@UlrikeFischer trouble is I think people just don't think of the filename at all, they download some file and it gets saved as whatever (2).tex and just expect it to work as it opens in an editor and they select some menu option to run it

Hello LaTeX fans! Is anyone up for a philosophical (LaTeX) question?

@UlrikeFischer probably thinks a commandline like epstopdf some file with spaces and no weird quoting .eps should work as well

It's in relation to a fellow U&L user's question on TeX (tex.stackexchange.com/questions/513418/…)

Ahhh! Here's the blog I was thinking of, re: the above discussion on spaces in filenames: dwheeler.com/essays/fixing-unix-linux-filenames.html

(section 6, mainly)

@DavidCarlisle probably. I wished him good luck ;-)

If anyone's interested in tackling my philosophical question, it's whether I'm forcing a square peg through a round hole by using LaTeX to solve a template-type situation, where the amount of incoming text is unknown. Paragraphs may be long or short, items may be long or short, and yet there's still a desire to apply some formatting constraints.

I get the sense that most people input a known amount of text and then fiddle with the (La)TeX to make it look perfect.

@JeffSchaller you can use latex for that sort of thing, but in that generality there's not a lot that can be said about what to do

ok; I'm building up the courage to ask a couple questions on the site about particulars that I'm running into, but I don't want to lead me (and others) down a dead end if it's a Really Bad Idea

@JeffSchaller well the description above would apply to more or less every document type

Ok; forgive my ignorance - I last used LaTeX to typeset my resume a couple decades ago, and so I'm not an active user. I'll search the site for template-type questions before going too much further, and will build up a couple specific typesetting questions, maybe over the weekend

heh! 6,666 results for "template" questions!

@JeffSchaller I don't know what you mean by template type, but "the amount of incoming text is unknown. Paragraphs may be long or short, items may be long or short, and yet there's still a desire to apply some formatting constraints." is basically what latex does, you set up the styling for (say) a thesis but the incoming thesis text may have all that variation. But I suspect that you have a more specific thing in mind.

@DavidCarlisle I do, so I'll ask specific questions on the site; I was just afraid (after reading around some more) that I was swimming upstream with my general approach, and wanted a sanity check

I enjoy oddball challenges, and Wildcard dropped this idea into the U&L chat the other day, and it struck my fancy. (They're trying to replace a proprietary document system with something free)

(brb, putting kids to bed)

@JeffSchaller "template" isn't a defined term in latex, so people use it to mean anything, usually just means some kind of starter example taht can be edited to contain your real text, s

@JeffSchaller the indesign question you linked to? That really needs more information to be answerable, eg a screenshot of the indesign layout you want to emulate. I wouldn't know where to start to answer with just the information there

@DavidCarlisle yes, that's the one, and yes I'm crazy for tackling it. I've made enough progress to feel optimistic -- the old "80% done, 80% left to do" :)

and yes, by "template" I meant a general structure that would be used repeatedly with various text

I've started a shared Google Doc with Wildcard where they provided an example layout, so I've had an inside track, so to speak

@JeffSchaller which in a latex context is essentially meaningless as that's how latex works, you set up the structures in a document class and that is used for multiple documents

@DavidCarlisle hmmm, good! We're approaching the edge of my ignorance. by "structures in a document class" do you mean (what I think of as) a plain old latex.tex file that has formatting commands in it along with example text, or do you mean a whole new type of document (beyond "article", etc) where the end-user creates their own new file.tex that used commands from that new ... type

"documentclass" == "new type of document"

@JeffSchaller every latex document starts off \documentclass{something} where the something document class sets up the structures, so article has things for a typical basic journal article, sections, enumerated lists, title etc, beamer has structures for powerpoint replacing presentations, fixed size slides, gradual reveals, etc, and so it goes on.

Yes; I've (simplistically?) started this effort with \documentclass{article}, so I wasn't sure if you meant to continue in that direction, or to create a new documentclass

(based on some inertia suggested at tex.stackexchange.com/questions/513418/…)

@JeffSchaller start off with article as its easy place to start, but if you end up with article and a 10000 line set of definitions in the preamble to customise that then it is trivial to move the definitions around a bit so the individual documents just have \documentclass{mycustomclass} where the mycustomclass file inputs article and sets up what you need behind the scenes

@DavidCarlisle that sounds like it may end up being a good idea, and also beyond my current skillset. Like I said, I'm probably misguided by my progress so far with the various tweaks on article

@JeffSchaller it is really trivial actually, if you have \documentclass{article}\usepackage{graphicx}\def\zzz{123} then don't make every document have that make them have \documentclass{wibble} where wibble.cls is nothing more than \LoadClass{article}\usepackage{graphicx}\def\zzz{123} that's all there is to it.

