TeX, LaTeX and Friends

12:01 AM

@DavidCarlisle -- Probably, but I'll let you handle that. (I'm busy reading/correcting a guidebook on monuments to illustrious figures associated with printing.)

David Carlisle

@barbarabeeton am I in it?

barbara beeton

@DavidCarlisle -- Nope. O haven't seen a reference to anyone born later than about 1920, and even that is only an artist or sculptor.

David Carlisle

@barbarabeeton shocking

barbara beeton

@DavidCarlisle -- The two most frequently represented individuals are Gutenberg and Ben Franklin.

David Carlisle

@barbarabeeton never heard of either of them:-)

barbara beeton

12:09 AM

@DavidCarlisle -- I guess we'll just have to educate you.

yo'

@DavidCarlisle Now what the heck is "POTUS"?

barbara beeton

@yo' -- "POTUS" = "President Of The United States"

Similarly, "SCOTUS" = "Supreme Court ..."

barbara beeton

2:04 AM

5

bp2017

3:51 AM

After browsing packages and running into ptex engine on ctan, I wonder whether it's widely used for Japanese typesetting. By the looks, it hasn't been updated for a while. What do they use to typeset Japanese anyway?

P.S. I thought xetex can be used for any language.

yo'

7:22 AM

@barbarabeeton Ah ok thanks!

BrainStone

7:56 AM

For a package to be submitted to CTAN, I don't need a dtx and ins file, right? A sty and a tex file (plus the generated pdf) or enough, if I understand the docs correctly.

Ulrike Fischer

@BrainStone no, dtx is not required. I have packages without them.

BrainStone

@UlrikeFischer Also is there any real benefit in using them over just having a separate sty and tex file?

Stuffing everythin in one file seems kinda weird to me

Ulrike Fischer

@BrainStone well if you want to document code, it is better if the documentation is near the code and not in some other file. The main problem I had with dtx is that you constantly had to unpack them before you can try if the code works. But with l3build is much less painful, so I'm starting to get used to it.

BrainStone

Alright. Thanks :)
I think I'll take a pass on it for the first time. I find it to be fairly confusing at the moment. I'm sure I'll get back to it at some point.

Paulo Cereda

8:32 AM

quack

@barbarabeeton LOL

David Carlisle

9:51 AM

@bp2017 yes ptex (and uptex) arethe dominent tex engines in Japan as far as I can tell

@bp2017 not particularly good at Japanese I think luatex (and luatexja packages are the main alternative to uptex)

@BrainStone look at latex itself, the user manual is a book, written as a separate document, the dtx files make source2e.pdf the typeset line by line documentation of the source. dtx files are optimised for the latter but for smallish packages it is often convenient to put the user manual at the start of the code documentation rather than have two separate files.

Rmano

@BrainStone you can also look at my question here: tex.stackexchange.com/questions/512982/… --- I finally decided to go for separate .sty and .tex/.pdf for my two simple packages, but the makedtx can help you get started...

BrainStone

11:14 AM

Regarding the bug with the spaces in file names, has that been fixed? And if so when can I expect that being on texlive?

David Carlisle

@BrainStone it's in texlive sources now so you should get it today or tomorrow most likely. Do you really have spaces in your main document file???? That is so weird:-)

BrainStone

Yes I do

Not happy about it either, but that's the way it is

David Carlisle

Jan 21 '14 at 12:33, by David Carlisle

@NicolaTalbot people who put spaces in filenames deserve no sympathy.

That said, it was supposed to work, you need 2019-10-01 Patch Level 3

BrainStone

@DavidCarlisle What do you mean with 2019-10-01 Patch Level 3? I don't really understand...

@DavidCarlisle people who don't handle special chars in filenames correctly deserve no sympathy imho

David Carlisle

@BrainStone when you run latex, the terminal output (and lig file) should start like:

$ latex \\stop
This is pdfTeX, Version 3.14159265-2.6-1.40.20 (TeX Live 2019) (preloaded format=latex)
 restricted \write18 enabled.
entering extended mode
LaTeX2e <2019-10-01> patch level 3

BrainStone

11:18 AM

Ah ok

Just patch level 2

Since i just updated I hope it's live tomorrow

David Carlisle

@BrainStone spaces in filenames make scripting even the simplest thing like "copy the file $1 to zzz.txt " an impossibly difficult thing to do, and that's just copying...

