TeX, LaTeX and Friends - 2018-10-04

« first day (2895 days earlier) ← previous day next day → last day (2040 days later) »

2:01 AM

If I use \read<#> to \sth and it works for the all but the last instance where it gives you an error in the log, and running from the console requires me to input \sth by hand, namely '\sth=<blinking prompt>'. Hitting return or entering anything to make it continue yields the whole code, building in sublime on the otherhand, i.e., not in the interactive console, breaks off at and I don't get the rest.

From the TeXBook 'If the hnumberi is not between 0 and 15, or if no such file is open, or if the file has ended, input will be from the terminal' - however, the log says \periods=\read3

2:18 AM

Okay, so if you read off of a file and test with \ifeof, it'll be true only once, after which it'll give an error or prompt you to manually input.

2 hours later…

3:51 AM

Hey @CarLaTeX, what's going on? I think today I am not able to play Pinturillo because I am doing homework :(. I'm sorry, I promess you we would play another day c:

1 hour later…

4:56 AM

@manooooh No problem <3

2 hours later…

Harald Hanche-Olsen

6:33 AM

@DavidCarlisle Indeed that's what I thought, but then I could not find that bit of information when I went looking for it. Thanks for confirming it. I guess if I do this, it may be best to use just one kind of float consistently, and always to use \captionof – even it that loses some flexibility.

@HaraldHanche-Olsen well it's the same if you use longtable or [H] or other non-floating inserts, so mostly people manage, but it's something to watch out for

@vlg why are you reading off the end of the file? The intention is to do the \ifeof test first and only \read if that is false.

2 hours later…

9:05 AM

@HenriMenke oh I follow both. :) I just followed samcarter because samcarter8 started to follow. :)

@samcarter ^^

9:27 AM

@PauloCereda after I saw the arara readme I got jealous and added badges here too: github.com/u-fischer/luaotfload ;-) (the other repos will have to wait, no time now ...)

@UlrikeFischer ooh :)

@PauloCereda You're just a bit early :) I recently gained access to this user name and am currently considering moving my repos there. Just have to figure out what the best method is to do this.

9:42 AM

@samcarter ooh stealth ducks

1 hour later…

11:00 AM

@DavidCarlisle It supposed to \read from the file not every time, because because value would get overwritten. However, there is a bool 'isRead' in place, and even putting in an extra `\ifeof#' in its false branch before doesn't work, it still wants manual input.

It's weird because there's an even number of rows in the to-be-read file, and values always get read in pairs

Putting the read value in the last cell of the row before it breaks yields the correct value, too

@vlg impossible to guess what you mean, Make a proper example and post a question on site, but you never need to read past the end of a file as you can always do \ifeof\zz something \else \read\zz to \foo\fi

11:23 AM

Hi everyone.

nvm

I was explaining some TeX commands and I used verbatim for that. Now, if I want to explain verbatim itself, this is not possible:
\begin{verbatim}
\begin{verbatim}\end{verbatim}
\end{verbatim}

Can I escape this?

11:36 AM

@ThomasWeller Just write it out 'by hand', e.g. \textbackslash\{verbatim\}

@DavidCarlisle May I please ping you separately for a quick non-TeX question?

@ThomasWeller verbatim package has something for that (I forget the exact syntax)

@Krishna here or you can find my email easily enough (don't promise an answer though:-)

11:56 AM

@DavidCarlisle thank you. Sent you a message in a private chat. I can also email if you prefer that.

Yes, \textbackslash kinda works, but looks different than all other commands now.

@ThomasWeller for your exact example you could use \begin{verbatim*}, but if you have space they would be visible. In general the best is to use some \verbatiminput command, or eg. listings.

