| Re: PNGs without X? [message #34030 is a reply to message #33837] |
Fri, 14 February 2003 11:21   |
Dr. Sven Geier
Messages: 2
Registered: January 2003
|
Junior Member |
|
|
Boy, I hope top-posting is alright...
A while ago I posted the thing appended below -- and got a few responses. I
dug through the verious methods and none of them really work. I had noted
before that postscript isn't really an option and the same reasons really
apply to plotting into the Z-buffer: neat trick that, and I may well use
that in the future; but the output I get is nowhere near the output that I
get when plotting to an X window. The fonts are entirely different, the
colors are differnt, pretty much everything is different.
The color part, I suppose, is an outcropping of another problem, that I
never managed to find a solution to: I find it very hard (to say the least)
to use a certain given color in IDL. It seems like the simplest thing in
the world to have a graph with four lines and I want one of them to be red
and one green and one yellow and one blue. However there doesn't seem to be
a mechanism in IDL(?) that accomplishes such a thing. Before I knew of
24-bit displays, I had a self-made colortable (with tvlct()) in which I
knew index 1 is red and index 5 is blue and such, but the first time I
tried this on a true-color visual, I learned something.
The problem is that I need to be certain what the colors are before the
plot, since there are some comments added to the data (via xyouts) that
refer to "red line = such-n-such" and such. For the same reason I must be
able to rely on a certain plot geometry (which I am consistently not
getting when plotting in the Z-buffer) because the positions of things are
carefully computed. The entire thing has been evolved over years to work
just right and if you suddenly run it with a larger character size (like
the Z-buffer seems to do) it'll all blow up.
As an example, here's some random data and label:
data = tan(findgen(5))
plot,data,/nodata
oplot,data,color='ff'x
oplot,data,psym=5
xyouts, 1,1.53,' <- triangle'
On my 24-bit display this draws a few red lines, white triangles and labels
one of the triangles with "triangle". If I do this on a
"window,xsize=350,ysize=350" and then I do the same thing into the Z-buffer
following David's website (great site, by the way. Or terrible site, as you
may look at it: Many idl programmers lose much productive time browsing
through it... :) I get something that is "kinda like" the original, except
that the plotting area has a different size, the fonts are different and
the color is gone - pretty much as different as a plot can be while still
showing the same thing.
!P.font = -1 or 0 give the same result in the Z-buffer, !p.font=1 gives a
slightly prettyier (but less legible) one, but all three are different from
their counterparts that are plotted into the X-buffer and not one of them
is as clear and legible as the p.font=0 in X-windows.
If I plot into the PS device and turn the result into an image like this:
gs -q -sDEVICE=ppm -r100 -sOutputFile=- idl.ps | pnmflip -r90 | pnmscale \
-xsize=350 -ysize=350 | pnmtojpeg > test.jpg
I do not only get no colored line (because all colors 255 and up are just
"white" to postscript, apparently), but the result is transparent (i.e.
white background) and the fonts are unreadable. It's easy to fix the color
in this simple example by searching for 'setrgbcolor' in the postscript
file, but once you have a bunch of colored lines and text, that gets pretty
tricky to say the least.
I am writing this all down in a last attempt to see if someone has the magic
trick "how to get consistent colors and something like an X-buffer without
X" - but if there is none, I'd like to say thanks to those who posted
before anyways; there were a few good ideas...
-- S
-
Catch a couple Z's before responding...
Reimar Bauer wrote:
> Dr. Sven Geier wrote:
>
>
> Dear Sven,
>
> did you have thougt about the trick printing into a ps file and then using
> convert or pstoimg (which I prefer) to get the ps file converted in
> whatever you want. (You have an unix system so you can use these routines
> with a spawn too)
>
> This method if it's suitable for you has another effort too. The
> postcriptfonts are much better readable as the image fonts.
>
>
> regards
>
> Reimar
>
>
>
>>
>> Heya all
>>
>> I have a routine that gathers data from various places, then plots it,
>> then uses tvrd(/true) to read the plot into an array and then writes it
>> out as a .png graphics file. This all works fine so far.
>>
>> Since I rarely ever look at the graphs as they're plotted (the pngs go to
>> a web-directory) I switched the plotting to a pixmap, i.e. the
>> idl-process now opens its own window with the /pixmap parameter (and also
>> xsize and ysize and such) and thus the whole shebang works quite nicely
>> without ever showing anything on the screen.
>>
>> Now in the process of automation I'd like to move the whole operation to
>> a server box that does not have X running. Unfortunately it turns out
>> that these "pixmaps" are *X*-pixmaps and that I can't open them without
>> X.
>>
>> So now I'm looking for some way to plot data into/onto a PNG (or GIF or
>> JPG or whatever) without having an actual graphic server running. I
>> played around with plotting into a PS device and using ghostscript to
>> convert that into a JPG, but not only do the colors not come out right
>> (the usual PS stuff) but also the fonts are all off (the !P.font=0 for
>> the 'x' devive is designed to be readable on a 400x400 window, the
>> ps-fonts become illegible when I reduce a PS-plot to that resolution) and
>> similar problems.
>>
>> Anybody know how to do a standard run-of-the mill X-type plot into a file
>> or into an array or into RAM or such?
>>
>> Thanks in advance...
>>
>> -- SG
>>
>>
>
--
Catch a couple Z's before responding...
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