BrainStone

@DavidCarlisle Not really...

I mean I've been scripting for a good decade and the shell handles that stuff quite well if used correctly

David Carlisle

@BrainStone well you can surround $1 by quotes which works most of the time, unless $1 already contains quotes (which incidentally why it fails in PL2) so you have to check if $1 contains any quotes and remove them and then surround by quotes (which is what PL3 does) but doing that everywhere that you ever refer to a filename just so people can use spaces is a real pain

@BrainStone can you show a one line bash script that does that "correctly" and then say how many bash scripts in the wild actually do that?

BrainStone

If I need to expect a file name that might already contain quotes, I do the handling and store it in a new variable and use that as expected.

@DavidCarlisle I have never even seen a filename with quotes being passed to a script

Which is silly

That's why you have the shell

Rmano

@DavidCarlisle ...I use vimtex, and I think that the author went almost crazy doing the "quoting filenames" correct across operating systems... and then people start putting quotes in file names (and I even saw a new-line, once).

BrainStone

11:25 AM

Though in short if you're expecting a filename and it comes with quotes when it shouldn't then the caller messed up

David Carlisle

@BrainStone the reason spaces in the top level file fails in PL3 is because we added quotes everywhere specifically to allow spaces in \input and \includegraphics, but forgot to remove quotes in \jobname, if the top level file is one two.tex web2c tex passes "one two" as \jobname so we ended up with ""one two"".aux and it fails, so need to explictly loop through the filename and remove quotes, first.

BrainStone

Don't get me wrong, I know dealing with that stuff is annoying and I tend to not handle it when working with internal scripts, POC, or WIP stuff. But anything production ready has no excuse not to

David Carlisle

@BrainStone why spaces? do you assert that everything production ready should allow newlines and quote characters in filenames, sure any restrictions have to be documented but most systems have restrictions somewhere.

BrainStone

restrictions are ok. But they need to be documented

And having spaces as a restriction is just ridiculous. That's easily 10% of user files restricted

David Carlisle

@BrainStone I don't see why filenames with spaces are less silly than spaces with quotes. Both seem pretty weird to me. But as I said the issue here was that previously top level filenames could have spaces (as the system quoted them) but in 2019-10-01 every other file could have spaces (as we quoted them) but top level files could not (as both latex and the tex system quoted them) so that was a bug, it happens, but fixing it just made tex slower for everyone

Rmano

11:32 AM

I had real-life filenames generated with single quotes, like example_1.png, example_1'.png, example_1''.png --- no idea if they would work or not, I told the student to rename them (they used the underscore because I told them "no spaces!" before ;-))

David Carlisle

@Rmano single quotes work actually. web2c tex systems can not handle files with double quotes at all, double quotes are always used for space quoting and there is no escape mechanism to specify a literal " character.

Joseph Wright

@DavidCarlisle Well neither can a major OS ...

Rmano

@JosephWright Nor can use ":" in the names, which I think it's even more surprising than spaces (think timestamps).

Joseph Wright

@Rmano Oh, there's quite a long list, it's just that quote marks are notable as they are used to quote filenames for dealing with spaces

@DavidCarlisle We should just use the Secret Plan code ... it already works for this

David Carlisle

@JosephWright The OS can, it just displays them weidly on the commandline

C:\tmp>type ab*ef.txt

abcdef.txt


hello

Sebastiano

11:46 AM

@DavidCarlisle Hi, excuse me for yesterday. I have seen your message last tonight and after I have gone to the bed. I thought that the problem was only on cased and not on another complete .tex document that I then downloaded from the link placed by the user.

David Carlisle

generated by echo hello > 'ab"cd"ef.txt'

Joseph Wright

@DavidCarlisle docs.microsoft.com/en-us/windows/win32/fileio/naming-a-file

The following reserved characters:

    < (less than)
    > (greater than)
    : (colon)
    " (double quote)
    / (forward slash)
    \ (backslash)
    | (vertical bar or pipe)
    ? (question mark)
    * (asterisk)

Sebastiano

In the meantime, good morning, everyone.

David Carlisle

@JosephWright yes I know, but that says that you can't use a " in the cmd but if (eg by using bash) you get a file with such a name, the ntfs filesystem does in fact preserve that name internally, it just scrambles it as exposing it would confuse the cmd bat scripts

Joseph Wright

@DavidCarlisle That doesn't have " in the name ...