@DavidCarlisle thank you! Seems within my grasp, although the preamble isn't The Mess right now; it's all the things I've tried to get the formatting for the body :)

question - Are users using a tag when posting to automatically build provided LaTeX and produce an image in their posts? Or are they just using the blockquote tags and then posting a picture? That's what I've been doing, but wondering if there is an easier way...

@delrocco For the code you can select it and press CTRL+K, for the rest no easier way. See Section 3 here: tug.org/TUGboat/tb39-3/tb123duck-format.pdf

thx

@Skillmon ooh secret rabbit list sorting

@delrocco most operating systems let you cut a fragment of your pdf view then you can just paste an image in with ctrl-v or your general browser paste key shortcut and teh site will add the markup needed, or you can drag and drop images (you don't need blockquote for images)

@PauloCereda :) It has two variants currently, one of them is faster than l3sort even for worst possible input (correctly sorted list) for lists of 20 elements and fewer (and taking about 1.7 times longer for 40 elements). For optimal input (reverse-sorted list) it is faster than l3sort for 160 elements. And for random input it is faster for up to 40 elements. I'm not sure how robust it is compared to l3sort however, and it doesn't strip spaces from the list elements.

@PauloCereda all of this benchmarking has a big error margin however, as l3benchmark isn't the most precise benchmarking possible (but probably close to the best possible in TeX).

@Skillmon I've looked a bit and in fact \clist_set:Nn takes as much time as your sort on short lists (5 elements). I think space trimming unavoidably slows things down. If I only benchmark sorting itself (doing \clist_set:Nn outside the benchmark and suitably copying the clist to avoid starting from a sorted clist when the benchmark code loops), l3sort catches up around 15 entries. I agree it could be better to use merge sort only for large lists, but I'm hesitant to add more code for [...]

[...] something that doesn't seem so central to the kind of tasks LaTeX is used for. I may be wrong about how important it is to optimize here.

@BrunoLeFloch I currently have two versions (see github.com/Skillmon/tex_insertionsort), the simple one being considerably faster on simple data for a well defined comparison test. I agree that optimizing the sorting has very low priority, and combining the mergesort with an insertionsort on small sublists would add a lot of code just to achieve that bit of performance gain. [...]

[...] I've started this just because I was contacted by someone who wants to write a TikZ library for 3d drawing with automatic object depth coverage, who needed a sorting algorithm and didn't want to use l3.

And since he doesn't expect large lists to be sorted (talking about an estimated 20-50 elements list) I thought mergesort or another n log n sorting algorithm wasn't worth the code complexity.

@BrunoLeFloch oh, and on an l3seq variable my sorting algorithm would be considerably faster, too, as I could just redefine \__seq_item:n and wouldn't need the argument grabbing (but I haven't tested how much performance that would gain)...

@BrunoLeFloch and don't worry, I still think you're much better in coding in TeX than I am (as well as in probably any other language) :)

@Skillmon I'll have to look more carefully. It might be possible to do insertion sort without too much more code. Using this at the lower levels of merge sort would certainly speed up sorting at all list sizes.

Sorry, I have to go.

@BrunoLeFloch ok, bye!

@BrunoLeFloch hi!

@daleif no rush but if memoir input the new iftex instead of ifvua/xe/pdf... then this could be a lot shorter:

(/usr/local/texlive/2019/texmf-dist/tex/latex/memoir/memoir.cls
Document Class: memoir 2018/12/12 v3.7h configurable book, report, article docu
ment class
(/usr/local/texlive/2019/texmf-dist/tex/generic/iftex/ifpdf.sty
(/usr/local/texlive/2019/texmf-dist/tex/generic/iftex/iftex.sty))
(/usr/local/texlive/2019/texmf-dist/tex/generic/iftex/ifetex.sty)
(/usr/local/texlive/2019/texmf-dist/tex/generic/iftex/ifxetex.sty)
(/usr/local/texlive/2019/texmf-dist/tex/generic/iftex/ifluatex.sty)

@DavidCarlisle it it now include all those system tests? Then clearly I'll change it to that. I'm preparing a new release, just have to clean a chapter in the manual. Then hopefully not too many miktex users will have issues with that switch

@daleif yes all the others just input iftex (ifvtex defines a few extra ones after that like \ifvtexhtml but the ifpdf,ifetex,ifxetex and ifluatex packages now all just input iftex.sty and do nothing else.

