00:00 - 17:0017:00 - 00:00

user19161
12:23 AM
Hey guys, I was retagging all my old questions and have finished doing so. Sorry if I undid some of your edits on my posts. I just wanted a minimum number of tags on my own questions.

user19161
I feel that having too many tags diminish their specificity. If you really feel the burning need to retag again, of course you are free to do so.

7:38 AM
Hi all! I have completely reformulated a question I asked yesterday tex.stackexchange.com/q/68030/16865 I belive that now the question is much clearer and also much more interesting for other people. Unfortunately, Heiko Oberdiek's answer now seems to miss the point. Is that a problem? I have adjusted my own answer to the new question and unaccepted Heiko's answer and added a note that the question has been edited. Hope that's OK.
@Paulo Cereda: By the way, have you ever ordered from squeakyducks.com? :-D I like the classic line quite a lot ;-) If I was American I would only consider the "Hatched in the USA" line, of course!
Note that it's only 5 of the daffodil ones squeakyducks.com/product_p/rd_daffodil_bd.htm and shipping will be free!

8:27 AM
@Canageek MathML is an output format. LaTeX is an input format. My guess would be that they needed an output format rather than an input one.
@lpdbw You should add a comment to Heiko's answer explaining that you've edited the question. You should also phrase it in such a way that it's clear to a passerby that Heiko's answer was to the original form of the question - this is to stop people voting it down thinking that it is an answer to the current form of the question.

@AndrewStacey OK, I will do that!
I hope Heiko will take notice of the new question and be kind enough to adjust his answer.

8:47 AM
@lpdbw That's another reason why you need to comment on his answer. If you don't do that, he won't see the change.

9:02 AM
@AndrewStacey damn and we pour mathml into a machine and (eventually) some of it gets formatted by latex. I must be running my life backwards.

@DavidCarlisle You know what I mean! For others, I mean that no-one ever writes MathML: you write something else that gets translated to MathML. MathML is an output format in the sense that PDF is an output format: you still need a program to translate it into "pixels on paper". It might be that tex is involved in that, but that's because tex can be both input and output.

9:25 AM
hi!
Is there any easy way to put a small number in the margins of latex? I tried \newcommand{\number}{ \begingroup \thefoo \hspace{-2em} \endgroup }
but this also moves the text after the number into the margin. (\thefoo is just a simple counter)

@AndrewStacey yes but.. we used to have sgml (then xml) with tex fragments for the maths but when we (I) switched it to have mathml fragments (just entered directly in emacs) our (human) editors liked the change, if the rest of the document and authoring tools are xml aware (and the mathematics is all gobldygook anyway) then editing mathml is easier than editing tex
@N3buchadnezzar are you sure you want to redefine \number (think carefully before giving a more than two letter answer)
@N3buchadnezzar \marginpar{\small\thefoo}

9:42 AM
@DavidCarlisle In my actuall document I use a norwegian names for my commands, so... No problems there

@N3buchadnezzar you have redefined the entire latex kernel to have different names including tex primitives? that is a project on the scale of latex3, I'm impressed!

Gah! Bloody Englishmen and their sneaky interpretations!

Name of contribution: The chkfloat package
Author's name: Tomas Hejda
Package version: 0.1
Location on CTAN: tex-archive/macros/latex/contrib/chkfloat/
Summary description: Warns whenever a float is placed too far from its origin.
License type: lppl

@DavidCarlisle Possible to get this effect on the left side of my document and not the right?

@tohecz: Congrats! :D

9:44 AM
@N3buchadnezzar what I mean is that \number is a TeX primitive and if you redefine it things will go wrong.

@DavidCarlisle I know, like I said. In my document I use \tall, which causes no problems.

@N3buchadnezzar really honestly, believe me, you want to answer "no" to the earlier question.

\reversemarginpar did not work as expected

@N3buchadnezzar if your numbers are at beginning of paragraphs then simpler is to use \noindent\llap{\thefoo\hspace{something}} although it ought to be a redefinition of \section or \paragraph or something of course really

@DavidCarlisle Yeah, memoir in combination with a hanging style I guess.
But this is for mathematics hand ins, there are no sections nor paragraphs.

9:57 AM
@N3buchadnezzar \section and \paragraph in latex mean "numbered logical unit" and that is (I assume) some kind of numbered logical unit or enumerated list or...

@N3buchadnezzar I recognise that course code!

@AndrewStacey ;) Trying to set up a good system for doing the coursework.

10:28 AM
grr why can't people use google

10:46 AM
@PauloCereda well use bing then because microsoft is pure as the driven snow

@DavidCarlisle LOL how about the duck engine? :)

@PauloCereda yep that works too duckduckgo.com/?q=hmisc+center+tabular

@DavidCarlisle ooh! :D

11:43 AM
hello

Hi @Raphink! How are you? :)

fine thanks
you?