Sebastiano

11:48 AM

@DavidCarlisle In short, I understood something else.

Joseph Wright

@DavidCarlisle Oh, are you cheating and not using Windows echo?

David Carlisle

@JosephWright I generated the file using bash so it is there (and the underlying filesystem can cope, it's just the windows command shell that that then hides the quotes as the command language can not deal with the file otherwise

@JosephWright that was a bash echo and redirect yes.

Joseph Wright

@DavidCarlisle I guess I tend to think 'Windows' = "cmd.exe" as that's the official CLI

At least at present, still ...

Ulrike Fischer

@DavidCarlisle what happens if a file name has two consecutive spaces?

Joseph Wright

@UlrikeFischer Then it does: TeX tokenization is a little tricky, but once that's done, it's OK

David Carlisle

12:08 PM

@UlrikeFischer :

$ pdftex '5space\space\space\space\space\space here.txt' \\bye
This is pdfTeX, Version 3.14159265-2.6-1.40.20 (TeX Live 2019) (preloaded format=pdftex)
 restricted \write18 enabled.
entering extended mode
(./5space     here.txt) [1{/usr/local/texlive/2019/texmf-var/fonts/map/pdftex/u
pdmap/pdftex.map}]</usr/local/texlive/2019/texmf-dist/fonts/type1/public/amsfon
ts/cm/cmr10.pfb>
Output written on "5space     here.pdf" (1 page, 10330 bytes).
Transcript written on "5space     here.log".

@UlrikeFischer but I thought you were more interested in hearts than spaces

$ pdftex ❤❤❤❤.tex \\bye
This is pdfTeX, Version 3.14159265-2.6-1.40.20 (TeX Live 2019) (preloaded format=pdftex)
 restricted \write18 enabled.
entering extended mode
(./❤❤❤❤.tex) [1{/usr/local/texlive/2019/texmf-var/fonts/map/pdftex/updm
ap/pdftex.map}]</usr/local/texlive/2019/texmf-dist/fonts/type1/public/amsfonts/
cm/cmr10.pfb>
Output written on ❤❤❤❤.pdf (1 page, 10330 bytes).
Transcript written on ❤❤❤❤.log.

sadly not pink

Rmano

@DavidCarlisle (if I may, @PauloCereda) ooh :-):-)

Paulo Cereda

@Rmano ooh

Ulrike Fischer

@DavidCarlisle worth a feature request for luahbtex ;-)

@DavidCarlisle nice, does it works in \input too? (I can try later). But I think we handled this wrong, we should have setup a github project "enable spaces in file names" and wait for sponsors - @JosephWright would probably already have his mac ;-).

Paulo Cereda

Er... guys...

I have a bug report...

David Carlisle

@Rmano you are 6833 behind @PauloCereda

@PauloCereda It's not my fault

2

Paulo Cereda

12:21 PM

@DavidCarlisle ooh it's close

David Carlisle

@PauloCereda bug is?

Paulo Cereda

@DavidCarlisle I have a filename with spaces in it. XeLaTeX fails with it, complaining there's no \begin{document} in it. Using a new filename without spaces works.

Let me try a MWE.

David Carlisle

@PauloCereda your latex is too old

Paulo Cereda

@DavidCarlisle I updated today...

David Carlisle

@PauloCereda it's still too old:

1 hour ago, by David Carlisle

@BrainStone when you run latex, the terminal output (and lig file) should start like:

Paulo Cereda

12:24 PM

@DavidCarlisle Oh

@DavidCarlisle so it's technically my fault?

David Carlisle

@PauloCereda read my comment just above the linked one.

Skillmon

@DavidCarlisle what's a lig file?

David Carlisle

@Skillmon it's within acceptable typing error of a log file

Ulrike Fischer

@PauloCereda you have a file name with space?? Booooo.

2

Skillmon

@DavidCarlisle is it used to save ligatures?

Paulo Cereda

12:26 PM

@UlrikeFischer <3

@DavidCarlisle :)

David Carlisle

@PauloCereda so which comes first, PL3 or dinner time

Paulo Cereda

@DavidCarlisle oh no

@DavidCarlisle I should have not fixed tex.stackexchange.com/questions/155444/… long time ago. :)

Skillmon

This is pdfTeX, Version 3.14159265-2.6-1.40.20 (TeX Live 2019) (preloaded format=latex)
 restricted \write18 enabled.
entering extended mode
LaTeX2e <2019-10-01> patch level 2
No pages of output.
Transcript written on texput.log.

