TeX, LaTeX and Friends

CarLaTeX

05:15

Happy Birthday, @PauloCereda!

6

Skillmon

06:25

@JosephWright why are my braces so loose? tex.stackexchange.com/a/122842/117050

Joseph Wright

@Skillmon :)

Skillmon

06:54

@FaheemMitha sorry, late to the party, indeed your question is a dupe (even though I answered it -- but my answer contains a list of packages that have predictable brace stripping; in which one package is deliberately missing: ltxkeys has predictable (but wrong) brace stripping as well: It doesn't strip any braces, meaning it's impossible to pass in a value that contains a comma that isn't surrounded by braces, oh and it used to break LaTeX completely but nowadays doesn't even load).

@FaheemMitha oh, and there used to be a package (pgfkeyx iirc.) that used to "fix" the wrong brace stripping and category code fragility of pgfkeys by the author of ltxkeys, but it isn't maintained and no longer compatible with pgfkeys.

David Carlisle

@CarLaTeX roast dinner to celebrate!

Faheem Mitha

07:14

@Skillmon Thank you for your answer. I haven't looked properly at it. But if pgfkeys has such issues, has anyone actually reported it? And if so, do you know the link?

CarLaTeX

@DavidCarlisle You're mean!

David Carlisle

@Skillmon I think it's best to treat keyval brace stripping as correct by definition and judge imitations of that package accordingly

Jun 29, 2017 at 16:15, by Paulo Cereda

@DavidCarlisle you are not mean :)

Joseph Wright

@FaheemMitha It's a feature

@DavidCarlisle :)

@DavidCarlisle @Skillmon and I both have what I suspect we would call 'reference' implementations :)

Ulrike Fischer

07:37

8

@PauloCereda ^^^

Rmano

07:52

¡Feliz cumpleaños! @PauloCereda

4

Skillmon

@FaheemMitha the maintainers are aware, but there is nothing they can do for backwards compatibility.

mickep

Grattis på födelsedagen @PauloCereda!

5

Joseph Wright

@Skillmon 'Tell people to use l3keys/expkv' ;)

Skillmon

@JosephWright which I did :)

Joseph Wright

@Skillmon Sounds like a plan to me

@DavidCarlisle I see textcase is no more ;)

Skillmon

07:55

@JosephWright got the update yesterday

Joseph Wright

@DavidCarlisle Answer tex.stackexchange.com/questions/647525/…?

@Skillmon Yup - I just updated this morning

@Skillmon Going to be part of my TUG talk

Skillmon

@JosephWright yeah, saw your abstract here.

David Carlisle

@JosephWright well it's hopefully still there, just different

Skillmon

@DavidCarlisle is the new code a line longer than indentfirst?

Joseph Wright

@Skillmon You considered a talk? Keyval-related, perhaps?

David Carlisle

07:58

@JosephWright I was going to close as fixed in current release but I could answer?

Joseph Wright

@Skillmon My one will be the TUG2015 one + 'now it works in pdfTeX, and it's what's being used for \MakeUppercase'

@DavidCarlisle That's an answer, no?

samcarter

@PauloCereda Alles Gute zu deinem Geburtstag!

Joseph Wright

@DavidCarlisle Yes, but realistically we could for example start updating answers to say '2022-06-01 onward: just use \MakeUppercase'

Skillmon

@JosephWright I guess the most important keyval-related changes are yours to present, aren't they? expkv is just a small package only few use, changing the package option code is bound to have much more influence.

David Carlisle

@Skillmon yes it's two \let so 6 tokens so 50% bigger than indentfirst

Joseph Wright

08:00

@Skillmon That's true, but I think a talk surveying, looking at performance, etc., might be good - there are reasons for picking keyval, l3keys, expkv and perhaps pgfkeys, depending on the context

@DavidCarlisle :)

@Skillmon I guess I could talk about the option handling: hadn't thought of that

Ulrike Fischer

@JosephWright I wanted to mention it in my second talk (and even tested it yesterday a bit). How do one get a documentation for it?

Skillmon

@JosephWright well, my biggest interest is how to hack into it so that other option packages also profit from the new mechanism to circumvent option clashes :P

Joseph Wright

@UlrikeFischer I think at the moment it's only in source2e - I guess should be in clsguide

@Skillmon :)

Ulrike Fischer

@JosephWright texdoc source2e is fine.