Fine. :)
We miss you. :)

haha
I'm still a bit active on tex.sx
@paul
@PauloCereda you who are an expert in naming software projects
I need to name a piece of code :-)

@Raphink ooh I have some wacky project names. :D
@Raphink tell us moar! :D

11:50 AM
haha
ok
there's a program called facter which provides facts (key/value pairs)
this program takes plugins to provide additional facts
I wrote a generic plugin based on augeas to easily provide new facts by adding to a conffile rather than writing code
that's the piece of software I need to name
the plugin which generates new facts on the fly

ooh cool :)
How about some [fictional] inspector/detective name? :)

could be
so far, I thought of mythomaniac ;-)

holmes, closeau, poirot. :)

or simple augeasfacter, but that's very simplistic

ooh let's find one!
how about artemis? Hunting for facts. :)

11:55 AM

thank you @DavidCarlisle

@DavidCarlisle yes, yes, YES!!!!! :D

@Raphink @PauloCereda has a thing about ducks at present...

@DavidCarlisle No I don't. :P /ducks
oopsie

@PauloCereda you're suffering of anatidaephobia?

11:59 AM
@Raphink LOL

12:54 PM
Can be closed as TL, according to OP's last comment:

1:23 PM
@JosephWright: can we interview you for the next TeXtalk? :)

1:33 PM

@PauloCereda you just want to interview a duck.

@PatrickGundlach :)
@DavidCarlisle Duck, where?!
:)

@PauloCereda Sigh, I guess if there is no-one else

@JosephWright We can always interview @DavidCarlisle. :)

1:51 PM
Thanks @Paulo . Now my name is finally carved into the CTAN marble floor :)

@tohecz We should have a party. :)

@PauloCereda Sure!

@tohecz :)

damn, I travel to Paris in 3 weeks and I still don't know where will I be accomodated there. Is that a reason to be nervous?

@tohecz Oh I have no experience. :(

2:08 PM
Just answered a question migrated from SO. If I keep answering only those and collect about 200 of them, I can get the "unsung hero" badge!
5

@StephanLehmke that is THE PLAN ;)

@StephanLehmke epic plan! :)

@tohecz I assume your host will care for that. Maybe you can get a place in student's accomodation? I once got one when I was in Toulouse as a visiting researcher, and it was a memorable experience ;-)

@StephanLehmke I go there for PhD. I have government scholarship, but I have not been given any confirmation letter yet. The deputy attache of French Embassy has just written me that he now forwards my case to Campus France and they will "organize me an accommodation". My only option now is to wait and hope...

@tohecz If all else fails, Paris has a really large system of underground tunnels (metro, catacombs) ;-)

2:17 PM
@StephanLehmke Lol, no it's not that bad. My "plan B" is that my supervisor offered me to stay at his place until we arrange the accommodation.

Subway is a 1985 French Comedy drama film directed by Luc Besson and starring Isabelle Adjani and Christopher Lambert. The film is part of the Cinema du look movement. Plot Having stolen some compromising documents from a powerful and successful entrepeneur/gangster at a party, a man known as Fred (Lambert) escapes from the police and takes refuge in the underground world of the Paris Métro. There he integrates with the dwellers and befriends several colourful characters, some of which are living under the subway to avoid police arrest like him. While the gangster's henchmen look for hi...

1

I am trying to put a line over a paragraph in LaTeX. I tried this code: \noindent\makebox[\linewidth]{\rule{\textwidth}{2pt}} \noindent\colorbox{gray}{\parbox{\dimexpr\textwidth-2\fboxsep\relax}{\textbf{Text}}} But there is a small space between the line and the paragraph as shown in this imag...

The Catacombs of Paris or Catacombes de Paris are an underground ossuary in Paris, France. Located south of the former city gate (the "Barrière d'Enfer" at today's Place Denfert-Rochereau), the ossuary holds the remains of about 6 million people and fills a renovated section of caverns and tunnels that are the remains of Paris's stone mines. Opened in the late 18th century, the underground cemetery became a tourist attraction on a small scale from the early 19th century, and has been open to the public on a regular basis from 1874. Following an incident of vandalism, they were closed to th...
@tohecz Accepted without upvoting, which seems to be normal at SO proper and is the very precondition for the BADGE WHICH WILL NEVER BE AWARDED AT TeX.SE.

Or this one (not Parisian, it's Hungarian, but it's about a guy who lived in the metro system):
Kontroll is a Hungarian comedy-thriller released to theatres in 2003. Shown internationally, mainly in art house theatres, the film is a darkly comic thriller set in a Hungarian Metro system. "Kontroll" in Hungarian refers to the act of ticket inspectors checking to ensure a rider has paid their fare. The story revolves around the ticket inspectors, riders, and a possible killer. The film was written and directed by Nimród Antal and starred Sándor Csányi, Zoltán Mucsi, and Csaba Pindroch. The film was entered in a number of film festivals in Europe and North America. It won the Gold ...
@StephanLehmke but well, having the badge on an SX site means something is wrong, doesn't it?

@tohecz Darn, I also need to keep the TeX.SE folk from seeing the answer. FAIL.

