@JosephWright the problem is: peple misuse certain constructs, so sometimes.it's hard to guess. I've for sure seen \operatorname{where} and \text{log}.

@UlrikeFischer I would appreciate it.

@DavidCarlisle -- And happy birthday to your Mum!

@UlrikeFischer -- I'm back. Here are some things that the Unicode guys said would never be added to Unicode, but would require markup: open-ended sub- and superscripts (low-value single digit numerals are okay as superscripts, used for footnote markers, but letters and other symbols, no; in other words, no math -- markup needed); limits on sums, integrals, etc.; fractions beyond what might appear in a cookbook (i.e., no letters or symbols); different sizes of fences;

@barbarabeeton We might not be able to write the only language that is common to the whole world (mathematics) in Unicode, but at least we have Jurassic emoji: unicode.org/L2/L2016/16072-jurassic-emoji.pdf

accents that can extend over multiple letters/symbols/other characters; different sizes of arrows except to distinguish between "single-size" and "extensible"; letters./digits/symbols used as embellishments for such arrows; a method for expressing a root; ... I think you can get the idea.

@HenriMenke -- Yes, they certainly have allowed that sort of "extension". I think MathML is probably their preferred approach for math markup.

@barbarabeeton So math needs extra markup, but emojis (which are practically all implemented as SVG in the fonts) do not. The Unicode consortium is an absolute joke.

Mandatory XKCDs: xkcd.com/1726 and xkcd.com/1953

@HenriMenke -- Unicode is essentially linear. I'm inclined to agree that they've gone overboard with emoti. (And the suggested images look much too happy in that proposal, Also, one wonders what's the reason for the spike showing extreme interest in octopi in late 2010?) You really have no idea what a struggle it was to get a script/calligraphic alphabet recognized; I'm extremely grateful to Murray Sargent for providing the Hamiltonian equation as a convincer.

@HenriMenke -- Love the xkcd with the detour sign.

The day that the UTC admitted that mathematics is a language was a breakthrough.

@barbarabeeton I really don't get the inclusion of emoji symbols. How is rendering SVG in a font different from rendering it on a HTML canvas? Why do they waste so many code points on stupid faces?

@barbarabeeton Are there any scientists on the UTC?

@HenriMenke -- Blame it on the ubiquity of the iphone.

@barbarabeeton That is actually part of the reason for the proliferation of emojis. People are much more likely to install software updates if they get exciting new emojis instead of only boring security updates.

@HenriMenke -- Murray Sargent is/was a practicing physicist. There are quite a few computer folks. At least one trained linguist (Ken Whistler). Asmus Freytag has a strong scientific background, although I'm not sure in what field.

@HenriMenke -- True. I think the first emojis were proposed by Japanese representatives to the UTC. I was off the STIX project by the time the first emojois surfaed.

@barbarabeeton I'm still amazed that it took a long time for them to recognize mathematics as a language then. Do you know the reason? Whether the language is spoken or not should be irrelevant for script.

@barbarabeeton Does it make sense to have math symbols as Unicode symbols?

@HenriMenke -- Why did it take so long? Not enough pressure from the people who wanted to communicate in "their own language" on the web. And the reluctance of the "interested" parties to pay for the effort to compile the list of symbols. The initial "guesstimate" for how long it should take to compile such a list was 35 hours. In reality, it took many hundreds, and by the time Unicode 4.0 came out with the additions, I'd spent over 3,000 hours on the project.

@barbarabeeton How much effort is it to add extra symbols?

@FaheemMitha -- If there weren't math symbols in Unicode, you wouldn't be able to read most math/technical material on the web in html form. It would be impossible to copy and paste passages from math/technical documents. There wouldn't be a common base for reliable communication. Lots of reasons.

@barbarabeeton I'd really like to see \hbar in Unicode because it is ubiquitous in physics. Also \lambda with a bar is very common in quantum optics.

@HenriMenke -- I think \hbar is there, and a good chance that \lambda with a bar as well. Finding them may not be so easy. I'll look.

@barbarabeeton There is LATIN SMALL LETTER H WITH STROKE but that is not \hbar.

@barbarabeeton Ah, ℏ

@barbarabeeton U+210F PLANCK CONSTANT OVER TWO PI

@HenriMenke -- I was going to suggest that \hbar might be one of the symbols designated by a "variant selector". Unfortunately, I don't have access to the sources of the Unicode submission since I retired. That's all back at the AMS.

