Re: Yet Another Plotting System for Octave: 0.3.0

[Shai Ayal]

On 10/6/07, Søren Hauberg <address@hidden> wrote:
> Shai Ayal skrev:
> > This is basically what is used by all OpenGL systems -- octplot &
> > jhandles included. It is a highly non-trivial taks to "render" 3D
> > OpenGL in postscript, and gl2ps is the best tool to do it. All the
> > problems actually stem from OpenGL, not the conversion to ps. As I
> > noted elsewhere, using OpenGL & gl2ps will give OK results, but not
> > 100% professional -- e.g. lines will not always join nicely, and some
> > numerical glitches may be noticeable -- OpenGL uses no more than
> > floats to do all it's calculations. This is most noticeable in 2D
> > plots, less so in 3D plots.
> > However, all that being said, the ps results are quite nice, and are
> > truely resolution independant vector graphics.
> I'm curious: do you believe that gl2ps provides good enough output to be
> used as the only way to create postscript output in Octave in some
> distant future? Or can it only be a temporary solution for the GL based
> projects?

I think the problem is OpenGL. It is not a high quality hardcopy
producing library -- it is a high speed screen 3D scene renderer. As
an example -- OpenGL has no concept of how to join lines -- take a
look at:
http://cairographics.org/samples/fill_and_stroke2/
In OpenGL, all the places where lines join at not 90 deg, it would
look bad (and probably even at places with 90 deg joins). A "real" 2D
hardcopy oriented library has facilities for this :
http://cairographics.org/samples/set_line_join/.
However as much as I like cairo (could you guess?) it has no support
whatsoever for 3D, and the people want 3D, and you must give the
people what they want ...

I am far from a graphics expert, but I am not aware of any 3D library
which does this (actually they are all based on OpenGL)

Shai