2:22 PM
Take a picture of the Eiffel tower. :)

Apparently I cite just as I vote. Is there any update on editors which can read biblatex syntax and offer auto-complete lists? I can switch to Linux just for this detail if something is available. hell, I'll buy a mac too and write two blog comments daily about how awesome it is, my kingdom for a horse!
3

@PauloCereda Isn't it strange that my Wikipedia link got a picture of the Arc de Triomphe which doesn't feature anywhere in the Wikipedia article itself ;-)

@percusse emacs can obviously do this, it'll just need a bit of lisp....

@StephanLehmke LOL true! :)

@AndrewStacey No no: They use it as an input format. You hit input equation and it has you type mathML. I think I still have some documents I wrote in high school where I wrote page after page of equations in it.
@percusse Isn't the biblatex syntax the same as LaTeX? Emacs will highlight it. I think notepad++ will do autocomplete once you've written one entry. What are you citing?

2:36 PM
@Canageek openoffice doesn't make you type mathml it makes you use a visual editor or type starmath (which is more like eqn than anything as I recall)

@percusse ottobib.com helps a lot. And once you have one entry you can just copy and paste the syntax and fill in the feilds
@DavidCarlisle Are you sure on that?

@Canageek well it was true a few years ago when I last looked at openoffice.org. Not sure I have it or libreoffice at the moment (I know I have emacs:-) I'll have a look...

@DavidCarlisle I've not used it since I learned LaTeX: Isn't "X + {7} over {4} = 2" mathML?

@Canageek no that's star math mathml is xml that would be <mi>X</mi><mo>+</mo><mfrac>...

@Canageek If you put \addbibliography{mybibfile} in the preamble for biblatex (which is already bad to use it as such but have to for the BibTeX spoiled editors) TeXnicCenter, WinEdt and maybe others no more collect the citation keys in the project. So I have to look up for the keys in my bib file from JabRef.

2:39 PM
Oh, I also need a 'LaTeX for Word users who have never seen markup' stat! I've got intrest from someone in my lab!
@percusse Hmm, I could write you a script that will give you a cheat sheet. Would that help?

@Canageek word can accept mathml just by pasting it in

@Canageek Thanks a lot for the offer. But I populate the bib file on-the-fly if I find something else. I really overuse JabRef's IEEE search a lot, so the cheat sheet needs to be updated too for such use.
@DavidCarlisle I recall Zawinski's words: "Linux is only free if your time is worth nothing"

@percusse Well, I can give you a program that would autocomplete each citation after you have used it once
@DavidCarlisle Huh, wonder where I got hte idea that was MathML from? Probably a mistake in wikipedia about 10 years back

@Canageek That's precisely the problem, I need to reach for the unused keys.

@percusse I don't add them until I use them, so I can't really help you. The one I did for my thesis I just made a 'Primary Authour Date' system, so 'percusse2012' so I could guess the key from the paper
@percusse I disagree. Linux for basic use is no more time wasteful then Windows. It is just that people who like Linux are more likely to think spending time on it has value in itself.
@percusse I've also seen that quote used to describe LaTeX, since it takes so long to learn.
I need a 'LaTeX for Word users who have never seen markup' stat! I've got intrest from someone in my lab!

2:47 PM
@Canageek the ODF format that openoffice sort of uses stores its mathematics in mathml but the openoffice math editor predates that and uses an eqn like linear text format and internally converts that to mathml and stores both in the odf file. so if you give the odf file to some other odf system (like word) it probably reads the mathml but openoffice itself actually reads the star math (as you can see if you edit the odf file in emacs and make the two forms non-equivalent

@DavidCarlisle .....I'm still wondering why they didn't use LaTeX syntax. Also, ah, I probably mistread that when looking it up.

openoffice was called star office before oracle bought sun who bought star and open sourced it, hence the name of the math format
@Canageek because latex is just rubbish as a format for a word processor. It has \newcommand which means that you think you have a wysiwyg interface to "latex" and someone tries to load in half a million lines of tikz code....

@DavidCarlisle I meant just the stuff between the .
"X + {7} over {4} = 2" vs "X + \frac{7}{2} = 2"

@Canageek yes same there, so you define an interface to \frac and then the latex ends up being ams align of mhchem something or any random answer from this site.

@percusse Complete rubbish. Linux only costs something if you regard learning something useful as a waste of time.

2:52 PM
@DavidCarlisle I disagree it's only in the order of hundred thousands.

Star math doesn't scale very well. "
{}=1- overstrike 1 over overstrike 3 + 1 over 2 - 1 over 4 + overstrike 1 over overstrike 3 "..." - 1 over n - 1 over {n+2} newline newline"
Is a line from an old document of mine.

@Canageek well using a wysiwyg word processor doesn't scale well either, so it is a better fit.

