Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.graphics.apps.gnuplot > #440 > unrolled thread

Postscript fill pattern broken?

Back to article view | Back to comp.graphics.apps.gnuplot

Contents

  Postscript fill pattern broken? Ingo Thies <ingo.thies@gmx.de> - 2011-07-17 15:00 +0200
    Filledcurves with manual color also broken? Ingo Thies <ingo.thies@gmx.de> - 2011-07-17 18:00 +0200
      Re: Filledcurves with manual color also broken? sfeam <sfeam@users.sourceforge.net> - 2011-07-17 09:33 -0700
        Re: Filledcurves with manual color also broken? Ingo Thies <ingo.thies@gmx.de> - 2011-07-17 18:39 +0200
      Re: Filledcurves with manual color also broken? Ingo Thies <ingo.thies@gmx.de> - 2011-07-17 18:36 +0200
    Re: Postscript fill pattern broken? sfeam <sfeam@users.sourceforge.net> - 2011-07-17 09:32 -0700
      Re: Postscript fill pattern broken? Ingo Thies <ingo.thies@gmx.de> - 2011-07-17 18:59 +0200

#440 — Postscript fill pattern broken?

Hi,

I have noticed some strange behaviour with fill patterns in Postscript. 
I don't know since when it is present, but at least gnuplot 4.4 or later 
(CVS version from July) seems to be affected.

When I try to use pattern fill in Postscript the pattern for fillecurves 
appears to be extremely pixeled, rather than smooth as in the style test 
page.

I am wondering how the patterns on the test page (i.e. with the command 
"test") are actually created. But obviously not using the actual 
commands for pattern-filled rectangles.

Here is an example script (note that the current gnuplot syntax only 
works with data-based curves rather than functions for fill between two 
curves; for this reason 'set table' is used. If there is an easier way, 
please describe this in the manual, if not already done so):

reset
set term push
set term postscript eps size 15cm,10cm "Helvetica" 20
set output 'testfill.eps'
set samples 100
set xrange [-2.*pi:+2.*pi]
set yrange [-1.1:2.4]
set table 'tmp.asc'
plot sin(x)+0.5
unset table
set style fill pattern 4
plot 'tmp.asc' u 1:($2-0.5):($2+0.5) w filledcu
unset output
set term pop

I would have expected that "fill pattern" would draw lines, but actually 
a pixeled pattern is used (use the magnifying function in ghostview). 
But please note that the sample pattern in the key looks OK, only the 
plotted pattern looks chunky.

Is there a workaround, or is this an old and not yet resolved bug?

Best whishes,

Ingo

[toc] | [next] | [standalone]

#441 — Filledcurves with manual color also broken?

Additional issue:

manual selection of fillcolor seems to be broken, too. The following 
script returns a monochrome image although a dark red tone one has been set:

reset
set samples 100
set xrange [-2.*pi:+2.*pi]
set yrange [-1.1:2.4]
set table 'tmp.asc'
set mxtics 2; set mytics 2
set format y "%.2f"
plot sin(x)+0.5
unset table
set style fill solid 1.
plot 'tmp.asc' u 1:($2-0.5):($2+0.5) w filledcu lc rgb '#777777'


Best wishes,

Ingo

[toc] | [prev] | [next] | [standalone]

#443 — Re: Filledcurves with manual color also broken?

Ingo Thies wrote:

> Additional issue:
> 
> manual selection of fillcolor seems to be broken, too. The following
> script returns a monochrome image although a dark red tone one has
> been set:
> 
> reset
> set samples 100
> set xrange [-2.*pi:+2.*pi]
> set yrange [-1.1:2.4]
> set table 'tmp.asc'
> set mxtics 2; set mytics 2
> set format y "%.2f"
> plot sin(x)+0.5
> unset table
> set style fill solid 1.
> plot 'tmp.asc' u 1:($2-0.5):($2+0.5) w filledcu lc rgb '#777777'

rgb "#777777" is grey, not dark red

[toc] | [prev] | [next] | [standalone]

#445 — Re: Filledcurves with manual color also broken?

