1:16 AM
@UlrikeFischer No, the message is the same in my current version:
The hook 'env/document/before':
Hook is not declared!
The hook is empty.
@UlrikeFischer What it says is correct (albeit weird, for a generic hook). Do you think a different message would be better for a generic hook?

5 hours later…
6:33 AM
@PhelypeOleinik well I do find the message irritating, as it is the same you get for an unknown hook, say \ShowHook{blub}, but not for e.g. \ShowHook{env/quote/before}. I'm still unsure about some hook names, and I used \ShowHook {env/document/before} to check if I got the name right and was hoping for something like "hello, yes this is the at end preamble hook you are looking for" that confirms that this is an existing kernel hook ;-)

7:15 AM
@JosephWright Travis CI for latex2e is broken.
Running checks on
github-0023 (1/384)
/tmp/texlive/texmf-dist/scripts/l3build/l3build-check.lua:91: ../build/test/github-0023.log: No such file or directory
The command "l3build check -q -H --show-log-on-error" exited with 1.
I'm just building the latest master branch here, no custom additions.

@HenriMenke Strange

@JosephWright I can also reproduce locally with latest TL2020.
@JosephWright Hm, the last successful build on upstream latex3/latex2e was on March 17 and there has been quite some activity on l3build in the meantime.

@JosephWright @HenriMenke I can reproduce that here

21 hours ago, by David Carlisle
@PauloCereda It's a device that allows you to blame Joseph for any errors in the upload.
Even more relevant now.
:D

7:50 AM
@HenriMenke good plan
@HenriMenke I probably can't look seriously until later today

8:03 AM
@DavidCarlisle @JosephWright I get an error that the lualatex format was written by luatex (and not luahbtex).
@JosephWright I think commit 6149c4582 is missing on master.

8:39 AM
@UlrikeFischer Quite possibly: we don't copy anything to master unless we have to
@HenriMenke master is just for releases: probably you wanted develop

@JosephWright I don't understand. Release is what matters, right? Because in the CI l3build coming from TeX Live which only ships releases, I hope.

@HenriMenke The master branch on latex2e only gets updated when we actually do a release. In between releases, we commit to develop unless there is a patch needed. Then we'd cherry-pick stuff from develop if required. As we've not needed a patch release since March, master has not changed, and in particular as @UlrikeFischer says it doesn't have the LuaHBTeX commit

@JosephWright Ah, I see. You meant the develop branch of latex2e and not the one of l3build. Sorry, that wasn't clear to me.

@HenriMenke Yes, that'sit
@UlrikeFischer For xcolor I was thinking something like
\documentclass{article}
\usepackage{l3color}
\usepackage{xcolor}
\usepackage{etoolbox}
\makeatletter
\ExplSyntaxOn
\cs_set_eq:NN \@expl@color@set@@nnn \color_set:nnn
\ExplSyntaxOff
\protected\def\extract@xcolor#1{%
\ifcsname\string\color@#1\endcsname
\expandafter\extract@xcolor@aux
\else
\expandafter\@gobble
\fi
{#1}%
}
\protected\def\extract@xcolor@aux#1{%
\begingroup
\def\xcolor@##1##2##3##4{{#1}{##3}{##4}}%
\edef\@tempa{\csname\string\color@#1\endcsname}%
\expandafter\endgroup
which just requires I extend HSB and Hsb support to hsb too
@UlrikeFischer The latter step is trivial (it's coded just not exposed)

8:54 AM
@JosephWright it doesn't cover \colorlet yet. Why are you appending to \XC@definec@lor instead of prepending as I did? Doesn't that loose mix models definitions like as \definecolor{abc}{rgb/cmyk}{1,1,1/1,1,1,1}?