David Carlisle

08:21

0

A: MakeTextUppercase broke for cyrillic

In LaTeX releases starting from June 2022, \MakeUppercase is an improved version of the \MakeTextUppercase command provided by textcase: it avoids math in the same way but has improved Unicode support, being based on the expl3 case changing functions. A new version of textcase has been relased th...

Skillmon

@DavidCarlisle please say that you didn't change the documentation, so that it now has the best documentation/code length ratio of all!

@JosephWright should be clsguide, documentation on kernel stuff is currently very fragmented unfortunately

David Carlisle

@Skillmon I added to the doc:-)

Joseph Wright

@Skillmon I've raised on the team list - I will see where people want it

@Skillmon You could volunteer to put in a load of PRs to tidy up ;)

David Carlisle

@JosephWright clsguide3 would match usrguide3 (but I think usrguide3 would be better merged into usrguide I'll look at email...

Joseph Wright

@DavidCarlisle Yes that's a possible

@DavidCarlisle One for a team meeting I suspect

Skillmon

08:50

@JosephWright and when should I do this? Currently I don't even manage to get a rewrite of expkv's documentation done (I intend to bundle the different packages up and provide a single documentation for them, plus adding some auxiliary package to reduce code duplication -- and the rewrite is ongoing since I think January...)

I also didn't start working on the code PRs I still got open (and most likely forgot about quite a few which were on my imaginary todo-list)

David Carlisle

@Skillmon schedule it for after the longtable v5 release

Faheem Mitha

@JosephWright Is that the official word on the topic?

:61367324 I thought PGF/TikZ was on Sourceforge.

And it's shown as fixed, too. Is this the same issue? It looks like it to me, but I often get these things wrong.

Paulo Cereda

@CarLaTeX awww thank you <3

@DavidCarlisle oh no

@UlrikeFischer awww thanks so much <3

David Carlisle

birthday breakfast

Paulo Cereda

@Rmano muchas gracias, señor squirrel <3

@DavidCarlisle oh no

@mickep awww thanks <3

@samcarter aww dankeschon! <3

Faheem Mitha

08:58

@samcarter The current version of the manual mentions GitHub. Specifically, it says

David Carlisle

@FaheemMitha see support and issue links at ctan.org/pkg/pgf

Faheem Mitha

> If that does not solve the problem, try having a look at the GitHub development page for pgf and TikZ (see the title of this document). Perhaps someone has already reported a similar problem and someone has found a solution.

On pg 29. Further down it talks about bug reports, but does not mention where it should go. At the end of that section it says

> As a last resort you can try to email me (Till Tantau) or, if the problem concerns the mathematical engine, Mark Wibrow.

It seems likely that's out of date. Is Tantau even working on this project now?

@DavidCarlisle Yes, I see. OK.

David Carlisle

@FaheemMitha in practice, no, but he has not formally stopped

Faheem Mitha

@DavidCarlisle Oh.

mickep

Reading ltnews35 about all great news. Wonder if "\usepackage{doc}[=v2]" is a typo, or if the key really should be entered that way.

Rmano

09:13

@mickep It's correct, it's the "rollback" thing in action...

latex-project.org/publications/…

mickep

@Rmano Oh, then it is just me not being used to nothing being before the =.

@Rmano Thanks! It is everywhere! :)

Marcel Krüger

09:25

@mickep It's not supposed to be a keyval with empty key, it's supposed to indicate that you are requesting an exactly the specified version and not just a minimal version.

mickep

@MarcelKrüger Yes, I understand. I just never before saw those = without anything in front of them.

Paulo Cereda

@MarcelKrüger ooh

maybe we could go to the future

\usepackage{siunitx}[=v4]

Skillmon

@PauloCereda well, in theory we can, the labels for the =<something> syntax are author-defined, so I could release a package that is currently v3 but also include a prerelease and exploit the rollback syntax for a rollforward

Paulo Cereda

@Skillmon OOH

Skillmon

@PauloCereda you could also give meaningful names to your version-keys, e.g., you could say \usepackage{animal}[=duck] for the duck-release and \usepackage{animal}[=anteater] for the anteater-release.