Super

@cis arara doesn't use kpathsea for the lookup, see the documentation about configuration files.

@cis did you check the documentation? The second page even has an example for a local path for rules.

@cis No, either you should create an arararc in your home directory specifying the rule paths or make that project specific. But as @UlrikeFischer said the documentation is quite good.

@cis The first example in section 4.2 shows a paths setup. Put that into a file starting with !config saved at some location specified in section 4.1 and you are good to go. It's not harder than having a local texmf.

@UlrikeFischer what a thing to say

@DavidCarlisle easier than trying to remember where I put my arara rules ;-)

@DavidCarlisle xspace finally bit me. I'm done with it now. :)

@barbarabeeton ltugboat compatible iftex.sty gone to CTAN

@AlanMunn it seemed like a good idea at the time.

@AlanMunn what did it do wrong?

@DavidCarlisle And I haven't been using it for years actually, but I was modifying a old file that had some macros that used it, and I was trying to write their values to a file. It took a little while to figure out why I was getting incomplete \iffalse errors...

@DavidCarlisle So as per usual, 'it' was 'me'.

@DavidCarlisle Or in your sentence, 'it' was 'you'. :)

@AlanMunn not my fault then, that's OK.

@DavidCarlisle It never is.

@AlanMunn er don't mention shellesc

@DavidCarlisle -- Thanks. (Now, I just have to figure out why you're being so nice to us and treating us as special.)

@barbarabeeton Perhaps it's because I am nice to everyone and make them feel special, even if I don't really care?

@barbarabeeton Did you see this? mathoverflow.net/questions/345383/…

Can someone confirm for tex.stackexchange.com/q/514498/35864 that

\documentclass{article}
\usepackage{etex}
\begin{filecontents}[force]{ipsum.tex}
  dolor
\end{filecontents}
\begin{document}
  \input{ipsum.tex}
  sit
\end{document}

produces an error?

@moewe I get an error but not the one the op describes but a "no room for a new write error".

@AlanMunn -- I hadn't seen it. Thanks. (The only Greek I know is modern, so I couldn't translate it accurately, but rumors did exist among AMS employees. It's not really "approved" by the current management -- they want something more friendly and inclusive. I'm not sure you'll see the old logo on anything official any more. But the new one doesn't say "mathematics" to me. On the other hand, I'm now retired and shouldn't care.)

@moewe I get ! No room for a new \write . with TL19 (not quite up to date) on the second run,

@moewe I get no error. I haven't updated super recently.

@UlrikeFischer I get that too and if I press on it then tries to overwrite \jobname as the OP describes.

@AlanMunn did you compile twice.

@moewe yes just saw but my reflexes are geared towards the first error ;-)

@UlrikeFischer Ah, no. Yes on second compilation I do get the error (no room for new write) (I hadn't read the question.)

@UlrikeFischer Thanks for confirming that. Do you perhaps recall if

\documentclass{article}
\usepackage{etex}

\begin{document}
  blub
\end{document}

@barbarabeeton Well in these days of "inclusion", it must be offensive to non-geometers. :)

used to produce a warning about etex being obsolete?

@DavidCarlisle did you see @moewe's example?

I think it did at one point, but I don't get that warning any more with the newest LaTeX kernel.

@moewe you get the warning if the extended registers are already in use. So e.g. if you already allocated lots of counters, but not if you load it quite early.

@UlrikeFischer Ah OK. Thanks for the explanation.

@moewe and so something like this lets the document compile again:

\documentclass{article}
\usepackage{expl3}
\ExplSyntaxOn
\int_step_inline:nn{300}{\newcount\blub}
\ExplSyntaxOff
\usepackage{etex}
\begin{filecontents}[force]{ipsum.tex}
  dolor
\end{filecontents}
\begin{document}
  \input{ipsum.tex}
  sit
\end{document}

@UlrikeFischer Ooh, that is an interesting workaround.

@moewe removing etex would be better. But I will open an issue, we should look at this.