@UlrikeFischer The reasoning was to leave parsing to xcolor and take what it actually sets: my worry with grabbing the arguments is there may bit cases we don't cover or where the different in maths means the outcomes vary
@UlrikeFischer xcolor doesn't actually use both definitions, so I was thinking we should respect that (and take the one it actually does use)

@JosephWright hm, I thought it would be a nice way to setup an user interface to define l3colors without confusing users with new commands. I mean if a user wants the multimodel what should they do if \definecolor ignores it? Beside this: we are not changing the colors of xcolor, if the user calls \color{abc} they still get what ever xcolor thinks right.

@UlrikeFischer If we go that way, we need to cover all of the models xcolor provides: I've just sorted hsb, wave is also on the list I guess

9:10 AM
Which commands are allowed inside \write18? It seems pdflatex is not

@Gaussler cat /usr/local/texlive/2020/texmf-dist/web2c/texmf.cnf. There is a list of shell escape commands.

@JosephWright Imho l3colors should be set by l3colors code / math. It wouldn't be good if \color_select:n gives different results depending on if the color has been defined by \color_set or \definecolor (from color) or \definecolor (from xcolor) or ....
@JosephWright well it would be nice if all models work, but imho it would be also okay to issue a warning if a model is not supported. I mean, people can always do an additional \color_set for l3color for odd colors. We don't need to ensure that every color definition in xcolor works (e.g. I don't think that there is a way to handle series definitions for now).

@UlrikeFischer Indeed: one option is to have a 'doesn't error' version of \color_set:nnn
@UlrikeFischer I'll look at the models and extend, plus cover the : syntax (by removing it)

@JosephWright yes, that would be a good idea. with message redirection one could even decide if it warns or is silent.

@UlrikeFischer Yes, provided it's not a kernel error ... probably I can just change that and we can use redirection (no new code needed)

9:22 AM
@JosephWright I also think that we should not try to setup something for "global" colors. That looks rather odd in xcolor.

9:32 AM
> Richard Stallman Discusses Privacy Risks of Bitcoin, Suggests 'Something Much Better'
@DavidCarlisle ^^ emacs coins

@PauloCereda That's probably going to be just plain cash. Everything else is not privacy-friendly.

@HenriMenke indeed. :)

Why does
tlmgr install sinhala
not work?

I have a box made with tcolorbox, how can I place it wherever I want in a beamer slide? Is there a simple TikZ method?

@fomin Is this a tcolorbox with special features? Because beamer contains boxes as well.

9:42 AM
I want to put in the top left corner sometimes

@fomin Maybe you are searching for sth. like that:
4

My solution leads to: Since I'm not very expert with expansions I tried to re-create the block by means of the rectangle split and put it in front of the content by means of a new background. To do so, two libraries are needed: backgrounds and shapes.multipart. Here is the code: \documentcla...

thanks!

@JosephWright I just emailed Frank the exact same patch.

@cis because there is no package for texlive, you will have to install the fonts manually.

@HenriMenke :)
@UlrikeFischer I'll have to look at it

9:52 AM

@Gaussler anything if you use --shell-escape of course

^^ @JosephWright get this keyboard for your new laptop

@Gaussler By default, pdflatex (or any tex engine really) is disallowed because you could otherwise invoke it with --shell-escape and then execute arbitrary commands.

Aha, that (the manual font installation) was kind of a complicated process. I did that or managed to do that once.
Strange, there is a sty file in the download folder.

@cis it is not so difficult, you only have to unpack the tar.gz and move the files into the right place. Or if you only want to test, move your test file in the folder.

10:07 AM
@cis font installation at the system level is usually just a matter of dragging the file into your fonts folder these days or perhaps running fc-cache

@UlrikeFischer @DavidCarlisle Oh, oh, oh.... I just download the sihala-dowload-folder and run a teX-test-file inside, what works...