I'm experimenting with \begin{alltt} now

Maybe I should have use the listings package in the first place

12:11 PM

@DavidCarlisle, @egreg A question needing the e-TeX \leavevmode definition :)

0

Q: Using a macro with parameters in \everypar fails

Out of curiosity, I wonder why this fails: \documentclass{article} \usepackage{xparse} \NewDocumentCommand{\bla}{mm}{#1.#2!} \begin{document} \begingroup \obeylines \everypar={\bla} {C}{G} skdjfs;ldjkf;sjkf {ERT}{ERTERT} djkfsjdfl;ksjdf;l \endgroup \end{document} with the error: Argument ...

@JosephWright did you put the Op up to it? :-)

@Krishna got it thanks (note that chat on this site is always public though)

@DavidCarlisle alright. Then, maybe I shall move it to email. I didn't realise that

@DavidCarlisle :)

12:34 PM

@JosephWright voted!

1:09 PM

@JosephWright luatex tests die on travis again I see :(

2 hours later…

2:52 PM

@DavidCarlisle Drat

@JosephWright the old duplicate font name issue it looks like.

@DavidCarlisle Oh great :(

@JosephWright I need to run it locally to see what's changed, I ran a full build here this morning and it all passed, so doing a tlmgr update to see what comes up...

@DavidCarlisle perhaps the change in font.each()?

@UlrikeFischer I assume so but where's that changed has that been pushed to texlive already?

2:59 PM

@DavidCarlisle yes, it is in the fontloader. Hans added something after the discussion with Marcel on the list and it got naturally pulled when I pull the other changes.

@UlrikeFischer so perhaps it's now safe to re-insert the code used prior to 1.07 to reset the fontname logic

@JosephWright @UlrikeFischer I'll try a check run this evening

@DavidCarlisle Cool

@JosephWright and we can blame @UlrikeFischer

@DavidCarlisle Sound good

3:13 PM

user image

4

@DavidCarlisle I'm already running ...

@DavidCarlisle I'm very surprised of you. :-P That silly comment costs you at least an upvote.

@egreg I have too many upvotes already

@egreg I'm sure I can find a display math env somewhere that's not inside $$ :-)

@DavidCarlisle Not even the plain TeX ones.

@egreg and all the answers on site here suggesting \item $\begin{aligned}[t]... ?

@DavidCarlisle Well, that would be inline math mode, wouldn't it?

3:23 PM

@egreg you might say that, a reader might wonder why \sum etc are in display mode but the mathrel spacing is cramped

@DavidCarlisle Assuming the OP really wants to go on with that heresy.

@egreg meanwhile people have such rude question titles tex.stackexchange.com/questions/453824/…

1 hour later…

4:28 PM

@DavidCarlisle You never answered why \ineof doesn't work.. Having your suggested \ifx\readstufff\endpar to \booltrueinstead of any ifeofs works somethow

1 hour later…

5:40 PM

@vlg sorry I thought that was clear, you only did an ifeof test before the first read so that one never runs off the file, but you left the second \read unguarded (I could have used an ifeof there instead of of the ifx endpar test

@vlg I hadn't seen your comment under the answer (was driving home then eating:-)

However, if they get read in pairs, calling the func 3 times would result in 6 reads, with 6 rows, wouldn't that be the end of the file or is the new line the end?

Does \ifeof only check if the new line is a \par, or? Why do you know so much about TeX??

@vlg no eof is true _after _ the final line (which becomes \par) is read.

@vlg I have use TeX a few times before.

a few you say

@vlg actually very few times this century other than answering questions here.

@vlg I can't remember the last time I actually wrote a tex document for my own use. Probably only two or three since I moved here (in 1998)

"you see, back in the stone age, we didn't have no computers- no, no. back then tex ran on stone, and was programmed entirely with sand- don't ask how"

So do you help out others for fun? How has your expertise or knowledge not subsided with time?

Are you secretly studying to become a texgod and denyign that you do to seem cool :D

5:56 PM

@vlg yes more or less, mostly the answers don't change, I just recycle answers that I could have posted to comp.text.tex in 1995.

