[Tex/LaTex] Searching for a better workflow to get a cropped PDF and EPS graphics – Stage 1

dvipsepstoolepstopdfpdfcroppdftops

I am using TeX Live 2010. I don't use MikTeX anymore because of many troubles I found in the past.

Input files:

gridon.tex:

% gridon.tex
\documentclass{minimal}
\usepackage{pstricks}

\pagestyle{empty}
\begin{document}

\begin{pspicture}[showgrid=true](3,3)
\pscircle(1.5,1.5){1}
\rput[tr](3,3){3}
\end{pspicture}

\end{document}

gridoff.tex:

% gridoff.tex
\documentclass{minimal}
\usepackage{pstricks}

\pagestyle{empty}
\begin{document}

\begin{pspicture}[showgrid=false](3,3)
\pscircle(1.5,1.5){1}
\rput[tr](3,3){3}
\end{pspicture}

\end{document}

main.tex:

% main.tex
\documentclass{minimal}
\usepackage[hiresbb]{graphicx}
\begin{document}

\includegraphics[scale=2]{gridon}
\vspace{5mm}

\includegraphics[scale=2]{gridoff}
\end{document}

Methods:

MethodA.bat:

echo off
latex %1
dvips %1
ps2pdf %1.ps
pdfcrop --hires %1
pdftops -eps %1-crop.pdf

MethodB.bat:

echo off
latex %1
dvips %1 -E -o %1.eps
epstool --copy --bbox %1.eps %1-crop.eps
epstopdf --hires %1-crop.eps

Results:

enter image description here

Best Answer

I would recommend method A, because the EPS files will in general be much smaller.

You noticed "bad" cropping with method A. The reason for this is subtle: your use of pstricks produces a PDF where the font used for labelling the grid (eg, the figure "3") is Helvetica and it is not embedded in the PDF. Therefore ghostscript and adobe reader have to use a subsitute font. Most viewers (including the many viewers based on ghostscript, also xpdf) will use a version of Helvetica. Adobe's viewer will substitute ArialMT (on windows at least), which is a slightly different shape and therefore gets cut off. You can check this with Adobe reader on the "fonts" tab of the "document properties" dialog. You can solve this problem by producing a file with the fonts embedded. One way to do this is to change your method A to use ps2pdf -dPDFSETTINGS#/prepress %1.ps %1.pdf.

Why does method B seem to be superior? Actually I was quite confused for a while because on Ubuntu the two methods do produce the same bounding box. Which is to be expected, since pdfcrop and epstool both use gs -sDEVICE=bbox to generate it. So why do you see a difference? Well, I found that the miktex version of epstopdf has a bug in it: it rounds the bounding box to an integer number of points. In this particular case this had the effect of cancelling the clipping, but in general it could make the clipping worse. I'll file a bug against miktex tomorrow, but in the meantime this is another reason to prefer method A if you are using miktex. (The miktex version of epstopdf is a reimplementation in C of the bugless original perl script).

Related Solutions

[Tex/LaTex] When converting PDF to EPS, how to prevent pdftops from rasterizing fonts and transparent regions

It can't be done. The PostScript language does not support arbitrary opacity (only fully opaque and completely transparent). See this wikipedia reference.

The Ghostscript language, however, does support arbitrary opacity, as an extension to the language (extra commands such as .setopacityalpha). See here for details. This is how pstricks made an apparently-transparent eps file.

In all your dozens of questions about figure conversion, I don't recall that you have ever explained why you want to produce eps versions of your figures (which makes it much harder to answer all those questions, by the way). If your reason is only to use the figures in a latex->dvips->ps2pdf workflow, where you can guarantee that the conversion to pdf will use ghostscript, then the .setopacityalpha method is appropriate (although if this is your goal, why not simply use pdf images with pdftex in the first place? or at least add -Ppdf to your dvips invocation to make pdf-optimized eps?). However, the reason people usually want eps figures is because they are giving the figures to some publisher who does not accept pdf, in which case the publisher will almost certainly also not accept ghostscript-specific extensions and the .setopacityalpha method will fail also. If you happen to know that the publisher uses Adobe Distiller, then there is another way to produce transparent-extended EPS, via the pdfmark extension, described here. You can ask pstricks to use the pdfmark method for opacity by replacing pstricks.con by distiller.cfg (distributed with pstricks). (You will have to ask your publisher to set /AllowTransparency true in their joboptions file).

If you need a strictly standards-conforming EPS file that will work with all PostScript engines, then you need either to avoid transparency altogether or use something like the ps2pdf followed by pdftoeps method as given in your batch file, which will rasterize the transparent parts of the image. (I think the next version of of pdftops will at least allow you to specify a rasterization resolution). Due to the rasterization, the EPS will in general be larger, of course.

[Tex/LaTex] Can’t get the pdf and .ps files to use type 1 fonts

If pkfix and pkfix-helper are not able to help, the easiest way is to use a tool that can take the document with text in embedded fonts and convert them to curves/outlines. That way there is no embedded fonts of any kind that can be of trouble. It also saves the effort of trying to find suitable replacement fonts etc.

Since your .ps files contain DocumentMedia DSC comments let's use them to save some work.

$ grep DocumentMedia class_1_best_describer_mention.ps
%%DocumentMedia: 555x608 555 608 0 () ()

Open the .ps file in almost any major vector-based drawing program. For example in CorelDraw either select Import text as curves when opening .ps file or select Export text as curves when exporting as .eps. In others it may be called Create Outlines or something similar.

Edit the exported .eps file and add the line above starting %%DocumentMedia: somewhere near the line starting %%BoundingBox:.

If you want, you can change the .eps ending to .ps.

To create the .pdf file. From the DocumentMedia line above take the dimensions and put them to command below into the corresponding DEVICE*POINTS options.

$ ps2pdf -dDEVICEWIDTHPOINTS=555 -dDEVICEHEIGHTPOINTS=608 class_1_best_describer_mention-2.ps class_1_best_describer_mention-2.pdf

Now you have files that will pass.

However, the best way is to follow the directions in Avoiding Type 3 fonts in matplotlib plots.

I have converted the example files you provided and have them available if there is still need for them.