@cis are you sure you want that it would be much much easier (and better output) to use an opentype font and xetex or luahbtex. I wouldn't consider using a legacy 8bit font for such a language
18 hours ago, by David Carlisle
@cis I don't think so but with xelatex and lualatex you can use a system font or free ones like noto see examples at https://tex.stackexchange.com/search?q=sinhala I think the only indic font actually in texlive is shobhika for Devanāgarī but I may have missed something

10:29 AM
@DavidCarlisle There is a user, he said that he does not want to use Xe-/LuaLaTeX. So I want to test with the sinhala-package.
@DavidCarlisle

a) With sinhala.sty it is a syntax like  {\SHa\char213a}

b) With a sinhala ttf-font it is a Unicode-syntax like: \mysinhalafont\symbol{3461}

I think, this is his problem: He would have too much to change using b).
But that's speculative

@cis But with the ttf font you could also actually use the Unicode character instead of using symbol.

This is the MWE of the user. He would have much to change, if he would use a sinhala-ttf-font, I think....
\documentclass[12pt]{article}
\usepackage{sinhala}
\usepackage{mathtools}
% \usepackage{times}

\begin{document}

Normaler Text (Times).

Singhalesischer Text:

{\SHa\char213a}{\SHa\-\char5\char253a\char0} \textbf{{\SHa\char189a}{\SHb\-\char69}{\SHb\-\char69a}}.

\end{document}
@TeXnician @TeXnician

@cis users are not always right.
3

10:44 AM
@DavidCarlisle Because of me. But he wrote a complete book in the internal syntax of the sinhala package. So it would be too much work for him to change everything.

@cis it's a legacy non-unicode font for pdflatex so you can not use any natural input form and you will get non of the font shaping that harfbuzz wld do with a real sinhal font.
@cis {\SHa\char213a}{\SHa\-\char5\char253a\char0} \textbf{{\SHa\char189a}{\SHb\-\char69}{\SHb\-\char69a}}. is a painful way to write the text, compared to හෙලෝ වර්ල්ඩ්

@DavidCarlisle that is from a question on texwelt. And the font works more or less fine, what doesn't work is the \textbf, as the new latex bold series code interferes.
@DavidCarlisle the user wrote he used a preprocessor for this.

@UlrikeFischer I suspected that your conversation with Frank is related:-)

@DavidCarlisle ;-)

@DavidCarlisle ooh Deustchetalk

10:48 AM
@PauloCereda Enten sollten gegessen und nicht gehört werden

@DavidCarlisle oh no

@DavidCarlisle I suspect that the only way to get [rm context] some text \somescriptfont some text to behave is to ensure that the \somescriptfont supports the same series and shapes as the latin font.

@UlrikeFischer or persuade Frank to stretch the code a bit?

Oh, it can be advantageous to use an abstract input format. I write Japanese worksheets and I also use the syntax \symbol{<unicode>} for single signs or example words.
If I had to write an entire novel, I would probably want to type everything in directly.

@cis agreed for single symbols, but also (as far as I understand) Japanese doesn't lose a lot if you enter the characters via \char (\symbol) but for these languages where whole runs of letters get replaced in complicated ways, you lose almost all the font features

10:53 AM
@DavidCarlisle I'm not sure. This bx/b stuff is already complicated enough. The alternative is the "if in bold context code", but imho declaring a fd-entry is cleaner.

@UlrikeFischer yes probably right

@DavidCarlisle

What I like best is the input through roman transliteration, e.g.

In: Makudonarudo
Out: マクドナルド

(also works for other [Asian] languages)

But for this we need the parser quak.sty from @Skillmonlikestopanswers.xyz
And we don't know when it will come.

@cis you can do that in xetex via a texkit mapping (there was a separate pre-processor for Devanagari It seems that the package you are using is intended to be used with a transliteration scheme (rather than explicit \char ?
@cis see the first column of tex.stackexchange.com/questions/2576/… which is using transliteration for devangari obviously you'd need a different mapping file but same basic principle.

11:15 AM
I prefer to wait for the parser quak.sty from @Skillmonlikestopanswers.xyz
Here I have a direct line to the author.

