Various questions about plotutils..

[Theodore Papadopoulo]

        Hi,

I wanted to use libplotter to display an image. While I know that it 
is not made for this purpose, my final goal is to plot some text and 
symbols over the image, for which libplotter is an option. Having the 
portability of plotutils against various device decided me that it 
should be nice to use the plotutils to do all the job. Unfortunately,
I ran into several problems:

First, given the following code:

#include <iostream>
#include <plotter.h>

using namespace std;

int
main(int argc,char *argv[])
{    
    //  Initialize the plotter.

    PlotterParams params;
    params.setplparam("PAGESIZE",const_cast<void*>(static_cast<const 
void*>("A4")));
    params.setplparam("X_AUTO_FLUSH",const_cast<void*>(static_cast<const 
void*>("no")));
    
params.setplparam("USE_DOUBLE_BUFFERING",const_cast<void*>(static_cast<const 
void*>("yes")));

    XPlotter plotter(cin,cout,cerr,params);
    if (plotter.openpl()<0) {
        cerr << "Couldn't open Plotter\n";
        return 1;
    }

    plotter.fspace(0.0,0.0,538,634);
    plotter.flinewidth(0.0);
    plotter.pencolorname("black");
    plotter.erase();

    for (unsigned i=0;i<539;++i)
        for (unsigned j=0;j<634;++j)
            plotter.point(i,j);
        
    //  Close the plotter.

    if (plotter.closepl()<0) {
        cerr << "Couldn't close Plotter\n";
        return 1;
    }

    return 0;
}

First, it is quite slow and worse I get white lines in the images due to
some rounding effects with the coordinates after transform. This bug 
disappears if I'm using boxes instead of points, but then it is even 
slower...

While the bug can be corrected, I do not think that the speed can be 
improved significantly using the primitives available. Then, 
realizing that most vector page descriptions (at least postscipt and 
SVG) provide a way to embed images into the description I wonder 
whether such an extension would be considered. Basically, I believe 
that something like:

image(data,sizex,sizey,urx,ury,ulx,uly)

where data is an array of (sizex x sizey) pixels, (urx,ury) and 
(ulx,uly) are the respective coordinates of the box in the viewport 
in which the image should be mapped would be nice to have...

If this is acceptable, I can certainly work out something generic 
(along the lines of the above code) + versions for X windows and 
Postscript. I know too little of the other formats to be sure what I 
can do/what can be done.

To summarize, I guess this mail ask for answers to the following 
questions:

- Can be white line bug be corrected (I suspect that it is difficult 
as it seems to be some rounding problem, thus leading to the idea 
that the image mapping should be done in the device coordinate space).

- Is the proposal for an image primitive acceptable (modifications to 
what I proposed are of course possible).

- If yes, is it possible to get some help to really implement it for 
all the supported devices.

        Thank's a lot,

        Theo.

 --------------------------------------------------------------------
 Theodore Papadopoulo
 Email: address@hidden Tel: (33) 04 92 38 76 01
 --------------------------------------------------------------------