@UlrikeFischer That goes without saying. Thanks for opening the issue.

@moewe the issue is here github.com/davidcarlisle/dpctex/issues/23.

@UlrikeFischer no..

@DavidCarlisle I just added an issue to etex ^^^ it is perhaps a latex bug, but as latexbug complained ...

@UlrikeFischer yes just seen it, I'll look, what's it doing?

@DavidCarlisle you get a ! No room for a new \write . error.

@UlrikeFischer yes I mean why, hang on I'll try it....

@UlrikeFischer @moewe why is the question closed?

@DavidCarlisle I think it was unclear, but the OP just added an update with the etex example.

@DavidCarlisle It was closed because the MWE that was given when it was posted did not reproduce the example.

@moewe ah, OK

The OP edited the question today and I saw it in the reopen queue.

Voted to reopen and so did a few other people, so I guess it will be reopened soon.

@moewe I voted.

@UlrikeFischer @moewe \def\ch@ck#1#2#3#4% isn't he same as \gdef\ch@ck#1#2#3

@DavidCarlisle yes I saw this too. But I have no idea how to resolve this.

@UlrikeFischer we could make the fourth token after \ch@ck be \relax or anything that wasn't \if@filesw or we could just say tough don't use etex or ....

@DavidCarlisle we could reserve enough counters in the kernel so that etex is never used ;-).

@UlrikeFischer my original plan was that etex.sty was empty, but got too many issues at the time, the current dance was a delicate compatibility balancing act. But that is still an option if it starts causing issues.

%        [1997/08/12 v0.1 eTeX basic definition package (DPC)]
%        [1998/03/26 v2.0 eTeX basic definition package (PEB)]
%        [2015/03/02 v2.1 eTeX basic definition package (PEB,DPC)]
%        [2015/07/06 v2.2 eTeX basic definition package (PEB,DPC)]
%        [2015/07/08 v2.3 eTeX basic definition package (PEB,DPC)]
%        [2015/09/02 v2.4 eTeX basic definition package (PEB,DPC)]
%        [2016/01/07 v2.5 eTeX basic definition package (PEB,DPC)]
%        [2016/01/11 v2.6 eTeX basic definition package (PEB,DPC)]

@UlrikeFischer I'm trying to remember why the new format allocation re-used \ch@ck name rather than use a new name like \e@alloc instead of \@alloc I vaguely remember that was intentional and made something slightly more compatible but...

@DavidCarlisle perhaps some other packages with allocations are involved too?

@UlrikeFischer can you look at the definition of \newinsert in etex.sty?

@DavidCarlisle sure, what should I check?

@UlrikeFischer sorry got called away..

@DavidCarlisle no problem.

@UlrikeFischer etex sets ch@ck back to have 4 arguments where the 4th is supposed to be the second digit #1<#2#4 but in \newinsert the 4th argument is the nested test, isn't it?

@UlrikeFischer oh I see, it works out OK in the end,

@DavidCarlisle I thought #4 is always for the nested test?

@UlrikeFischer maybe it was, although I thought it was digits like 1#1 where #1 is the second digit, maybe I mis-remembered

@DavidCarlisle in etex it seems to be used only in newinsert, so if the #4 should have some extra function it must be in latex or some package.

I've been trying different fonts. I like most so far the following. Using baskervaldx for text and Asana Math for math. What do you think of this choice? do you also like it? Been trying it and see no problems so far. Here is a MWE and after that a screen shot of the result

\documentclass[12pt]{article}

\usepackage{amsmath}
\usepackage{unicode-math}
\defaultfontfeatures{Scale=MatchLowercase}
\setmathfont{Asana Math}
\usepackage{baskervaldx}

\begin{document}

\textbf{Simplest form of the \textit{Central Limit Theorem}:} Let
$X_1$, $X_2,\cdots$ be a sequence of iid random variables with mean $0$
and variance $1$ on a probability space $(\Omega,\mathcal{F},\mathbb{P})$. Then
\[\mathbb{P}\left(\frac{X_1+\cdots+X_n}{\sqrt{n}}\le y\right) \to\mathfrak{N}(y)
\coloneq\int_{-\infty}^y \frac{\mathrm{e}^{-t^2/2}}{\sqrt{2\uppi}}\,

@DavidCarlisle the only problem that I would have with dumping etex is that mtpro2 still has the faulty allocation command and would break if loaded to late and without etex:

\documentclass{article}
%\usepackage{etex}
\usepackage{expl3}
\ExplSyntaxOn
\int_step_inline:nnnn {1}{1}{100} {\newcount\a}
\ExplSyntaxOff
\usepackage[lite]{mtpro2}
%\makeatletter
%\alloc@0\count\countdef\insc@unt\pointcount@ %from mtpro2.sty 2009/4/27 v2.3
%\makeatother
\begin{document}

\end{document}

But perhaps we should dump it to put a bit pressure on them ...

