agree that Gtk2 is much more powerful, but I 
> am not sure we need all its power for 99% of the simple GUIs stuff 
> required on top of R.
> 
> We should end the widgets war, and start looking at gWidgets by John 
> Vezani, which intend to provide an interface to GUI widgets from R 
> independent from a particular implementation. At the moment, there is 
> only an implementation of gWidgets interface for RGtk2 (and one for 
> Java?),... but an implementation for tcltk (/tcltk2) should not be too 
> difficult to do. I may work on it once I have some time, that certainly 
> not in the near future!!!
> 
> Anyway, I will continue to improve tcl/tk support through tcltk2 when 
> time permit, and since I have now a Mac with MacOS X + Windows + Linux 
> installed on it, I will work on more portable code than before (I was 
> only working under Windows).
> 
> Really, the effort should be targetted to make R GUIs independent from a 
> particular GUI widget and start coding or recoding stuff using gWidgets 
> and work on both RGtk2 and tcltk/tcltk2 implementations. Currently, I 
> should consider this, personnally:
> - Linux: RGtk2 for sure,
> - Windows: tcltk+tcltk2 for ease of installation and better Windows 
> installation, or RGtk2 if you need more features than those provided by 
> tcltk/tcltk2,
> - MacOS X: nothing really satisfying currently.
> 
> Best,
> 
> Philippe
> ..............................................<°}))><........
>   ) ) ) ) )
> ( ( ( ( (    Prof. Philippe Grosjean
>   ) ) ) ) )
> ( ( ( ( (    Numerical Ecology of Aquatic Systems
>   ) ) ) ) )   Mons-Hainaut University, Belgium
> ( ( ( ( (
> ..............................................................
> 
> Peter Dalgaard wrote:
> > Michael Lawrence wrote:
> >> GTK+ has many more features than tcl/tk (tile included), though tcl/tk does
> >> have some features that GTK+ lacks, like a canvas widget.
> >>
> >> The API of the underlying toolkits, as well as the R interfaces, are quite
> >> different. I've heard many people say that they prefer the GTK+/RGtk2 style,
> >> though there are probably many people that feel the opposite. The tcltk
> >> package is a binding to the Tcl language, while RGtk2 is a binding to a
> >> collection of C libraries.
> >>   
> > Minor quibble: tcltk (mostly) binds directly to the Tcl *engine*, 
> > bypassing the parsing layer, $-substitution, quoting hell, etc. The 
> > Tcl/Tk engine is just another collection of C libraries. The major 
> > difference to Gtk+ is that the Tcl/Tk engine was designed to support 
> > interpreted languages (mainly, of course, Tcl).
> > 
> > Apart from that, I pretty much agree that RGtk2 could well be the 
> > toolkit of the future. It's main detraction at this point is that it is, 
> > at the tutorial level, mainly documented by demos and that it seems to 
> > be a bit sensitive to changes in the underlying toolkit and have a 
> > couple of rough edges in general. But it can certainly do amazing stuff; 
> > all it needs is a dedicated effort to take care of a number of little 
> > issues (esp. portability). That and an introductory document that covers 
> > the concepts involved.
> > 
> > (It needs to be no secret that I have been approached about the 
> > possibility of a book on R/tcltk, and ended up resolving that there 
> > would certainly be a market, but of the 3 or 4 candidates to write it, 
> > probably none would want to. However, in due course, when the obvious 
> > author completes his academic obligations, a book on RGtk2 should be a 
> > very interesting prospect.)
> > 
> >> RGtk2 also binds libglade, meaning that one can design a GUI graphically
> >> using the Glade tool and display it using R. Glade may or may not be useful,
> >> depending on the application. The Rattle tool seems to have used Glade
> >> effectively, but I personally prefer to construct the interfaces
> >> dynamically.
> >>
> >> Since GTK+ is object-oriented, RGtk2 is able to support the creation of new
> >> widget classes that override the functionality of their parents. This is a
> >> neat feature, but is rarely necessary.
> >>
> >> For graphics, RGtk2 binds several useful libraries, including Cairo (vector
> >> graphics) and GdkPixbuf (image manipulation).
> >>
> >> You might also consider the gWidgets package. It is a convenient API for
> >> creating simple GUIs in R. It abstracts away the underlying toolkit.
> >> Implementations exist using RGtk2 and rJava (swing). You could always start
> >> using a GUI this way and then convert to RGtk2 or some other toolkit later
> >> if necessary.
> >>
> >> I hope this helps,
> >> Michael
> >>
> >> On 7/28/07, morphwj at comcast.net <morphwj at comcast.net> wrote:
> >>   
> >>> Why use R tcltk/tcltk2 instead of R GTk2, or why use R GTk2 instead of R
> >>> tcltk/tcltk2?
> >>> Bill Morphet
> >>>         [[alternative HTML version deleted]]
> >>>
> >>> _______________________________________________
> >>> R-SIG-GUI mailing list
> >>> R-SIG-GUI at stat.math.ethz.ch
> >>> https://stat.ethz.ch/mailman/listinfo/r-sig-gui
> >>>
> >>>     
> >> 	[[alternative HTML version deleted]]
> >>
> >> _______________________________________________
> >> R-SIG-GUI mailing list
> >> R-SIG-GUI at stat.math.ethz.ch
> >> https://stat.ethz.ch/mailman/listinfo/r-sig-gui
> >>
> > 
> > _______________________________________________
> > R-SIG-GUI mailing list
> > R-SIG-GUI at stat.math.ethz.ch
> > https://stat.ethz.ch/mailman/listinfo/r-sig-gui
> >
> 
> _______________________________________________
> R-SIG-GUI mailing list
> R-SIG-GUI at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-gui
-- 
Dan Putler
Sauder School of Business
University of British Columbia