:( I'm lacking behind :(

Paulo Cereda

@Skillmon /duck hug

Skillmon

@PauloCereda aaww Ɛ><3

Rmano

12:34 PM

So I better /hide...

[romano:~] % pdftex
This is pdfTeX, Version 3.14159265-2.6-1.40.18 (TeX Live 2017/Debian) (preloaded format=pdftex)
restricted \write18 enabled.
**

Skillmon

@Rmano AAAHHH

Rmano

Will create a portable private install somewhere, if I find a) instrunctions and b) time...

Paulo Cereda

@Skillmon oh no

Skillmon

@Rmano tug.org/texlive/doc/texlive-en/texlive-en.html#tlportable

Rmano

@Skillmon yes, thanks. I'll oblige...

Paulo Cereda

12:38 PM

Mar 8 '17 at 11:52, by Paulo Cereda

Q: Why did the L3 chicken cross the road?
A: togettotheothersideohnoiforgotspacesaregobbled

David Carlisle

@Rmano vv

This is pdfTeX, Version 3.14159265-2.6-1.40.20 (TeX Live 2020/dev) (INITEX)
No pages of output.
Transcript written on texput.log.

Skillmon

@DavidCarlisle ooh, a time traveller.

Paulo Cereda

@Skillmon timelord

He's English, after all

Rmano

Just a question: if I install a portable TeXLive in a subdir (I am on Ubuntu Linux), will I be able to keep it completely independent from the distro-provided version? I need to have that, because is what my students/collegues have... I know it's a waste of space, but well...

Or should I setup a different user? (bothersome...)

Skillmon

12:57 PM

@Rmano you should be able to install indefinite independent versions in parallel, I have 2018 and 2019 installed on my system. You then need to change your path to use the different versions, but that is easy, you can just add a symlinked directory to your path and change the symlink if you want to change the version.

Rmano

@Skillmon thanks! I'll try.

Joseph Wright

2:06 PM

@DavidCarlisle, @UlrikeFischer I needed Frank's \SHOWFILES test for siunitx v3 :) (testing PDF bookmarks)

Ulrike Fischer

@JosephWright what does it do?

Joseph Wright

@UlrikeFischer Shows a file :)

@UlrikeFischer Here, I wanted the .out file to check the PDF bookmarks without needing a full PDF-based test

@UlrikeFischer github.com/josephwright/siunitx/commit/…

@UlrikeFischer \showoutput truncates \write lines, so it's not always usable for things being passed to secondary files

Ulrike Fischer

@JosephWright interesting, I didn't know that this exists, looks handy.

Joseph Wright

@UlrikeFischer Yes, I really didn't fancy a binary-based test: this way I just have to verify visually that my Unicode stuff is what it's supposed to be (it is)

Ulrike Fischer

@JosephWright I have been thinking about the bookmarks - I would love to extract the pdfstringdef stuff from hyperref and get it in a separate style.

Joseph Wright

2:10 PM

@UlrikeFischer Yes, that sounds good

@UlrikeFischer Something for l3str in a way: the conversion stuff is there, just need the list of exceptions. l3pdfstr?

Ulrike Fischer

@JosephWright yes, but I think the main work would be to extract the current code from hyperref - I fear it is not done by copying line x to y ;-)

Joseph Wright

@UlrikeFischer I was thinking I can do the test with only one pass: I'll probably adjust to get the ,out file right at end-of-ru

David Carlisle

@UlrikeFischer and make unicode-math stop using \let :-)

Will Robertson

@UlrikeFischer I've changed it in Github already!

@DavidCarlisle Fine by me... just need to detangle the whole thing! :(

Ulrike Fischer

2:25 PM

@WillRobertson does that mean that all the answers I made with Renderer=Harfbuzz are wrong now? ;-(

David Carlisle

@WillRobertson just don't put \ifvtex guards in the wrong place and break everything....