Paulo Cereda

09:33

@Skillmon oh my, this is awesome

When Adele meets TeX, rolling in the heap :)

samcarter

@PauloCereda Only @JosephWright is allowed to use this - he is a timelord

Paulo Cereda

@samcarter ooh

@samcarter and has three hearts

samcarter

@PauloCereda <3 <3 <3

Paulo Cereda

@samcarter ooh

Skillmon

09:51

@PauloCereda I thought it was two?!

Paulo Cereda

@Skillmon oh I actually don't remember, is it two? I might be thinking of that alien bloke from 3rd Rock From The Sun....

Skillmon

@PauloCereda I'm not sure, too. It's been a while...

Paulo Cereda

@Skillmon just googled it! You are absolutely right: two hearts!

Rabbits are very good at Doctor Who. :)

yo'

10:16

@MarcelKrüger Oh cool, thank you! (Good that I have The TeXbook at home :-) )

Paulo Cereda

@yo' ooh

gusbrs

10:54

@PauloCereda Parabéns, meu caro! Felicidades!

Paulo Cereda

@gusbrs Obrigado, meu amigo! <3

gusbrs

@PauloCereda I've been using arara these days. :)

mickep

@yo' This is the one I carry around everyday. It is a bit lighter:

2

Paulo Cereda

@gusbrs oh no :)

gusbrs

@PauloCereda Oh yes. :) Making my life easier with makeglossaries. And I also learned how to make it play better with AUCTeX.

Paulo Cereda

10:57

@gusbrs ooh :)

David Carlisle

@gusbrs and you managed to get some output?

gusbrs

@DavidCarlisle Yes, but mostly in the log file. :)

Btw, what I learned to appease AUCTeX log parsing of the log file is to run all but the last round in draftmode.

Paulo Cereda

11:14

@DavidCarlisle no ducks were hurt in the making of the document. :)

Skillmon

@PauloCereda /insert ducktape joke here

gusbrs

Ops, I meant batchmode above.

Paulo Cereda

@Skillmon ooh :)

gusbrs

@Skillmon They only get hurt when ripping the tape off. :)

mickep

Wouldn't it be better to use the new fancy \mathcolor here? tex.stackexchange.com/a/647839/52406

David Carlisle

11:30

@mickep yes

mickep

@DavidCarlisle Oh, everybody should read ltnews :)

David Carlisle

@UlrikeFischer re your texlive list message, you are better off without that package. It causes all sorts of trouble.

2

Ulrike Fischer

11:50

@DavidCarlisle ;-) I should avoid to open documentations, that causes all sorts of trouble too.

Paulo Cereda

12:00

overleaf.com/blog/new-feature-ready-set-detach

7

Joseph Wright

@PauloCereda I see your duck has a name ...

mickep

@PauloCereda Very nice feature!

Paulo Cereda

@JosephWright ooh :)

@mickep :)

gusbrs

12:16

@JosephWright \DeclareRobustCommand\DuX{D\kern-.0667em\lower-.25ex\hbox{u}\kern-.125emX\@}

Ulrich Diez

12:38

LaTeX and external databases, e.g. .csv files - there are packages like datatool and csvsimple that can be used to read/process .csv files directly in the LaTeX run. Are there any papers about what to consider when intending to have read databases directly with LaTeX - e.g. in the design of the data structure, so that workflows both with and without the involvement of (La)TeX do not cause problems?

Or more exotic scenarios like creating a database that is supposed to consist of snippets of LaTeX code, where you have to keep an eye on tokenization and timing and chronological order of expansion - also with regard to hooks and output routines? Probably also covering some basics of database-management and relational algebra? I'm considering writing something about this myself. But if there is already something better than what I could write, my consideration is obsolete.

daleif

13:07

@UlrichDiez Instead of trying to read stuff directly from LaTeX it self, why not just use an exaternal programme to generate the LaTeX code for you. That might be a lot safer.

Faheem Mitha

13:22

@UlrichDiez I found datatool very slow. I use LuaTeX and a Lua database adaptor library. LuaSQL, I think. And SQLite.

Converting from CSV to SQLite can be done by using SQLite's own builtin tools.

