TeX, LaTeX and Friends

7:59 AM

@JosephWright why does \text_expand:n wrap a * inside braces?

And why is a macro defined with \NewExpandableDocumentCommand not expanded by \text_expand:n?

9:03 AM

@JosephWright I'm pretty sure that the \group_align_safe_begin: inside of \__text_expand:n isn't needed since the contents are always inside a set of braces inside of \__kernel_exp_not:w during the processing.

David Carlisle

9:14 AM

@Skillmon but the braces in ` ... &\unexpanded{.....}` wouldn't stop \unexpanded being expanded by the \omit scan would they? (admitedly pre-expanding \unexpanded probably doesn't hurt in most cases

Skillmon

@DavidCarlisle since the macro uses \unexpanded\expandafter{\romannumeral<inner-loop>} that wouldn't be an issue. During every step of inner processing that would use undelimited arguments the outer brace group of \unexpanded is still there, and only when the complete inner processing is done the \unexpanded will be expanded and remove the brace group.

Joseph Wright

@Skillmon Test case?

Skillmon

@JosephWright \NewDocumentCommand\something{sm}{\IfBooleanTF{#1}{foo}{bar}}\text_expand:n{\something*{abc}} will expand to \something{*}{abc}.

Joseph Wright

@Skillmon That's \NewDocumentCommand

Skillmon

@JosephWright ?

Joseph Wright

9:25 AM

@Skillmon It's a non-expandable macro in a text context, for which the 'standard' set up may well be sub-optimal

@Skillmon They are

Skillmon

@JosephWright hm, something very strange is happening here, will have to investigate...

@JosephWright thing is, that in my MWE to test things everything works out, the document I encountered this in however doesn't behave correct.

Ulrike Fischer

@Skillmon in case you are trying to redefine \t, better don't ;-)

@MarcelKrüger I think we need to switch the repo in the luaotfload setup from the pretest to something normal?

Skillmon

@UlrikeFischer I don't want to redefine \t...

David Carlisle

@UlrikeFischer \renewcommand\c{Carlisle}

Ulrike Fischer

@Skillmon I added an answer here with a text command version: tex.stackexchange.com/a/594246/2388. But I actually don't have the time to get involved here.

@DavidCarlisle that would be an improvement, you would get famous ;-)

David Carlisle

9:38 AM

@UlrikeFischer especially in France

Ulrike Fischer

@DavidCarlisle ;-)

Ulrike Fischer

9:52 AM

@DavidCarlisle soulutf8?

David Carlisle

10:08 AM

@UlrikeFischer blub

Ulrike Fischer

@DavidCarlisle ;-) But what should we answer? I don't understand the structure of this repo, why it contains files from soul, or why there is no issue tracker.

David Carlisle

@UlrikeFischer it was set up as a fork of the main soul repo and forks don't hav eissues. But I can't remember if I did that or if it came that way from Heiko, but I didn't think his files were on github until we aquired them, I'll see if I can reconstruct the history. We could elete the fork and make it a normal repo

@UlrikeFischer oh you can enable issues on Forks I'll do that

Ulrike Fischer

@DavidCarlisle A normal repo is perhaps better (but it should probably remove the soul files shouldn't it?) Should I suggest to simply describe the issue for now in oberdiek?

David Carlisle

@UlrikeFischer I think I'll enable issue and move it we can always un-fork later

Ulrike Fischer

@DavidCarlisle ok.

Skillmon

10:21 AM

@JosephWright ok, the issue is that the token at hand is part of \l_text_accents_tl and hence the braces are added. All that redefining and strange *-usage in tipa which now carries over to tipauni is really problematic...

David Carlisle

@UlrikeFischer that's your fault I hadn't noticed the email at all until your ping and now I go and type l3build check and two thords of the tests fail...

! Package atveryend Info: Empty hook `BeforeClearDocument' on input line ....
! [54]
! Package atveryend Info: Empty hook `AfterLastShipout' on input line ....
! (soulutf8-test2.aux)
! Package atveryend Info: Executing hook `AtVeryEndDocument' on input line ....
! Package atveryend Info: Empty hook `AtEndAfterFileList' on input line ....
! Package atveryend Info: Empty hook `AtVeryVeryEnd' on input line ....

Captain Hook strikes again.

Marcel Krüger

@UlrikeFischer Done, now we just have to make the tests pass again.

David Carlisle

@UlrikeFischer you may need your sword

@MarcelKrüger join the queue: I want her to make my tests pass first!

