@DavidCarlisle Now that is a real longtable!! :-)

@MartinScharrer Been a while since I've used xkeyval: could you post a demo (non)working example?

@JosephWright I will try. I couldn't yesterday night. A minimal form actually worked, but as adjustbox key it doesn't. I have to peel some layers off first.

@MartinScharrer OK

@JosephWright A high-level MWE would be:

\documentclass{article}
\usepackage{adjustbox}
\begin{document}
\adjustbox{clip={a}}{Test}

\adjustbox{clip={a,b}}{Test}
\end{document}

Here I use the clip key which ignores this value.

However in the second case I get clip={a,b} as content while the Test disappears.

@MartinScharrer Yup, see that. Taking a careful look now :-)

I will try to zoom into it.

It looks like xkeyval does \@onelevel@sanitize to the argument: pretty much a disaster

@JosephWright Yes, but apparently the finally used code is not sanitized, because I got an error when I used a macro which is only defined for the key, but not for the macro.

@MartinScharrer Yes, I'm trying to trace through the full route. Not helped by the internal macro names here (for xkeyval)

@JosephWright Indeed. It seems this is only done to check for the existence of internal commas.

@MartinScharrer Got it! In \@adjustbox` you use \toks@, but between assigning and using you have \setkeys. What will be happening here is that xkeyval is setting \toks@ 'unsafely' when there is a comma in the value, which is why you see a problem.

Conclusion: either send a fix to Hendri for xkeyval or use a private toks. The second is much easier to do :-)

@JosephWright Thanks!

I just figured out that this works: \adjustbox{clip={a,b},}{Test}

Apparently the "is-there-a-comma-with-a-different-catcode"-test gets confused when the comma is in braces.

First it tests for a sanitized comma, then using normal scan techniques. If the second form doesn't find the comma it assumes the comma has a different catcode. However, the braces hide it and it branches the wrong way.

@JosephWright \toks@ is also used by graphics, that's the reason I used it as well.

@MartinScharrer Possibly. I abandoned xkevyal some time ago as there are various issues with the code and the interface for creating keys is not great

@MartinScharrer Yes, done properly you should be fine, as it should only ever be set within a group

@JosephWright Well, I started from graphicx which uses keyval but needed some of the advanced features of xkeyval.

@MartinScharrer Yes, I went much the same way: 'I want an easy way to do X' ...

In my last contact with Hendri and he said he isn't active with the package or with LaTeX in general.

For simple things, I'd now use kvoptions/kvsetkeys, or for more advanced stuff l3keys (pgfkeys is also a good choice, but it's in many ways a similar 'payload' to expl3)

@MartinScharrer Yes, I'd seen that

@MartinScharrer Will you try to fix this?

@JosephWright I will try to find a solution, at least for a little while.

@MartinScharrer As I said, 'use a private toks' is the easy fix for you :-)

@JosephWright I don't think so, because this only fixes the content issue, the inner keys are still not processed correctly.

\begin{adjustbox}{bgimage={rotate=45,}{example-image},} doesn't rotate the background image.

@MartinScharrer Ah, hadn't thought of that

@JosephWright Ok, changing the toks actually causes the keys applied to bgimage to stop working because I still use graphicx code for the images which uses the different toks.

@JosephWright: I figured now a way to avoid it by locally disabling that check of xkeyval, however now the , and = must have the correct catcode.

@JosephWright How does l3keys handles catcodes of = and ,?

@MartinScharrer Same method as Heiko uses in his code: we do a search-and-replace for active = and , tokens

We use \tl_replace_all:Nnn to do that, which is all built-in to LaTeX3

% \begin{macro}{\keyval_parse:n}
%   The parsing function first deals with the category codes for
%   |=| and |,|, so that there are no odd events. The input is then
%   handed off to the element by element system.
%    \begin{macrocode}
\group_begin:
  \char_set_catcode_active:n { `\= }
  \char_set_catcode_active:n { `\, }
  \char_set_lccode:nn { `\8 } { `\= }
  \char_set_lccode:nn { `\9 } { `\, }
\tl_to_lowercase:n
  {
    \group_end:
    \cs_new_protected:Npn \keyval_parse:n #1
      {
        \group_begin:

@MartinScharrer I considered making keyval support other catcodes at the time, but I think I didn't as I was worried about token usage back then, but you can just delay defining KV@do KV@split until use time so they pick up the catcodes current. (I don't see the need to support multiple catcodes in a single instance

@JosephWright in L3 shouldn't we just outlaw document-level catcode changes so the issue of normalising them doesn't exist?

