Include column information into synctex?

[user202729]

I think this is not too difficult, the (1-indexed) column can be computed as loc - start + 1 provided state != token_list.

I already have a proof of concept implementation at user202729/luatex@synctex-column , which appears to work reasonably for simple documents.

I'd like to ask:

• is there sufficient interest in this idea?

• while I implement this, I don't really understand the memory layout, which leads to me packing the line and column into a single int like a bit mask. This leads to a limitation of file length less than $2^{24}$ lines and column less than $2^8$.

  How can I modify the memory layout to add an extra field column into the box?

[jlaurens]

I am not sure whether column numbering significantly helps.
Concerning memory management, things are very different on pdftex, xetex and luatex. I am already thinking of using the memory actually dedicated to tag and line numbers to store more than 2 integers.

[user202729]

That's also a good idea. (which is what I did in my proof of concept implementation, but packing the column number in top 8 bits of the line number)

In retrospect, given that the tag number is more likely to be small than the line number, it's probably a better idea to pack the column number into the tag number instead. (Let's say, 16 bits for the tag number and 16 bits for the column number? Is that safe enough --- surely a typical implementation can't use up to 65536 files?)

Quick test for how long it takes to input 65536 files, doing nothing else:

%! TEX program = pdflatex
\documentclass{article}
\begin{document}
\makeatletter
\immediate\openout13=b.tex\relax
\immediate\write13{\@percentchar}
\immediate\closeout13
\def\a{\ifnum\count@<65536 \input b.tex\relax\advance\count@ by 1\relax\expandafter\a\fi}\a
\makeatother
1
\end{document}

Result: Around 14 seconds on my machine.

[jlaurens]

Some remarks

• One good practice is to let the user choose between column support or not...
• I need another number to tag some boxes
• I have difficulties to foresee the real enhancement of column support mainly due to macro expansion mechanism.