On Sat, Feb 27, 2010 at 10:18:23AM +0100, Peter Dalgaard wrote: > Jens Elkner wrote: ... > >Yes - still a little bit big (most are usually not bigger than 50K, but > >some exceptions hav 80..100K), but much better than 2M and sufficient > >IMHO, too. ... > I suppose I should put it somewhere for others to play with. It should
Yes, good idea! > be possible to reduce the number of colour layers and get down to the > 50 K or so, but it is quite fiddly. I found that if you put inkscape in > "outline mode", then you can manipulate the Bezier curves directly and > get rid of the artifacts from the bitmap tracing. Currently, it has > about 64 paths of about 60 control points each, so it might be a better > strategy to start over with a coarser trace. Well, as said, for my taste it is good enough. However, to get a real small good one, I think it should be made from scratch and not based on a scanned pixmap. BTW: What do you think about Baptiste's work ( http://www.maths.lancs.ac.uk/~rowlings/Graphics/Logo/R/logos.svg )? IMHO looks cool, sharp, modern and simple enough to be used in any vector drawing program efficiently. > >http://dev.cs.uni-magdeburg.de/lnf/i386/R/LNFr-desktop/ ) -> looks > >update-desktop-database ; update-mime-database $instdir/share/mime ... > Any chance of a couple of screenshots? Yes, can make them on monday at work. @home I have to use a macbook right now :( (my Linux WS died few month ago and 've had not yet the time to look for a new motherboard and stuff ...) > >Currently the last thing I'm thinking about is: "with '-g Tk' start > >Rcmdr if available, otherwise tkStartGUI()". Would you mind to integrate > >that into the official R tree? I know, its more or less easy to modify > >library/tcltk/exec/Tk-frontend.R, however admins as well as package > >creators/maintainers don't like, when one package modifies the content > >of another one and have to keep track of changes, i.e. need to do more > >or less "sophisticated" things, when they get removed ... > > This sounds a bit far-reaching. The "-g Tk" route is a bit of a kludge > already, so everything is in for a review and I'd rather not build > things on top of a crumbling foundation. Yes, actually I though about something generic as well: e.g. an etc/gui directory where everybody may put the startup files for the GUI like etc/gui/ Tk.R Rcmdr.R ... So "R -g {Tk|Rcmdr}" would be valid ... However, AFAIK these are the only GUIs, which can be lunched from R. All others are embedding R, so the add. work might not be worth it. > You might be able to do something with environment variables and > .Rprofile files instead. That's not an option for desktop integration - having to manipulate each user's profile is certainly a really bad thing. > (I was about to suggest 'R -e something' but > Rcmdr refuses to start in a non-interactive session). Yes, echo "library('Rcmdr')" | R --interactive... doesn't work either, because it than goes crazy (looks like a kind of BATCH mode). BTW OT: wondering why one just can't pipe its script into R. IMHO it should be possible and after having the code processed it should stay in interactive mode (giving an R prompt) so that one may continue to work. Only if CMD BATCH or something like that is given, it should auto exit. The "Fatal error: you must specify '--save', '--no-save' or '--vanilla'" is IMHO completely non-sense, since everybody is able to start simply 'R' (i.e. no option at all) and enter stuff. Actually normal progs don't really care, where the input comes from ... Anyway, will be back on monday with some screenshots ... Regards, jel. -- Otto-von-Guericke University http://www.cs.uni-magdeburg.de/ Department of Computer Science Geb. 29 R 027, Universitaetsplatz 2 39106 Magdeburg, Germany Tel: +49 391 67 12768 ______________________________________________ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
