@Skillmon good night!

@JosephWright wasn't there a \tl_case:nn? It's gone now...

Ah, found my mistake, ignore that ^^^

@DavidCarlisle @egreg @cfr mail to Michael Sharpe sent.

@cis the code shown there runs on texlive.net

@cis if you want a low-level optimised \ResP for maximum performance I'm afraid I'll have to ask for money, I don't feel like implementing an optimised subset of \fp_eval:n... :P

(and l3fp is already extremely well optimised, so I'm not even sure how much μs I could shave of the runtime of the current implementation)

@cis 'A website'?

@Skillmon Only by going LuaTeX-only I think

@JosephWright well, I'm sure I could find spots in which there could be micro optimisations... But otherwise, yes, Lua would be the go-to solution

@JosephWright for instance every \expandafter\nextstep\numexpr...; could be made faster using \expandafter\nextstep\numexpr...\relax;

@Skillmon If you do that, take all of the ; out ...

@cis ask them to update or use a newer tex system. 2019 is pretty old now, especially if trying to run new code.

@DavidCarlisle Yes, I asked. I hope they will update to TL2024 (25?)

@cis do you not have tex locally on your machine?

@JosephWright nope, that would make it slower, because then you'd have to use \expandafter\nextstep\expandafter{\numexpr...\relax}. The ; act as the right-delimiters of the macro.

@Skillmon LuaMetaTeX ...

@JosephWright I know, you could still do this, but with a performance degrade because of \expandafter\nextstep\numexpr...\expandafter;\numexpr...; turning into \expandafter\nextstep\numexpr...\expandafter\relax\expandafter;\numexpr...\relax; (just realised I omitted every \the in my previous messages)

@Skillmon I see the performance argument, but if we make a change here, I'd rally rather be able to handle ; in LuaMetaTeX

@Skillmon Is \numexpr expandable? I guess you want \expandafter\nextstep\the\numexpr...

@egreg obviously, see my comment in that message in ().

@Skillmon but you could presumably use (something) instead of ; (it's only an internal separator isn't it ? (@JosephWright)

@DavidCarlisle Yes - it's just it shows up like 1500 times!

@DavidCarlisle yes, but there was a quick and dirty test in the past to replace it with another separator, but that didn't work, iirc.

@JosephWright well get @Skillmon to use a reasonable editor then that's not a problem:-)

@DavidCarlisle wasn't I who tested that, so blame @UlrikeFischer

@Skillmon I would never blame her for anything

@DavidCarlisle oh no ;-(

@UlrikeFischer but at least @Skillmon will be happy to hear that there are rumours that you have been using a good editor recently.

@DavidCarlisle she used VIM?

@DavidCarlisle I used an OS with in-built xml-validation ...

@Rmano thanks. I would be interested to know what response you get. but \mathscr is not configured as a maths alphabet, so I'm not sure it is really a bug. at the same time, I haven't followed through what it's doing in enough detail to really understand what it's trying to do. so maybe it could be made to behave a bit more as if were a maths alphabet, even though it isn't one.

@DavidCarlisle so is set TEXINPUTS=".;dpi300;;" the correct command?

i would actually prefer linux, as you say :-p. it is just that my main laptop containing linux has broken down, and i am using a spare windows 11 laptop

@ApoorvPotnis my sympathies ;). you could try the texmf.cnf method David suggested.