3

6:07 PM

@vlg @DavidCarlisle is a TeXgod already

@DavidCarlisle e-TeX ;)

@DavidCarlisle \pdfstrcmp

@DavidCarlisle LuaTeX (but perhaps 'blame Hans' was valid then too ;)

@JosephWright proves my point I hardly ever use such newfangled things in my answers:-)

@UlrikeFischer On \leavevmode: yes, the specific example doesn't need \protected, but the definition I've used is only generally safe with it. (I'm pushing the team for a change, but not everyone is keen ...).

his beard and overall countenance give it away, he's too nice AND knowledgeable

@DavidCarlisle ^^^ :)

Is there an easy way to test how inefficient your commands are defined? Like cycletime or something in the log?

6:12 PM

@DavidCarlisle, @UlrikeFischer (@egreg) Would it be worth me posting a question/answer about \leavevmode and the fact that the DEK definition leaves a token after \everypar, which can be solved with e-TeX? It's rather esoteric but it's come up recently ...

@JosephWright unproven assumption that my avatar gives more indication of my appearance, than say yours

@DavidCarlisle :)

@DavidCarlisle Quack

@PauloCereda ^^^

@vlg \tracingall :)

@vlg use \tracingall and count how many lines of log isn't bad (but don't try that with tikz or beamer

@vlg You can use the timer macros in pdfTeX (etc.)

@JosephWright group-think

6:13 PM

@DavidCarlisle I could have gone for \loggingall I guess :)

just stick* it in somewhere, or?

@DavidCarlisle Probably we should release l3benchmark ...

@vlg or make a loop that repeats the code enough times that you can use clock time, here for example comparing 0 and \z@ it wouldn't help to use \tracingall as neither expand but:

84

A: What does \z@ do?

latex.ltx says \newdimen\z@ \z@=0pt % can be used both for 0pt and 0 so as it says it is short (and efficient) way of getting 0. You should always have a copy of the latex source file latex.ltx in a text editor window while reading package code:-), or perhaps, if you prefer, the typeset versi...

@DavidCarlisle cl.cam.ac.uk/~rf10 :(

@JosephWright they finally noticed that he retired?

6:22 PM

@DavidCarlisle Looks that way, or perhaps (like us) they've retired support for personal pages entirely

@JosephWright the old code to preserve luatex font names doesn't work. I may leave it to look again at the weekend or I may need to ping @UlrikeFischer as I'm not sure I understand the luatexfont table well enough anymore

starting to wonder if I should've input all the file for the tracing all.......

@DavidCarlisle Sounds about right :)

@vlg it makes you realise how fast TeX is, when you see what it's doing:-)

yeah, I should forgo all the fanciness of xelatex, and go for the grandaddy of them all

A scaling factor supposedly for some guy with a laptop, of 25, using other engines than the base just makes it run moreso like molasses. After how many minutes should I call it quits?

6:48 PM

worst is not know how much more time could be left.. log's at 162mb already :D

6:59 PM

@DavidCarlisle wouldn't it be easier in the long run to load the 10pt first in any case so that all engines show this instead of trying to tweak luatex?

@UlrikeFischer I don't see how that can help, if you load the same font under two names luatex shows one name and everything else shows the other, so if you load the fonts in the other order the problem doesn't go.

@JosephWright ooh a duck

@UlrikeFischer try this:

\font\z=cmr10
\font\y=cmr10
\tracingonline2

\setbox0\hbox{\z abc}
\showbox0
\bye

pdftex shows \y and luatex shows \z

@UlrikeFischer the lua font callback code in latex2e/support/test2e used to adjust the ids so that luatex reported the same as pdftex but I can't get that to work now

Should I just compare the log files with and without tracing all? Never mind, I should compare the size of log files with different definitions

@DavidCarlisle I meant something like this:

\ifcsname directlua \endcsname
\font\z=cmr10
\font\y=cmr10
\else
\font\y=cmr10
\font\z=cmr10
\fi