Ulrike Fischer

@JosephWright regarding luahbtex: can I add a new luahbtex tag and make the tag "harftex" be a synonym?

daleif

@DavidCarlisle added to the next memoir version will require the iftex from 2019/11/07

David Carlisle

@daleif thanks

Joseph Wright

@UlrikeFischer Sure, good plan

Ulrike Fischer

2:28 PM

@JosephWright ok lets try, never added a tag before ;-).

barbara beeton

@BrainStone -- I doubt you're trying to script for several wildly different operating systems at the same time, including usually several "legacy" versions of those same systems. It's simply a marvel that there aren't more problems. Consider how many patch updates are issued for those same operating systems; and most operating systems are maintained by large commercial organizations, not a handful of devoted volunteers.

Ulrike Fischer

@JosephWright hm, I think it doesn't work that way. If I want luahbtex to be the "master" tag shown in new questions, I will perhaps have to rename the existing harftex tag and then create a new synonym harftex.

David Carlisle

@WillRobertson \let\fooa=a (or \chardef\fooa=.. or any other non-expandable token would be OK if the "make plain text string/make bookmark" function recognised those and replaced by the character token, which doesn't sound impossibly difficult, but I don't think l3 currently does that does it (@JosephWright) ?

Paulo Cereda

@UlrikeFischer oh no

Joseph Wright

@DavidCarlisle Er, no, not at present, but we could do it for a 'PDF string' function

David Carlisle

2:42 PM

@UlrikeFischer given there is only a finite list of renderer names can't you do a case-insensitive comparison?

Joseph Wright

@DavidCarlisle \str_fold_case:n?

David Carlisle

@JosephWright might be easier than getting every package to specify a pdfstring version of every math character command it defines

@JosephWright tell Will, it's his package:-)

Ulrike Fischer

@DavidCarlisle it is not us, it is the keyval processing of fontspec. Currently it accepts [Renderer=Harfbuzz] but complains for [Renderer=HarfBuzz].

David Carlisle

1 min ago, by David Carlisle

@JosephWright tell Will, it's his package:-)

Paulo Cereda

@DavidCarlisle ooh

Joseph Wright

2:45 PM

@DavidCarlisle Yes, that would likely be best

@UlrikeFischer That's l3keys

Ulrike Fischer

2:57 PM

3

@DavidCarlisle My husband gave me this cartoon ^^. He named it "David, Ulrike and the oberdiek monster"

7

Joseph Wright

@UlrikeFischer Well Bruno and I are writing the L3 replacement bits ;)

David Carlisle

@UlrikeFischer you shouldn't be cruel to monsters:-)

Ulrike Fischer

@JosephWright you are probably currently in the castle ;-)

barbara beeton

@UlrikeFischer -- sigh. Dragons always get the fuzzy end of the lollipop.

Paulo Cereda

@UlrikeFischer ooh das Deutschedragon

bp2017

4:19 PM

@DavidCarlisle, thank you.

I am trying to figure out how \chardef works. In particular, whether it behaves like \newdimen when the same name is assigned a value, creating new register instead of reusing old one.

But after examining what \show prints when used on a \chardef name, I now know that it's not a register. I wonder if \chardef is a macro and something like
\chardef\testChar"54
\chardef\testChar"55
is akin to
\def\testMacro{T}
\def\testMacro{U}

David Carlisle

4:31 PM

@bp2017 it's neither a macro nor a register allocation it is a specific token type a chardef token

@bp2017 in particular it is a <number> the popular \tw@ shorthand for 2 is defined by \chardef for example. as is \active used as \catcode`\z=\active

Alan Munn

4:53 PM

Apparently winter is here.

Paulo Cereda

@AlanMunn ooh

CarLaTeX

5:17 PM

@AlanMunn Oooh snow!

Paulo Cereda

@CarLaTeX Jon?

bp2017

5:55 PM

@DavidCarlisle, thank you. Not sure if reusing \chardef name (either for different or same values) results in unwanted behavior, like accumulation (somewhere) of previous values instead of the values being discarded.

CarLaTeX

6:10 PM

@PauloCereda No, Jon is Snow, it's case sensitive :)

Paulo Cereda

@CarLaTeX it's a regression bug, we should blame someone

egreg

Nice: fancyvrb has been updated and the new version sports \gdef\FancyVerbSpace{\tt }

Paulo Cereda

@egreg uh-oh

David Carlisle

@bp2017 it is a local definition like \def or \newcommand so if done in a group the old definition is placed on the save stack, but that's just general tex behaviour, nothing specific about chardef

cis

Hey, I need some help!

Does somebody have a link to this "LaTeX-Tiger"?

I need that picture.

For a lecture.

@PauloCereda Does only have ducks....

