I could do with interface suggestions for github.com/josephwright/siunitx/issues/554

Setting up so each entry in a set of quantities has a separately-calculated unit is fine, but I'm not sure how best to provide an interface

I have at least two possibles: keep range-units as-is (so the issue only applies to repeat) and have a setting for 'what does repeat mean'

Or have a new range-units = <something> setting

Either way, it only makes a difference when prefix-mode=combine-exponent, so it's ... tricky

This is the last thing I'll tackle for v3.3, so once I know what interface to use, I can ship

\documentclass{article}
\usepackage{siunitx}
\begin{document}
\sisetup{
    range-units=repeat,
    ,exponent-mode=engineering
    ,prefix-mode=combine-exponent
    ,drop-zero-decimal
%    ,<something>
    }


\qtyrange{1e3}{2e9}{\m} % => $1\m\mathrm{km}$ to $2\,\mathrm{Gm}$.
\end{document}

I need to work out <something>

@JosephWright This is getting more and more complex/powerful.

@JosephWright I do find the requested syntax rather confusing and I'm not sure if it is a good idea to provide a simple interface. Perhaps you could offer a \qtyrangesep command, then people could do \qty{1}{Bär}\qtyrangesep\qty{2}{Duck}?

@mickep It's what v2 did, but that didn't really make sense (and was not reliable, and was slow) - I'd say that if you are using a semantic \qtyrange you'd normally expect the two units to be the same

@UlrikeFischer I see the point - there's a reason I changed this in v3 (at the code API level it's quite easy to do - all I'd have to do is make the various separators public, then you can use \seq_use:Nnnn with each entry as a quantity)

\cs_new_protected:Npn \@@_quantity_auxi:nn #1#2
  {
    \@@_print_boundary:V \l_@@_boundary_close_tl
    \seq_clear:N \l_@@_tmp_seq
    \tl_map_inline:nn {#1}
      {
        \seq_put_right:Nn \l_@@_tmp_seq
          { \siunitx_quantity:nn {##1} {#2} }
      }
    \seq_use:NVVV \l_@@_tmp_seq
      \l_@@_separator_pair_tl
      \l_@@_separator_tl
      \l_@@_separator_final_tl
    \@@_print_boundary:V \l_@@_boundary_close_tl
  }

^u

@UlrikeFischer ^^^ Current implementation, as you'll see not very tricky

@mickep github.com/josephwright/siunitx/issues/496 ;) (Down for v3.4)

@DavidCarlisle You really should get funding for texlive.net!

@JosephWright Feels very natural that they are the same, yes.

@JosephWright Oh, even more complex/powerful then :))

@mickep I try to live up to my tag line

@mickep Personally, I'm not so keen on parsing free text, but I understand that users would like it (I'm still undecided on m s^-2 or similar, I think Hans' syntax does need to be covered)

@JosephWright I don't see exactly why to support the context syntax. Easier to copy code between the formats?

@mickep 'A comprehensive (SI) units package' - one of my inspirations from day one was that it should cover all previous/alternative approaches, and ConTeXt is clearly one alternative appraoch

@mickep I've had a few requests for some form of 'free text', and Hans' all-words approach has the advantage that you are not dealing with e.g. ^-2 and trying to pick out a terminating char. But I've not really looked at how hard it will be do to just yet

@JosephWright OK, very ambitious!

@UlrikeFischer I'm thinking that if I can't find a good interface, then something is probably wrong and I might kick to v3.4 and give some more thought

@mickep It comes up, and it's nice to have a feature request for the units code rather than for numbers/tables (most people seem to see siunitx as about numbers, but it's really a units package at heart)

@JosephWright Oh, I think of it as a unit package rather than a package for typesetting numbers.

@JosephWright yes sounds better.

@UlrikeFischer @Rmano has just suggested something (or inspired something, at least)

@JosephWright let me read the thread here... I was stuck with trying to color calibrate a new monitor, what a bunch of broken software...