As for the user and sinhala: I have no idea. The sinhala package is too poorly documented for me.

11:32 AM
@DavidCarlisle, @UlrikeFischer thoughts on the answer to tex.stackexchange.com/questions/555325/… welcome

@JosephWright well we didn't change the filecontents package and made it inactive, that was a decision of the author. But I wonder why this lead to an allocation error.

@UlrikeFischer I'm not sure why there is an allocation error either
@UlrikeFischer Ah, it does some low-level stuff
  \chardef\reserved@c15 %
\ch@ck7\reserved@c\write%
\immediate\openout\reserved@c#1\relax%
2

@JosephWright but isn't that only used with an older kernel? Then overwrite shouldn't work?

@UlrikeFischer Indeed, only used on older kernel
@UlrikeFischer Hm, something not quite right, yes

12:10 PM
@JosephWright yes, but I don't see what should break here, the only way I could reproduce the error was by loading etex.

@UlrikeFischer Ah, right, that does make sense
@UlrikeFischer The question is not totally clear, so you could well be right: I'm not really seeing why filecontents should cause the issue

@JosephWright well if they never realized that files are not overwritten then it could change a lot. But what should anyone do here, if they don't look in the log?

@cis don't bet too much on me :) I'm currently rather occupied and the bit of TeX time I have I'm currently spending on opening PRs for l3kernel that will be rejected anyways :P

Is there a maximum for the name length of a TeX macro?

@Skillmonlikestopanswers.xyz I hope we can useful discussion and get them accepted

12:27 PM
@JosephWright well, the l3tl PR is pretty conservative in its changes (and sometimes the effects are not really that big, e.g., simply switching out \if_catcode:w with \if:w is around 0.5-1% faster according to my benchmarks, if the test is more like \if:w \if_false:w { \fi: \something_weird:w #1 } you can imagine the effect being rather small -- \tl_if_empty:nTF is really measurable faster now [...]
@JosephWright [...], and \tl_if_head_is_N_type:nTF is around 15% faster for empty arguments or arguments with a leading space, and roughly the same for arguments starting with a brace or an N-type)

@UlrikeFischer Hm, yes, that's a good point. Perhaps env/document hooks should be declared in the code, and other generic hooks could have a better message...

@JosephWright and the check for the end-marker in \__tl_act: is more than 15% faster and used on each iteration with an N-type, but the check is not really the performance critical factor.

@PhelypeOleinik what are "generic hooks" here?

@JosephWright and the second l3keys PR might get rejected in its current form, which I'd understand (the \cs_set_nopar:Npx usage is not style-conformant, Frank is right about that, still the currying part...)
Note to myself: Write that email to the team...

නමුත් විග්‍රහකය හොඳයි, මන්ද එය වෙනත් රොමැන්ටික නොවන භාෂාවලට ද ව්‍යාප්ත කළ හැකි බැවිනි. වඩා වැදගත් විය හැක්කේ කුමක්ද?

namut vigrahakaya hon̆dayi, manda eya venat romænṭika novana bhāṣāvalaṭa da vyāpta kaḷa hæki bævini. vaḍā vædagat viya hækkē kumakda?

12:34 PM
@UlrikeFischer Any you don't need to do \HookNew before adding code to them. The current list is hooks that start with env/, file/, package/, class/, and include/

@cis the problem is not what is more important, but what is less stressful :) The architecture for the parser is still not thought through (totally my fault for not spending thoughts on it....) and it doesn't make sense to start much coding before that is the case (I currently have about 40 lines of code, which are so generic that they can't be considered to further quack at all (those lines are about implementing a stack to save the parser's state -- and are poorly written))

@Skillmonlikestopanswers.xyz Would be good: not something that has been discussed

Yes, But you forget: If the parser can be extended to any non-Roman languages, you will become world famous.