Joseph Wright

@Skillmon Ah, right: log an issue?

Skillmon

10:37 AM

@JosephWright no, that's not an issue of \text_expand:n, imho, but an issue of packages like tipa that provide a starred version of accent-macros.

@JosephWright I'll locally do \tl_remove_all:Nn \l_text_accents_tl { \t } inside the scope in which I know that should change.

Joseph Wright

@Skillmon Ah, right

@DavidCarlisle, @UlrikeFischer Thoughts on github.com/josephwright/etoolbox/issues/37?

Ulrike Fischer

@MarcelKrüger I will check in the afternoon if things have settled down a bit, but if there is still a latex difference between latex and latex-dev we will to decide which tests we want to pass.

@JosephWright tell koma not to set it to relax but to \@empty or whatever?

@DavidCarlisle just checking ;-)

David Carlisle

@JosephWright the fact that \begin{document{ leaves \do let to \noexpand seems a bit suspect, it's undefined in the preamble so uses of ettolbox \do there would need newcommand as well. I don't suppose it's politically correct to say I'd use \def

@UlrikeFischer no I think koma is right and the format and etoolbox are wrong (@JosephWright)

Ulrike Fischer

@DavidCarlisle so document as \def\do? Yes, it is probably better. Actually I think there was a discussion on one of the german groups where I said just this.

David Carlisle

@JosephWright ah no it's etoolbox that leaves it let to \noexpand, I think that's probably wrong

Joseph Wright

10:48 AM

@DavidCarlisle Hmm

@UlrikeFischer Can't really do that in etoolbox: it's not LaTeX syntax. I guess \providecommand then \renewcommand

David Carlisle

@JosephWright \csdef{do}

Joseph Wright

@DavidCarlisle Hmm, true

David Carlisle

but given that etoolbox provides \undef using \def wouldn't be so bad

Ulrike Fischer

@JosephWright well the whole idea of use \do as a user command is not very LaTeX-like, I mean this is a command that is used internally in latex and packages a lot.

Joseph Wright

@UlrikeFischer Yes, true: it's really got the wrong name up-front (should be \do@ or \@do or something)

David Carlisle

10:55 AM

@JosephWright separately from how you document the definition I don't think the package should leave it globally defined especially not to \noexpand as the standard usage is \do{aaa}\do{bbb}\do{ccc} and \noexpand is not good there

Ulrike Fischer

What do you get --recorder in the fls file ?

Joseph Wright

@DavidCarlisle OK, I'll look at that

David Carlisle

@UlrikeFischer ?

Ulrike Fischer

@DavidCarlisle how often a typical file is recorded?

David Carlisle

@UlrikeFischer im confused

Ulrike Fischer

10:58 AM

@DavidCarlisle well after verb it is \noexpand too:

\documentclass{article}

\begin{document}
\verb+blub+
\show\do

\end{document}

David Carlisle

@UlrikeFischer oh I thought latex only did local settings OK @JosephWright is forgiven, it's just rubbish interface

Ulrike Fischer

@DavidCarlisle there is a discussion about slower compilation, and Christian just mentioned that files are touched more often now: github.com/MiKTeX/miktex/issues/782#issuecomment-826070524

David Carlisle

@UlrikeFischer oddly enough I don't follow the miktex list I thought you were talking about etoolbox or soulutf8 ...

Ulrike Fischer

@DavidCarlisle ;-)

Joseph Wright

@UlrikeFischer On siunitx, I'm going to add those Symbol functions back in for emulation, and I'm wondering if I need a document-level interface for \siunitx_print:nn (which I also think needs renaming)

David Carlisle

11:01 AM

@UlrikeFischer that looks odd why is book.cls pinged so often?

Ulrike Fischer

@DavidCarlisle that is the question. I was wondering if it is our fault, in tl19 it is touched only twice.

@DavidCarlisle one for the team list?

David Carlisle

@UlrikeFischer book.cls got opened twice in tl2017 (once for the \openin test and then the \input)

@UlrikeFischer yes

Ulrike Fischer

@DavidCarlisle my current guess regarding the soul issue is "load xcolor".

David Carlisle

@UlrikeFischer oh did he post the real issue, let me see

Ulrike Fischer

@DavidCarlisle well he tried hard to avoid to give sensible info.

@DavidCarlisle oh not I misread the issue. Need to recheck

@DavidCarlisle no it is this: tex.stackexchange.com/a/520315/2388

David Carlisle

11:21 AM

@UlrikeFischer hmm this issue didn't get a lot of feedback github.com/drehscheibe/soul/issues/1

Ulrike Fischer

@DavidCarlisle yes, so I guess we should fix soulutf8 at least.

