Re: [Groff] Using pic/groff/gs to produce gif/jpeg

[joerg van den hoff]

David Griffiths wrote:
> Hi, I'm new to pic/groff and I've been having some problems producing 
> either gif or jpeg images using them. The basic problem seems to be that 
> pic/groff/postscript all have the concept of a page that they are 
> drawing on whereas I want the resulting image to have all the whitespace 
> cropped from the margins. This is to go in a web page so the idea of a 
> page doesn't make sense.
>
> Anyway I started out following some instructions here: 
> http://www.spinellis.gr/sw/umlgraph/doc/faq.html#antialias but the 
> resulting command was very slow. Here are my findings as pasted from my 
> internal blog. I wondered if they made sense or if I'm missing some 
> simpler approach to the whole thing (btw I need the gs stage because 
> using -T gif in groff produces very poor text, no antialiasing).
>
> Thanks!
>
> Dave
>
>  From my blog:
>
>  I missed the hint in the FAQ and after a bit of messing around got the 
> following command to work:
>
> pic test | groff | ps2eps | gs -dDEVICEHEIGHTPOINTS=1000 -q -r360 
> -dNOPAUSE -sDEVICE=pnm -sOutputFile=- - -c quit | pnmcrop | pnmscale 0.5 
> | ppmtogif > test.gif
>
> I had to add the -dDEVICEHEIGHTPOINTS=1000 option because without that 
> it chopped off the top of the picture.
>
> Anyway, I think I prefer this approach if only because the pnmscale 
> command gives me the option to scale the size of the image. Plus 
> ppmtogif gives me control over transparency.
>
> Posted by D. H. Griffiths on May 04, 2005 at 07:02 AM EDT
> I've been playing around with this some more because the command above 
> was just so slow (eg 17 seconds on my T40 for a trivial diagram). Turns 
> out that most of that time is spent in pnmcrop. The image is far larger 
> than it needs to be (as specified by the resolution argument -r360) and 
> is then shrunk back down again with pnmscale (for my tests I'm using 
> pnmscale 0.25). Why is that? Well if you use -r90 and get rid of the 
> pnmscale, the text looks ragged as though it's missing antialiasing. But 
> I found that you could overcome that by specifying -dTextAlphaBits=4 in 
> the gs command. So you end up with a command like this:
>
>     pic test | groff | ps2eps -q | gs -dTextAlphaBits=4 
> -dDEVICEHEIGHTPOINTS=1000 -q -r90 -dNOPAUSE -sDEVICE=pnm -sOutputFile=- 
> - -c quit | pnmcrop | ppmtogif > test.gif
>
> now that's about four times faster because the intermediate pnm image is 
> four times smaller. This command took about 4 seconds on my Thinkpad. 
> But I found a still faster way.
>
> Why bother to go to pnm format in the first place, why can't the image 
> be created directly by ghostscript? Well the answer is that can, but 
> ghostscript can only produce jpeg images, not gif. Which is fine for me. 
> The next issue relates to the pnmcrop command. What that does is chop 
> blank space from around your image. The problem is that a lot of what 
> groff and postscript do revolves round the concept of writing to a page 
> (typically letter size). Pic for instance has its coordinate origin at 
> the bottom left but positions your image at the top of the page. If you 
> play around you run into all sorts of issues with the image having white 
> space above or below or the image top gets chopped off because it's 
> above the top of the imaginary page and so on. After a lot of messing 
> around I came up with this version of the command:
>
>     pic test | groff | ps2eps -q -g | fixbox | gs -dTextAlphaBits=4 -q 
> -r90 -dNOPAUSE -sDEVICE=jpeg -sOutputFile=- - -c quit > test.jpg
>
> the -g option to ps2eps is important because that tells it to use the 
> "internal bbox device of ghostscript" (whatever that means!), but anyway 
> it prevents ps2eps from setting the top of the bounding box to the top 
> of the default page. It's important because without it the diagram goes 
> above the bounding box set by ps2eps and the image gets cropped.
>
> fixbox is a small python script I wrote that is a much simplified 
> version of epstopdf. What this does is to move the image to the bottom 
> left corner and set the page size to just include the bounding box. 
> Using this means that there is no longer a need for cropping. The script 
> is here:
>
> #!/usr/bin/python
> # This script converts the BoundingBox of a postscript file so that the 
> origin is at
> # the bottom left corner (0, 0). It also sets the pagesize which avoids 
> the need for
> # cropping blank space from the image.
>
> import sys
> import os
> import re
>
> inlines = sys.stdin.readlines()
>
> for line in inlines:
>     if line.find("%%BoundingBox: ") != -1:
>         tokens = line.split()
>         minx = int(tokens[1])
>         miny = int(tokens[2])
>         maxx = int(tokens[3])
>         maxy = int(tokens[4])
>         new_maxx = maxx - minx
>         new_maxy = maxy - miny
>         print "%%BoundingBox: 0 0 " + str(new_maxx) + " " + str(new_maxy)
>         print "<< /PageSize [" + str(new_maxx + 10) + " " + str(new_maxy 
> + 10) + "] >> setpagedevice"
>         print "gsave -" + str(minx) + " -" + str(miny - 1) + " translate"
>     else:
>         print line,
>
> print "grestore"
>
> And finally, the best news is that this takes just 1 second to run!
>
>
>
> _______________________________________________
> Groff mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/groff