@DavidCarlisle Possibly, but here we are L3-in-2e I guess: we could alter this definition using the appropriate DocStrip guards

@JosephWright \regex_replace_all:nnN {\cA=}{=} \l_keyval_sanitise_tl

@DavidCarlisle That is broadly the plan, of course: pick a fixed set of active characters

@egreg Too slow

@JosephWright I guessed so. :)

@egreg Also written much later than l3keyval (which got merged into l3keys)

The regex stuff is clever but where things can be done 'directly' it's still easier

@JosephWright One might think to add a hook in the code for treating other active characters that are used in a specific document.

@egreg Yes, that is true, but as @DavidCarlisle point's out we really want to go the other way. For working with LaTeX2e this might be needed and can be added if required.

@JosephWright Consider French and babel: colon, semicolon, question mark, and so on.

In XeLaTeX one can avoid activating them, with the interchartoks feature.

@JosephWright You replace all normal = with active = and then use active = as separators?

@DavidCarlisle Different catcodes can occur if you define some macro in the preamble and use it in the body. \def\mykeyval{a=1,b=2} ... \expandafter\mycmd\expandafter{\mykeyval,c=3}

@MartinScharrer No, the other way around. I wrote this in a slightly odd way (the 8 and 9 have catcode other, and so represent the 'normal' versions)

@MartinScharrer when I rule the world I'll ban packages from changing catcodes

@DavidCarlisle This would allow the = and , to have different catcodes. However, I agree that this is not really something you need to support.

@egreg Yes, I know :-(

Actually I was thinking about both myself, support active and normal characters or pick up the currently used catcode.

checksum mismatch in local font rpxmi (-2147483648 != -567552317) in virtual fo
nt pxmi.vf ignored.
checksum mismatch in font pxmi.vf ignored

Should I be worried?

@AndrewStacey No.

@AndrewStacey However on my machine the two checksums agree.

@AndrewStacey Try "vftovp pxmi.vf|grep CHECKSUM". What's the output?

Mine is

(CHECKSUM O 37432210015)
   (FONTCHECKSUM O 33612753303)
   (FONTCHECKSUM O 16045310566)

@egreg Mine too.

@MartinScharrer I'm sure I could find some documents with catcode 5 = for you to test with...

@AndrewStacey Then it's the other font: try "tftopl rpxmi.tfm | grep CHECKSUM"

It should output (CHECKSUM O 33612753303)

@egreg It did.

Why are the checksums in the error message negative?

@DavidCarlisle I was afraid for something like this ;-)

Any other catcodes?

@AndrewStacey It's quite strange. Can you show a minimal document?

@egreg Give me a minute ...

@AndrewStacey 2147483648 is octal '20000000000, so I suspect the usual cs tricks where 1111111 is just -1

@MartinScharrer just for you

\documentclass{article}
\usepackage{pxfonts}
\AtBeginDocument{
\newlength{\widthofcdot}
\settowidth{\widthofcdot}{$\cdot$}
}
\usepackage{unicode-math}
\setmathfont{xits-math.otf}

\begin{document}
hello
\end{document}

@egreg (lualatex) Okay, now I know what you're thinking ... but most of the bits you see there are going on inside packages that were being loaded by the document I was compiling. Stripping out all the unnecessary stuff, I ended up with the above. Fortunately the genius who wrote the stuff that was automatically loading pxfonts provided a nopxfonts option so my solution is to invoke that.

@AndrewStacey I don't get any warning

@AndrewStacey Oh, lualatex! I believe it's not doing good when comparing checksums for virtual fonts.

@egreg Okay, so really nothing to worry about then.

@AndrewStacey I'm pretty sure I've seen those spurious messages with luatex before. Check on luatex-l

@egreg No hits for "checksum" on the mailing list at tug, but I found this on context: archive.contextgarden.net/message/… (Conclusion for me: don't load pxfonts automatically if I'm not going to use them)

@percusse: what would be the possible copyright issues if we create a pinkfloydlipsum? :)

@DavidCarlisle ;-) Not all catcodes are possible. Like, 0, 1, 2, 9, 14 and 15 should not appear in this context. I also think 5 as well.