@JosephWright How do you decide when to use which prefix? E.g lets assume that you want the function also for lists, and someone wants to create a list 1mm, 10cm, 0.5m, 1km by inputting everything in centimeter: \qtylist{0.1,10,50,10000}{cm}

@DavidCarlisle looking at showkeys, is there a way to locally disable showkeys in a macro? I have a macro in which \label is redefined such that I can set two labels with different data. The latter is only accessible using special macros and I do not want it to show up if showkeys is used. Is that doable?

hmm, \let\SK@\@gobbletwo inside a group seems to work

@daleif yes probably or since in newer formats it adds itself via \AddToHookWithArguments{label}{\SK@\SK@@label{#1} you can probably use a higher level hook thing to adjust, but \let\SK@... will certainly stop it

@UlrikeFischer That's the existing settings prefix-mode=combine-exponent means any exponents are turned into prefixes, exponent-mode=engineering means that values are converted to exponents with power-of-three values

@DavidCarlisle Because I'm setting two labels thye key mark ends up on top of each other. But the user does not need to know about this extra label.

@JosephWright can't you reuse this interface? After all a range is simply a list with two elements.

That means it seems I can now redefine the labels used by memoir for \subcaption, \subbottom and \subtop via kernel \label instead of defining personal macros, thus keeping the hook (at least it seems to work in my tests, even works on TL21 and TL22)

@UlrikeFischer That's where the current issue starts ... with the existing code, I look at the first entry, convert the prefix there, then re-use exactly the same prefix and unit with all of the other entries

@UlrikeFischer Don't worry about ranges/lists: internally, they are all handled by a generic 'compound' routine

\documentclass{article}
\usepackage{siunitx}
\begin{document}
\sisetup{
    range-units=repeat,
    ,exponent-mode=scientific
    ,prefix-mode=combine-exponent
    ,drop-zero-decimal}

\qtylist{0.1;10;50;10000}{\cm}

\end{document}

@UlrikeFischer ^^ Try with and without exponent-mode = scientific

@JosephWright I don't get different units in the list, only cm in one case and mm in the other.

@daleif in simpler times you could have used \SK@label for your internal label which is the saved original version before showkeys patches it. But now it uses the generic command hook for label rather than patching the command so it applies to the original version as well (@UlrikeFischer)

@DavidCarlisle I know, I had a look. And the more I play with \ExpandArgs the more useful I find it. Sadly I also want to support TL21, so cannot really rely on it.

@DavidCarlisle something like this:

\documentclass{article}
\usepackage{showkeys}

\begin{document}
xxx\label{duck}

\def\showkeyslabelformat#1{}
x\label{dinner}

\end{document}

Regarding the question about latexrelease, I've set memoir to require a format from 2021/06/01 onwards. I thought 2022/06/01 was too early

@UlrikeFischer I have no tex memory after 2000 so \changes{v3.13}{2006/01/09}{Added command} is too young for me

@UlrikeFischer that was probably even better, as \let\SK@\@gobbletwo requires you to know what \SK@` is and where it comes from.

@DavidCarlisle you could have checked the documentation (I looked into the code ...).

@WillRobertson hi

I am lurking while trying to get kids to sleep

i forget I can stop by to read the latest gossip

I used Joseph’s S column today and was very thankful for it

@WillRobertson :)

@UlrikeFischer Er, no, that's the point: the current behaviour assumes that if you use \qtyrange or similar, the range is made up of values which need to be 'compared', so have a single prefix to the unit. But people want to use it for other stuff ...

@WillRobertson Yay!

\IfFormatAtLeastTF { 2020-10-01 }
  { }
  {
    \RequirePackage { xparse }
    \providecommand \ExpandArgs [1]
      { \cs_if_exist_use:c { exp_args:N #1 } }
  }

@WillRobertson hello

@daleif ^^^ With possibly a definition for \IfFormatAtLeastTF also

@JosephWright yes I got that, that's why I asked about lists. If you want to extend the syntax to support more than one prefix, then imho this should apply to general lists and not only to ranges.