Paulo Cereda

6:21 PM

@cis quack

@cis I think it's a lion

David Carlisle

@cis which tex tiger?

Paulo Cereda

@DavidCarlisle Tigger

cis

@DavidCarlisle Or the lion

I don't know....

David Carlisle

@cis which lion? I was expecting a link

@cis if you mean any tex lion ctan has one with very open useage licene

cis

Maybe, this is a DANTE-picture with this tiger (lion).

David Carlisle

6:22 PM

@cis ctan.org/lion?lang=en

cis

@DavidCarlisle Ahh very good!

Paulo Cereda

@DavidCarlisle ooh wiggly tail

cis

@DavidCarlisle Mr. Carlisle, if you able to read german, you can do my two LaTeX-exams in a style like "Who Wants to Be a Millionaire?"

Part II: https://matheplanet.de/matheplanet/nuke/html/wwm.php?fsid=382

But attention! German language.....

David Carlisle

6:51 PM

@cis @UlrikeFischer will confirm that my German is excellent, strangely on a par with my Japanese.

cis

7:02 PM

Ulrike Fischer

@DavidCarlisle should the grffile stubs handle package options?

David Carlisle

@UlrikeFischer as in silently do nothing with them? Not sure. I'm on a conf call until 8 (here) so only half here

Ulrike Fischer

@DavidCarlisle currently it errors.

David Carlisle

@UlrikeFischer you have write access so I could blame you for the errors?

Sounds like we should add a * option that just accepts them all.

Ulrike Fischer

@DavidCarlisle yes, sounds ok.

David Carlisle

7:15 PM

@UlrikeFischer or we could try harder to emulate some of the options I suppose, but I'd rather not, I'd rather it was a simple step to let people remove the package.

Ulrike Fischer

@DavidCarlisle I was wondering if letting it errors would force people to remove it, but then it would probably better if it errors with a more meaningful message.

David Carlisle

@UlrikeFischer I could do it later, or do you want to do it?

Ulrike Fischer

@DavidCarlisle I can do it, but not now, I have to make dinner ...

David Carlisle

@UlrikeFischer :-) I'll ping you later see where we are. As I say I'm on a mathml call at present

bp2017

7:39 PM

@DavidCarlisle, that doesn't say much to me. What I need is to know whether unneeded values clog up the stack (or whatever place they're kept in) in this case (as they clog up registers when \newdimen is used on the same name). And if so, how can I prevent it or remove them (unneded values)? (I used \ifcsname to check for existing \newdimen name, but I'm not sure what can be done with \chardef's since they're not kept in registers).

David Carlisle

@bp2017 like any local definition the old definition goes on the save stack so it can be restored at end of group.

@bp2017 you can't stop that other than doing \global\chardef

bp2017

@DavidCarlisle, but there is no group. I do \chardef in the same place (in a loop).

David Carlisle

@bp2017 everything might be in a group surely?

bp2017

@DavidCarlisle, what I meant to say is my \chardef's are in the same place (in the same group).

David Carlisle

@bp2017 if it is in a group then each definition causes the old definition to be saved on the stack

@bp2017 but that is true of everything in tex, \def count assignments, ...

@bp2017 what are you doing with the chardefed tokens, using them as characters, or using them as numbers, or something else?

David Carlisle

9:09 PM

@UlrikeFischer I added something to grffile

Ulrike Fischer

@DavidCarlisle you beat my by a few minutes ;-), I was just looking.

David Carlisle

@UlrikeFischer sorry, should have pinged you first, but it's only a line. You will be impressed by my in depth testing in the test3 that I added

Ulrike Fischer

@DavidCarlisle ;-). I hadn't thought about a test file ;-)

David Carlisle

@UlrikeFischer wrote the test first, then made it pass. test driven development... :-)

Paulo Cereda

9:25 PM

@DavidCarlisle ooh

Skillmon

Today I used l3benchmark and learned that \protected macros are slower than unprotected ones. I therefore started to do something really bad: I removed \protected from every internal macro which isn't called in an \edef environment (in this case, every internal macro) and only left the user-level macros \protected. Is this super bad, or only minor bad?