@MartinScharrer = can have any catcode 0-15, not all of them result in a character token of that catcode that is true, but if you use the current value of \catcode\=` to define a command then you might pick up such nonsense. Or you just state that the catcode of syntax characters may not change. LaTeX optional arguments don't work if you change the catcode of [] and very few problems have been reported because of that.

@DavidCarlisle I wonder how basic LaTeX deals with active , wrt. option lists.

@StephanLehmke It doesn't deal with them. syntax characters are catcode 12

@StephanLehmke \long\def\@forloop#1,#2,#3\@@#4#5{...

Heya!

I tried searching, but could someone please provide me with a link where I can read about how to add items from enumerate into the table of contents automatically? =)

@PauloCereda I don't think there will be any issues since we are just mentioning the lyrics. But that package can be handy for verses and quotations I think :-)

@percusse Ah. :)

@PauloCereda We should make a multimedia lipsum using PF to test for animations and sound clips.

@percusse Wow! That would be epic. :D

@PauloCereda Then we are in deep trouble with copyright infringements unless we keep it under 5 seconds.

:-)

@percusse Poor Gilmour, no guitar solo for him. :P

@PauloCereda We can put the is there anybody out there? part for debugging purposes.

@percusse :)

@N3buchadnezzar It's quite a bizarre request, I'd say. The "whole" item?

Just the number/Label =)

@N3buchadnezzar It's not very clear how this would help the reader.

@JosephWright, @DavidCarlisle: I now added some code which replaces active = and , as well as the ones with the current enabled catcode with the normal catcode 12 (other) versions first.

So you can mix normal, active and one other catcode easily. It's easily extensible to more catcodes to be handled by default.

@MartinScharrer Yes, no doubt it can be done (you only need a 'replace all in' macro and then the appropriate decisions on what to replace). As @DavidCarlisle said, the reason for being wary of overdoing this in the LaTeX3 code is more that we'd rather have fixed catcodes in documents. That of course means a bigger set of actives (likely ~, ", $, _, ^, &, I think), plus I guess a stronger presumption n using UTF-8 engines (still an area to get sorted!)

We may of course need some more cases, but if so what is needed is a better way of tracking things

@egreg: one professor of mine suggested me a very nice book a few years ago. Once in a while, I like to re-read it. :) I think you might like: The Lamb's Supper

@PauloCereda I don't have a Kindle. :)

@egreg Oops, my bad. I sent the ebook link! :)

@egreg: I think it might have an Italian version. I have the Portuguese one. :)

@PauloCereda Oh, yes, there's also the printed version. And there is also in Italian. I'll look in a bookshop.

@egreg It's a wonderful book. :)

Anybody interested in getting a free gnuplot book? There's a contest - write a small article on the topic of LaTeX and graphics and post it on the LC site, two winners will receive a book. latex-community.org/component/content/article/92-contests/…

@MartinScharrer I wonder: can adjustbox perform "manual page break of its contents" - i.e. I give the max. height and you return series of vertical boxes...

@DavidCarlisle That's the road to ActiveTeX, isn't it? All catcodes are the same...

@tohecz Let's say you have a normal paragraph; I'd typeset it in a \hbox and measure its width, so that I have an estimate of the necessary line width; then I'd typeset the paragraph in a \vbox increasing the \hsize until it fits.

@egreg mis-understanding. The width is given. I'm really just interested in a way to break something on "pages" so that I can manipulate with them afterwards...

@egreg Unless the paragraph is large and you get a "dimension too large" on the hbox :)

@StephanLehmke Like Joyce's flows of consciousness?

maybe I'm unclear, my apologies, I just try to disturb myself from being nervous before tomorrow's state exams

@tohecz That's already available: \vsplit to <dimen>

yes, I know, I just thought that Martin might have implemented some manupulation like this

@egreg How many lines might it take? 200? Doesn't seem that large, especially if it contains \\

@StephanLehmke 35 with Plain TeX default \hsize. Probably some juggling with a \vbox with predefined width is better, in this case. :)

One overruns the \maxdimen when the \vbox contains 26 pages (default Plain TeX \vsize). But we could increase the \hsize to \maxdimen ...

One can accommodate 45000 lines in a \vbox, 1800 pages (default values). :)

@egreg Interestingly, that's less of a problem than it might seem. I've made documents with hundreds of pages which are built by putting the whole content into one large table in a \vbox and then splitting of page-size chunks with \vsplit. TeX complains about dimension too large (so you can't find out how many pages you'll get :) but incremental \vsplit still works.