@UlrikeFischer Sure: I'm using ranges as an example, any extension will be in the generic code which covers ranges, lists, powers

Just curious, but how many people here primarily stick with plain TeX? In grad school I used the TeXBook as a fruitful field for procrastination and ended up kind of falling in love with plain.

@B.Wilson Very few for 'real' work

@B.Wilson (To be clear, I do mean non-zero)

Okay. I guess that's about where my priors sit.

@B.Wilson I don't, I only use it if I have to make a small example for bug reports.

@UlrikeFischer and Christmas carols

Hrm. Okay. Assuming decent TeX familiarity, I think it's pretty reasonable to code up non-trivial things. I've made complicated, annoying government forms and the like before.

Granted, it's also not reasonable to ask everyone to read a giant tome just to get started. lol

@B.Wilson more than that plain tex is "appendix B" "Basic macros" that is, it is examples for the texbook, never really intended as a production document format. It is of course possible to build a format above that (eg Knuth's manmac or amstex or whatever) but typesetting directly in plain isn't usually reasonable, although if you are writing in English you may get away with it for some classes of document

@B.Wilson that is not the point, I have read the tex book, but plain is simply not powerful enough for my needs. E.g. something similar to nfss or inputenc is missing (as a german I need sensible support für curious chars with dots on top.).

Is non-ascii not just an engine issue? Most stuff I've done is CJK which eptex handles quite nicely.

@B.Wilson but by then you are not really using plain tex in any real sense

Sorta. But the gulf between TeX and LaTeX is way larger than TeX and plain+eptex.

I dunno. My question certainly isn't precise. Mainly, I just want to know how many people use TeX "close to the metal" per se, instead of relying on large stacks of macros.

@B.Wilson but you didn't ask "are there alternatives to latex", you asked, "do people use plain"

@B.Wilson ^^ that is what I get with eptex from grüße \bye, which is not quite the wanted result.

@B.Wilson Well, there are definitely plain TeX users, and there are people using OpTeX (plain + utilities on LuaTeX), as well as people using plain constructs in LaTeX and ConTeXt, and ...

@JosephWright That's interesting!

@B.Wilson Like I said, a small number compared to the number of LaTeX users (ConTeXt user base is also small)

@B.Wilson Of course, you asked for a sense of the scale: I think we don't have any regulars in the chat who use mainly plain, we have largely LaTeX users with some ConTeXt users

@UlrikeFischer Yeah, ptex/eptex is specifically for CJK. What about luatex or something?

In other news: github.com/josephwright/siunitx/commit/… (@UlrikeFischer - take a look?)

@B.Wilson I would guess "small number" means possibly a few thousand compared to a few millions or tens of millions of latex users.

@B.Wilson ^^ with luatex

@B.Wilson With the right font loaded, it would be fine - but then you need to load luaotfload (or roll your own), and I'm not sure that would be in the 'no macros' style either (granted, that code involves functions not macros)

@JosephWright Thanks. That's helpful to know.

@B.Wilson For scale, we know that Overleaf has >10 million users (of which at least 10% are active), and whilst you can use a non LaTeX interface, it's built on LaTeX, so that's 1 million LaTeX users before you look at those not using Overleaf

@B.Wilson I'd say the majority of the regulars do know how to use plain (e.g. likely have read The TeXbook, TeX by Topic or similar) - coding for the LaTeX kernel, for example, I need to know how all that stuff works

Maybe I'm just stuck in a weird preference hole and should just learn LaTeX proper. I'm sure there are people who fluidly use LaTeX without feeling mired in a myriad of one-off solutions.

@B.Wilson It depends what you want: there is no 'right' approach

@B.Wilson No one is saying you can't be happy using plain, just that most people find that the limitations outweigh any benefits

My (naive) understanding is that it's mostly a tradeoff between needing to roll your own stuff but having a relatively small, grokable language versus having lots of pre-made solutions but little flexibility once outside of pre-determined parameters.