@HenriMenke -- You can access a (now rather old) version of the source list via a link on the page ams.org/STIX . Looking there, one finds that \lambdabar is indeed a variant (using the math variant selector) of U+019B . Regarding the collection of information on that page, I know some of the links are broken, and there's no one interested in maintaining that information; I also don't know whether anyone will be interested in keeping that page in existence.

@barbarabeeton So there are already math symbols in Unicode? Sorry, I haven't been paying attention.

I just added a tikzpicture environment in a scrartcl document after \begin{document} but before \title, and created a separate page before the actual main document.

Does anyone know if this is normal/expected behavior, or do I keep to come up with a MWE?

If I put the environment a bit later, then it appears on the first page of the document.

I did this so I can put some text above the title. It feels like an inelegant way to do this, but it's obvious to me how to do so with just scrartcl.

@barbarabeeton There might be complaints in the logs. I haven't looked at them.

Though I thought that TikZ was supposed to write on the existing page, but create a separate one.

Assuming, of course, the page starts with begin{document}.

@FaheemMitha -- As I said, I'm not really familiar with the KOMA classes, but amsart, and I believe article, has code to explicitly prohibit insertion of a float at the top of the first page. So it would not be surprising if scrartcl has a similar trap.

@barbarabeeton Oh. Is tikzpicture a float? I thought with the overlay option it just writes on top of whatever is there. I suppose I could try to read the documentation.

Or I could ask a question. I might learn something.

quack

@PauloCereda quack, quack!

@CarLaTeX ooh a morning duck

(it's 4:18AM here)

@PauloCereda OOOOOOOHHHH!!!!

@PauloCereda You should go back and sleep

@CarLaTeX I am okay :)

@PauloCereda sleep is very important for ducks!

@CarLaTeX :)

@FaheemMitha No, a tikzpicture is just a box. With the overlay option, it forces the box to have sides of length zero. Page 256 in the current(?) pgf manual.

@DavidCarlisle quack <3

@PauloCereda breakfast!

@FaheemMitha yes it is expected, a tikzpicture is always a box. Sometimes a quite small box (e.g. with overlay), but nevertheless a box.

@DavidCarlisle oh no

very nice package for structural engineers!

ctan.org/pkg/stanli?lang=en

And today I again realized how bad the \scantokens implementation is :(

@Skillmon Hah!

@Skillmon \scantextokens

@JosephWright defined where how?

@JosephWright ok, it's part of LuaLaTeX. How long will it take for it being available in the other engines?

@Skillmon 'Never' or at least 'not until Joseph does the work'

@JosephWright you see, I'm not really a LuaTeX user, most of the time I use pdfTeX (and never XeTeX).

@JosephWright :( ok, I'll go into my corner and cry.

@Skillmon What are you using \scantokens for? Even when 'fixed', it's still a tricky primitive

@HaraldHanche-Olsen @UlrikeFischer So even a small box (with overlay) would force an extra page?

@cis The expl3 code is rather newer than the pgf one, and there are performance issues when you want to do masses of calculations, as for plotting

@JosephWright e.g. ducksay uses it for its verbatim input-mode (though currently in a bad manner, that seems to not be completely correct... I thought doing only \group_begin: \tex_everyeof:D { \exp_not:N } \exp_after:wN \group_end: \tex_scantokens:D { <stuff> } would suffice, but that's still not completely safe, maybe I'll switch to using \tl_rescan:nn, but I think there was some problem for me back when I tried)

@cis very unlikely, the TikZ maintainers don't want to use expl3 code.

Because: The expl3-syntax is hard, pgfmath is simplier.

@cis this would require massive rewrites of perfectly fine working code, I don't think they'll.

@cis \usepackage{xfp}\fpeval{10*sin(30)/(cos(30)+5)} doesn't look too hard?!

@Skillmon I havesomething like this:

%\pgfmathsetmacro{\ModII}{mod(\Square,\Divisor)}% old
\pgfmathsetmacro{\ModII}{\fpeval{trunc(\Square-(\Divisor*trunc(\Square/\Divisor,0)),0)}} % new

Can I use xfp.sty for that?

@cis \edef\ModII{\fpeval{...}} instead.

@Skillmon mod(\a, \m) is a simplier syntax at all...

@JosephWright spotted spelling error in interface3: description of trunc function in l3fp: [...] largest absolute value less that that of x [...] instead of less than that.

@cis expl3 has \int_mod:nn.