@Jake seems like the question you linked to makes this one a duplicate
2

I would like to make a 3d plot containing the 3d axis and a convex polyhedron or better a Dodecahedron on the positive side for which one of its boundary edge points on its top is annotated e.g. \hat{x} "chosen" or "selected" and draw a line segment from that labeled point towards the origin. Th...

what do you think?

@DavidCarlisle I didn't mean put LaTeX inside it. Just put a similar, minimal markup based on it. Also there is a program to embed LaTeX into OpenOffice

@AndrewStacey Time is taken in the context of switching to another system and still keeping productivity, in other words, not stalling too much until you get the basics and find your way around etc.
Linux is not the best thing to fiddle with if you don't know what you are doing.

2:56 PM
@Canageek things are always possible (see how mathjax copes with an ever expanding subset) but the issue is always basically there that a language like latex that gives you the flexibility to define everything down to the lexical categories of the characters isn't an ideal format for a word processor to handle. I wasn't there, the real reason might have been they just preferred nroff to tex.

Let alone choosing the right distro...

@percusse You've not used it for a few years, have you? These days unless you have really odd hardware it comes all set up and ready to use.
@DavidCarlisle Sure, for laying out a full document. But what about just the default math that goes between $and$? That supports all the same stuff star math does, but has less oddities.

@Canageek It's not the Windows/OSX like UI I'm talking about.

I'm speaking as someone who learned and was quite dedicated to star math for a long time, then moved to LaTeX: Speaking with someone whom has used both to typeset formulea: The LaTeX syntax is better. Just copying the input syntax in the same way that say, codecogs.com/latex/eqneditor.php does would be a wonder.

Yay, @BrunoLeFloch is here! :)

3:00 PM
mathjax only supports math and the number of requests to support one package after another is never ending. you can put a blockarray inside math in latex but blockarray is... strange.. and I wouldn't want to support its syntax in any non-tex system. I suspect (since wysiwyg editing is what they really interested in) that they assume most ppeople use the template driven visual math editor

Hi guys, a question that has nothing to do with the current discussion: I see many people using the fp package and its macros. Would it be useful to provide a package defining the fp macros by translating them to corresponding l3fp functions?
Hi @Paulo@

@Canageek well you could use that click the mathml button and then stick the mathm, into the openoffice docuemnt

(Benefit = increase range and precision, robustness against weird inputs, speed)

@DavidCarlisle Probably, didn't know about that back then though.
I will not be surprised if this is a duplicate, but I'm in a rush (Sorry)
0

I've got interest from someone in LaTeX, but she's not a computer person. She is smart, very smart, but not a computer person. i.e. She didn't know that typing something on the command line ran a program, she thought they were all part of the command line. Is there a really basic introduction to ...

@BrunoLeFloch fpu library of Christian Feuersanger has a similar objective, would that be helpful to reduce the doubling the effort? See section 36 here texdoc.net/texmf-dist/doc/generic/pgf/pgfmanual.pdf

3:05 PM
@Canageek small2e too small? (not so short intro too long?)

@DavidCarlisle What is that? The Not So Short Introduction to LATEX2ε is what is recommended so far.

@BrunoLeFloch I like the idea. :) Will it be just a high-level interface?

@cmhughes I think the new question isn't really a duplicate because it also asks for ways to highlight the intersections of a line and the polyhedron. To which the answer, unfortunately, has to be "Can't be done with reasonable effort in TikZ", I believe.

@Canageek texdoc small2e texdoc lshort (the latter is the one you mentioned)

@Jake ok, just wanted to check :)

3:11 PM

51

This is a community wiki since there is no "one true answer"- if you find an answer that contains a lot of your choices, but is missing something you feel is incredibly useful. Feel free to add it. A short description of what topics the reference covers would be nice as well. Organize by subje...

@percusse Ah, I suppose something useful would be to provide a TikZ library which would change pgf from using its own computations to using l3fp. It should be roughly as fast as pgf, and much more precise (basically, fp's accuracy with pgf speed). I'll look into that.
But that would be a different project than just defining \FPadd and friends.

@PauloCereda Which file are the slides?

@PauloCereda Yes, probably just \def\FPadd#1#2#3{\tl_set:Nx #1{ \fp_to_tl:n {#2 + #3} } }.

@BrunoLeFloch That's interesting too! I was thinking something like generalizing that library functionality (without using it of course) as a standalone tool l3xxx.

3:13 PM
@percusse I don't understand what you have in mind.

@cmhughes Good stuff, but very few introductions.
@DavidCarlisle ...why is there random physics in the middle?

@Canageek slides.tex :)

@Canageek in the middle of what?

@DavidCarlisle small2e

@BrunoLeFloch In other words, yes that would be great! and my comment should read as maybe you can benefit from that library to ease the pain of recoding everything :)

3:15 PM
@PauloCereda I only see one slide there....

@BrunoLeFloch Cool, I love the idea. :)

@Canageek not looked at it this millenium:-) I think the text is same as in the latex2.09 small.tex so ask Leslie L...

@percusse Thanks for the tip. I'll look at that by the end of this week, I hope.