I usually convert the postscript output from groff to EPS with `ps2epsi` 
(George Cameron), a script which drives `gs' and most of the time makes 
the bounding box really fit the picture. don't now if this is in the 
standard distribution together with `gs' (see attachment). the 
eps-output more often - in my experience - has the correct bounding box
than obtained with `ps2eps' (Roland Bless)
and something (ps2epsi does not like pipes ...) like

groff -p file.pic > file.ps
ps2epsi file.ps
convert file.epsi file.jpeg

(`convert' from the ImageMagick package)

produces the desired result.

joerg

#!/bin/sh
# $Id: ps2epsi,v 1.9 2002/02/21 21:49:28 giles Exp $

tmpfile=/tmp/ps2epsi$$

export outfile

if [ $# -lt 1 -o $# -gt 2 ]; then
        echo "Usage: `basename $0` file.ps [file.epsi]" 1>&2
        exit 1
fi

infile=$1;

if [ $# -eq 1 ]
then
        case "${infile}" in
          *.ps)         base=`basename "${infile}" .ps` ;;
          *.cps)        base=`basename "${infile}" .cps` ;;
          *.eps)        base=`basename "${infile}" .eps` ;;
          *.epsf)       base=`basename "${infile}" .epsf` ;;
          *)            base=`basename "${infile}"` ;;
        esac
        outfile=${base}.epsi
else
        outfile=$2
fi

ls -l "${infile}" |
awk 'F==1       {
                cd="%%CreationDate: " $6 " " $7 " " $8;
                t="%%Title: " $9;
                f="%%For:" U " " $3;
                c="%%Creator: Ghostscript ps2epsi from " $9;
                next;
                }
        /^%!/   {next;}
        /^%%Title:/     {t=$0; next;}
        /^%%Creator:/   {c=$0; next;}
        /^%%CreationDate:/      {cd=$0; next;}
        /^%%For:/       {f=$0; next;}
        !/^%/   {
                print "/ps2edict 30 dict def";
                print "ps2edict begin";
                print "/epsititle (" t "\\n) def";
                print "/epsicreator (" c "\\n) def";
                print "/epsicrdt (" cd "\\n) def";
                print "/epsifor (" f "\\n) def";
                print "end";
                exit(0);
                }
        ' U="$USERNAME$LOGNAME"  F=1 - F=2 "${infile}" >$tmpfile

gs -q -dNOPAUSE -dSAFER -dDELAYSAFER -r72 -sDEVICE=bit -sOutputFile=/dev/null 
$tmpfile ps2epsi.ps $tmpfile <"${infile}" 1>&2
rm -f $tmpfile

(
cat << BEGINEPS
save countdictstack mark newpath /showpage {} def /setpagedevice {pop} def
%%EndProlog
%%Page 1 1
BEGINEPS

cat "${infile}" |
sed -e '/^%%BeginPreview:/,/^%%EndPreview[^!-~]*$/d' -e '/^%!PS-Adobe/d'\
        -e '/^%%[A-Za-z][A-Za-z]*[^!-~]*$/d' -e '/^%%[A-Za-z][A-Za-z]*: /d'

cat << ENDEPS
%%Trailer
cleartomark countdictstack exch sub { end } repeat restore
%%EOF
ENDEPS

) >> "${outfile}"

exit 0