@Skillmon Has xfp.sty something similar?

I will look into the manual.

@cis All xfp does is \NewExpandableDocumentCommand \fpeval { m } { \fp_eval:n { #1 } } and \NewExpandableDocumentCommand \inteval { m } { \int_eval:n { #1 } }.

@cis the documentation is currently pretty minimal, you won't find anything.

@cis instead look in texdoc interface3 and the description of the l3fp module.

@JosephWright not sure who was responsible for xfp, but why didn't they just do \cs_new_eq:NN \fpeval fp_eval:n or \cs_new:Npn \fpeval { \fp_eval:n }? Why the huge overhead of xparse for just a single short mandatory argument?

@Skillmon OK.

@Skillmon It's not huge, really

@FaheemMitha if there is a \clearpage after it yes.

@Skillmon Policy thing: really, we want all document commands to have xparse interfaces, as that over time wll allow us to track them

@Skillmon Compared to doing the fp work, xparse is not heavy

@Skillmon @egreg is working on it

@JosephWright I just need to describe the conditionals and it should be ready

@Skillmon It's the whole 'layers' business. The lesson of LaTeX2e is you need 'this is a document interface', 'this is the implementation of a feature taking a known set of arguments', with some way of mapping between the two

@egreg Great

@JosephWright @egreg @DavidCarlisle any comments on this output? vvv

T1 + rmfamily looks odd ...

@egreg ooh

@JosephWright I know the reasoning why you want xparse to handle things, but I still think that for only mandatory braced arguments it is over the top.

@UlrikeFischer With T1 you get the requested glyph. With TU, the standard Ligatures=TeX enters the scene; one might argue that U+0022 > U+201D should be omitted.

@UlrikeFischer I don't know if there is a \clearpage after it or not.

@PauloCereda twitter.com/Duck_page/status/1203399324937445379

@PhelypeOleinik awww

@FaheemMitha Not necessarily, of course. But it could. Like all LaTeX boxes, it is forces the start of a paragraph if not already in horisontal mode, and that paragraph could wind up on a new page.

@FaheemMitha And by the way, please use the site's mechanism for linking to the post you're replying to. The chat being so very asynchronous, it may be hard to follow the discussion otherwise. (It's hard even with proper links, but even harder without.)

@HaraldHanche-Olsen I do normally do that. In fact I regularly tell other people to do so. But when responding to multiple replies, it's less clear what to do.

@FaheemMitha I see. Perhaps reply to one, then reply to the other pointing at the first reply?

(pgfmath ---> Dim. too large....)

Can anyone tell me if there is a limit to how much tex input one can create from lua using tex.print(…) from within a single call to \directlua? Of course, sooner or later the process will run out of memory, but short of that? I am asking because it seems easier to write a lua loop than writing the loop in tex, repeatedly calling out to lua to check if there is more data to be had.

@cis \fpeval{##1<0 ? : ##1}

@HaraldHanche-Olsen Yes, I could do that.

He does not take \fpeval{##1<0 ? "" : ##1}

@cis Yes, there must be an expression between ? and :

@egreg Hehe, this does not work in this way:

every column/.code={
\ifnum\pgfplotstablecol>0
%\pgfkeysalso{column type/.add={|}{}},%
\pgfkeysalso{postproc cell content/.style={
@cell content={
\fpeval{##1<0 ?  : ##1}
}
}}\fi
},

@cis You can use another comparison test, if you want to return different things.

\ExplSyntaxOn
\NewExpandableDocumentCommand{\fpcompareTF}{mmm}
 {
  \fp_compare:nTF { #1 } { #2 } { #3 }
 }
\ExplSyntaxOff

Then you can use \fpcompareTF{##1<0}{}{##1}

@egreg Puh, this is very advanced. But it works. Thx!

@UlrikeFischer more or less what I'd expect. You get straight " but ' is apostrophe so you don't get a straight ' from '

@Skillmon it's only really the overhead of loading the package if not already loaded. The overhead of the actual definition if there are just mandatory arguments is very small.

@DavidCarlisle You want another example?

I'll open a bottle of champagne, if I have one. Since I have none, stay with German beer.

@cis I'm sure you could do it with pgfplotstable but nothing you have shown so far suggests that package is helping in any way.

@DavidCarlisle Yes, I just can not expl3 etc. This helps me in the long run more if I master pgfplotstable; although the way is stony and cumbersome. :)

@ChristianFeuersänger

@ChristianFeuersaenger

:)

@DavidCarlisle yes but is it sensible that T1 and TU differ on the double quote?

