Techniques to make tabular contents copyable

Question

Take the following table:

\documentclass{article}
\begin{document}
\begin{tabular}{|c|c|}
11 & 21 \\
12 & 22 \\
\end{tabular}
\end{document}

What are the techniques to make this table text-copyable (as much as possible)?

For example, Evince sees this as text (note that order and lines are lost):

11
12
21
22

(Acrobat does slightly better job in this case but still can get confused in larger tables).

One alternative I explored is to use accsup. However I don't know how to enforce tabs (between cell) and new lines; this doesn't work:

\documentclass{article}
\usepackage{accsupp}
\begin{document}
\BeginAccSupp{ActualText=11\t 21\n 12\t 22\n} %this line produces errors
\begin{tabular}{|c|c|}
11 & 21 \\
12 & 22 \\
\end{tabular}
\EndAccSupp{}
\end{document}

What code I have to put to get this as copyable text 11 \tab 21 \n 12 \tab 22 \n? Is there other technique to help the PDF viewer at lest follow (flow) the copyable text in a more sensible way?

---

EDIT: I made some mild progress using accsup by defining a line-by-line alternative text:

\documentclass{article}
\usepackage{accsupp}
\begin{document}
\begin{tabular}{|c|c|}
\BeginAccSupp{ActualText=11 21} 11 & 21 \EndAccSupp{} \\
\BeginAccSupp{ActualText=11 21} 12 & 22 \EndAccSupp{} \\
\end{tabular}
\end{document}

Still I don't know how to add the tab characters but the copying has improved

---

EDIT 2: I can produce a tab charactar with \BeginAccSupp{method=hex,unicode,ActualText=0009} but then I can't combine it with normal text. (As in ActualText=a ?0009? b).

---

EDIT 3: A line-by-line variant from Heiko's answer:

\documentclass{article}
\usepackage{accsupp}
\begin{document}
\begin{tabular}{|c|c|}
\BeginAccSupp{method=plain, ActualText=11\string\t 21}
11 & 21
\EndAccSupp{}\\
\BeginAccSupp{method=plain, ActualText=12\string\t 22} 
12 & 22 
\EndAccSupp{}\\
\end{tabular}
\end{document}

Answer 1

Method pdfstringdef

If hyperref is loaded, \pdfstringdef is available and can be used for package accsupp via method=pdfstringdef:

\documentclass{article}
\usepackage[pdfencoding=auto]{hyperref}
\usepackage{accsupp}
\begin{document}
\BeginAccSupp{
  method=pdfstringdef,
  ActualText=11\unichar{"0009}21\unichar{"000A}12\unichar{"0009}22\unichar{"000A},
}
\begin{tabular}{|c|c|}
11 & 21 \\
12 & 22 \\
\end{tabular}
\EndAccSupp{}
\end{document}

\textHT can be used instead of \unichar{"000A} and \textLF instead of \unichar{"000A}.

Method pdfstringdef is the most comfortable method, because it supports a wide range of TeX code. \pdfstringdef is the very same macro, which is used for the bookmarks.

Method plain

The most riskiest method is plain, here the string is given, as it should appear in the PDF file after expansion. Requirement is a good knowledge about string syntax of PDF or PostScript.

\t and \n can usually not be used directly, because they do not expand to themselves, \t is an accent command and \n is undefined usually. Here \string (or \detokenize) can be used to convert \t and \n to two character tokens \ and t or \ and \n:

\documentclass{article}
\usepackage{accsupp}
\begin{document}
\BeginAccSupp{  
  method=plain,
  ActualText=11\string\t 21\string\n 12\string\t 22\string\n,
}
\begin{tabular}{|c|c|}
11 & 21 \\
12 & 22 \\
\end{tabular}
\EndAccSupp{}
\end{document}

Method escape

The risk of plain is, that backslashes at the wrong places or unmatched parentheses can lead to a broken/invalid PDF file. Method escape avoids this by escaping the problematic characters. Therefore \t and \n can't be used the same way as with method plain, because the backslash would be escaped itself. Instead the macros should expand to the character tokens:

\documentclass{article}

% define \HT as horizontal tabular
\begingroup
  \lccode`\0=9\relax
\lowercase{\endgroup
  \def\HT{0}%
}

% define \LF as line feed
\begingroup
  \lccode`\0=10\relax
\lowercase{\endgroup
  \def\LF{0}%
}

\usepackage{accsupp}
\begin{document}
\BeginAccSupp{  
  method=escape,
  ActualText=11\HT 21\LF 12\HT 22\LF,
}
\begin{tabular}{|c|c|}
11 & 21 \\   
12 & 22 \\   
\end{tabular} 
\EndAccSupp{} 
\end{document}

Answer 2

I'm not an expert for PDF. Your belief, that the PDF contains something like a table / tabular, seems questionable. I don't think there is a tabular in the PDF, just rules and glyphs.

First step: you copy the content of the PDF. What do you copy? Just the content, whatever it is in the PDF: glyphs, graphics. Some software is able to enrich the content, which will be transfered to the clipboard, with information. My old version of Acrobat Pro e.g. offers to copy text "as a table". But this depends on the viewer of the PDF. To my knowledge there is no way to label at table in the PDF as "TABLE".

Now, you paste the content of the clipboard into another software. Maybe this software analyses the content of the clipboard and reckognises the outlines of a table.

In some cases it seems an improvement to use the command \pdfinterwordspaceon, but my own tests showed poor results.

So: If you wish to share tables, don't print them into a PDF.