TeX, LaTeX and Friends

Kim Jong Un

5:54 AM

For those who maintain the tikz package about super heroes, don't forget to add the new one. Here it is.

Paulo Cereda

8:35 AM

@DavidCarlisle ooh

@DavidCarlisle quack

@Skillmon counter... ooh? I am confused

David Carlisle

8:58 AM

@PauloCereda 早餐桌

Paulo Cereda

9:18 AM

@DavidCarlisle oh no

@DavidCarlisle Joseph will learn Chinese in no time. :)

David Carlisle

11:01 AM

shooting-range

3

1

Q: Aligning ducks and their labels in a tree

I'm building a tree of labelled tikzducks. An MWE would be \documentclass{scrartcl} \usepackage{tikz} \usetikzlibrary{ducks} \begin{document} \begin{tikzpicture} [level distance=4cm, sibling distance=4cm] \node { \begin{tikzpicture} \duck \node [xshift=8, y...

Paulo Cereda

@DavidCarlisle OH NO

Ulrike Fischer

@DavidCarlisle the alias idea sounds quite good.

David Carlisle

11:20 AM

@UlrikeFischer just needs Will to make it work:-)

@UlrikeFischer originally tetex came with a standard alias file so that longtabl.sty (which emtex users tended to put in documents) got aliased to longtable.sty (so it worked)

Ulrike Fischer

@DavidCarlisle and find something to build the alias file.

David Carlisle

@UlrikeFischer I was thinking of finding @Marcel

Paulo Cereda

@DavidCarlisle ooh

Ulrike Fischer

@DavidCarlisle yes I got it that you had some advantage here. Who else could know about this alias file?

Paulo Cereda

@UlrikeFischer ooh a secret

David Carlisle

11:25 AM

@UlrikeFischer everyone who read the kpathsearch documetation?

Paulo Cereda

@DavidCarlisle ooh documentation

Skillmon

12:30 PM

@PauloCereda :) Hi, Mr. Duck!

Paulo Cereda

@Skillmon Hi. mr. rabbit! Miss you! <3

Skillmon

@PauloCereda Miss you, too! And thanks for the carrot. Here have one, too: <3

@PauloCereda how's it going at Overleaf?

Paulo Cereda

@Skillmon Yay! <3

@Skillmon quacking and making a lot of noise. :)

@Skillmon It's been great! <3

Skillmon

@PauloCereda Very good to hear!

Paulo Cereda

@Skillmon How are you, mr. rabbit?

Skillmon

12:34 PM

@PauloCereda It's weekend, so fine. My workweeks are a bit stressful lately, as I have to supervise a lot of student experiments each week.

Paulo Cereda

@Skillmon ooh things explode in the experiment?

Skillmon

@PauloCereda if they do I did my job poorly, no. It's electrical engineering and measurement experiments.

Paulo Cereda

@Skillmon ooh

Jonas Stein

12:50 PM

Hi, I try to debug why Inkscape does not render LaTeX formulas here. In the temp files I saw that the pdf which is rendered by the extension contains only Type3 fonts

when inkscape imports this pdf with the formula, it shows pixeled pictures

dpaste.com/4QQHQLZUU

this is the tex file which is generated by the script to render the formula

manually rendering the tex file has the same result as the script has

I get a .pdf with Type3

improved paste: dpaste.com/3P634ZY7H

David Carlisle

@JonasStein you can paste code fragments here, no need for external links. also don't use minimal class for such tests it does not set up fonts and s can cause issues that do not occur in a real class. What happens if you save the svg as pdf from inkscape and include the pdf in latex do you still get type3 fonts?

Jonas Stein

%% processed with pdflatex.py
\documentclass{minimal}
\usepackage{amsmath}
\usepackage{amssymb}
\usepackage{amsfonts}

\begin{document}
x=$\frac{1}{E}\left(\frac{1}{T}- \frac{1}{T_{\text{MF}}} \right)$
\end{document}

@DavidCarlisle Perhaps minimal is the reason

which class should I use instead? The Inkscape Plugin uses this snipped to render and import formulas

this is how formulas look now in inkscape. However it worked months ago.

Ulrike Fischer

@JonasStein article or so, but your files compiles fine, so the class is not the reason for the wrong fonts.