@StephanLehmke no:-)

@StephanLehmke You can say \setbox0=\hbox{a\hskip\maxdimen b\hskip\maxdimen c\hskip\maxdimen} and \unhbox0; TeX will complain about Overfull \hbox (15944.80058pt too wide)

\box0 is just a list, the calculations are performed only when requested.

So TeX is able to go down your gigantic list and to find the break points, but not to tell you the height of the full box.

To get an estimate of the number of pages you have to split the box.

For those familiar with the movie "Groundhog Day" we seem to be having this experience with Vidhya and his table questions. :)

@AlanMunn Yes

@JosephWright I've counted three.

@egreg I suspect the forum approach: he wants us to write his table exactly, and won't take a duplicate as an answer

See comments on tex.stackexchange.com/questions/58672/…

@JosephWright He's already got that!

@egreg I'd seen that

Regarding people like Vidhya: I'm always curious what drives them to ask the same question again and again. Did one of you ever have a real life conversation with a troll?

@KeksDose I've seen a few of them in action in newsgroups; but I guess that they don't show off in public.

@tohecz Not at the moment, but I might try to add this. There is an answer which shows how to do this with a \vbox.

@StephanLehmke you don't get that ... you get a silent overflow. if you measure the hbox it will simply show a small value after a while

@FrankMittelbach It showed up in connection with the famous bug in tabularx where 35 X columns triggered a Dimension too large error.

@egreg sure you can produce the error but not by simply typetting material into an hbox for ages, right?

@FrankMittelbach If you're a "philosopher or modernistic novelist who writes 200-line paragraphs" … Hint for the casual reader: look for "Joyce, James" in the index of the TeXbook :)

I still believe catcodes are some sort of feline plan to take over the world. :)

@egreg I think we are talking about different things ... if you try to typeset Joyce and then you may end up with a problem if you try linebreaking on such a paragraph but if you just typeset the whole Joyce into \setbox0\hbox{"Joyce..."} then everything should happen without TeX noticing.

Who's Joyce? :)

funnily enough I'm right now just playing around with an old idea of my on that topic to enable rebreakable paragraph material

@PauloCereda dunno someone index in the TeXbook

@FrankMittelbach Ah got it. :)

Hm carrot cake! :)

@PauloCereda especially the active ones ... the passive ones are ok ... right @David?

@FrankMittelbach Phew, I'm pretty safe here. :) My cats are lazy. :)

@PauloCereda Riverrun past Adam and Eve's ...

@egreg got that (more or less). What I'm trying to do is to typeset a number of paragraphs in a data structure that allows me to retypeset it afterwards in different measures without doing any macro level expansion happening

@JosephWright :P

@PauloCereda Dubliners is OK, could not get on at all with Finnegans Wake

@JosephWright I couldn't go beyond page 10 of Ulysses (in Italian).

@egreg I've not tried

@FrankMittelbach That's the same: \unhbox :) But of course you lose anything that needs vertical mode.

@FrankMittelbach I expected something like this. :) Different measure might result in different macro expansion. Think to align.

@egreg, @MarcoDaniel Did you see the LaTeX-L 'discussion' e-mail?

@egreg That would eb the restriction. You can't do that for everything without completely going over to active TeX and implement everything outside TeX stomach. But for a lot of situation this is enough. E.g., think of captions etc

@JosephWright I haven't yet recovered from the shocking news that I'll have to modify kantlipsum because it uses \prg_stepwise_function:nnN :)

@egreg Oh no!

@egreg We really are trying to get l3kernel really stable: @FrankMittelbach has done a very good job of keeping us honest

@FrankMittelbach People put any kind of things in their captions. :)

@egreg Don't update CTAN until after I do, of course (I do the LaTeX3 CTAN releases: somehow I seem to have ended up as 'release manager' by accident)

I'll have to update siunitx, too :-)

@egreg there is a 3-6 month grace period so ...

@FrankMittelbach We've gone for 6 months on anything used elsewhere in TeX Live, I think

@egreg sure they do including \footnote ... but we'll see how far i get with my ideas first

they are only about 10 years old so quite new

@JosephWright yes, this is why i said 3-6 month

@JosephWright From my point of view of "package writer" this shouldn't make much difference, I guess.