David Carlisle

@PauloCereda only hide pdf or editor. Not both :(

Paulo Cereda

@DavidCarlisle ooh

David Carlisle

13:44

@JosephWright lwg this evening?

Joseph Wright

@DavidCarlisle My diary has that next week .. am I out-by-one?

Marcel Krüger

@FaheemMitha Using native libraries in LuaTeX requires shell escape and therefore significantly more trust in the used packages and documents. Especially if you want to share documents with other people it's often better to avoid this. CSV is pretty straight forward to parse directly in Lua and there should also be some libraries available.

Faheem Mitha

@MarcelKrüger There are always tradeoffs, yes. I just use LuaTeX for my own personal use.

David Carlisle

@JosephWright my calenda also has next week but email from Besy on the 13th says "15th June) which was why I asked...

Faheem Mitha

@MarcelKrüger Perhaps the LuaTeX people could add LuaSQL to LuaTeX? They seem to have added lots of other things. Most of which I was not aware of till someone pointed them out. And which aren't document or even mentioned in the LuaTeX manual, AFAIK. E.g. (some version of) LPEG.

Alternatively, perhaps use a more fine-grained version of shell-escape which just trusts certain binaries?

David Carlisle

13:50

@FaheemMitha that is the default setting

Faheem Mitha

@DavidCarlisle Third party Lua binaries, to be clear.

Or some suitably blessed Lua binaries, perhaps.

David Carlisle

@FaheemMitha lpeg gets prominent listing in the luatex manual

Faheem Mitha

@DavidCarlisle It does? Let me take another look.

OK. I see it's mentioned in "4.3 LUA modules". There's also the document that comes up with texdoc lualibs.

Marcel Krüger

@FaheemMitha It's also mentioned in the Introduction. The lualibs document is about Lua libraries. While they are loaded by default in LaTeX, they are not compiled into the binary and don't contain native code.

Faheem Mitha

14:06

You mean:
> We currently use Lua 5.3.*. There are few Lua libraries that we consider part of the core Lua machinery, for instance lpeg. There are additional Lua libraries that interface to the internals of TEX. We also keep the Lua 5.2 bit32 library around.

Ulrike Fischer

@DavidCarlisle I think yes @JosephWright

Skillmon

14:28

@DavidCarlisle now I want to know which package!

mickep

15:25

@Skillmon Curious like a dodo.

David Carlisle

15:37

@Skillmon tug.org/pipermail/tex-live/2022-June/048215.html

@FaheemMitha that plus:

4.3 LUA modules
Some modules that are normally external to Lua are statically linked in with LuaTEX, because
they offer useful functionality:

‣ lpeg, by Roberto Ierusalimschy, http://www.inf.puc-rio.br/~roberto/lpeg/lpeg.html. This
library is not Unicode-aware, but interprets strings on a byte-per-byte basis. This mainly
means that lpeg.S cannot be used with utf8 characters encoded in more than two bytes, and
thus lpeg.S will look for one of those two bytes when matching, not the combination of the

barbara beeton

15:50

@PauloCereda -- Joyeux anniversaire!

Paulo Cereda

16:02

@barbarabeeton Merci! <3

samcarter

16:13

A mistyped url just lead me to tug.com/shop/displaykites/Bear :)

mickep

@samcarter You went for displaykittens, didn't you? ;-)

Faheem Mitha

@Skillmon Looking at your pgfkeys answer now. As someone who is more familiar with the various key packages than I am, are any of them comparable in functionality with pgfkeys, and if so, which?

samcarter

@mickep :)

Faheem Mitha

@Skillmon pgfkeys seems to have quite extensive functionality. That's one reason I'm using it. The other is that, since it's part of PGF/TikZ, it certainly gets plenty of use, so is probably not too buggy. Also, the documentation is part of the PGF/TikZ manual, and is comparatively extensive as these things go. Often TeX package don't have much by way of documentation at all.

Ulrich Diez

17:29

@daleif With my own stuff I usually have MariaDB generate files that the LaTeX-side can load via \inputand which provide expl3-property-lists or macro-definitions. But from time to time I see people using datatool/csvsimple and creating csv-files via filecontents* or during the LaTeX-run or the like, whereby in some cases basic rules of database-management/relational algebra

are not obeyed and expansion- and catcode/tokenization-issues arise due to mixing values of quantities, e.g., measured sizes, with LaTeX-code for representing the values of these quantities within the to-be-created LaTeX-document, ....