@UlrikeFischer oh no, it isn't I'd say " ought to be straight

Last expample (because this is a solution, I am searching for)

@cis have you said anywhere what the algorithm is, I can guess most of the rows and could probably work out what the rest are but a spec would be interesting, I'd think you wouldn't need a lot more than a basic latex tabular.

@cis @UlrikeFischer wird dir sagen, dass ich gut in Deutsch bin

(The further gaps would make the algorithm easier to overview. But I do not have implemented them yet.)

@cis "gaps" you mean empty cells? that doesn't require anything special surely, a tabular cell is empty by default. Or do you mean something else?

See my temporarly code, if you want. It is clear, that you could make it better. For me it is good and needfull in the first solution.

\documentclass[a4paper, landscape]{article}
\usepackage{amsmath, amssymb, amsfonts}
\usepackage{pgfplotstable}
\pgfplotsset{compat=1.16}

% tex.stackexchange.com/questions/348799/…
% tex.stackexchange.com/a/34425/46023
% OLD:
%\usepackage{expl3}
%\ExplSyntaxOn
%\cs_new_eq:NN \fpeval \fp_eval:n
%\ExplSyntaxOff

% New:
 \usepackage{xfp}

 \ExplSyntaxOn
\NewExpandableDocumentCommand{\fpcompareTF}{mmm}
 {
  \fp_compare:nTF { #1 } { #2 } { #3 }
 }
\ExplSyntaxOff

@DavidCarlisle would we be able to change it? For lualatex probably, but who maintains the teckit mapping of xelatex? And if we can, do we dare? ;-)

@DavidCarlisle grabbing a single argument with xparse takes about 100 times longer than with a simple \def...

@Skillmon Yes, we know: we have experimented with 'shortcut' versions, but that leads to consistency issues

@UlrikeFischer Arthur would be first person to ask, but I'm not sure we can change it really

@Skillmon actually I was thinking about the short circuited versions (but I forgot they are not used currently) but even so given the overhead of the arithmetic, 100 times the speed of def isn't really an issue for a top level command is it?

@DavidCarlisle I'm not sure either. As a german I'm not really affected - we can't use " as a glyph anyway ;-).

@UlrikeFischer do you know what hyperref is doing here?

    \edef\{{\ltx@backslashchar\ltx@leftbracechar}%
    \edef\}{\ltx@rightbracechar}%
    \edef\\{\ltx@backslashchar\ltx@backslashchar}%

why no \@backslashchar for \} ?

(looking at \ltx@... use)

@DavidCarlisle depends on the frequency of use. One of my recent documents contains 1162 macro calls in its chapters alone (no preamble, no inner calls to other stuff counted).

@Skillmon yes but even if 500 of them were \fpeval how much longer would the document take doing xparse for that compared with doing toing the actual arithmetic, or doing filesystem operations to load a couple of packages? (that said it's probably possible to short circuit the definitions of simple macros if that does prove to be an issue)

@DavidCarlisle no idea, looks curious.

@DavidCarlisle of course that's not the bottle neck. I just wonder why some code is written as it is.

@Skillmon inner calls likely should be code-level ...

@JosephWright most of the times, maybe. What if I write package A that needs stuff of package B? Shouldn't package A stick to the documented interface where possibly and not rely on some internal code of package B?

@Skillmon Yes, but the expl3 approach is that the code-level interfaces should be documented too, not just the document-level ones. For example, my siunitx re-write will do exactly that: xparse for \num, etc., with a way that another programmer can use the interfaces

@JosephWright Guess I'll have to rewrite some of my packages, especially drawing a sharper border between internal code level and top-level code level (that sounds stupid, didn't know a better way to describe it)

@Skillmon I'm doing the same: the ideas have developed over time, and for siunitx I didn't do that to start with

@Skillmon the model has always been that the code level should be more or less self contained and not use document level commands, hence interface3 documenting that level, and xparse being one possible top level interface giving a latex2e-style user interface, at one point I think we assumed that an xml-style top level declaration would also be needed, but user-level xml interfaces haven't proven to be as popular as the promise of the start of the xml era.

@DavidCarlisle @JosephWright Hi, David, thank you very much because you wrote my name :-) about the unicode characters. This Siverapp application as recently discussed with you, does not support anything: images and formulas is a huge problem. The only better solution is to use unicode characters.