David Carlisle

@JonasStein try with article. Note I wrote minmal class when we were designing the latex2e class/package system to answer the question "what is the minimal set of commands that you need to add to a file to be a latex class without generating errors" It wasn't designed to make a class that did anything useful.

2

Jonas Stein

Oh I see.

I think the Plugin uses minimal, because we want a PDF which just renders the formula without a DIN A4 Paper and so on

this was the Bug I reported upstream, but I need more investigation before reopening the ticket gitlab.com/inkscape/inkscape/-/issues/2278

Joseph Wright

1:04 PM

@MarcelKrüger (@DavidCarlisle, @UlrikeFischer) I think I can switch all LuaTeX testing in 2e to use PDF mode, not just the TU tests. I'm wondering if we should (finally) bite the bullet and switch from etex to pdftex for the main tests

David Carlisle

@JonasStein minimal sets sizes for US Letter, why is that better than A4 ?

Jonas Stein

oh... I thought it would make a page with the dimension of the plain formula

is this available too?

Ulrike Fischer

@JonasStein you can use standalone.

Marcel Krüger

@DavidCarlisle Found me. What do you need alias files for?

David Carlisle

@JonasStein standalone but that is a rather massive dependency.

Jonas Stein

1:06 PM

ok I will backup the extension and test different classes now in the python script

will be back in a few minutes and report

David Carlisle

@MarcelKrüger long acrimonious discussion on texhax list. on the need to configure fontconfig to look in the texkive tree if you want to access fots by internal font name in xetex. I suggested an alternative that doesn't require changing teh syste fontconfig settings would be to put something like this

Alegreya-Regular.otf Alegreya
Asana-Math.otf Asana-Math
Asana-Math.otf AsanaMath
FreeSerif.ttf FreeSerif
NotoSerif-Regular.ttf NotoSerif

as aliases at the top level of texmf-dist so that kpathsearch would find the otf files given some approiximation to the internal name. This could possibly be constructed from the luaotfload cache

@MarcelKrüger see tug.org/pipermail/texhax/2021-May/025082.html and tug.org/pipermail/texhax/2021-May/025092.html

Marcel Krüger

@DavidCarlisle Interesting, I'll write a draft.

David Carlisle

@MarcelKrüger doesn't work currently with fontspec though as (I think) \setmainfont{Alegreya} sees there is no extension so never tries a filename lookup,

@MarcelKrüger thanks you don't have to do things just because I randomly ping you in chat:-)

Jonas Stein

For some reason I get a larger bounding box when using article, but the pixel result is the same

Skillmon

@PauloCereda I totally forgot to counter-ooh.

Paulo Cereda

1:20 PM

@Skillmon <3

David Carlisle

@JonasStein I don't quiote follow how you are getting pdf from that, if you get tyoe3 fonts forom that with pdflatex you don't have the type1 cm fonts or the cm map files installed, bith of which should be in any tex distribution, also the expression x=$\frac{1}{E}\left(\frac{1}{T}- \frac{1}{T_{\text{MF}}} \right)$ is very odd why is x= not in math mode? If it was all in math and you wanted a pdf clipped to that you coul do...

Marcel Krüger

@DavidCarlisle Maybe even bigger issue (since it seems basiclly unfixable): The font won't be found with NFSS's \UnicodeFontName for non fontspec users.

Paulo Cereda

ooh @MarcelKrüger was found

David Carlisle

@MarcelKrüger yes it wasn't a fully worked solution, I was just "musing out loud" :-)

Jonas Stein

Inkscape ships plugins in https://gitlab.com/inkscape/extensions
here we use https://gitlab.com/inkscape/extensions/-/blob/master/pdflatex.py from the menu "extensions"

x not in math mode is just a typo. It does not influence the debugging

David Carlisle

1:29 PM

Jonas Stein

perhaps my local latex installation is broken and the type 1 fonts are missing....

David Carlisle

@JonasStein ^^

\documentclass{article}
\usepackage{amsmath}
\begin{document}

\sbox0{$x=\frac{1}{E}\left(\frac{1}{T}- \frac{1}{T_{\text{MF}}} \right)$}
\pdfpagewidth=\wd0
\pdfpageheight\dimexpr\ht0+\dp0

\hoffset=-1in
\voffset=-1in
\topskip=0pt
\shipout\hbox{\box0}

\end{document}

Jonas Stein

\pdfpagewidth=\wd0 sound like a good solution for the plugin

David Carlisle

@JonasStein note that you can only do that for inline math display constructs like align are always \textwidth wide

Paulo Cereda

@JonasStein do you have a vanilla or a system TL?