David Carlisle

@daleif any idea about coloneqq? ?

Paulo Cereda

@DavidCarlisle colonel Q

David Carlisle

@PauloCereda in the dinning room with the hammer

@daleif @barbarabeeton github.com/wspr/unicode-math/issues/590

Paulo Cereda

17:44

@DavidCarlisle oh

Ulrich Diez

@FaheemMitha Depending on the engine in use etc one might indeed prefer tools/libraries other than datatool etc. What I am after is texts providing guidance in organizing data/databases so that issues due to the peculiarities of TeX/LaTeX are minimized. E.g., from time to time I see people using filecontents* or the like for creating csv-files where within the same data-field values of quantities, e.g., measured sizes, are mixed with LaTeX-code for representing them...

... in the document created via LaTeX. E.g., this morning I witnessed a scenario where a .csv-file produced via filecontents* was copied via macros of the verbatimcopy-package. The copy was edited manually in order to remove all LaTeX-commands for formatting so that the copy could be loaded/processed by gnuplot. (I might have chosen TikZ.) This seemed to work out for the one who did that, but I saw the need of providing guidance in how to organize data when it comes to invoking LaTeX.

daleif

@DavidCarlisle do you mind registering an issue on mathtools on it, else I'll probably forget

David Carlisle

@daleif will do, probably also at newtxmath

secret meeting first

barbara beeton

17:59

@samcarter -- Interesting, but I prefer non-inflatable kites. And have an interesting, well-worn collection. One of my favorites is a Valkyrie, by the Nantucket Kiteman. A yellow delta, with black keel, made of cotton sailcloth, it was very well behaved, and we once had it out well over 600 meters (the creel of kite string held 650 meters). We used to have a wonderful kite-flying site in the city, but Brown built it out with athletic facilities. Sigh.

David Carlisle

@UlrikeFischer lwg?

@JosephWright ^^

Ulrike Fischer

@DavidCarlisle I can't connect I always get an error. Will try with the laptop.

David Carlisle

@UlrikeFischer some people managed after two or three attempts there seems to be issues

barbara beeton

@DavidCarlisle -- I'm sure, although I no longer have access to the archive that would prove it, that the regular 2-line = is what would have been requested from Unicode. (And that's what's shown by unicode-math.) I wasn't aware of the assignment in mathtools. In my opinion, should be fixed to agree with unicode-math. What's "recognized" by MathML? And what would be a good name for :-?

Ulrike Fischer

@DavidCarlisle I'm in but don't hear anything ;-(

David Carlisle

18:10

@barbarabeeton Unicode and mathml have nothing corresponding to :-

barbara beeton

@DavidCarlisle -- But if it's in mathtools, it must have a recognized meaning. I guess this is a question for @daleif (or Morten).

David Carlisle

@barbarabeeton mathtools documents it as copied from txfonts and newtxmath documents txfonts was wrong.....

barbara beeton

@DavidCarlisle -- So just an error in the original txfonts, never corrected? I'm not at all sure how one might track down existing uses in legacy documents.

David Carlisle

@barbarabeeton thats the question really, deciding it is wrong isn't necessarily same as deciding to change it

barbara beeton

@DavidCarlisle -- Well, detexify doesn't seem to have either a one-line or two-line "colon-lines" (which I find surprising), so that's no help.

samcarter

18:27

@barbarabeeton I know the problems of vanishing kite-flying sites. As a kid, there was a large field just down the road which was perfect for kite-flying after the harvest in autumn; but then they build houses there. No more kite-flying in the vicinity :(

Faheem Mitha

@Skillmon You can add some (or most) of this to your answer for context.

barbara beeton

18:40

@samcarter -- Try a smaller kite. We've got one called "scrapflake", a hex box about 15cm high. It will lift in a very gentle breeze, and can almost be flown from a spool of nylon or poly thread. (Regular kite string is too heavy, and monofilament tangles, although I've seen one flown from a lightweight flyfishing rod.) The fabric is bright red ripstop, and the struts are delicate bamboo splints. Fun! But too fragile to fly on a beach.

