November 2024 - dev-context - NTG Mailing Lists

Re: LuaMetaTeX accepts `;` as part of `\numexpr`, which breaks LaTeX3
by Hans Hagen 30 Apr '25

30 Apr '25

On 11/8/2024 9:53 AM, Joseph Wright wrote: > Hello both, > > On 06/11/2024 15:28, Hans Hagen wrote: >> On 11/6/2024 12:37 PM, Vítek Starý Novotný wrote: >>>> Anyway, today I'm not in the mood to ponder semi colons, as it's the >>>> least of my worries, >>> >>> I am very sorry to bother you and other ConTeXt devs with this. If it >>> makes it feel any better, the LaTeX folks seem equally annoyed with me. >>> >>>> I'm not sure what "just %" means here. >>> >>> These were Joseph Wright's words. I assume Joseph was referring to the >>> fact that in e-TeX and LuaTeX, including a comment after a `\numexpr` >>> causes the comment to be ignored by TeX, similarly to how in LuaLaTeX, >>> including a semicolon causes parts of the following text to be ignored. > > It's not a comment - it's that the standard e-TeX implementation for > \numexpr only accept a very limited number of characters, and any > characters that are non-expandable and not in that set are treated as > end-of-expression. sure, that's what i thought but it's good to have that thread fixed >> Well, that's just numexpr then, but how about: >> >> \def\fool{\hskip10pt plus 0pt} >> >> oeps \fool me >> >> oeps \fool minus me >> >> This whole lookahead scanning is fragile by design and just like one >> should define >> >> \def\fool{\hskip10pt plus 0pt\relax} > > Most of the time I agree, but \relax is a lot longer than (say) ; and if > you are writing a complex parser that can be handy. > > [In the end, if ; stays as part of the extended \numexpr syntax for > LuaMetaTeX, at some stage I will have to re-work the code causing the > issue here to use another token in place of ; - \relax (in our case > \scan_stop:) really is a bit long, so probably something like @ is the > viable change.] sure, but this expl stuff si pretty verbose and long winding anyway (or that's what i remember from some BT presentation) >> one should also make sure an expression doesn't scan ahead, if only >> because it will expand whats coming next; basically \relax or matching >> () should have been mandate and it's a design flaw. > > Arguably yes but e-TeX was finalised in 1999 - that was an discussion > for a previous millennium (at least from a plain/LaTeX point of view) :) as much of tex .. >> As mentioned there, we also cleaned up some function names (or moved >> some to another namespace); there's a reason why luatex was frozen and >> luametatex was the follow up. > > Of course - but one day one might imagine LuaMetaTeX will more something > that makes sense for non-ConTeXt users too. (Plain and OPTeX as well as > LaTeX, I imagine.) i can add \etexexprmode=1 that ignores : and ; operations (not much overhead) as long as it doesn't trigger all kind of "can you also do this or that" demands to make < 1999 inspired code happy) if course \etexexprmode defaults to zero Hans ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | www.pragma-ade.nl | www.pragma-pod.nl -----------------------------------------------------------------

4 9

Re: LuaMetaTeX accepts `;` as part of `\numexpr`, which breaks LaTeX3
by Hans Hagen 09 Nov '24

09 Nov '24

On 11/8/2024 6:05 PM, Joseph Wright wrote: > True but when you want half a dozen tokens in an argument, length counts > :) We use quite a lot of ";" currently to implement our expandable > floating point stuff, and that likely has to stay in macros even in > Lua(Meta)TeX for ever - I fear subtle differences from pdfTeX if we > tried to implement differently. sure. (btw, fwiw using \relax as sentinal is faster because there is no pushback needed and and internally - as macro body - a ; or \relax takes the same amount of memory and i suppose most goes into macros but performance and memory consumption is a topic on its own and probably very different between macro packages) >>>> one should also make sure an expression doesn't scan ahead, if only >>>> because it will expand whats coming next; basically \relax or matching >>>> () should have been mandate and it's a design flaw. >>> >>> Arguably yes but e-TeX was finalised in 1999 - that was an discussion >>> for a previous millennium (at least from a plain/LaTeX point of view) :) >> >> as much of tex .. > > Indeed - I know you have a different view on how to handle that than the > LaTeX one :) the fact that there was never a fundamental follow-up etex (language wise) kind of tells a story i guess but let's not waste time on that Hans ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | www.pragma-ade.nl | www.pragma-pod.nl -----------------------------------------------------------------

1 0

LuaMetaTeX accepts `;` as part of `\numexpr`, which breaks LaTeX3
by Vítek Starý Novotný 06 Nov '24

06 Nov '24

Dear ConTeXt developers, since TeX Live 2024, LaTeX3 can no longer be used with LuaMetaTeX. [1] The issue seems to be that unlike e-TeX and LuaTeX, LuaMetaTeX accepts `;` as part of `\numexpr`. Since `;` is commonly used as a delimiter in LaTeX3, this is a significant issue. The issue would be best solved in LuaMetaTeX by sticking to just `%` like e-TeX and LuaTeX do. Would that be an option? [1]: https://github.com/latex3/latex3/issues/1518 Best, Vitek

2 3