@PauloCereda Oh, there we go. This does look good.

@PauloCereda Any further thoughts about the idea of a make-like Lua-based tool?

3:18 PM
@BrunoLeFloch I swear, I was writing a message for you here about it right now! I have some ideas on what we could do, can I send an email to you later on? :)

@PauloCereda Sure, take your time. I've got way too many ideas right now (LaTeX3 ones, but also writing a C compiler, a Lua virtual machine...).

@PauloCereda I'd add something explaining that \emph stands for emphasis, and show the output. You write a lot of code; I think you should show what it does in the slides.

@BrunoLeFloch Wow! :)

@BrunoLeFloch xor
@BrunoLeFloch xfss, l3ldb, ...

@JosephWright zor
:P

3:23 PM
@BrunoLeFloch ....why are you writing a c compiler? Don't we have rather a lot of those written already?

@Bruno: So far, I have the name of our testing framework. :)
2

@PauloCereda excellent point!

@PauloCereda How goes in the methods section at the end of the paper; the why goes in the introduction.

@Canageek Mostly to learn how compilers work. Also because then anyone can use any language from within TeX, since every language has a compiler written in C. Really, I just need to write an x86-assembly-to-TeX compiler.
But C is funner.
@JosephWright Yep. The C compiler idea is not LaTeX3-related :)

@BrunoLeFloch But then you couldn't run it on an ARM system ;)

3:25 PM
@Canageek :)

@Canageek No, I mean: compile from <any language> to x86 assembly using your favorite compiler, then run that within TeX ;-)
2

I am currently thinking about implementing TeX in go, just as an exercise to learn both better.

@PauloCereda [re the name] which is?
@PatrickGundlach That would be cool. Do you plan to open-source the code?

@BrunoLeFloch Ah, I see. What about langauges that compile to bytecode like Python?

@BrunoLeFloch of course, MIT license or so (but it will be mainly for myself / educational purpose)

3:27 PM
@Canageek The byte-code is eventually interpreted, and the interpreter is written in some language, which can typically be compiled to assembly.

@Canageek: I stopped asking myself why I was reinventing the wheel. Sometimes I think we need some crazy stimuli to develop our skills. That's why I decided to focus on the how part. It's far funnier. :P

@BrunoLeFloch dandelion. :)

@PauloCereda ... how?? ...

@BrunoLeFloch Mainly going from TeX the program (tex.web) to go

3:28 PM
@Paulo well, really, I wanted to ask "why?"

@BrunoLeFloch LOL

... then start factoring out some code

@BrunoLeFloch I thought you couldn't make a static assembly code of a python program?

@PatrickGundlach Are there automated tools for converting from C or Pascal to Go?

@BrunoLeFloch I don't know, but that's not part of the plan, because I'd lose the education part.
but Go has many similarities with Pascal

3:29 PM
@BrunoLeFloch Because it's from dentdelion. :P Literally the lion's tooth. Lion = LaTeX, teeth = packages. :)

@Canageek No, but the bytecode is interpreted by an interpreter. And I believe that the interpreter is programmed in C.
@PauloCereda Nice!

@BrunoLeFloch C and python, yes.

@BrunoLeFloch I'm crazy. :)

@PatrickGundlach So you're planning to write a Web-to-Go tool?

@BrunoLeFloch No, a manual translation from tex.web to tex.go. No web2go

3:31 PM
@PatrickGundlach Ok. That makes sense. Only TeX, or also later engines?

@BrunoLeFloch The hardest part (IMO) would be translating the "strange" Pascal record accessors to Go code

@PatrickGundlach If I understand correctly, Go has good support for paralellisation. Would it make sense to make TeX more paralell. [Disclaimer: I really haven't thought about this much.]

@BrunoLeFloch Well, if I had enough time, I would also integrate PDF(TeX), and nice fonthandling, but I know that I should be happy if one day I could run plain.tex on top of it

I want someone to write up TeX in Haskell.

3:34 PM
@BrunoLeFloch I think that there can by quite some interesting things. For example: calculate paragraphs with different parameters in parallel or such.
.. or dvi/pdf writing should be done in another thread

@BrunoLeFloch Because it is as radically different from C as you can get. The review of Go I read thought it was too similar to C and C++ to really do anything radical

... another possibility would be to preload images in another thread while doing the typesetting of material before the image is used

"Despite the large amount of enthusiasm for language design, modern mainstream programming languages don't fall far from the C tree. The best that Microsoft, Sun, and Apple have to offer are just variations on that theme, with the addition of predictable object models and conveniences like garbage collection. The slim minority of language geeks who have rebelled against bracist tyranny and stumbled over to innovative languages like Haskell and Erlang are doomed to toil in relative obscurity."
[...]
"When I learned that Google was going to announce a new programming language, I was hopeful that the search giant would bring something truly novel to the table. They haven't, but the result isn't bad. Although Google's new Go programming language is yet another take on object-oriented C, it's got some nice features."