\tracingonline2

\setbox0\hbox{\z abc}
\showbox0
\bye

7:13 PM

@UlrikeFischer blurg I suppose so but to make a generally usable test2e (or regression-test.tex) I'd have to do that in the test suite for every font that might be loaded under two names, so every font that might be substituted by NFSS fd file handling

@UlrikeFischer prior to 1.07 it "just worked" and luatex reported the same names as pdftex on any tfm based testfile

@UlrikeFischer but you might be right that doing that is less intrusive, and a better test than installing a non-standard font callback (@JosephWright)

@DavidCarlisle Yeah, does sound safer: no invasive Lua changes

@DavidCarlisle A bit of an effort but probably a one-off

@UlrikeFischer @JosephWright actually I'm not sure I can make that work:

\font\z=cmr10
\font\y=cmr10
\tracingonline2



\setbox0\hbox{\z ab \y ab}
\showbox0
\bye

pdftex shows \y (the second name) for both but luatex shows \z for the first two and \y for the second

@UlrikeFischer so I'm not sure I can arrange to make the declarations such that I get the same log for both

@DavidCarlisle grumph ;-(.

@DavidCarlisle where is the code you used to reset the font name?

@UlrikeFischer test2e.tex but it is:

% disable this for now on development builds of luatex
\ifx\directlua\undefined\else
\ifnum\luatexversion<1080
\directlua{
local original_fontloader=font.read_tfm
if (luatexbase==nil) then
  callback.register('define_font',latexDefineFont)
else
  if(luatexbase.in_callback==nil) then
    error('update luatexbase')
  else
    local cbl=luatexbase.callback_descriptions('define_font')
    if(cbl[1]\string~=nil) then
      original_fontloader=luatexbase.remove_from_callback('define_font',cbl[1])
    end

@UlrikeFischer the version test is 108 in the checked in version it's 1080 in the above to switch it back on

@UlrikeFischer but despite that if we concentrate on getting the latex2e/base to pass and not worry about any other l3build tests giving teh same output with luatex then you are probably right that some combination of pre-loading tfm fonts in a different order and a limited number of luatex.tlg files would get the build passing again.

7:35 PM

@DavidCarlisle when I enable the code I get \z for both luatex and pdftex. Doesn't it work for you?

@UlrikeFischer I thought it didn't, let me try again ah hang on, which luatex do you have?

@DavidCarlisle I tried with texlive and the dev (something 1.9).

I just remembered I reverted to 1.07 (stock texlive 2018) (I was thinking I had 1.09)

@UlrikeFischer let me try again..

@UlrikeFischer oh works. Now why didn't the test suite pass, OK I'm going in circles I'll remove those extra print again and try rebuilding the test suite format.... Thanks:-) (can I still blame you if it fails again)

8:04 PM

So, 95% of the log is just the stupid math in place to decide how many empty rows a row with text(that is, a p/m type column) in a certain cell would take; facilitated disgracefully by

`\setbox1=\hbox{\noindent#8}% `
`\pgfmathparse{int(ceil(divide(\wd1,(28.4*\myd))))}`

@vlg Just do it using classical TeX dimen work?

Division by 1, multiplication a \myd by 28.4, sure, but a \ceil (Rounds x up to the nearest integer)?

It spits out an int that only once goes beyond 5, I gues I could compare the result to a counter, and when it's above the counter take the value of the counter plus 1?

@vlg don't use odd numbered boxes for local assignments, also \noindent isn't doing anything in the first line and you don't need pgf for the second, just something like \numexpr\numexpr\wd1\relax/\numexpr28.4\myd\relax\relax

@vlg actually not quite the above but some combination of dim and num arithmetic woud do the right thing:-)

I'll try and suss it out, for sure it'd be better than pgf

@vlg why don't you just set the text in a box of the right width and see how many lines it takes rather than setting it in a hbox and then guessing that it will break to a certain number of lines based on its natural width (ignoring any strtech or shrink in the white space) ?

8:17 PM

Why no odd-num'd boxes?

