\pdfglyphtounicode with XeTeX

Question

One well known problem with using the font Linux Libertine (presumably other fonts too), is that its many ligatures are not recognized in the resulting pdf-file, so that it's not possible to search for or copy a ligature such as 'Th'.

One way of fixing this is addressed in the question Make ligatures in Linux Libertine copyable (and searchable)

But the fix addressed there doesn't work with XeTeX. Compiling Ulrike's MWE will only result in an error message

> Undefined control sequence.
l.7 \pdfglyphtounicode

So what is the best fix for XeTeX?

Answer 1

XeTeX simply doesn't support the \pdfglyphtounicode primitive, so this route is not available. To understand why, remember that pdfTeX is 8-bit while XeTeX is natively UTF-8. AS such, it's perfectly reasonable for XeTeX to expect fonts to be 'properly' constructed to have glyphs in the appropriately-named slots and therefore to work correctly in the PDF output with no 'trickery'.

Taking the 'what to do' side of things, the problem is most likely with the font. For example, some fonts have the correct glyphs but in the 'wrong' places (e.g. in the Private Use Area). That's really not a 'TeX problem': all that can be done is to talk with the font designers and try to get them to fix it.

Answer 2

I realize that this post is old, but for future readers' sake, I thought I'd point out that I was able to create ligatures using Linux Libertine 5.3 and XeTeX 3.14159265-2.6-0.99991 (TeX Live 2014/W32TeX) (preloaded format=xelatex 2014.7.31) that are copyable and searchable when viewed in Adobe Reader 11.0.08.

\documentclass{article}

\usepackage{fontspec}
\setmainfont[Ligatures=Historical]{Linux Libertine O}

\begin{document}
    fluffier soufflé fisticuffs fb fh ffh fj ffj fk ffk ft fft tt Qu Th ch ck ct
\end{document}

You can check my output PDF to verify that the ligatures are indeed copyable and searchable.

Does anyone know what might have changed to make this possible since the original question and answers here? Linux Libertine 5.3 (the version I used) was available at the time of the original post.