@Canageek ooh nice idea! :) I love Haskell. :)

....Worst. Sentance. Ever. "To a solution of 3 in dry THFunder argon stirred at −78 °C was added dropwise tert-butyllithium (1.35 mL of a 1.5 M solution in pentane, 2.02 mmol) over 40 min." ......The writer BETTER not have English as their first language.

3:39 PM

@PatrickGundlach Great! I thought that since TeX had to read the document linearly, parallelisation would not help, but you are right.
@Canageek I've never coded in Haskell yet :(.

@BrunoLeFloch The challenge is to break the massive spaghetti-TeX-code into smaller parts to be parallelized.

@BrunoLeFloch I wonder if you could integrate the multipass system of latexmk into it, and have the second pass start before the first one is done. So the second pass always says one page (or more) pages behind the first, but runs at the same time.

It is really interesting to read the Pascal code. For example the strings are removed from the Pascal code and written to an external file, only to be loaded again during runtime.

@PatrickGundlach ...that sounds painful. Easy to edit the text later without having to go into the source code though, as long as you can document the external file. Could see that leading to really easy localization.

3:45 PM
@BrunoLeFloch I am not really sure if reading the input file and formatting the output file need to be in the same thread.

I mean, a lot of projects do that now anyway, so the people writing the text don't need to understand code.

@Canageek TeX expects strings at fixed byte positions, so no editing ...

@PatrickGundlach .....TeX is written in Pascel?

@PatrickGundlach Just find some really clever translators with the added constraint to use the same number of bytes :).

@Canageek Yes

3:48 PM
@PatrickGundlach I thought it was Web (LaTeX), C (PDFLaTeX) or Lua (LuaLaTeX)?

@Canageek Most of the code is like this (after translating from WEB)
c:=tfmfile^;
qw.b2:=c+0;
get(tfmfile);
d:=tfmfile^;
qw.b3:=d+0;

fontinfo[k].qqqq:=qw;
end;

if a<>0 then begin begin if(a<bc)or(a>ec)then goto 11 end;

qw:=fontinfo[charbase[f]+a].qqqq;
if not(qw.b0>0)then goto 11;
end;

@PatrickGundlach Looks like BASIC with some Cisms added.

@Canageek WEB is a special form of Pascal (hides the fact that TeX is mainly one big while loop)

@PatrickGundlach Ah. I see why no one uses it then.

@Canageek Remove the "La" and you're almost right. LaTeX, pdfLaTeX, XeLaTeX and LuaLaTeX are the LaTeX macro package (written in TeX) using different engines (TeX, pdfTeX, XeTeX, LuaTeX).

3:50 PM
@BrunoLeFloch Ah, ok. So LaTeX3e is all in TeX?

@Canageek Yep. Macros are fun. Actually I realized that my l3fp code looks a lot like data-flow programming.

@BrunoLeFloch ... and the engines are written in C (plus the WEB part of TeX converted to C)

A big part of the work was deciding where (in which order) I'd put the data in the input stream, since I only use macro expansion.

All this great programing and I still can't insert svg images into my documents :(

@PatrickGundlach Except TeX itself, which is in Pascal, I presume.

3:52 PM
@BrunoLeFloch the TeX we use is probably also WEB converted by web2c + kpathsea

@PatrickGundlach That sounds...inefficient. Why not just compile the pascal directly? Doen't GCC support pascal?
Huh "GCC now uses C++ as its implementation language [2012-08-14]" That must have been a big (and controversial) change.

@Canageek There are many things missing: file lookup, cross plattform issues,

@Canageek What about converting SVG to TikZ code?

@Canageek I am not sure if the GNU Pascal would handle the Pascal code Knuth used > 30 years ago

@BrunoLeFloch No idea how hard that is. I want something as easy as when I use PNG. PDF and Postscript are bad since they are page layout formats, so you have to mess around with page size and bounding boxes and such.

3:56 PM
@Canageek pdfcrop should take care of this (PDF size)

@PatrickGundlach Ah. I wonder how much faster LaTeX would run if someone went and did a manual optimization of it?

@Canageek good question, but I wouldn't expect much.

@PatrickGundlach What is that, and what systems does it run on? It just strikes me as really cludgy, since if I'm putting it on the web I can use SVG quite nicely, but I can't use PDF.
@PatrickGundlach How fast does PDFTeX run vs. TeX? Wasn't PDFtex written in C from scratch?

@Canageek If you have a PDF that is paper size but only parts of the page is used, you can call pdfcrop to set the crop box of the PDF. After that the PDF has the size of the material on the page
@Canageek AFAIK PDFTeX is also automatically converted WEB code

@PatrickGundlach Huh. Bug fixing in that must be painful, since it would be undocumented, wouldn't it?

4:01 PM
@Canageek glad it's not my job....

@PatrickGundlach Now there would be a great way to learn a lot of TeX and C. Rewrite all of TeX in C.

@Canageek AFAIK The LuaTeX has converted many parts of TeX in C.