@B.Wilson I'd usually describe it the other way round, the extra complication in latex macros is because they try to supply generic solutions, whereas plain offers simple macros targeted at a single use (or a single language) with the intention that you change the source if you have any kind of different use.

That sounds like a fair take.

@B.Wilson there is also the aspect of a shared language for communication. If you code up your own plain tex macros for your own syntax, that seems simple, but you can only easily share your document with yourself. If you use latex markup it works across many journal formats, in mathjax, and other places. That is why essentially no journals accept plain tex macros, they want a consistent known macro set they can use (or convert to their in house typesetter) so they specify latex submission

You've basically hit on the biggest pain point I've had with (spiritually) plain TeX so far. It shoves one off in some weird corner, increasing friction outside of its own little world.

Anyway, thank you all for entertaining my question. I was half hoping to hear there was a nice thriving community of plain TeXnicians or something, but everything you say kind of does line up with my experience, whether I'd like to admit it or not.

@B.Wilson the language doesn't grow with latex. At the end you have the same set of primitives. And you don't have to load tikz or hyperref or biblatex or any premade solution, if you want you can write your own macros. But at the end it is lots of work to do all of this yourself. Do you e.g. know how to create links and bookmarks or tikzducks in plain tex?

@B.Wilson what we said is true but you were discussing this with three members of the core latex development team, who some would say may not be the most unbiased observers:-)

@DavidCarlisle :)

@B.Wilson For many of us, the attraction to LaTeX comes from the fact that there are great packages that do specific things in our fields. Many of them don't have plain TeX equivalents.

Ah, @DavidCarlisle, God sends rain on the just and the unjust. And sometimes it falls in the final innings of an Ashes test. In England. In the summer.

@DavidPurton Rain, in England. Who would have predicted that?

@MarcelKrüger are you there?

It wouldn't happen in Adelaide

@DavidPurton but you can't even get the summer at the right time of year

@DavidCarlisle well it's not raining here today...

@DavidPurton fly them all back and finish the game there?

Anyone know latex4technics.com?

Came up in a google search of something tikz

Lol, it is using TL17. Nevermind

@daleif :)

It was interesting though that it runs both latex and mathjax. Might be useful for some

@daleif one reason for keeping texlive.net interface aggressively simple is that updating it remains essentially trivial, just tlmgr update --all every now and then.

Internal implementation of labels for \subcaption and friends in memoir is now down to a 1/4 (number of code lines, 59 to 15) and uses the kernel \label macro to set the two labels even if we locally redefine it. No more manually writing \newlabel to the aux.

As far as I can see it still gives the same output in (TL21, 22 and 23)

@DavidCarlisle exactly

@daleif nice ;-)

@UlrikeFischer I needed a copy of the kernel \label. I just using \AtBeginDocument{\letmem@kernel@label\label} is there something better?

@daleif why do you need a copy?

@UlrikeFischer The macros internally sets \label=internal macro in order to set two labels (it has always done that, not going to change that) with different values for \@currentlabel.

Problem here is also that these macros are always a part of the title, always been like that, so \nameref cannot pick up the title. Is not going to change that.

@daleif sorry I don't understand, what a simple copy changes here. What is the difference between \label{a}...\label{b} and \label{a}...\mem@kernel@label{b}?

@UlrikeFischer inside say \subcaption in memoir, \label is set to \memsub@label, and we want \label{key} to run \mem@kernel@label{key} and \mem@kernel@label{sub@key} (with different values of \@currentlabel). I cannot of course use \label as it gives a loop.

So it is basically one \label that ends up running two \label commands

I was just wondering whether the kernel had a \kernel@prestine@label macro somewhere. Looking at the nameref code I conclude it does not as \NR@label@copy is hard defined in the file.

@daleif in which sense is \label not pristine? And if isn't, why does it help do make a copy at begin document?

@daleif which means that I still don't get, why \memsub@label doesn't simply call \label{key} and \label{sub@key}.

@daleif don't assume that such a copy will stay. There were needed as nameref patched \label, but at some time it will disappear.