Actually, this is a miniature of a meter-wide "snowvlake", which is a good beach kite.

CroCo

18:56

Hi everyone, I'm using IEEE style and it seems passing [t] [h] to "figure*" has no effect and figures appear in the middle of a page (on their own)

any suggestion?

mickep

@DavidCarlisle Oh, secret meetings... :-)

Joseph Wright

@CroCo STATUS BYDESIGN

CroCo

@JosephWright not sure what do you mean.

Joseph Wright

@CroCo It's a class for journal submission: if they ignore float placement, it's because it's a design decision

CroCo

19:12

@JosephWright this is what I get

it doesn't look good.

I've seen alot of papers in IEEE place the figures in the top of current page.

Also, there is a huge gap between the two figures* as you can see

mickep

@CroCo Are you inserting the figures in the same way? Looks very different. (I know nothing about the IEEE style...)

CroCo

@mickep \begin{figure*} \end{figure*}

inside it, I'm using subfloat{}

In the preamble, I see
\ifCLASSOPTIONcompsoc
\usepackage[caption=false, font=normalsize, labelfont=sf, textfont=sf]{subfig}
\else
\usepackage[caption=false, font=footnotesize]{subfig}
\fi

any suggestion!

Ulrike Fischer

20:49

@DavidCarlisle why does the second \@removeelement error?

\documentclass{article}

\makeatletter

\@removeelement {blub}{}\@unusedoptionlist

\@removeelement {pagecolor={0.5,0.98,0.94}}{}\@unusedoptionlist

\begin{document}
Foo
\end{document}

David Carlisle

@CroCo for any class figure* comes at the earliest on the next page. So what you show looks expected to me. You can not use [h] with a double column float, but in any case it's usually a bad idea to try to customise float positioning in a journal class, it just annoys the journal production staff and delays production.

@UlrikeFischer you can't make a delimited argument using {

Ulrike Fischer

@DavidCarlisle so "don't use braces in class options ..."?

David Carlisle

@UlrikeFischer @UlrikeFischer \expandafter\@remove@eq@value\ tries to maake it a bit safer, so it jus records the opion as pagecolor not pagecolor={0.5,0.98,0.94}

@UlrikeFischer or use classes using the new option handler

Ulrike Fischer

@DavidCarlisle trissto?

David Carlisle

@UlrikeFischer modulo a vowel or two

Ulrike Fischer

20:55

@DavidCarlisle the problem is hyperref. It has an option pagecolor and tries to remove it from the \@unusedoptionlist.

David Carlisle

@UlrikeFischer but (since last year?) the unused option list should have pagecolor not pagecolor={0.5,0.98,0.94}

Ulrike Fischer

@DavidCarlisle the list is empty, but hyperref is processing the key. It is from this question: tex.stackexchange.com/q/647885/2388

David Carlisle

> \@unusedoptionlist=macro:
->pagecolor.
l.4 \show\@unusedoptionlist
                           ]
?

\documentclass[pagecolor={0.5,0.98,0.94}]{article}

\makeatletter
\show\@unusedoptionlist]
\begin{document}


\end{document}

the \@unusedoptionlist is now sanitized for this problem so never has the value part

@UlrikeFischer it should probably use \@remove@eq@value now. I'll look.

Ulrike Fischer

@DavidCarlisle yes but the content doesn't matter. \@removeelement {pagecolor={0.5,0.98,0.94}}{}\@unusedoptionlist errors regardless of the actual content of the list.

@DavidCarlisle but does that mean that hyperref should not try to remove pagecolor={value..} but only pagecolor? Then it would be an error in kvoptions?

David Carlisle

@UlrikeFischer yes but pagecolor={0.5,0.98,0.94} will nver be an lemen so can' be removed,need \@removeelement \@remove@eq@value {pagecolor={0.5,0.98,0.94}}{}\@unusedoptionlist with some \expandafter

Ulrike Fischer

21:05

lemen?

David Carlisle

@UlrikeFischer welcome to my world, why we gave up gradually adding keyval to the existing option handler, and made a completly new one

@UlrikeFischer element (of the list)

Ulrike Fischer

@DavidCarlisle ah. Well I don't mind if it isn't removed, currently I only want that it doesn't error when trying to remove it.