David Carlisle

@UlrikeFischer since it's a fork we could make a pr to soul:-)

Ulrike Fischer

@DavidCarlisle or can we merge both? As utf8 is the default a dedicated soulutf8 doesn't make much sense does it? (not that I want to inherit soul ;-()

David Carlisle

@UlrikeFischer I can't rember what this is I seemrd to have approved the PR but is this version used anywhere? github.com/circumscribe/soul/pull/1

@UlrikeFischer ctan has maintainer inactive for soul so I guess simplest thing is to just adopt soul in that repo and bundle both together (that's probably what I had in mind when I made it a "soul" repo as a fork rather than a "soulutf8" repo.

Ulrike Fischer

@DavidCarlisle well soul still breaks without the change, so imho this never made into ctan. I don't remember if the guy actually contacted the soul author, perhaps a search in the chat can unearth something, but after lunch (no ducks involved).

David Carlisle

11:31 AM

@UlrikeFischer he made a PR at the original soul repo but no response

David Carlisle

11:56 AM

@UlrikeFischer naturally adding a few z fixes it

Skillmon

12:44 PM

@JosephWright I think there is a bug in \text_expand:n. Both \c_group_begin_token and \c_group_end_token will throw a missing number error as in these cases a { or } is left in the input stream in front of the \romannumeral-expansion.

egreg

1:01 PM

@Skillmon Yes, I can confirm

Skillmon

@egreg already opened an issue

Alan Munn

5:21 PM

@yo' It is possible to run pandoc in Overleaf?

Joseph Wright

@Skillmon Well, they should not really be there at all, but yes they should give a nicer result

David Carlisle

5:59 PM

NOT MY FAULT

% \changes{v1.1g}{2004/01/23}{Use kernel version of
% \cs{@ifnextchar} (pr/3501)}

yo'

@AlanMunn good question :) I'm not sure

Skillmon

@JosephWright there might be even more problems, e.g., \exp_not:n is handled sup-optimal currently.

Ulrike Fischer

@DavidCarlisle ?

David Carlisle

@UlrikeFischer bm answer on site

Alan Munn

@yo' Specifically, pandoc foo.docx -o foo.tex (don't worry about how well it will work).

David Carlisle

6:02 PM

@UlrikeFischer it's only been broken for 17 years.

Alan Munn

@yo' I did a quick test using \immediate\write18 but it produced nothing, although no error either.

samcarter_looks_forward_TUG'21

@AlanMunn If you need some online thingy to run pandoc, it should be available in rstudio.cloud (but often outdated versions)

Alan Munn

@samcarter_looks_forward_TUG'21 Thanks, that's helpful. As usual, I don't really need it, but some poor soul on the internet does.

David Carlisle

@AlanMunn no overleaf.com/read/jwcnqfcnbzwp

samcarter_looks_forward_TUG'21

@AlanMunn :) docx is involved - this means that it is a really poor soul

Alan Munn

6:09 PM

@samcarter_looks_forward_TUG'21 Yep. tex.stackexchange.com/q/594276/2693

@samcarter_looks_forward_TUG'21 But the pandoc solution actually works really well since the conversion really only needs to deal with italics and maybe boldface.

Stefan Kottwitz

@AlanMunn Sounds like Latex vs. LaTeX.

Alan Munn

@StefanKottwitz You mean rubber vs latexmk :)

samcarter_looks_forward_TUG'21

@AlanMunn Oh, I'm not sure if I should be impressed or pity them for doing all the citations manually!

David Carlisle

@UlrikeFischer I wonder who I could blame, it seems unfair to pick on individuals ~~2004-01-31 Frank Mittelbach <Frank.Mittelbach@latex-project.org>~~

yo'

6:40 PM

@AlanMunn sorry I'm not at a PC. You can try \input{|"which pandoc"} (I hope I have the syntax right)

David Carlisle

@yo' which is more or less exactly what is at the link above overleaf.com/read/jwcnqfcnbzwp :-)

Alan Munn

9:27 PM

@samcarter_looks_forward_TUG'21 I have friends who have a single Word document containing all the references they've ever used, and they just cut and paste into their documents at the end. Others are a bit smarter and use Endnote or some other reference manager. I tried Endnote once, found a bug and never tried it again. :)

Transcript for

$TeX - LaTeX$ TeX, LaTeX and Friends