@Canageek Taco Hoekwater did that, I believe

@PatrickGundlach That is downright impressive. Why does no one use his implementation?
@BrunoLeFloch Yeah, but I suspect the speed difference between Lua and C is a lot worse then good handwritten C and automated C. Or does it run noticeably faster?

@Canageek I believe much of his code has gone into LuaTeX

4:06 PM
@Canageek LuaTeX tends to be a bit slower than pdfTeX simply because characters take up more memory. I don't know whether the change from automated C to handwritten C changed the speed. Presumably a reason was to make changes easier.
In particular, LuaTeX provides access to many parts of TeX, but it can probably add such hooks in hand-written code.
I believe that if you run your usual document through LuaTeX, no (or very little) Lua code is run. Lua is only used if you (or packages) call \directlua.

@BrunoLeFloch Huh. Interesting. Why do characters take up more memory if it isn't using Lua?

@Canageek Because of Unicode. TeX and pdfTeX only know about bytes. In particular, there are only 256 catcodes to keep track of, only 256 lccodes, uccodes, etc. LuaTeX (and XeTeX) have to keep track of 100000s characters. They may be clever about storage of the catcode info, for instance saying "the range ^^^^0123 to ^^^^0234 is all letters", but it still takes a bit more time to start.

@BrunoLeFloch Ahhh, so XeTeX would have the same slowdown?

@Canageek Presumably yes. Open a terminal, type pdftex, then <enter>, then \relax (or whatever), then <enter>. This will very quickly go from the ** line to a * line. If you try the same with luatex or xetex, you'll note a delay.

@BrunoLeFloch Now if we wanted it really fast, someone could write a hand-optimized version of it in x64 assembly..... Who needs portability, or maintainability? ;)

4:22 PM
@Canageek TeX itself is so damn fast. Several hundred pages per second is no problem. The problem is the extremely slow macro language on top of it. Implement more functions in the core, and you get back your high speed.

@PatrickGundlach Besides, we have the I/O limitation. TeX per se is fast. :)

@PauloCereda So an SSD would have a huge impact on TeX speed? Or just storing the PDF in memory till the end and outputting it while I'm reading it?
@PatrickGundlach Interesting. However, I've looked at pure TeX, it scares me.

@Canageek Suggesting to write hand optimized x64 assembly but being scared of pure TeX?

@PatrickGundlach :)

@PatrickGundlach The hand optimized assembly was a joke. It is done for video codecs though; it seems C isn't fast enough for real-time decoding yet. The last bastion of speed matters.
@PatrickGundlach Also, there are people here rewriting compilers for fun. I figure one of them might make something awesome I can use ;)

4:27 PM
@PatrickGundlach Well, for a macro language, TeX is not slow.
The problem is that macros are not meant to write elaborate programs. They are good as shorthands in your document, though.
@Canageek Not sure I'm ready for hand-optimized assembly.

@Canageek I don't think so. I think we might end up with the same problem: you have a IO limitation somewhere, even talking about memory. IIRC TeX flushes pages as they are done, and I'm not sure if keeping them on memory will suffice - for our documents, maybe it's ok, but think of some of Patrick's manual, which might reach a +1Gb final PDF.

@BrunoLeFloch Could be, I have no real comparison. But it really adds up. I guess you could give some estimates how much faster floating point calculus would be if taken directly from the processor (by calling c routines)

@PauloCereda I have 4 GB of ram on my laptop. I saw an input file yesterday with the command line %MEM=30GB

@Canageek :D

4:33 PM
@PauloCereda You could definitely be working on page 2 while page 1 is still writing to disk at least.

@PatrickGundlach Basically, TeX fp is 2000 times slower than Python at computing sines.

@Canageek you could but would you want to?

@DavidCarlisle Why not? Would increase RAM usage, but would remove the output bottleneck, which is the whole point of RAM.

@PauloCereda Yes. @PatrickGundlach We had this idea with Paulo to build an analog of make in Lua.

@BrunoLeFloch Is there a lot of floating point math in TeX?

4:34 PM
@Canageek TeX had a state of the art linebreaker in 1982 which is still state of the art but its page breaker was so-so back then and really is rubbish now mainly because it is designed to ship out pages as fast as possible so can make no global optimisations of breaks over a chapte.

@BrunoLeFloch call it "lake"

@PatrickGundlach I sense some inner duck joke going on. :)

@Canageek no, as far as you don't use tikz, pstricks or similar

@BrunoLeFloch I know, I know, I wanted to write a Make clone in Lua as well.

4:36 PM
@DavidCarlisle Right, but it still can't go back pages right? So unless we rewrite TeX to look at several pages at once (which would rock), this wouldn't CHANGE anything TeX is doing.

@PatrickGundlach The big advantage is that with lfs it would be platform independent.

@Canageek in tex-the program almost none, just in glue stretching mostly and any fp calculations are deliberately made inaccessible from the tex macros to ensure portability

@DavidCarlisle Would it be possible to simply keep the whole document as one big box, then \vsplit it at the end?