@daleif ah, looking at the code, I understand. You want the user label command to set two labels

@UlrikeFischer exactly, and I need a copy or I get a loop

I think the implementation was taken from subfig/subfigure back in the day.

@daleif but do you really want a full label including everything that is perhaps in the hooks?

@UlrikeFischer that is what the old implementation did by hand. I can just get away with all that manual \newlabel stuff that are in current memoir with this method. I'm not entirely sure if we want the whole thing. To me it only seems to be a simple method of getting just the subnumber.

When something like zref is in the kernel, we could just store the information there. For now nameref and autoref seems to agree on the data and then memoir has \subcaptionref to access the extra label to get the number.

I think I'll leave it like this for now. Then it can be changed later. It does seem to do the same thing with and without hyperref loaded. and also works for TL21-23.

@daleif Since you are into this stuff, I'm still hoping for that subcaption hook. :-)

@daleif how do you retrieve the number? With a normal \ref etc?

@daleif There's l3ref-tmp, but @UlrikeFischer says its a secret.

@UlrikeFischer \subcaptionref is just \ref of sub@userkey

plus the bonus of hyperref support.

github.com/josephwright/siunitx/releases/tag/v3.3.0

The only difference in the old implemtation between the two labels is the value of the currentlabel. Same with the new implementation.

@daleif I think a more manual solution is better. I would do something like \AddToHookWithArguments{label}{\mem@sub@label{#1}} and then (re)define locally \mem@sub@label to write the additional \newlabel command. Perhaps we could provide a kernel command which writes only the newlabel.

@gusbrs did you send me an email with MWE about that?

@UlrikeFischer that hook is not in TL21.

Why should the standard \label with its hooks be bad here

@daleif There was no MWE for that, since there was not arguable misbehavior. Just a feature request for a "handle" so that I can add a zref property in the right place and support memoir subcaption references. I can provide a MWE for you to play, but that would be a zref-clever MWE rather than a memoir one.

@gusbrs as long as I have some code and an explanation ...

It is getting a bit late at work, I'm outta here.

@daleif Ok, I'll send a MWE for this later this afternoon.

@daleif well it feels overdoing. And I'm using the hook also for tagging and wondered if you then get the correct reference.

hmm what can go wrong here: if tonumber(luaotfload.version) < 3.19 then ;-(

@UlrikeFischer Er ...

@UlrikeFischer Where's that?

@JosephWright luatexja, it broke the luaotfload tests as in dev we have version 3.24-dev ;-)

@UlrikeFischer i think it might be an idea to have a way to add bookkeeping labels without the hooks. The stuff to strictly check a page (don't remember the macro and is no longer at pc), but there is code to strictly see if say a marginpar is on an odd or even page and those are label based. Though I'm not sure if they use \label

@daleif it depends on how you want to reference. If you do it only with an internal command I would use zref or l3ref once it exist, that is what I use e.g. in the tagging code. If you want \ref and \pageref to work, you need a "normal" \newlabel in the aux.

@UlrikeFischer I'm here now.

@MarcelKrüger I was looking at lualibs/luaotfload and was wondering why luatexja failed, but already found the reason ("not our fault"). If the tests works, do you have time to upload luaotfload?

@UlrikeFischer I'll think a bit about the system font PR and then upload once that is resolved.

@MarcelKrüger the stuff from xdvipsk? We can request changes if their additions are odd, this dvips variant should in the long run go into texlive and that mean it should make sense also for other users.

@UlrikeFischer They had two PRs. The DVI backend is now in DEV but with a somewhat different implementation, but remaining one is not directly xdvipsk related but related how their setup works: Basically they want to completely ignore system fonts and not even have them in the cache (because they want tp share the cache across systems)

@MarcelKrüger I think it wasn't the first time people wanted to use fonts only from local directories or from the texmf, so an option to configure that isn't bad imho.