@UlrikeFischer yes I was just wondering if I can rename it there...

@UlrikeFischer can't we get mtpro2 changed, seems to have my name on it...

@UlrikeFischer bbc news showing Berlin Walll...

@DavidCarlisle yes 30 years. A big change. I still remember it, we drove to Berlin and marched through the Brandenburger Tor with lots of other people and everyone was quite in awe that it had really happened.

@DavidCarlisle I don't know. I wrote in 2017 to the pctex support and suggested to replace the line by \newcount\pointcount@. I got an answer, but then nothing happened.

@DavidCarlisle Yes, thank you very much. I've done that for years on SE. I was only asking because some of you are so quick with your replies, complete with LaTeX and appropriate image, that I thought... "maybe this SE has some special tags for pasting in MWE LaTeX and the server builds it into a standalone image and posts it, kind of like the web C++ compiler on the cplusplus.com". I was wrong :)

@UlrikeFischer I was in Berlin for 4 days when I was 18, when the wall was very much there, growing up in the UK where you never see a gun, not even the police are armed, it was a very strange experience, we went through to the east for one day...

\documentclass{article}
\makeatletter
\let\zzz\ch@ck
\usepackage{etex}
\let\ch@ck\zzz
\makeatother

\begin{filecontents}[overwrite]{addfile.tex}
    \centering{Hello World!}
\end{filecontents}
\begin{document}
    Just a test.
\end{document}

@UlrikeFischer the worst that happens is that #4 happens afterwards rather than before the \else so in non-error cases it should be the same.

@DavidCarlisle yes, we were for a day in the east too and it was rather curious to have to go through strict security. We were also once in Ireland in a year when there was more or less peace and drove over to North Ireland and seeing the fortifications were quite strange too.

@Nasser The space between the integral sign and its limits look a bit wrong(/strange) to me.

@Nasser should use \symxx not \mathxx (usually) with unicode-math

@DavidCarlisle Isn't that the standard text from some "show math and text fonts together"?

@mickep don't know, it's still mostly wrong unicode-math markup, wherever it comes from:-)

@DavidCarlisle doesn't it matter then that the last clause is then in a group?

@DavidCarlisle Well, correct. I just meant that one should probably not blame @Nasser...

@UlrikeFischer yes if I did it in etex.sty I'd stick a firstofone somewhere but I think I'm tempted not to touch it

@mickep as long as no one blames me I don't mind who we blame.

@mickep we could blame @UlrikeFischer

@DavidCarlisle Oh, poor Ulrike...

@UlrikeFischer happy anniversary!

I wanted to go to Germany... :(

@DavidCarlisle young lad

@PauloCereda but probably not at this time of the year - it is getting cold here ;-)

@PauloCereda To injure Rudi Völler so it could have ended differently 1990?

@mickep you see @PauloCereda as some kind of Terminator-Duck travelling back to change history?

@DavidCarlisle ohh, skills!!

@DavidCarlisle I read it as he wanted to go to Germany 1989...

@mickep I know, but for some reason your reply changed my interpretation:-)

@DavidCarlisle I see. Maybe the real question here is if @PauloCereda would prefer West Germany or Argentina to win a game in soccer.

@UlrikeFischer I just sent hyperref to ctan (using iftex, not using hobsub)

@DavidCarlisle oh good. I wanted to make a luaotfload update but there is an error somewhere in the script handling that I haven't tracked down yet.

@UlrikeFischer blame Marcel

@UlrikeFischer on splitting oberdiek I chatted with Petra, she's not keen to have to update 90 catalogue entries in one go so is happy if we do as did for grffile, take packages out one at a time as needed and just keep re-submitting the main oberdiek as it shrinks.

@DavidCarlisle sounds like a sensible plan.

@DavidCarlisle Sounds good to me

@JosephWright splitting is still a pain, as I found with grffile, they all cross reference each other and do version checks so get very cross when you break the chain...

