An IDL cron job, true color plots, Xvfb, Z-buffer, and all sorts of

Hi folks,

This newsgroup has been a huge help to me with these sorts of problems
in the past, and I am hoping once again to draw on your expertise.

My problem is a little complicated.  I have a particular plotting
routine that uses a color scale to represent signal response over a
grid of detectors.  I use this a lot and it's really handy.  It also
took me a really long time to get it to work and to be able to
write_png files from it without getting goofy colors.  What I would
like to do is automate this routine and run it with a cron job,
outputting png files to include in a daily web page.

I have noticed that quite a few people have tried to do similar things
and run into problems.  I have problems with all of the suggested
solutions.  Any one of them would be fine but I can't get them to
work.  So here goes:

option 1) use the Z buffer:  Here I think I am running into problems
with my understanding of true-color display issues in IDL.  My plot
routine includes a line  'Device, decomposed = 0' and the Z-buffer
doesn't work the same way as the X buffer.  That's problem number
one.  Problem number two is that I can't figure out how to take the Z-
buffer plot and make a png file out of it with correct colors.

option 2) write to a ps file and convert it to png after the fact.  I
have similar problems here with my Device, decomposed = 0, and I don't
understand how to translate my plotting routine so that it will do the
same thing with the ps device that it does with standard X output.

option 3) use the /pixmap option with an X window, and then use Xvfb
to run a virtual X session so that I can run this from cron.  I have
Xvfb installed and with some fiddling, I can get the authentication to
work out so that I don't get an immediate Xlib: connection to ":1.0"
refused by server, but even then I still get WINDOW: Unable to connect
to X Windows display: :1.0 errors, even though the DISPLAY variable
looks to be set correctly to me.   I really hoped I could get this to
work, because doing window, 1, /pixmap, running the plotter, and then
calling write_png, 'file.png', tvrd(/true) works wonderfully and gives
exactly what I want... just not through cron.

I think the main reason I'm having so much trouble here is I'm trying
to get something to work with color tables and true-color display
without really understanding them.  A stripped-down version of the
plotting routine is below.  I would invoke it with something like
>plot_test, [0,2,3],[0.2,0.7,1.0].  Before I run this, I have quite a
few graphics setup calls in my idl_startup file.  Between running it,
I typically do
loadct, 0
!p.color = !black
!p.background = !white
to make the display look right each time.

Sorry for the long-winded question.  Sample plotting routine follows:

pro plot_test, indices, color_scale,$
                title=title

if n_elements(title) eq 0 then title=''

; example simple grid of 4 points:
x_pos=[1., 2., 1., 2.]
y_pos=[1.,1.,2.,2]

;plot, the center points for each detector:
plot, x_pos, y_pos,  xrange=[0,3],yrange=[0,3],/xstyle,/ystyle,/iso,$
  psym = 3, title = title
Device, decomposed = 0
LoadCT, 33
; determine the color scale to use
color_scale=float(color_scale)

color_values = ((color_scale - min(color_scale))/(max(color_scale)-
min(color_scale)))* 255

;plot circles filled with the colors in the scale
for i=0,n_elements(indices)-1 do begin
    polyfill, circle(x_pos[indices[i]],y_pos[indices[i]],0.25),
color=color_values[i]
endfor

colorbar, /vertical, format='(F6.2)',$
          range=[min(color_scale),max(color_scale)]

end


- Kathryn