@DavidCarlisle I mean, right now isn't it doing 'process page 1, write page 1, process page 2, write page 2'?

@DavidCarlisle And portability is one reason I wrote a TeX macro package for floating points rather than for instance relying on Lua to do computations for us (plus, there is the fact that we don't assume LuaTeX).

4:37 PM
@DavidCarlisle Ah, so no harnessing the GPU for us.

@BrunoLeFloch yes but vsplit doesn't have all the feature of the page breaker. notably it doesn't do inserts. latex2e doesn't use insertion much but does a bit so it depends how far you want to go

@DavidCarlisle Are insertions a necessity, or just helpful for some tasks that could be done differently?
If so, we may be able to revive some galley ideas that Joseph/Frank had, by opening one big box for the document.

@BrunoLeFloch one or the other. You can always move the descision maker off tex-the-program in to the macro layer but as noted already in this thread there is a time/complexity price to pay. At some point you have to ask how much of a price you want to pay for sticking on an unextended tex program.

@DavidCarlisle Ok, so a more precise question: what uses insertions?

@BrunoLeFloch in 2e footnotes use the insertion mechanism in a non-trivial way (in particular because footnotes themseleves can split) other float types dont really althoiugh they use \newinsert as a convenience to get a bunch of connected registers

4:43 PM
@DavidCarlisle Ah, so an intersting question is to know if packages which extend the footnote mechanism also use inserts.

@BrunoLeFloch and \enlargethispage
@BrunoLeFloch most do I think as that is the easiest way of getting the page breaker to leave space (and getting the notes onto the correct page, er usually)

I think the discussion is a bit "one way". While I agree that it might be hard to change TeX it is now and keep compatibility, "we" should think in more directions. Using TeX's algorithms to typeset paragraphs (boxes and glue) without sticking to the input language and processing model.

@PatrickGundlach Suggested by Jonathan Fine several times :-)

There are so many PDF generators out there that produce ugly PDFs, TeX would be a perfect drop in for these generators. But the way TeX is now, it is almost impossible to embed this in any program.

@PatrickGundlach There's also the question of whether there are better paragraph breaking methods anyway: there's a link on one the river questions to an alternative multi-variable model for 'goodness'

4:49 PM
@JosephWright I know, and while we can argue about a lot of things, I really agree with him in this point (/me ducks and runs away)

@PatrickGundlach I've said several times to him that the issue is that someone would have to do the non-TeX part, and that's not easy and I see no volunteers

@JosephWright LuaTeX is getting in this direction. I only use the LuaTeX for generating PDF with TeX's awesome algorithms. No more. No format file, no tex files, no sty, no whatever, only Lua code
While this is an ugly hack, it works wonderfully

@PatrickGundlach But the challenge is to write a general input format for this: I suspect your case is rather restricted in terms of input

@JosephWright The problem with TeX is now that it can only be used with the \backslashish input

@PatrickGundlach For example, I asked Taco about the verbatim-in-arguments challenge, and whether ConTeXt was going to move to argument parsing in Lua not TeX. Doesn't seem to be the case

4:53 PM
no other input possible.

@DavidCarlisle Right, the big-unbroken-vsplit model has a big flaw that there is no way (with current TeX) to know on which page a given piece of code ends up.

@PatrickGundlach ConTeXt seems to manage XML okay

@JosephWright my case is restricted to XML.
@JosephWright but it still needs TeX, the input language
There is currently no way to write general applications that use TeX's algorithms

@PatrickGundlach I've not used the XML input mode for ConTeXt, so I'll take your word for this

One needs a ".tex" file and always gets a ".log" and ".pdf" or ".dvi" file out
@JosephWright ConTeXt hides this very well, but context is a massive, huge thing

4:55 PM
@PatrickGundlach It may be possible to take TeX, and change the get_next procedure so that it takes input from some other source.
Then you only need to feed TeX the right stream of primitives.

There is no way to embed TeX in other programs and make, for example, a PNG of the result

@PatrickGundlach As I've said to Jonathan, I have sympathy with the idea of using an alternative input methods, 'TeX as a service' as he calls it, but don't have the necessary skills to tackle such a project.
Of course, he thinks LuaTeX is the spawn of Satan

@BrunoLeFloch But we are still thinking in "primitives" and "macros". While this is good for our input, it is incompatible to the rest of the world (for example word processing tools)
@JosephWright LOL!
@BrunoLeFloch Imagine Libre Office including TeX library and making use of boxes and glue model and using that for aligning things.
This is currently impossible

@PatrickGundlach True, but how do you then for example deal with page building

I think TeX is so great and we miss a huge opportunity to get TeX into other software.

4:59 PM
@PatrickGundlach Or the fact that TeX's paragraph algorithm will alter the whole paragraph on one change

@JosephWright It should be up to the application. Use TeX for paragraph building and do the rest (page building) in your application.

00:00 - 17:0017:00 - 00:00