Jonas Stein

1:34 PM

I have TL2021 from the gentoo distribution plus TL2021 with tlmanager locally insalled in userspace

The systemwide TeXLIve is pulled in as dependency since some people in Gentoo thought the user should not be allowed to use the "packages.provided" feature anymore. This enabled to install TeXLive manually and tell the packagemanager that TeX is already there

Paulo Cereda

@JonasStein Hm interesting... out of curiosity, is the issue reproducible if you run inkscape from a user shell?

David Carlisle

@JonasStein ignoring inkscape to start with, if you run pdflatex on the above what fonts do you get in the pdf, it should be something like

$ pdffonts aa237.pdf
name                                 type              encoding         emb sub uni object ID
------------------------------------ ----------------- ---------------- --- --- --- ---------
MHOLNP+CMMI10                        Type 1            Builtin          yes yes no       4  0
JHYTSG+CMR10                         Type 1            Builtin          yes yes no       5  0
EJASTL+CMR7                          Type 1            Builtin          yes yes no       6  0
MAFBCJ+CMMI7                         Type 1            Builtin          yes yes no       7  0

Jonas Stein

@PauloCereda yes:
inkscape --export-type="svg" input.pdf

pdffonts input.pdf
name type encoding emb sub uni object ID
------------------------------------ ----------------- ---------------- --- --- --- ---------
[none] Type 3 Custom yes no no 4 0
[none] Type 3 Custom yes no no 5 0
[none] Type 3 Custom yes no no 6 0

Paulo Cereda

@JonasStein hm thanks

Jonas Stein

@DavidCarlisle your example returns a PDF with
pdffonts test.pdf
name type encoding emb sub uni object ID
------------------------------------ ----------------- ---------------- --- --- --- ---------
[none] Type 3 Custom yes no no 4 0
[none] Type 3 Custom yes no no 5 0
[none] Type 3 Custom yes no no 6 0

so I should fix my TeX first

Paulo Cereda

1:41 PM

@JonasStein any different output if you include --pdf-poppler in the inkscape call?

Jonas Stein

how can I check, it the fonts are built systemwide?

Ulrike Fischer

@DavidCarlisle I tried to test this aliases, but it seems not to work for me. xetex doesn't find neither \font\asmb="[Alegreya]" nor \font\asmb=["Alegreya"], and kpsewhich Alegreya doesn't output anything either (should it?). I tried by adding a file "aliases" beside the ls-R.

Jonas Stein

perhaps our package is broken and does not build the fonts properly

David Carlisle

@UlrikeFischer worked for me on the one file I tried it on:-)

Ulrike Fischer

@DavidCarlisle what happens if you do kpsewhich on an alias name?

David Carlisle

1:48 PM

@UlrikeFischer hmm not a lot. It would need a format hint but I get:

$kpsewhich --format 'opentype fonts'  Alegreya
$kpsewhich --format 'opentype fonts'  Alegreya-Regular.otf
/usr/local/texlive/2021/texmf-dist/fonts/opentype/huerta/alegreya/Alegreya-Regular.otf

@UlrikeFischer so nothing from the alias

Marcel Krüger

@DavidCarlisle The aliases line has to include the extension. With Alegreya-Regular.otf Alegreya.otf the lookup works.

Ulrike Fischer

@DavidCarlisle this here works:

C:\Users\Nililand2020\Documents\Tests>kpsewhich blublongtable.sty
c:/texlive/2021/texmf-dist/tex/latex/tools/longtable.sty

David Carlisle

@UlrikeFischer to make it easier to spot the alias in the log I changed the tex to

\font\asm=["Azzz"]\show\asm \asm a

%\font\asmb=["Alegreya"]\show\asmb \asmb b


\bye

with a line Asana-Math.otf Azzz in aliases then you ca watch kpathsea find it:

xetex finding the file:

