How to keep indentations in Python code copied from LaTeX PDF?

Question

I tried to include Python 3.6 code in a LaTeX PDF document which should be easily be copied to save to a file or to try the Python code.

Although,

\begin{verbatim}
for row in range(1,9):
    for col in range(1,9):
        print(int(str(row)+str(col)))
\end{verbatim}

shows nicely in the PDF but if I copy it and paste it to a text editor it looks like:

for row in range(1,9):
for col in range(1,9):
print(int(str(row)+str(col)))

The indentations which are essential for Python are gone.

I also tried the suggestion here using the listings package:

How to highlight Python syntax in LaTeX Listings \lstinputlistings command

where I can simply give the file name of an Python file e.g.

\pythonexternal{Test.py}

And the source code will be included and colored. But the same thing with the missing spaces at the start when copy&pasting.

If I use the option "showspaces=true" I get the following:

for␣row␣in␣range(1,9):
␣␣␣␣for␣col␣in␣range(1,9):
␣␣␣␣␣␣␣␣print(int(str(row)+str(col)))

which is also not suitable for copy&paste. Well, I could replace all ␣ with space... not a really practicable solution.

There have been some weird hacks described in 2011 here:

How to make listings code indentation remain unchanged when copied from PDF?

Is there anything new since then...? Any ideas how to achieve copy&paste Python code in a LaTeX PDF? Thank you for suggestions.

Answer 1

One more solution that works well in both Acrobat Reader and Chrome, but is compatible with pdfTeX only (not XeTeX/LuaTeX):

\usepackage{transparent}
\makeatletter
\def\@xobeysp{\leavevmode\nobreak\texttransparent{0}{\char32}}
\makeatother

The \texttransparent is used to make ⎵ symbol invisible (actually, in some fonts the slot 32 is mapped onto the real space character, and this trick is not necessary, but the above code covers all cases).

Most likely, this code works because the generated file is not Unicode-based, and PDF viewer has no information about matching between code 32 and actual Unicode character, and so it copies it "as-is" as character code 32 (i.e. the space character). In favor of this hypothesis, there are failed XeTeX/LuaTeX-generated files and an attempt to set Unicode correspondence via /ActualText command (by replacing \char32 with \pdfliteral direct{/Span<</ActualText<FEFF0020>>> BDC}\char32\pdfliteral direct{EMC} in the above code).

Answer 2

The pdfTeX has command \pdffakespace to insert a space into the document. This space is invisible, but it is taken into account during text copying.

So, my first idea was to insert \pdffakespace after each space at the beginning of lines, but unfortunately, it results in spaces doubling in the copied text (one space comes from \pdffakespace and another is generated by Acrobat Reader automatically from TeX's \hskip).

Next idea was to count the number of leading spaces in the line and insert the necessary number of \pdffakespace after those spaces. Unfortunately, Acrobat Reader trims that spaces to a single one (in spite of the fact that the sequence of space commands [( )]TJ is actually presented in the generated PDF file).

But finally, I've found a tricky solution: group spaces into pairs and replace them with a sequence of \pdffakespace and \hskip of double space width. Starting with a standard preamble to make @ a letter

\makeatletter

we redefine the \@verbatim command by adding \hook@par at the end of \par command definition (to process text after line break):

\def\@verbatim{\trivlist \item\relax
  \if@minipage\else\vskip\parskip\fi
  \leftskip\@totalleftmargin\rightskip\z@skip
  \parindent\z@\parfillskip\@flushglue\parskip\z@skip
  \@@par
  \language\l@nohyphenation
  \@tempswafalse
  \def\par{%
    \if@tempswa
      \leavevmode \null \@@par\penalty\interlinepenalty
    \else
      \@tempswatrue
      \ifhmode\@@par\penalty\interlinepenalty\fi
    \fi\hook@par}% <=== HERE
  \let\do\@makeother \dospecials
  \obeylines \verbatim@font \@noligs
  \everypar \expandafter{\the\everypar \unpenalty}%
}

Then we create a counter to count spaces at the beginning of lines

\newcount\nspaces

reset this counter at the beginning of lines, and count the sequence of spaces

\def\hook@par{\nspaces=0\relax\check@space}
\def\check@space{\futurelet\@let@token\do@check@space}
\def\do@check@space{%
  \ifx\@let@token\@xobeysp%
    \advance\nspaces 1%
    \expandafter\skip@space%
  \else%
    \ifnum\nspaces>0%
      \print@spaces%
    \fi%
  \fi}
\def\skip@space#1{\check@space}

And finally, we print paired spaces (with a special treat for the odd number of spaces, in this case, we finally out \pdffakespace following a single-space \hskip that will be merged with the previous \hskip and results in a single space during text copying)

\def\print@spaces{%
  \leavevmode\nobreak
  \loop%
    \pdffakespace%
    \nobreak\hskip\dimexpr 2\fontdimen2\font\relax%
    \advance\nspaces by -2\relax%
  \unless\ifnum\nspaces<2\repeat%
  \ifnum\nspaces>0%
    \nobreak\hskip\fontdimen2\font\relax%
    \pdffakespace%
  \fi}

In the end, don't forget to restore @ back:

\makeatother

That's all. Voilà.

Answer 3

Alternative way is to use hyperref package and create multi-line text field:

\begin{Form}
  \TextField[multiline=true,disabled,borderwidth=0,
             width=\linewidth,height=3\baselineskip,
             value={%
  for row in range(1,9):\string\n%
  \space\space\space\space for col in range(1,9):\string\n%
  \space\space\space\space\space\space\space\space print(int(str(row)+str(col)))%
  }]{~}
\end{Form}

In this case you can copy text from Chrome too, but there is (seems to be) no way to adjust text font for text fields (it's always a default sans-serif font). And it doesn't work in SumatraPDF (though copying works badly in SumatraPDF even for usual text).

PS. This is just a proof-of-concept (as I guess nobody will type all those \space and \string\n for real code snippets).