I have added fontfamily to the "graphics context" information that can be passed from the graphics engine to graphics devices. Two rather important things remain to be done:
(i) the devices need to do something with this information. As Duncan Murdoch pointed out, Windows already allows something similar via a device-specific approach to the existing font parameter. I have suggested (but not yet tried to implement) a more device-independent approach in http://www.stat.auckland.ac.nz/~paul/R/gpars.html (linked off the "R Developer" web page). I haven't really had feedback on this suggestion yet and I intended to elicit more before hacking any code.
I'm quite fond of the new R_GR_gcontext approach. As of fonts, the fontdb idea sounds good to me, too. The alternative would be similar to what Java does: define generic types like "Serif" as strings and interpret anything else as device specific (but fontdb seems cleaner to me, because R can compute on the list).
The only thing I'm not quite sure about is the alpha handling. I don't know if I got this right, but wouldn't an alpha slot (or slots) collide with the RGBA specification? I mean: what is the result of col=0x80808080, colalpha=0.5 supposed to be? IMHO adding alpha as additional parameter to rgb and friends sounds appealing to me and preferable to additional gc entries. I would argue that col="red" is fully opaque red and if you need alpha, you need to specify it in an another way (not all devices support alpha anyway...). However col="#0000ff80" should be easy to support, too. Additional alpha flag in gc would have the benefit of easily obtainable "ghost" plots, but the question is whether this would not be better handled by the R side and not the device side...
(ii) the graphics systems (base and grid) need to allow the user to set this information. grid has a fontfamily gpar, but base has nothing yet. Again a suggested implementation is in http://www.stat.auckland.ac.nz/~paul/R/gpars.html and there is a bit more in http://www.stat.auckland.ac.nz/~paul/R/fonts.html
No objections here - it sounds like a sensible proposal to me...
Just my €0.02 ...
Simon
______________________________________________ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-devel