@vlg because you will trash the save stack and run out of memory. (see texbook, or we must have an answer on site somewhere. registers 0-9 are for scratch use but even are local odd global

@DavidCarlisle quack

@DavidCarlisle The box isn't getting printed, all this is internal, just to get the correct number or rows.

Apr 15 at 9:38, by David Carlisle

@PauloCereda dinner!

@DavidCarlisle oh no

8:19 PM

@vlg so as I said, why not ...

@DavidCarlisle you are mean

Back in March this was the only thing that worked, I don't remember clearly where I saw or thoguht of the invisible box of infitinite length divided by column width thing

Jun 29 '17 at 16:15, by Paulo Cereda

@DavidCarlisle you are not mean :)

@DavidCarlisle oh

And I don't get how all the boxing commands works, tex by topic and the texbook weren't too helpful, maybe I'm too low iq for this stuff.

Thought of a using a mono font, counting how many chars take up one row, get str length, and divide. Basically the same thing, but perhaps slightly less math than the pgf versoin.

8:24 PM

@vlg this shows you the first text takes 1 line if set to 3cm and the second takes two, no calculation required

\def\foo#1{%
\setbox0\vbox{\hsize=3cm
#1\par
\xdef\zzz{\the\prevgraf}}}


\foo{abc}\show\zzz

\foo{a a a a a a a a a a a a a a a a a a}\show\zzz

\bye

It says \bye is an undefined control sequence, oh well, guess you forgot to define it, texgod.

@vlg plain tex not latex (the body of the code would work in latex as well but I didn't bother with \documentclass etc so posted it as plain.

yeah, I'd never run tex, to be honest, it doesn't know what a docclass is. Interdasting

@vlg in plain tex you just need hello world\bye no preamble or other setup needed.

@vlg Demos of TeX fundamentals are normally given in plain

8:31 PM

Very* neat.

@vlg well not really as along with no preamble is no customisation, there are no standard commands to change the font size or font families or have automatically numbered sections etc. It is really a very basic example format useful for, er showing examples...

and calculating how many rows sth takes up. it's wonderful, xmas came early

@vlg actually apart from \bye I didn't use any plain tex commands there, all the rest are tex primitives.

the former* being more primordial, right? - I have no idea what I'm talking about, I'll be uqit

@vlg tex-the-program implements a certain number of typesetting primitives and a macro language. You can use the macro language to define more commands and save them in a format. plain tex is one format, and latex is another, and context is sort of a third but not really just a format any more. so formats like plain or latex are written in tex, whereas tex itself is written in pascal.

8:41 PM

@DavidCarlisle ooh Wirth

Woah, I remember I had turbopascal in school, and those are related.

@vlg yes noone has a pascal compiler anymore so tex is compiled via web2c which converts "web"==documented pascal to c for use with a real compiler.

@DavidCarlisle /ahem

[paulo@ravenna ~] $ fpc
Free Pascal Compiler version 3.0.4 [2018/03/06] for x86_64
Copyright (c) 1993-2017 by Florian Klaempfl and others
/usr/bin/fpc [options] <inputfile> [options]
 Only options valid for the default or selected platform are listed.
 Put + after a boolean switch option to enable it, - to disable it.
  @<x>   Read compiler options from <x> in addition to the default fpc.cfg
  -a     The compiler does not delete the generated assembler file
      -al        List sourcecode lines in assembler file

Quack. :)

8:57 PM

@PauloCereda people don't. I can't say about ducks.

@DavidCarlisle ooh

@PauloCereda so, can you compile tex without using web2c?

@DavidCarlisle never tried. :)

@PauloCereda I'm shocked:-)

@DavidCarlisle :)

9:01 PM

freepascal.org

@DavidCarlisle Yes, this one. :)

9:14 PM

@PauloCereda monperrus.net/martin/…

1 hour later…

10:37 PM

@PauloCereda If you find @DavidCarlisle's not interesting enough, here is an alternative. (Just kidding;-)

@marmot since the hypothesis is clearly false....