$ xetex --kpathsea-debug=-1 amath.tex 2>a.log
This is XeTeX, Version 3.141592653-2.6-0.999993 (TeX Live 2021) (preloaded format=xetex)
 restricted \write18 enabled.
entering extended mode
(./amath.tex
> \asm=select font "[Azzz]".
l.2 \font\asm=["Azzz"]\show\asm
                                \asm a
? x

and a.log has...

kdebug:kpse_find_file: searching for Azzz of type opentype fonts (from texmf.cnf)
kdebug:hash_lookup(try_std_extension_first.xetex) => (nil)
kdebug:hash_lookup(try_std_extension_first) => t
kdebug:variable: try_std_extension_first = t
kdebug:start generic search(files=[Azzz.otf Azzz.OTF], must_exist=0, find_all=0, path=.:/home/davidc/.texlive2021/texmf-config/fonts/opentype//:/home/davidc/.texlive2021/texmf-var/fonts/opentype//:/home/davidc/texmf/fonts/opentype//:!!/usr/local/texlive/texmf-local/fonts/opentype//:!!/usr/local/texlive/2021/texmf-config/fonts/opentype//:!!/usr/local/texlive/20…

@UlrikeFischer skip to the last line it finds it in the end:-)

kdebug:returning from generic search([Azzz.ttf Azzz.ttc Azzz.TTF Azzz.TTC Azzz.dfont Azzz]) => /usr/local/texlive/2021/texmf-dist/fonts/opentype/public/Asana-Math/Asana-Math.otf

Ulrike Fischer

2:27 PM

@DavidCarlisle \font\asmb="[Azzz]" works now for me. Spot the difference:

Alegreya-Regular.otf azzz # works
Alegreya-Regular.otf Azzz # fails

David Carlisle

@UlrikeFischer oh sorry you mean it has to be lowercase in the alias file?

Ulrike Fischer

@DavidCarlisle no the alias file requires lowercase.

David Carlisle

if I make the alias lowercase the uppercase does not work in xetex `! Font \asm=[Azzz] not loadable: Metric (TFM) file or installed font not found.
`

Ulrike Fischer

@DavidCarlisle for me both font definitions work, but probably one should then add both version to the alias file.

Marcel Krüger

2:43 PM

@UlrikeFischer When we generate them from the luaotfload database, we get normalized names anyway. I think it makes sense to just use them and adapt fontspec to do the normalization before trying the lookup.

Ulrike Fischer

@MarcelKrüger well the main problem will be the lookup of bold font versions etc. If "Aleygreya" actually does a file lookup, fontspec can't use Aleygreya/B etc.

David Carlisle

@UlrikeFischer I tried adding alias "Aleygreya/B" which didn't seem to work but maybe that can be made to work

Marcel Krüger

@UlrikeFischer @DavidCarlisle For now I only did fontname lookups, not family lookups but I thought that we might be able to remap them to use another separator character. E.g. Alegreya-B probably could be made to work.

Ulrike Fischer

@DavidCarlisle I tried that too, but I suspect that will need larger changes in fontspec to map this.

David Carlisle

3:18 PM

@UlrikeFischer luaotfload seems to have a new feature:-)

Ulrike Fischer

@DavidCarlisle ?

David Carlisle

@UlrikeFischer github.com/latex3/luaotfload/commit/…

Ulrike Fischer

@DavidCarlisle oh ;-).

Mike Renfro

4:58 PM

Welp, 2 years later (https://chat.stackexchange.com/transcript/message/50189574#50189574), figured out how to make a .torrent for MacTeX.

Could someone please verify that https://vps.mike.renf.ro/MacTeX.pkg.torrent is functional? If so, feel free to send it along to whoever manages the MacTeX download page.

Kim Jong Un

5:40 PM

I don't understand why my posted image in this chatroom cannot be loaded. I only see spinning progress bar. It does not seem to be normal.

Skillmon

@KimJongUn mayhaps a firewall intercepts imgur? That's the case for me when I'm at work

Kim Jong Un

@Skillmon But reposting the same image works.

Skillmon

@KimJongUn hm, I already saw your first one.

Kim Jong Un

@Skillmon OK. Thank you for confirming.

Mike Renfro

6:14 PM

Something's weird in the .torrent file I made.

- I copied over a bunch of the trackers from the regular TeX Live 2021 ISO,
- I can see that I'm able to talk to at least 10 of them,
- I know I've got the right ports open on the home router since I'm currently seeding TeX Live and a few other things,

but no go on the actual MacTeX.pkg.torrent file I linked a bit ago.