@egreg Well, in terms of using the kernel code yes

Assuming you're not using anything internal, of course :-)

@percusse I think you turn your comment into an answer :)

@JosephWright Never used anything more deep than \exp_after:wN (but I'll try to be a good boy and avoid it). Of course there's \tex_def:D in regexpatch, which is a different story.

@egreg yes and no. It is an "offer" meaning that nobody needs to makr in his packages anything as internal. However, we do hope that the convention is going to be picked up, because, frankly :-) speaking the 2e is a mess. I guess you can count the internal kernel commands that haven't been hooked into by one or the other package code is with your fingers

for makr read "mark"

and the same is true for many packages and their internals. As a result you can't even make slight changes to the order of some code without upsetting some use of "\patchcmd" or ...

@JosephWright nice writing Joseph.

@FrankMittelbach Thanks

@FrankMittelbach I had understood well, it seems: if I'm writing a package and use only non \__ functions, then my functions won't be \__. But if I need to implement something that can't be obtained with the "interface" functions (or I need a faster implementation), then the internals should be \__.

@egreg ??? not sure I parse that ...

@egreg if you write a package then you are not supposed to use anybodies internal function ever (in theory)

other than your own.

but you can of course (and are encouraged to) make a lot of the functions you define in your package internal too

to indicate that others should not use them but use only those that you indicate as interfaces to your code

@FrankMittelbach So I didn't understand well. :)

Maybe the concepts of "module" and "package" should be cleared up.

@egreg what is unclear about it?

serious question. I mean whatever is unclear should be cleared up .. right

@cmhughes Would you be willing to add that into your answer? It's really a minor thing :-)

@egreg The three files that Joseph converted are good examples to see what is meant. To give some example: in 2e you often have chain of commands (essentially implementing syntax, e.g., there is \@float \x@float ... and God knows what. Now each of those commands has been used by others package writers instead of always using, say \@float thereby making it impossible to do any changes/improvements without breaking one or the other package.

Instead we intend to mark all commands that we do not want to see being used outside our package as \__foo which means that only the interface commands remain. Now implicitly this is already the case through documentation but ...

@percusse you're too kind :) done.

@cmhughes \begin{blush}Thanks.\end{blush}

@FrankMittelbach So if I define a function \foo_func:nn that uses inside it two other ones, those should be \__ functions?

I don't really follow the reasoning. Functions with : in my packages are not for external use.

@egreg if you want me in my, say multicol package your function \foo_func:nn but not those that it calls "internally" then yes.

but if those functions are functions that you want to allow others in their code to directly use then you do not give them \__ names. It doesn't "prevent" anybody from using them technically it is just a convention. look at l3coffin as an example

@egreg Remember that for every document level function there is supposed to be one 'public' code-level function (OK, I've not sorted that for my own stuff yet, but ...). One of the issues that's already come up on the team list is that for a lot of packages almost all of the material is 'internal'. For example, siunitx probably needs to provide about 10 'public' functions in something like 6500 lines of code

@FrankMittelbach Well, until now I've used expl3 only for producing packages that provide user level macros, so I can't see the business too clearly. I don't expect that somebody hooks into regexpatch to use one of its internal functions: should they be all marked \__?

@egreg That's the concept, and also the reason why it's a point for discussion rather than something we've changed. It only makes sense if it works for enough people (in my opinion, anyway)

@egreg my experience with a good number of packages and the LATeX kernel indicates, yes they probably should because people use them over time

@JosephWright in my opinion it would work even if it is only for the kernel, but yes I hope it is something most people see a benefit eventually

@egreg but you are right, if you write a package that is not likely to be used other than as through their document level commands then it is less urgent to do so.

However, if you think about something like multicol or array ro what have you .. those packages also started out to implement document level commands, but then people used them as part of their packages. so they used what would be in expl3 terms some of the :nnn functions but instead of using only those that I thought would be interface functions for other packages they used many of those that where not meant to provide a "defined" interface by themselves

@egreg @JosephWright it also helps to think about what are those commands with "defined" semantics that you want to allow. For example, looking at Joseph's first attempt to convert l3coffins to this syntax he marked functions as internal that may or may be better interface functions. So conciously doing this helps to get a clearer picture of what your "official" programmer interfaces are.

For example, is \__coffin_if_exist:NT really internal? If so then the name is correct. But if one wants to allow other programmers to have this test then this would be a command to document and call \coffin_if_exist:NT

@percusse: Not sure what you mean by "black magic"? Haven't tested your solution yet (was too busy fixing mine) but will do so soon.

