Re: [PyX-user] embed font in eps?

[Chris v. D. <cn....@ho...>]

On Thu, 12 Nov 2009 10:25:51 -0600
William Henney <w.h...@as...> wrote:

> Hi Chris, Michael et al.,
> 
> On Thu, Nov 12, 2009 at 7:29 AM, Chris van Dijk
> <cn....@ho...> wrote:
> > Yes you are right, by looking into the eps file I can see that the
> > fonts are embedded in there (as a side note does anybody know of a
> > convenient tool like pdffonts for eps?). So the problem is obviously
> > not with PyX. Apologies for bothering you guys then...
> >
> > I will just need to figure something else out, as the text in my
> > figures does not show up in the merged pdf compiled from my tex
> > file by the system. I would prefer not to use a standard font for
> > the figures, as the rest of the article is in computer modern too.
> > I use the graphicx package to include the figures if it matters.
> 
> In my experience with most journals (with some honorable exceptions),
> their definition of an "acceptable" figure is one that works with
> their buggy, outdated, proprietary software. Figures produced by
> LaTeX, PyX, etc often do clever things with the EPS that, even though
> they may be perfectly standards-compliant, often throw a spanner into
> their inflexible workflow. Two recent examples that I remember off the
> top of my head are negative values in the bounding box (which
> apparently confuses the hell out of some old version of Adobe
> Illustrator), and that my fonts, although embedded, were also
> subsetted, which they did not like.
> 
> There is little point in fighting with them about this. The easiest
> approach is to simply post-process your PyX-created graphics to
> convert them into high-resolution, anti-aliased raster images or
> low-level vector postscript commands before you submit the figures.
> Which is more appropriate depends on the nature of your figures. A
> side effect of converting to low-level postscript (e.g., with
> ghostscript's epswrite device) is that all the text is converted to
> outlines, so there are no longer any fonts in your file at all.
> 
> Below I have pasted an example bash script that I wrote a few years
> back, which does a bit of both. You probably already have gs, but you
> will likely have to download http://imgtops.sourceforge.net, which is
> a (now old and seemingly unmaintained) python package that includes
> both the imgtops and epstoimg utilities. On the other hand, there may
> now be better ways of embedding raster images in EPS.
> 
> The script sanitizes each figure listed in figlist.txt according to
> the heuristics described in the comments. Maybe you will find it
> useful.
> 
> On a more positive note, I have noticed that journals that have gone
> over to an all-PDF workflow seem to be able to deal much better with
> PyX-generated figures. So these steps may become increasingly
> unnecessary as EPS dies out in scientific publishing.
> 
> Cheers
> 
> Will
> 
> #! /bin/bash
> #
> # convertfigs.sh - Will Henney (w.h...@as...) 11 Sep 2006
> #
> # Revised 28 Dec 2006 to do revised version of paper (one new figure)
> #
> outdir=../Revised
> i=0
> for origfig in $(cat figlist.txt); do
>     i=$(( i + 1 ))
>     newname=$outdir/f$i
>     if [ -f $origfig.eps ]; then
> 	# these are files written by PyX - need to pass through gs
> 	# to convert all fonts to outlines
> 	echo "$newname: converting fonts to outlines"
> 	gs -dBATCH -dNOPAUSE -dSAFER -q -dEPSFitPage \
> 	    -dNOCACHE -sDEVICE=epswrite \
> 	    -sOutputFile=$newname.ps $origfig.eps
>     elif [ -f $origfig.pdf ]; then
> 	# convert PDF to image with width 1600 pixels
> 	printf '%s' "$newname: converting to image..."
> 	epstoimg -v -w 1600 -o $origfig.png $origfig.pdf
> 	# then embed in a PS file
> 	printf '%s\n' "... and embedding in PS"
> 	imgtops -v -o $newname.ps $origfig.png
>     else
> 	# already an image, just embed in a PS file
> 	echo "$newname: embedding in PS"
> 	imgtops -v -o $newname.ps $origfig.jpg
>     fi
> done
> 
> 
> 
> 
> 

Hey Will,

Thanks for your input, that script was indeed useful. For now I
converted the eps figures to low-level paths, so that should alleviate
the problem of fonts for now. It is a good enough solution for this
article anyway. 

Thanks everybody for your help,

Chris