I have the following idea:
Go from a simple table with

trasliteration | Target sign

or similar and this is (once!) converted to an auxiliary file that quak.sty needs.

Much like the schematic done here: https://tex.stackexchange.com/a/557508/46023

12:52 PM
@cis non-trivial, what about strings which are also substrings of another symbol, e.g., in Japanese, how to input n-a instead of na? The current proof-of-concept dealt with all those "by hand" (by setting up something like an "end-of-symbol" token, iirc. that was ;). And it can not be extended to any non-Roman language, only to those which can be unambiguously transcribed.

1:09 PM
@Skillmonlikestopanswers.xyz In case of doubt I would prescribe {n}a versus {na}.

@cis oh, and in case you didn't realize yet, you don't become world famous for writing a LaTeX package.

@Skillmonlikestopanswers.xyz Well, then it will be the first time here. :()

1:28 PM
@Skillmonlikestopanswers.xyz @DavidCarlisle isn't world famous? Shocker. But he wrote the indentfirst package, which has 5 stars on CTAN with the comment Amazing package! Extremely useful! If that's not world famous, I don't know what it.

1:48 PM
@AlanMunn DPC rocks

@AlanMunn I wish @DavidCarlisle would write shorter code. He could've wrote a package with as many lines as stars on CTAN, but he screwed up without reason and needed a full line more!

2:06 PM
@AlanMunn Yes, the guys at the hairdresser's talk about it too.

@cis What's a hairdresser? It's been so long, I've forgotten.

@Skillmonlikestopanswers.xyz try texdoc indentfirst

@PauloCereda but the actual package does also \ProvidesFile etc.

@Skillmonlikestopanswers.xyz It must be the most well documented package in existence. 40 comment lines to 6 lines of code. 6.66 ratio... :)

2:17 PM
@AlanMunn :-)

@AlanMunn ooh

2:46 PM
@Skillmonlikestopanswers.xyz I prefer to class that as documentation, packages really shouldn't need more than 4 tokens of code.

3:25 PM
@AlanMunn
I cut my hair myself.

With the Flowbee!

@egreg do you know of any better answer to the rescale equation question: other than "no, that is not how we do things". The OP just commented the same on your answer ans they did on mine.

@daleif I have never used it but Heiko's resizegather probably is what the OP wants

3:41 PM
@daleif Not hard at all. But horrible.

@DavidCarlisle good to know, never heard of it. It is the same thing as all those users we see applying \resizebox to tables. I'd rather no tell them how to do it.

4:08 PM

Personally, I would have no problem if the syllables generally have to be separated by dots (or something similar).
Or that can adjust it if necessary.

As far as I know, this is what the (phonetic) Kunrei transliteration does.
In contrast to the Hepburn transliteration, which doesn't do that.

Look here:

4:43 PM
@egreg @daleif anyway, the tick justly awarded

I know e.g. not how a parser should break down this loan word correctly:

Makudonarudo (マ ク ド ナ ル ド)
----> Ma.ku.do.na.ru.do (not Ma.ku.do.n.a.ru.do)

Google can do it because it uses databases / dictionaries.

As I said, I would have no problem if the entry was mandatory
Ma.ku.do.na.ru.do or similar.

That should probably be much easier to program.

Either way, it will be necessary to make something possible:
Makudo{na}rudo

5:20 PM
@UlrikeFischer I've added the wave model to l3color, but I realise that it uses a different range for wavelength to the one I have: will have to update and adjust
@UlrikeFischer I'm pondering tHsb and the idea of custom ranges for the various models

5:49 PM
@JosephWright if you think it is not worth it, I don't see a problem to ignore some of the arcane model.

@cis with a greedy approach, ma.ku.do.n.a.ru.do wouldn't be the string the parser would choose, it'd always favour na over n.a, and if you want n.a you have to tell that explicitly (e.g., using {n}a or n{a} or {n}{a} or ...)

