Re: Octave graphics bugs and thoughts

[Daniel J Sebald]

Ben Abbott wrote:
> On May 6, 2009, at 6:00 PM, Daniel J Sebald wrote:
>
> > Ben Abbott wrote:
> >
> > > On May 5, 2009, at 11:01 PM, Daniel J Sebald wrote:
> >
> >
> > > In any event, regarding the issue with the fontsize for eps- output,  
> > > the gnuplot manual indicates that fontsize is scaled down  by a 
> > > factor  of 2. We can compensate for that in print.m
> > > Please comment on the attached changeset. I've compared it to ps   
> > > output and the fontsizes look consistent to me.
> >
> >
> > It basically emphasizes what you found, that the fonts aren't  
> > positioned correctly and land on top of the axis line.  (Actually,  
> > the tick labels should be right/left aligned appropriately depending  
> > on their relation to the graph, so that problem really should be  
> > addressed in a different way.)  My solution was to write an exterior  
> > version of print() that will scale down the paper size if EPS then  
> > call the normal print().  Fonts come out about right, but the tick  
> > marks are too big.
> >
> > But, placing conditional "if EPS" etc. scattered about isn't the  
> > desired way to go.
> >
> > We'll have to think about this a bit.
> >
> > Dan
>
>
> Dan,
>
> If I understand your conclusion we need to ...
>
> (1) Scale the paperposition by 2x
> (2) Scale the fontsizes by 2x
> (3) Scale the ticklength by 0.5x
>
> Do I understand you correctly?
>
> If so, I think that can be done without to much effort.

I set the "papersize" property to half the normal size.  (It's the same 
strategy as using a subplot, rather than fullsize plot, to generate graphs as that has 
larger relative fonts.)  Anyway, I agree that making those changes isn't difficult, but I 
think it would be better to put effort into a more robust method of achieving these 
things.  This weekend I will put together some simple scripts to describe what I mean.

Thanks Ben,

Dan