Re: [Patch]: Re: bug with plot

[David Bateman]

Michael Goffioul wrote:
> On 10/29/07, David Bateman <address@hidden> wrote:
>   
>>> This does not change the "look" of the dots. It still looks like a filled
>>> circle (I checked that the "pointsize" parameter is not defined in the
>>> commands sent to gnuplot).
>>>
>>> Michael.
>>>
>>>
>>>       
>> Huhh, a point by definition has no extent and so a point wouldn't be
>> visible. To make it visible the point in Octave adds a finite extent,
>> and logically the finite extent added should be the same in all
>> directions making a filled circle.. Why is this wrong? As for the look,
>> The gnuplot pointtype for points if pointtype=0 and not pointtype=7 as
>> used by Octave.. However, the pointtype=0 is so small that it is almost
>> invisible on the screen and so I think a filled circle is a better
>> option that using pointtype=0..
>>
>> The "pointsize" parameter will be sent if the Octave plot command is
>> called with the "markersize" option, and therefore the reason for the
>> patch I sent.. I'd therefore suggest that it is better to stick to what
>> we have with the patch I suggested.
>>     
>
> I know all this...
> Maybe I was not clear enough in my answer. What I meant is that your
> patch does not change the size of the dot (it uses default values of 1.0,
> which is the default marker size used by octave). So the patch has no
> visual effect. Setting the marker size to 0.5 makes markers more similar
> to dots. So you could define pointsize to 0.5 for dot markers.
>
> Michael.
>
>   

Your eyes must be better than mine.. This is effectively just the same
as a pointtype=0, which I find to be basically invisible.. The make the
pointsize of "." adjustable then a patch like the attached would be more
appropriate, though my eyes prefer a pointsize=1.0 ....

Ok, I just ran the same tests on your 2.9.15 binary and I see that the
markersize for 1.0 is much much larger than the X11 terminal on my linux
machine and that a markersize of 0.5 under your 2.9.15 binary
effectively corresponds to the markersize of 1.0 on my linux machine..
 
Given that gnuplot has different markersizes depending on the terminal
then this is a gnuplot bug in my opinion and I don't see what Octave can
and should do about it.

D.


-- 
David Bateman                                address@hidden
Motorola Labs - Paris                        +33 1 69 35 48 04 (Ph) 
Parc Les Algorithmes, Commune de St Aubin    +33 6 72 01 06 33 (Mob) 
91193 Gif-Sur-Yvette FRANCE                  +33 1 69 35 77 01 (Fax) 

The information contained in this communication has been classified as: 

[x] General Business Information 
[ ] Motorola Internal Use Only 
[ ] Motorola Confidential Proprietary

*** ./scripts/plot/__go_draw_axes__.m.orig16    2007-10-24 22:13:50.000000000 
+0200
--- ./scripts/plot/__go_draw_axes__.m   2007-10-29 12:05:21.263254191 +0100
***************
*** 1101,1107 ****
      style = "linespoints";
    endif
  
!   if (isfield (obj, "markersize"))
      if (have_newer_gnuplot)
        fprintf (plot_stream, " pointsize %f", obj.markersize);
      else
--- 1101,1116 ----
      style = "linespoints";
    endif
  
!   if (strcmp (pt, "7"))
!     if (have_newer_gnuplot)
!       fprintf (plot_stream, " pointsize 0.5");
!     else
!       if (! strcmpi (style, "lines"))
!       with = sprintf ("%s ps 0.5", with);
!       endif
!     endif
!     found_style = true;      
!   elseif (isfield (obj, "markersize"))
      if (have_newer_gnuplot)
        fprintf (plot_stream, " pointsize %f", obj.markersize);
      else