6:05 PM
Hi, is there a way to zip together l3seqs? I mean, if I have \l_a_seq, \l_b_seq, and \l_c_seq with the same number of items, can I obtain a new sequence with content that is {A1}{B1}{C1}, {A2}{B2}{C2}, ... so that I can apply to them a 3-argument function?

@Rmano We don't have that for 3 seq variables, only for two: \seq_mapthread_function:NNN <seq1> <seq2> <2-arg-function>. If you have an use case, consider filing a feature request

map/reduce thingy, people quack

6:20 PM
@UlrikeFischer I think we want to support as much of xcolor as we can, as the idea would be that most users will simply pick up l3color once we merge it

@PhelypeOleinik Ah, ok, it's in the l3candidates, I see. No especially use case, just learning --- I am trying to answer tex.stackexchange.com/q/557952/38080 as a learning exercise ;-). Should be possible to do what I need in two steps, sorting out the {} ;-)

@Rmano We almost certainly need to move that one to stable (@PhelypeOleinik)

@JosephWright 46 functions to remove until Christmas :-) (shall I make a PR for that one?)

@JosephWright would you say that using \__kernel_exp_not:w instead of \exp_not:n and \exp_not:o is fine?

@Skillmonlikestopanswers.xyz Yes, that's there for a reason :)
@Skillmonlikestopanswers.xyz If you want the primitive behaviour, it's fine

6:27 PM
@JosephWright well, it is non-uniformly in the current kernel. \exp_not:n is directly let to \unexpanded (which it really shouldn't, because of the "you can leave braces around single tokens out, if you want to" rule), while \exp_not:o grabs the second argument for that reason instead of being \cs_new:Npn \exp_not:o { \__kernel_exp_not:w \exp_after:wN }.

@Rmano As a side note, \l_mbc_listx should be \l_mbc_listx_seq (and the same for listp) and \__N should really not have that name (you seem to use it like an int, so maybe \l_mbc_cols_int and use \int_set:Nn \l_mbc_cols_int { \seq_count:N \l_mbc_listx } instead of \edef\__N...).

@JosephWright so you'd say changing \cs_new:Npn \tl_put_right:Nn #1#2 { \exp_not:o #1 \exp_not:n {#2} } to \cs_new:Npn \tl_put_right:Nn #1#2 { \__kernel_exp_not:w \exp_after:wN { #1 #2 } } is fine? (sorry, I just want to be sure)

@Skillmonlikestopanswers.xyz The 'you can leave out the braces' rule isn't universal: documented cases are things like \int_compare:nNnTF and \use:n

@JosephWright oh, ok, didn't know that. (then again, why is \exp_not:o defined as the slower version?)

@Skillmonlikestopanswers.xyz I mean that you are not allowed to do \exp_not:n \exp_after:wN and rely on that implementation detail, which is what \__kernel_exp_not:w is about

6:30 PM
@JosephWright yes, I'd then use the kernel variant.

@Skillmonlikestopanswers.xyz Goes back a long way: likely one we can't change now (some of this stuff goes back well before I was on the team, and there is a limit to the clean-up at this stage)

@PhelypeOleinik yes, you're right, got too lazy. Will fix the notation. Thanks!

@JosephWright don't take me wrong, I don't advocate for changing \exp_not:o, a lot of code would've to be rewritten.

@Skillmonlikestopanswers.xyz We've had several clean-up rounds for 'correctness', and \__kernel_exp_not:w was introduced so we 'stick to our own rules' w.r.t. \exp_not:n \exp_after:wN, not for performance
@Skillmonlikestopanswers.xyz Exactly: I've experimented in the past with a 'purist' format, using the things we've learnt, but at this stage that's a fun distraction not about delivering a product (cf. the fact I've just stripped out all of the initex guards)

@PhelypeOleinik changed, although I have to admit that the (lazy) \__N was handy, especially if you need to use it a lot, instead of \int_use:N \l_mbc_N_int... it's just a matter of getting used to the notation.