@UlrikeFischer Yes, I think the main question is on which level this should be done. This kind of system wide configuration can easily lead to really weird issues because it leads to documents behaving very different without changes in the document. Of course, system fonts aren't portable anyway so it's probably fine.

But mostly I think that the central part should be to change where lookup happens and that the caching effects should at most be a consequence of that instead of not caching system fonts and suppressing the lookup as a side effect since he cache happens to be empty.

@MarcelKrüger yes I agree, it should be possible to change the lookup without changing the cache. Perhaps one could have both? A configuration what is cached (for people who really don't want to cache system fonts) and one what the lookup should do?

@mickep if you have any insights about this, the math in chrome devs would appreciate it... github.com/w3c/mathml-core/issues/126#issuecomment-1635552830

@UlrikeFischer I mean you can already disable lookup in the actual document, do I'm considering to just make the default configurable and then disable the system font caching if the lookup is disabled at cache creation time.

@DavidPurton and the rain stopped in time

thanks

@MarcelKrüger you can?

@DavidCarlisle The parameter seems to be wrong in many fonts. See below the example with bonum, first with it set to 1900, secondly with the value from the font, 1250.

(The wrong horisontal placement of the lower limit a is due to other reasons)

@B.Wilson -- Some of the things I do for TUGboat are basically plain. For example, checking hyphenation for English text can't safely be done any other way.

@UlrikeFischer You can, the interface is just terrible (something like \directlua{fonts.names.set_location_precedence{'texmf'}} to restrict to texmf)

@DavidCarlisle @MarcelKrüger do ltluatex and luatexbase clash? (see question on main site)

@MarcelKrüger ;-). But we should document it (or add an interface).

@barbarabeeton ?

@DavidCarlisle -- Taking no chances, I use raw tex, not even pdftex, to be sure nothing extraneous creeps in. It's too much of a nuisance figuring out how to explain a misinterpretation when someone questions something in one of my exception lists.

@barbarabeeton as I suspected, so tex primitives rather than plain

@DavidCarlisle -- I think the tex engine I'm using has been built with plain, and already includes DEK's exceptions.

@barbarabeeton ah ok so you do the checks separately not inline in the author's latex article

@DavidCarlisle -- Yes. run "tex", enter \relax/return, then \showhyphens{...}. Very primitive, but trustworthy.

@barbarabeeton but you'd get same answer from pdflatex:-)

@DavidCarlisle -- That's what I use to set the exceptions article. But I do double check.

@barbarabeeton we could get brave one day and make latex default to 8 bit fonts (as we did with lua/xelatex defaulting to unicode) but until then it'll be the same barring bugs

@UlrikeFischer I posted an answer

@DavidCarlisle ah ok (I wondered if there were recent changes in ltluatex that got lost)

@DavidCarlisle -- Nah. The system used now is working. Please don't suggest anything that might slow things down. It's already more than complicated enough to try to get TUGboat out on time.

@barbarabeeton but you do get brillant reports, by lovely authors

@DavidCarlisle -- But I have to spend so much time making sure that "teh" doesn't sneak into print.

@barbarabeeton I checked that!

@UlrikeFischer ^^^^

@DavidCarlisle ;-)

@UlrikeFischer -- Thank you!

@PauloCereda blame you for abntex2? github.com/davidcarlisle/dpctex/issues/44

@daleif Sent. Let me know if something is not clear about it. And thanks!

ABNT in Word. Grounds for divorce? :)

@AlanMunn Isn't Word pretty OK these days? :)

@mickep Not if you no longer know how to use it...

@AlanMunn As with other tools...

@mickep Also, you may not be familiar with ABNT...

@AlanMunn That is indeed true.

@barbarabeeton another comment you could have made on my \not answer: tex.stackexchange.com/users/579/barbara-beeton

@DavidCarlisle we also inherited the centernot package ...

@UlrikeFischer ah that one I sort of remember

@DavidCarlisle your link for Barbara is wrong, I think you wanted to link to the comment?

@UlrikeFischer oops, she'll figure it out

@DavidCarlisle \caption{it out}