Skip to content

slow tcl/tk

7 messages · Adrian Dusa, John Fox, Brian Ripley +1 more

Original
Adrian Dusa
# 
Dear list,

Building my QCAGUI package using Rcmdr 1.3.0, under R 2.6.0 (using Kubuntu 
Gutsy) I noticed the library starts visibly slower than... somewhere in the 
past.

I used the same version of Rcmdr under previous versions of R and the tcl/tk 
window used to open instantly. Under the current version though, the window 
take seconds to open: first it builds the base window, then the menus and 
finally the complete window.

The whole process takes about five seconds.
I don't think it has anything to do with the Linux version, because I noticed 
the same behavior under the previous one as well.

Is there something changed about tcl/tk in R 2.6.0?

Thanks in advance,
Adrian
John Fox
# 
Dear r-devel list members,

As Adrian is aware, I've observed the same behaviour in R 2.6.0, in my case
under Ubuntu Gutsy. It's my impression that the slowness of dialogs to
display is true more generally of tcltk windows, but is particularly
apparent for the main Rcmdr window because, I suspect, of its complexity.

Since I'd like to insure that the Rcmdr works reasonably well under Linux
systems, it would help to know the source of the problem.

Thanks,
 John

--------------------------------
John Fox, Professor
Department of Sociology
McMaster University
Hamilton, Ontario
Canada L8S 4M4
905-525-9140x23604
http://socserv.mcmaster.ca/jfox 
--------------------------------
Peter Dalgaard
#
Adrian Dusa wrote:
Hmm, there's an event loop messup which was fixed in R-patched. The main
effect of that one was that scrollbars could go crazy. Not sure whether
it might cause your kind of symptoms, but you might check out the fixed
version.

It doesn't seem to make much of a difference here. Rather, I suspect
that Rcmdr itself got changed and a number of things that used to be
fixed are now generated programmatically. There's certainly a lot of
eval(parse(...)) stuff going on. (This sort of code always makes me a
little suspicious, but I haven't by far studied it well enough to say
whether there might be a more efficient way.)

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Brian Ripley
    

    

      
      #
    

  


  

      
On Wed, 31 Oct 2007, John Fox wrote:

        

            

      
      
      
    

        
We think it is solved in R-patched: can you try that?

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

        

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      John Fox
    

    

      
      #
    

  


  

      
Dear Peter,

            

      
      
      
    

        
It's possible that I'm missing something, but I don't think that there have
been recent changes to the Rcmdr that would account for this difference (and
note that I don't see it under Windows). In particular, I think that the
change coincided with R-2.6.0 (which seems to be Adrian's experience too).

            

      
      
      
    

        
The awkward eval-parses have always been there. There's probably a better
way to do at least some of this, but I haven't been able to think of it.

Regards,
 John

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Peter Dalgaard
    

    

      
      #
    

  


  

          

      
      
      
    

        
Yes, I realized that there could be other consequences of the change in
2.6.0. Moved to R-core for internal discussions.

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      John Fox
    

    

      
      #
    

  


  

      
Dear Peter,

Thank you for pursuing this.

Regards,
 John