@DavidCarlisle Yes, I wanted to write link after @DavidCarlisle's but somehow my claws didn't do that.... and you never know .... maybe some like free marmots better than free pascal ... ;-)

Would it be worth it trying to get an integer by doing a \dimexpr, advancing some count by it, and trying to deal with the left over floating ploint and dimension?

@DavidCarlisle I'm trying out different bad methods, yours, the best one likely, will be the final test, don't worry.

@vlg sounds about like the normal thing to do, most tex arithmetic goes via dimexp somewhere, but get an integer from what? all pgf arithemetic or scaling via graphics package for example uses dimensions

10:52 PM

@DavidCarlisle If you multiply by 0, you get an integer for sure. ;-)

@marmot well that depends on your type system..

But how do you get rid of the '.123131pt' that inevitably get printed?

@vlg latex defines \strip@pt for that reason

Without cheating by using some \StrBefore{\dimensions}{.}[\nodimensions]

these people have thought of everything, jesus

@vlg eg the \selectfont uses \edef\f@size{\strip@pt\@tempdimb}% so that \f@size ends up as 12 not 12pt

@vlg oh but I misread I think, strip@pt just gets rid of the pt to hide the fact that dim arithmetic was used, if you want to round then cast to a count register

11:00 PM

Cast? I'm no fisherman.
I tried \advance-ing an assigned counter, but and it breaks off at just before decimal point, so the rest would get printed

Turning a num into a dimen is so simple, yet the opposite not so.

@vlg \count0=\dimexp5.5pt makes it a number then just divide by \p@

Woah, it works.

@vlg don't sound so surprised:-)

Only becuase it's from you..
Makes me work how you can sort a 5.5pt in a count, however

I have done zero things that involve an @, so it's all arcane to me

@vlg unfortnately etex numexpr / is rounding div which isn't usually what you want for integers but if you use \divide it truncates (rounds doen) eg this gives 7 rather than 6

*\showthe\numexpr \dimexpr 6.6pt\relax / 65536\relax
> 7.

11:09 PM

nah, that's fine, I need a rounding up anyways, I'll add 1 and finito with it

@vlg all tex dimensions are just integers really in sp units. so going from a dimen to an integer is just expressing the length in sp where 65536sp=1pt

I thought pt was small!

@vlg tex uses (almost) no floating point arithmetic in its internal line and page breaking calculations it's all integer arithmetic in sp units done that way to get bit for bit machine independence at a time before ieee arithmetic when floating point models on different architectures were very different and could not give consistent results

user image

a giggle

@vlg beware the placement of % some are doing nothing some are wrong and some are Ok in that picture

11:23 PM

I'm paranoid so I put 'em everywhere. I've never seen them in on lines with IF, ELSE or FI, and some basic others

Could you elaborate where they're not okay and maybe why?

@vlg the % after numberic literals like 0 or 1pt are wrong they will force evaluation of code in the wrong order, the ones after command names like \p@ do nothing.

Got it, thanks

@vlg try:

\count0=5%
\ifnum\count0=5 \show Y \else \show N\fi

@vlg then try it without the %

more arcane magic!

@vlg no simply that the scan for a number is terminated by the white space or end of line, so if you comment out the eol then the number hasn't stopped:

count0=1%
2

@vlg sets count0 to 12, without the % it would set it to 1 and then typeset the 2.

putting a % at the end of every line is a really bad idea:-)

11:32 PM

neat, tested it myself, gotta go through everything I've ever written and check for bad ideas

11:44 PM

@DavidCarlisle other than \p@ are than such unit conversion for other units/dimensions?

« first day (2895 days earlier) ← previous day next day → last day (2040 days later) »

Transcript for

Oct '184

$TeX - LaTeX$ TeX, LaTeX and Friends

\begin{chat}

join this room
about this room

00:00

06:00

12:00

18:00

all times are UTC

$TeX - LaTeX$

site design / logo © 2024 Stack Exchange Inc; legal

mobile