David Carlisle

@UlrikeFischer is hyperref doing this by hand or is the standard opion handler failing? (I can check...)

Ulrike Fischer

@DavidCarlisle \ProcessKeyvalOptions is failing, so probably kvoptions

David Carlisle

@UlrikeFischer I'll look....

Ulrike Fischer

21:13

@DavidCarlisle this here errors too, and it uses the new handler:

\begin{filecontents}{testclasskey.sty}
\NeedsTeXFormat{LaTeX2e}
\ProvidesPackage{testclasskey}[%
    2022/06/14
    v1
    test class key]

\DeclareKeys
 {
   pagecolor  .code  = {\def\blub{#1}}
 }
\ProcessKeyOptions
\endinput
\end{filecontents}

\documentclass[pagecolor={0.5,0.98,0.94}]{article}

\usepackage{testclasskey}
\begin{document}

\blub
\end{document}

David Carlisle

21:28

@UlrikeFischer kv version:

\documentclass[pagecolor={0.5,0.98,0.94}]{article}

\makeatletter
\show\@unusedoptionlist

\usepackage{abc}
\begin{document}


\end{document}

where abc.sty is

% warning but no error
% \DeclareOption{pagecolor}{}
% \ProcessOptions

\RequirePackage{kvoptions}
\def\KVO@ProcessKeyvalOptions#1{%
  \let\@tempc\relax
  \let\KVO@temp\@empty
  \ifx\@currext\@clsextension
  \else
    \KVO@GetClassOptionsList
    \ifx\KVO@classoptionslist\relax
    \else
      \@for\KVO@CurrentOption:=\KVO@classoptionslist\do{%
        \@ifundefined{KV@#1@\expandafter\KVO@getkey
                      \KVO@CurrentOption=\@nil}{%
        }{%
          \@ifundefined{KVO@local@#1@\expandafter\KVO@getkey

Ulrike Fischer

@DavidCarlisle looks good. Now @JosephWright should do the same in the new handler.

Joseph Wright

@UlrikeFischer @DavidCarlisle can just change it :)

David Carlisle

@JosephWright \clist_remove_all:Nn \@unusedoptionlist {##1} I think?

Ulrike Fischer

@DavidCarlisle that is what breaks.

David Carlisle

@UlrikeFischer @JosephWright

\ExplSyntaxOn
\let\__keys_options_package:n \undefined

\cs_generate_variant:Nn \clist_remove_all:Nn {Ne}
\cs_new_protected:Npn \__keys_options_package:n #1
  {
    \clist_map_inline:Nn \@classoptionslist
      {
        \keys_if_exist:neT {#1} { \__keys_remove_equals:n {##1} }
          {
            \clist_put_right:Nn \l__keys_options_clist {##1}
            \clist_remove_all:Ne \@unusedoptionlist {\expandafter\@remove@eq@value##1=\@nil}
          }
      }
  }
\ExplSyntaxOff
\DeclareKeys
 {

Joseph Wright

21:42

@DavidCarlisle You going to change for PL1?

David Carlisle

testclasskey.sty is e expansion compleely safe here, think it is ....

@JosephWright probably \@remove@eq@value should be an expl3 version rather that coming back to 2e at that point? What would be the l3 key way to get the key name, junking any value?

@JosephWright oh what is \__keys_remove_equals:n ??

Joseph Wright

@DavidCarlisle What you said :)

@DavidCarlisle You remind me that Frank still needs to give the OK on the 'expand the keyname' business

Ulrike Fischer

@DavidCarlisle @JosephWright it I repair kvoptions then I get a similar error from the \ProcessKeysOptions { myclass } in Denis examples

David Carlisle

@JosephWright can that easily be refactored to only do \__keys_remove_equals:n once?

Joseph Wright

@DavidCarlisle Not sure what you mean ...

David Carlisle

21:51

\keys_if_exist:neT {#1} { \__keys_remove_equals:n {##1} }
          {
            \clist_put_right:Nn \l__keys_options_clist {##1}
            \clist_remove_all:Ne \@unusedoptionlist {\__keys_remove_equals:n {##1} }

@JosephWright ^^^ has __keys_remove_equals:n {##1} twice,

@JosephWright maybe that is OK, or could store it in a local or ...