6:43 PM
@Rmano I know what you mean. Quite often I was tempted to use (and do use in local test code) something like \__tmp: (or even shorter with l3docstrip: \@@). The problem is that if everyone starts using that we end up with the LaTeX2e problem of name clashes all over again, but this time with _ instead of @. Eventually your eyes get used to the longer names and ignore the prefixes and suffixes :-)

@PhelypeOleinik one thing I seem to do constantly while debugging a command is \tl_set:Nn\l_tmpa_tl{#1}\tl_show:N\l_tmpa_tl. Do we have shortcut for this, e.g. a \show_arg:n{#1}?

@UlrikeFischer \tl_show:n (or the \showtokens primitive, if you are \tracingall)

@PhelypeOleinik oh god, I knew I was missing something obvious ;-)

@UlrikeFischer :-)

@PhelypeOleinik but it would be handy, if it would show the argument number.

6:55 PM
@UlrikeFischer At the point that \tl_show:n does its thing the macro is already expanded, so TeX doesn't know about the arguments anymore. I usually do \tl_show:n {1: #1} or \tl_show:n {#1|#2|...}, depending the case

@PhelypeOleinik oh, that's a nice trick.

@JosephWright you realize how much time benchmarking consumes? (github.com/latex3/latex3/pull/789#pullrequestreview-464183344)

@UlrikeFischer In case you need it, when debugging expandable code I usually use \cs_new:Npn \eshow:n #1 { \exp_args:No \use_none:n { \SHOW <#1> } } (similar to what \msg_expandable_error:nn does) to show stuff expandably

7:21 PM
@Skillmonlikestopanswers.xyz I just want to make sure we have some solid record of the speedup
@Skillmonlikestopanswers.xyz It's in tl so there's a big impact

7:32 PM
@JosephWright I'm currently doing it... Still, do you know how much time this consumes?

@Skillmonlikestopanswers.xyz 'Quite a bit'?
@Skillmonlikestopanswers.xyz I do get the point, i'm just trying to head off questions down-the-line

@JosephWright it's fine (but don't ask my wife whether it's fine)

8:03 PM
Hello all!

@JosephWright I'm done.

I am trying to draw an arrow between the numerator of a fraction to the numerator of another fraction (the same with the denominator)
This is for making the terms of a recursive relation (or recurrence relation) using tikz or pgfplots. Do you know a link that answer this? Thank you!!

@AlanMunn luckily, docstrip strips (almost) all unnecessary comments to get a fast loadable file!

@manooooh tikzmark is your friend

@manooooh using remember picture and overlay might help. Also tikzmark as @DavidCarlisle says.

8:07 PM
@manooooh answer by a gold tikz badge holder so it must be right
3

You can mark points with tikzmark then draw any kind of tikz thing using those point, here a couple of arrows. \documentclass{article} \usepackage{amsmath,tikz} \usetikzlibrary{tikzmark,calc} \begin{document} \begin{gather} \tikzmark{a}12345 = x^3+x^2+x\\ a=b \quad \text{and} \quad b=a\ti...

@DavidCarlisle @PauloCereda thanks for the suggestions! I remember that I saw a question asking for drawing arrow between two numbers, but I can't find it. For example: \frac{2}{3},\;\frac{3}{4},\;\frac{4}{5},\dots draw an arrow from 2 to 3 (an add +1 above it), the same for 3 to 4 etc. Where do we find this?

@manooooh in the documentation. :)

@PauloCereda don't be silly

@DavidCarlisle quack :)

@PauloCereda dinner

8:08 PM
@DavidCarlisle oh no

@PauloCereda Ohh I see. Thanks Paulo

@manooooh TikZmark is well documented and full of examples. IIRC Andrew wrote exactly to help him write math notes.

Yes, he made a great work