Am 2011-07-17 18:33, schrieb sfeam:

> rgb "#777777" is grey, not dark red

Ok, sorry. But I found that colors even disappear if set by hex (e.g. 
the above appears white rather than grey), e.g. '#ff0000' (should be 
read, but is nothing).

Best wishes,

Ingo

[toc] | [prev] | [next] | [standalone]

#444 — Re: Filledcurves with manual color also broken?

Addition #2

Coloured filledcurves seem to work only if the colour is specified by 
name rather than by hexcode. Filed a bug (#3369407).

The pattern problem, in contrast, may be caused by the antialiasing of 
gv, since it disappears when antialiasing is switched off.

Best wishes,

Ingo

[toc] | [prev] | [next] | [standalone]

#442

Ingo Thies wrote:

> Hi,
> 
> I have noticed some strange behaviour with fill patterns in
> Postscript. I don't know since when it is present, but at least
> gnuplot 4.4 or later (CVS version from July) seems to be affected.
> 
> When I try to use pattern fill in Postscript the pattern for
> fillecurves appears to be extremely pixeled, rather than smooth as in
> the style test page.

Your test script produces an acceptable plot when run here.

I suspect you have hit the well known long-standing bug in ghostscript
that fails to correctly render pattern fill if antialiasing is enabled.
Try turning off antialiasing in your viewer, or try sending the
file to a real postscript printer rather than rendering it in 
ghostscript.

	Ethan

> I am wondering how the patterns on the test page (i.e. with the
> command "test") are actually created. But obviously not using the
> actual commands for pattern-filled rectangles.
> 
> Here is an example script (note that the current gnuplot syntax only
> works with data-based curves rather than functions for fill between
> two curves; for this reason 'set table' is used. If there is an easier
> way, please describe this in the manual, if not already done so):
> 
> reset
> set term push
> set term postscript eps size 15cm,10cm "Helvetica" 20
> set output 'testfill.eps'
> set samples 100
> set xrange [-2.*pi:+2.*pi]
> set yrange [-1.1:2.4]
> set table 'tmp.asc'
> plot sin(x)+0.5
> unset table
> set style fill pattern 4
> plot 'tmp.asc' u 1:($2-0.5):($2+0.5) w filledcu
> unset output
> set term pop
> 
> I would have expected that "fill pattern" would draw lines, but
> actually a pixeled pattern is used (use the magnifying function in
> ghostview). But please note that the sample pattern in the key looks
> OK, only the plotted pattern looks chunky.
> 
> Is there a workaround, or is this an old and not yet resolved bug?
> 
> Best whishes,
> 
> Ingo

[toc] | [prev] | [next] | [standalone]

#446

Am 2011-07-17 18:32, schrieb sfeam:

> I suspect you have hit the well known long-standing bug in ghostscript
> that fails to correctly render pattern fill if antialiasing is enabled.
> Try turning off antialiasing in your viewer, or try sending the
> file to a real postscript printer rather than rendering it in
> ghostscript.

I will try the printer soon. But indeed, gv indeed failed to render the 
pattern in antialiasing mode. I remember that the error appeared 
earlier, but than it appeared just as a solid block instead of a pattern 
fill, or a solid fill didn't appear at all.

I will look whether this might have caused the color fill issue; sorry 
again for the confusion I caused when mistakenly announcing a similar 
error with png or even standard terminal; it may definitely have been 
the postscript error. However, I could not reproduce this after some 
sort of panic (I have to submit a plot with such filled curves for a 
publication to the corresponding author of the paper as soon as 
possible). Please see my e-mail, too.

Best wishes,

Ingo

[toc] | [prev] | [standalone]

Back to top | Article view | comp.graphics.apps.gnuplot

csiph-web