3

I use the following command to convert a PDF to EPS:

gswin32 -dNOCACHE -dNOPAUSE -dBATCH -dSAFER -sDEVICE=epswrite -dLanguageLevel=2 -sOutputFile=test.eps -f test.pdf

I then use the following command to convert the EPS to another PDF (test2.pdf) to view the EPS figure.

gswin32 -dSAFER -dNOPLATFONTS -dNOPAUSE -dBATCH -dEPSCrop -sDEVICE=pdfwrite -dPDFSETTINGS=/printer -dCompatibilityLevel=1.4 -dMaxSubsetPct=100 -dSubsetFonts=true -dEmbedAllFonts=true -sOutputFile=test2.pdf -f test.eps

I found the text in the generated test2.pdf have been converted to outline curves. There is no font embedded anymore either.

Is it possible to convert PDF to EPS without convert text to outlines? I mean, to EPS with embedded font and text.

Also after the conversion (test.pdf -> test.eps -> test2.pdf), the height and width of the PDF figure (test2.pdf) is a little bit smaller than the original PDF (test.pdf):

test.pdf:

enter image description here

test2.pdf:

enter image description here

Is it possible to keep the width and height of the figure after conversion?

Here is the test.pdf: https://dl.dropboxusercontent.com/u/45318932/test.pdf

I tried KenS's suggestion:

gswin32 -dNOPAUSE -dBATCH -dSAFER -sDEVICE=eps2write -dLanguageLevel=2 -sOutputFile=test.eps -f test.pdf
gswin32 -dSAFER -dNOPLATFONTS -dNOPAUSE -dBATCH -dEPSCrop -sDEVICE=pdfwrite -dPDFSETTINGS=/printer -dCompatibilityLevel=1.4 -dMaxSubsetPct=100 -dSubsetFonts=true -dEmbedAllFonts=true -sOutputFile=test2.pdf -f test.eps

I can see the converted test2.pdf have very weird font:

enter image description here

that is different from the original font in test.pdf:

enter image description here

When I copy the text from test2.pdf, I only get a couple of symbols like: 

✕ ✖ ✗✘✙ ✚✛

Here is the test2.pdf: https://dl.dropboxusercontent.com/u/45318932/test2.pdf

I was using the latest Ghostscript 9.15. So what is the problem?

Changwang Zhang
  • 2,467
  • 7
  • 38
  • 64

2 Answers2

5

I just noticed you are using epswrite, you don't want to do that. That device is terrible and has been deprecated (and removed now). Use the eps2write device instead (you will need a relatively recent version of Ghostscript).

There's nothing you can do with epswrite except throw it away, it makes terrible EPS files. It also can't make level 2 files, no matter what you set -dLanguageLevel to

oh, and don't use -dNOCACHE, that prevents fonts being processed and decomposes everything to outlines or bitmaps.

UPDATE

You set subset fonts to true. By doing so the character codes which are used are more or less random. The first glyph in the document (say for example the 'H' in 'Hello World') gets the code 1, the second one (eg 'e') gets the code 2 and so on.

If you have a ToUnicode CMap, then Acrobat and other readers can convert these character codes to Unicode code points, without that the readers have to fall back on heuristics, the final one being 'treat it as ASCII'. Because the encoding arrangement isn't ASCII, then you get gibberish. MS Windows' PostScript output can contain additional ToUnicode information, but that's not something we try to mimic in ps2write. After all, presumably you had a PDF file already....

Every time you do a conversion you run the risk of this kind of degradation, you should really try and minimise this in your workflow.

The problem is even worse in this case, the input PDF file has a TrueType CID Font. Basic language level 2 PostScript can't handle CIDFonts (IIRC this was introduced in version 2015). Since eps2write only emits basic level 2 it cannot write the font as a CIDFont. So instead it captures the glyph outlines and stores them in a type 3 font.

However, our EPS/PS output doesn't attempt to embed ToUnicode information in the PostScript (its non-standard, very few applications can make use of it and it therefore makes the files larger for little benefit). In addition CIDFonts use multiple (2 or more) bytes for the character code, so there's no way to encode the type 3 fonts as ASCII.

Fundamentally you cannot use Ghostscript to go PDF->PS->PDF and still be able to copy/paste/search text, if the input contains CIDFonts.

By the way, there's no point in setting -dLanguageLevel at all. eps2write only creates level 2 output.

KenS
  • 30,202
  • 3
  • 34
  • 51
  • I tried your suggestion, but got very weird result. Details are updated in my question. Can you have a look? – Changwang Zhang Feb 20 '15 at 09:46
  • So the problem lies in the CID fonts used in the file? Is it because I have symbols (µ) in the figure that the figure embeded CID fonts? How can I get ride of the CID fonts and use normal ones? – Changwang Zhang Feb 25 '15 at 08:41
  • Its the CIDFonts causing the problem, yes. In order to embed those in a PostScript (EPS) file you need a higher level of PostScript than GS currently supports as output. The only way I can suggest to avoid getting a CIDFont is to not use a TrueType font. Whatever is producing the PDF file is embedding TrueType fonts as CIDFonts. – KenS Feb 25 '15 at 08:49
  • But really I have to say, round-tripping (PDF->PS->PDF) is always going to introduce some loss of functionality in the general case. Given that neither PDF nor (especially) PostScript was intended to be editable, search/copy/paste is always a matter of luck. – KenS Feb 25 '15 at 08:51
0

I used Inkscape To convert a .pdf to .EPS. Just upload the .pdf file to Inkscape, in the options to open chose high mesh, and save as . an EPS file.

ed24
  • 1