Skillmon

6:54 PM

@JosephWright, @PhelypeOleinik I have found a 20% faster \group_align_safe_begin: and a 6% faster \group_align_safe_end: implementation :) And it is only marginally more cryptic :) Unfortunately it still doesn't fix our little problem that a single align safe group doesn't work in \keyval_parse:nnn.

Phelype Oleinik

@Skillmon Is faster code ever less cryptic? ;-)

@Skillmon But nice! I guess we don't need lots of code readability there

Skillmon

\group_begin:
\tex_catcode:D `\^^@ = 2 \exp_stop_f:
\cs_new:Npn \group_align_safe_begin:
  { \exp:w \if_false: { \fi: `^^@ \exp_stop_f: }
\tex_catcode:D `\^^@ = 1 \exp_stop_f:
\cs_new:Npn \group_align_safe_end:
  { \exp:w `^^@ \if_false: } \fi: \exp_stop_f: }
\group_end:

@PhelypeOleinik ^^^

biggest drawback I've found so far: Any code that normalizes tokens without expanding them will break this, because then \exp:w will see either `{ or `}, and the result will not be empty.

Phelype Oleinik

@Skillmon Hm... That could be a problem... Does the test suite run fine with that?

Skillmon

7:09 PM

@PhelypeOleinik it does, but unravel for instance will not work with anything containing these.

Phelype Oleinik

@Skillmon Not good then... Maybe ask Bruno

Skillmon

@PhelypeOleinik using \exp:w -`{ \if_false: } \fi: \exp_stop_f: wouldn't have that issue and is still faster, not as fast as the other one, though.

Ulrike Fischer

7:51 PM

David Carlisle

@UlrikeFischer I was just about to ask a question, lucky I didn't

Ulrike Fischer

@DavidCarlisle thanks for being so considerate!

David Carlisle

@Skillmon why is `^^@ quicker than `} ?

Skillmon

@DavidCarlisle because with `} you also have to put a minus in front of it so that \romannumeral doesn't produce any output.

David Carlisle

@Skillmon sorry translation error, I read exp:w as \expandfter not romanumeral :-)

Skillmon

7:57 PM

Actually, I made a small mistake during benchmarking, it seems like \group_align_safe_end: is faster than my implementation (but the begin: part of my implementation is considerably faster, still).

David Carlisle

@Skillmon so losing the extra atokens with romannumeral is faster than using \else ?

Skillmon

@DavidCarlisle in \group_align_safe_begin: losing the extra tokens with \romannumeral is faster than with a \ifnum comparison, but in end: the \ifnum comparison also acts as the \iffalse to gobble the other token, and combined \ifnum is faster than \romannumeral and \iffalse.

(during my first benchmark I didn't use \c_zero_int, but 0 in my minimal test file)

David Carlisle

@Skillmon yes makes sense

Skillmon

@DavidCarlisle if comparing to a register instead of an explicit number, that is. Else the \romannumeral + \iffalse wins by roughly 1.5 % :)

Jonas Stein

9:10 PM

@PauloCereda inkscape --export-type="svg" --pdf-poppler test.pdf returns

inkscape --export-type="svg" test.pdf returns

evince test.pdf shows

Jonas Stein

9:31 PM

pdffonts test.pdf
name type encoding emb sub uni object ID
------------------------------------ ----------------- ---------------- --- --- --- ---------
MHOLNP+CMMI10 Type 1 Builtin yes yes no 4 0
JHYTSG+CMR10 Type 1 Builtin yes yes no 5 0
EJASTL+CMR7 Type 1 Builtin yes yes no 6 0
MAFBCJ+CMMI7 Type 1 Builtin yes yes no 7 0

I am out of ideas what to test next

oh there is a mistake in my tests...

Mike Renfro

vps.mike.renf.ro/MacTeX.pkg.torrent appears to be a valid Torrent file now. Only has one tracker in it so far, but verified it's seeding from a home system to a remote one.

Mike Renfro

10:03 PM

Edited to add more trackers, seems to still be working, emailed mactex-support@tug.org to see if it can get published on the regular download page.

Transcript for

$TeX - LaTeX$ TeX, LaTeX and Friends