@UlrikeFischer ?

Ulrike Fischer

@DavidCarlisle try example from tex.stackexchange.com/q/647885/2388 but remove hyperref. It will error at the class keyval handler.

David Carlisle

@UlrikeFischer OK I see it now., tracing....

@UlrikeFischer oh l3keys2e so almost biut not quite the new handler

Ulrike Fischer

@DavidCarlisle yes

David Carlisle

Apr 16 at 12:35, by David Carlisle

@UlrikeFischer not my fault

Ulrike Fischer

22:22

@DavidCarlisle ;-) I have prepared the ctan upload of kvoptions and will push it (we should rename the branch ...). I can sent it tomorrow to ctan.

David Carlisle

@UlrikeFischer error in the example pagecolor .code:n = { definecolor{myclass_pagecolor}{rgb}{#1} } } should be \definecolor

\begin{filecontents}[overwrite]{myclass.cls}
\ProvidesExplClass{myclass}{2022-06-15}{0.1}{Class of mine}
\NeedsTeXFormat{LaTeX2e}
\LoadClass { article }

\RequirePackage{l3keys2e}

\let\__keys_latexe_options_class:n\undefined
\cs_new_protected:Npn \__keys_latexe_options_class:n #1
  {
    \cs_if_free:cF { opt@ \@currname . \@currext }
      {
        \keys_if_exist:nnTF {#1} { unknown }
          {
            \clist_put_right:Nv \l__keys_latexe_options_clist
              { opt@ \@currname . \@currext }

Ulrike Fischer

@DavidCarlisle oh ;-)

David Carlisle

@JosephWright @UlrikeFischer l3keys2e version

^

Ulrike Fischer

@DavidCarlisle I wonder how many package options keyval handler there are in the wild. Perhaps @Skillmon has a list ;-)

David Carlisle

@UlrikeFischer \tracingall waded through dozens of lines of l3 code trace trying to work out where the \ was dropped then finally looked at the test example....

@UlrikeFischer oh didn't we master-main ho-tex at the time? hmmm

Ulrike Fischer

22:35

@DavidCarlisle not everything obviously, but I can do it tomorrow.

Skillmon

22:45

@UlrikeFischer not too many. I can think of KOMA's keyval based stuff, kvoptions, l3keys2e, options, pgfopts, expkv-opt, and I think that's it, the others don't work anymore (and I honestly hope I didn't forget one -- the information is only saved in my brain, which uses a lossy compression)

@FaheemMitha some of this is contained in the documentation of expkv, which includes a (non-exhaustive) list of the key=value parsers available on CTAN and a very brief comparison (among of which is a table that's really interestng -- it contains benchmark results of the different parsers as well as two stability measures, of which only four packages meet all, and of those one is unusable (ltxkeys if you wondered))

(the non-exhaustive criteria being: not based on another parser listed there, and I had to get it to work at some point in time)

@DavidCarlisle thanks, I could finally appreciate your commend :)

Skillmon

23:06

@FaheemMitha putting anything against pgfkeys feature wise is a rough ride, especially with the limited set of parsers that get things right™. From those three l3keys is likely to be the most tested and feature-comparable to pgfkeys. Compared to l3keys my own expkv lacks key filtering (I just didn't bother to implement an expandable key filter). Also expkv is made a lot more convenient by the accompanying packages expkv-def and expkv-cs (the former provides common key types [...]

[...] comparable to pgfkeys's key handlers or l3keys approach -- the latter provides means to set up expandable macros that take a key=value argument, and also is nice to define very simple key=value interfaces). But what's one of the features that makes pgfkeys so useful in many cases is the design choice to have the key-defining and the key-setting interface be actually the same. That's a choice that both l3keys and expkv don't share with it, [...]

[...] but in both cases you could define a key in your set that allows you to define additional keys. Regarding stability: I'm fairly certain that expkv is well-tested (but you have to trust a single person on this for the most part), and l3keys is used by some of the big packages (fontspec, siunitx, ...), and maintained by the kernel team (and I took some looks on it as well), so should be pretty stable. Oh and both are using unit tests, so that code changes shouldn't break too much.

Transcript for

$TeX - LaTeX$ TeX, LaTeX and Friends