@UlrikeFischer @JosephWright just recalled hyperref from ctan, will need to re-submit in a bit

@DavidCarlisle what happened?

@UlrikeFischer well initially I recalled as I realised I'd made the ctan zip with l3build, which is a wrong as the l3build.lua only sets up the tests, so it just built some generic zip file. then when I tried running ./mkctan I found that one of the vtex tests is broken so I have had to update the dtx :-) (running checks now you'll see the commit in a minute)

@UlrikeFischer because I just input iftex not ifvtex package the extra \ifvtexhtml tests were not defined so I changed them to \ifnum\OpMode=10 or whatever, there were several all were already guarded by \ifvtex so fine, except one where I missed it wasn't so got undefined command just added in \ifvtex around that clause....

I have a few questions on best practice regarding creating latex packages

Like I'd like to create a package for a font

And for now I'd like to just make it work only with fontspec. Is that ok?

@BrainStone whatever, if it's a single font then probably that, but usually there are a range of fonts foo-regular.otf, foo-bold.otf, foo-math.otf etc and the package might be called simply "foo" (or to be honest font specific packages are needed a lot less with otf fonts than they are for pdftex and tfm fonts)

@BrainStone what does the font package have to do over and above loading the font via fontspec?

Probably not much

It offers several variants

Hm and also a variant where you can fine tune with a number value

So it would make sense to provide a command for that

It's also a ttf font

@BrainStone do you need a package then? Packages are needed for pdftex as you have to build custom fonts with custom encodinfs, but the point fo fontspec is more or less that you can access any system installed font without any font-specific tex support.

@BrainStone ttf/opentype essentially same thing as far as fontspec cares.

@DavidCarlisle I mean eventually I'd like to add the files needed for plain pdflatex, once I know how that is done properly

Though the main point being that you can get that font through CTAN

@BrainStone no harm in providing a package but unlike the case for pdftex you don't have many existing examples for unicode fonts

And as mentioned there's a point in making a helper command for the fine tuned font

I think I'll figure it out

@DavidCarlisle imho all packages from Bob set up font for both fontspec and pdflatex ctan.org/author/tennent @BrainStone

@BrainStone sure, but the need for a package is by the sound of it based on features I don't know about rather than being based on the font existing, so do whatever seems right:-)

@UlrikeFischer yes thanks I was going to look up that, couldn't remember the name:-)

@DavidCarlisle The main reason I'm even considering to be adding it like that is because the font is great. Had lot of research going into it

In any case. Thanks for your swift responses

@BrainStone the link Ulrike gave is the packages from the leading exponent of setting up modern fonts for tex, so following some examples from there is probably better than taking hints from me:-)

Oh. I completely missed that

Thanks!

@DavidCarlisle Markus gave me a bit of a shock as he wrote that KOMA would badly patch hyperref ;-). But I checked, the worst parts are in the else branch of a \@ifpackagelater{hyperref}{2009/11/24}

@UlrikeFischer skype?

@DavidCarlisle ?

@DavidCarlisle need to start, wait a minute.

Comparison of lambda.sty, l3sort and my insertionsort (inso):

@Skillmon funky st ligature on a formal plot took my eye away from the actual data:-)

@DavidCarlisle who said the plot is formal? :)

@Skillmon no ducks, so it must be fairly formal

@DavidCarlisle the data isn't reliable anyway, I opened the window sometime close to the end of the benchmark, so the cooling of my laptop was way better during that time, hence the processor was faster, hence the data is falsified.

@DavidCarlisle I just wanted to suggest to use ducks for some of the marks ... ;-)

@Skillmon as long as it is falsified to benefit your version to the detriment of competitors, that is OK, that is how contemporary politics works.

@UlrikeFischer hyperref accepted this time. and I disabled l3build ctan so I don't forget next time.

@UlrikeFischer I could change the name, from insertionsort_simple.tex to ducksort_simple.tex, but I'm not sure whether that's a good idea...

added the number of elements for which l3sort is faster the first time.

@Skillmon ducksort is much easier to remember ;-)

@UlrikeFischer but it is an insertionsort...

Unfortunately a binary search isn't really applicable in TeX, since the indexing would most likely take too much time :( Would have way better performance with it, especially on huge lists and for sorted input.

@UlrikeFischer TL list ....

@DavidCarlisle just saw your mail ;-(