@PeterGrill I want to ask you how OP wants to draw from (1.5 and 4) without even defining it.

@JosephWright by the way I know what is wrong with __ syntax ... you have to escape it ion this site even in comments :-(

There is something not right with the question I guess.

@percusse Yeah, that was the point of the "Notes" section, where I pointed out that you can only access nodes you have defined -- perhaps I was not clear.

You're solution works fine.

@PeterGrill It's clear to me but i think OP is clogged with over-automation.

@percusse Well, gotta give the OP the benefit of the doubt as probably what he wants to do is more complicated and he was not able to figure out that he was trying to access nodes he had not defined -- tried to provide info on that as well.

@PeterGrill Sure I am constantly in that mode anyway. No disesteem about the power of a confused mind :)

@percusse Likewise... Twice this past this week I went all the way to composing the question and the MWE, and just before I hit the post button, I realized my mistake...

@PeterGrill I did that last week in a big project of mine. I was gonna even make a bug report of a well-known library. :P

@PauloCereda Good to know I am not the only one who goes that far before having the "Duhh!" moment.

@PeterGrill I do that a lot. :)

@percusse Heya, I am about to ask a question on the site! Eg in the making of a MWE and all that. In tikz/pgf, how do one declare numbers to represent numbers? Eg put a=5, then write (0,a) later.

I even registered to the mailing list in order to report the "bug" when I realized I did a capital mistake. :)

@N3buchadnezzar So you are publishing the trailer of your question, eh? Coming soon to a question near you! .... Unmistakenly some irrelevant explosions and noisy stuff follow

It also includes PF background music!

@percusse Don't look back when an explosion happens! It makes you look even cooler! :)

@N3buchadnezzar Oh yeah! :D

I think ... takes off the sunglasses ... the answer is \def\a{5} :P

Def initively!

I just feel bad asking a question where I have not been able to make any progress :p

@percusse Did not work! would not bang 0/10!

@MartinScharrer I suppose you thought of me as you needed some deep tikz insight? The document looks the same to me if I use pdflatex or latex/dvips/ps2pdf not sure what difference I am supposed to be seeing. Of the 6 coordinates in the aux file, the last 2 are unchanged but the first 4 are displaced by (exactly) 203.5pt

@N3buchadnezzar But but \begin{tikzpicture}\def\a{5}\draw (0,0) rectangle (\a,\a);\end{tikzpicture}

It worked Eureka!

@PauloCereda Too much Voltran when you were a kid eh?

@percusse LOL

@PauloCereda It's amazing that you got that out of the gray matter :).

@percusse Indeed. :)

tikz refuses to caclulate 0/5 hmm...

@DavidCarlisle No, I meant the output routine part. I have not much knowledge about it and I'm not sure if the principle way I do it is correct or if there is a much simpler way.

I guessed:-) Also I was wrong to say it looked ok via dvips of course I forgot to run it twice so my dvi run used the existing (pdflatex) aux file coordinates. If I use latex/dvips on a clean run the big cross is in the right place but the two small lines move off up right

I think I need this hometone.com/entry/no-snooze-alarm-clock-by-vivek-sasindran

Ahoy question ahead =)

@MartinScharrer not sure if that's a bug in tikz or not as I haven't looked what the claimed origin is for tikz ccordinates. Similarly hard to comment quickly on the output routine use as you don't touch it directly just via atbeginshipout (which I've not used)

@percusse My alarm clock beeps for at least 20 minutes. It's sufficient to put it beyond arm length. :)

@egreg I think I'm a sleepsnoozer

@N3buchadnezzar Do you want to "prove" that $2=\sqrt{2}$? :)

@egreg As usual I am writing solutions for exams ;)

@N3buchadnezzar It's a nice problem for showing that one must not rely on intuition when talking about convergence: if you increase the number of steps in the stair, it becomes indistinguishable from the diagonal. But each stair, no matter how many steps it consist of, has length 2, while the diagonal has length $\sqrt{2}$.

@egreg Indeed

@N3buchadnezzar You can do the same with smooth curves: $f_n(x)=sin(nx)/n$ in $[0,2\pi]$ IIRC, but computing the lengths is harder. However it's possible to find a lower bound that's greater than $2\pi$.

And the same can be done with "trollscience"

@egreg It is basically 4*times the same concept

@N3buchadnezzar I like more the one with smooth curves. :)

Good night.

Good night!