Got a perfect solution for tex.stackexchange.com/questions/702042/…

@Skillmon A new toy for the bunny?

@DavidCarlisle removing ltxcmds from hyperref is hard work ;-). It is not loaded directly in your branch, but by pdftexcmds and pdfescape, and replacing \EdefUnescapeString from the second package is a bit of a challenge.

@samcarter well, then my work PC restarted on its own (thanks group policies), and I lost the file... Maybe I'll revisit it. That was just the result of a custom-defined "mode" of pxpic, I could also write a more complicated parser, evaluating also neighbours, to get the * and + right, and to have the filled area touch the neighbouring lines instead of the strange gaps and missing pixels.

@Skillmon Oh no! What a mean pc!

@samcarter the real question is: Implement the entire logic in TeX, with an option to write results to file or to render them, or write the conversion in another language and only pick up the results?

@Skillmon Tricky question, indeed!

@samcarter and I really wonder how OP draws circles in ASCII-art. I could imagine several different approaches, but none is easy to parse (except for "o" and "O" for small and big circle)

@UlrikeFischer pdftexcomds should be gone and unescaping ought to be doable with l3backend stuff (although I haven't looked at that yet)

@Skillmon should really use picture of course

@Skillmon I'm not sure if it gets any better than o. The python tool @Marijn linked to, seems to use * for bullets.

@PabloGonzálezL \tl_set:ce { l_@@_the_cnt_#1_tl } { \use:c { themycnt#1 } }

(and if you don't want to expand \themycnti etc. but store the macro \themycnti, use \exp_not:c { themycnt#1 } instead of the \use:c bit)

@Skillmon Thank you very much, :ce and \exp_not:c was the line I was looking for :D

@PabloGonzálezL Why the :w signature? It should be :n

@egreg It is a temporary function and I use it several times by changing its definition (it is only to create variables or execute recursive functions)...is there anything wrong with using the :w signature?

@PabloGonzálezL w means a non standard argument such as a primitive delimited macro argument

@PabloGonzálezL if you don't like typing :nnnnnnnnn don't use expl3

@PabloGonzálezL hmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmm

@PabloGonzálezL Yes. If you want a temporary function taking one argument, name it __my_tmp:n, one taking two arguments \__my_tmp:nn and so on.

@muzimuzhiZ I'd normally use dvilualatex as the command, but it comes to the same thing. that's very weird.

@DavidCarlisle Yes, unless you set something else

@DavidCarlisle Same as just latex

@JosephWright yes but I'm not sure it's a good default for luatex (as it can't handle the default fonts) dvips: PK font [/usr/local/texlive/2023/texmf-dist/fonts/opentype/public/lm/lmroman10-regular.otf] not found;

@JosephWright but more importantly where did the - go??

@DavidCarlisle That's ... not my fault

@DavidCarlisle It's the font selection that's questionable, to my mind

@JosephWright no but couldn't we make it default to xdvipdfmx back end?

@DavidCarlisle We could, but that feels horribly inconsistent

@JosephWright matches xetex :-)

so A - B works and A-B doesn't.....

@DavidCarlisle Er, well yes, but that's just how XeTeX is: I meant that in DVI mode, LaTeX assumes dvips unless told otherwise - true for pdfTeX, pTeX, upTeX so makes sense for LuaTeX too

@JosephWright for some definition of make sense. You could probably more usefully say dvi based unicode engines default to xdvipdfmx the others are essentially 8bit and defaulting to tfm fonts, dviluatex and xetex are defaulting to opentype

isn't this supposed to work with plain (dvi)luatex? with dviluatex I get

luaotfload | db : Reload initiated (formats: otf,ttf,ttc); reason: Font "lmroman10-regular.otf" not found....ocal/texliv
e/2023/texmf-dist/tex/latex/base/ltluatex.lua:109:
Module luatexbase Error: Unable to register callback.
(luatexbase)
(luatexbase)             Correct usage:
(luatexbase)             add_to_callback(<callback>, <function>, <description>)
 on input line 4
.
<to be read again>
\rm
l.4 \rm

? x

@muzimuzhiZ the above error is failing to register pre_shipout_filter which also shows in your diff

@DavidCarlisle I did make some test with dvilualatex after tug and improved tagpdf to support it better. I think it did work ok then, but I need find the files ...

@UlrikeFischer The odd thing is A and B work but - fails, some weird handling of discretionary noads?

@DavidCarlisle ah just found it. the file used a package \usepackage{luafonts} they sent me. But I have to make dinner first, it is already late.

@muzimuzhiZ yes but it showed in the diff, but I'm not sure why you are getting a diff at all, the font handling should be more or less the same, I'd have thought. But I don't know, raise an issue at luaotfload?

@DavidCarlisle Done, see github.com/latex3/luaotfload/issues/270.