David Carlisle

@Skillmon really, I haven't timed but I'm surprised it's measurable, essentially just testing one internal flag in the generated C code, compared to everything else tex does while expanding a macro

@Skillmon @JosephWright will be cross with you, whether that is bad or not depends on your point of view

Skillmon

test macro \testprotected:
2.53e-8 seconds (0.0806 ops)
test macro \testunprotected:
2.39e-8 seconds (0.0759 ops)
2.38e-8 seconds (0.0759 ops)
test macro \testprotected:
2.53e-8 seconds (0.0804 ops)

@DavidCarlisle both are just \empty but one is protected the other isn't. benchmarking done with \fp_set:Nn \g_benchmark_duration_target_fp { 10 }

Joseph Wright

@Skillmon I'm not entirely surprised: there is a similar trade-off in the way tl data is implemented ..

@Skillmon Minor bad I guess, but it leaves you needing to track any potential 'escape'

David Carlisle

@Skillmon I must admit that normally I ignore the digits and just look at the order of magnitute, but if it took you 15 minutes to restructure your code not to use \protected versions you will have to use it a lot to get that time back...

Skillmon

9:38 PM

@DavidCarlisle just in one package (the insertionsort) for the macros inside the loop.

David Carlisle

@Skillmon ah I guess the inner loops could get run quite a lot in sorting so we may forgive you...

Joseph Wright

@Skillmon For me, the trade-off is worth it

David Carlisle

@UlrikeFischer should I release grffile? ctan will be missing me, I haven't released anything yet today

Joseph Wright

@DavidCarlisle It's a bit like tl being \edef\foo{\unexpanded{...}} though; yes, there is a speed hit, but it's so small as to be worth it for the clearer interfaces, etc.

David Carlisle

@JosephWright in general tl code then definitely less I agree but if you are writing the inner loop of a sorting algorithm and you have full control of all the macros and they are going to get executed thousands of times if used at all, the tradeoffs may be different

Skillmon

9:42 PM

@JosephWright which trade-off? Being faster while someone might trap, or being safer while it takes longer?

Joseph Wright

@Skillmon Yes, plus the fact that saying 'make things work by expansion or \protected' is a clear line that doesn't require careful examination of code paths

@DavidCarlisle Sure, but I suspect we'd have to be talking about tasks that are really outside of TeX's scope ...

David Carlisle

@JosephWright like writing a regex engine for example?

Ulrike Fischer

@DavidCarlisle I see no problem unless you want to give ctan a break ;-)

Skillmon

@JosephWright that was an either-or-question... You can't answer with "yes" :(

Joseph Wright

@Skillmon Well both: being faster makes it risker

David Carlisle

9:52 PM

@UlrikeFischer I could do but seeing all the weird issues being bounced between pandoc and R studio and who knows where else, probably best to make grffile be as silent and error free as possible

Ulrike Fischer

@DavidCarlisle yes, probably a good idea to avoid the error message (but the rstudio problem seems to be due to an outdated oberdiek, so standard miktex problem ...)

bp2017

@DavidCarlisle, here
\def\ii@SaveEachDigit#1%
{ \ifx#1\i@Terminator
\else
\stepcounter{i@SavedDigitCount}
% ---------------------------------
\expandafter\chardef
\csname
i@SavedDigit\the\value{i@SavedDigitCount}%
\endcsname
\the\numexpr
`\0+#1
\relax
% ---------------------------------
\expandafter\ii@SaveEachDigit
\fi
}
what happens if I call this macro 1000's of times and each of them (times) would at least assign to \i@SavedDigit0? Does it mean that there would be thousands of \i@SavedDigit0 macro names in the stack? What happens to them? Do all of them except the last one (which is the…

David Carlisle

{ \ifx#1\i@Terminator should probably be {\ifx\i@Terminator#1 certainly looks safer that way round (and without the space)

Skillmon

@JosephWright so you agree that it is worth it being less stable while being faster, great!

David Carlisle

@bp2017 and \stepcounter{i@SavedDigitCount} should probably be \stepcounter{i@SavedDigitCount}%

Joseph Wright

9:58 PM

@Skillmon No, that's not what I said: in general, for expl3 we've picked stable/predictable if we can; it's not always easy, but it's normally sensible

Skillmon

@JosephWright not my fault, if you only say yes, I pick what I want :)

David Carlisle

@bp2017 no that will not use the stack at all unless it is in a group, in which case the first use of each name would put the old definition on the stack.

David Carlisle

10:14 PM

@UlrikeFischer grffile gone to ctan

Ulrike Fischer

10:33 PM

@JosephWright: l3str-convert use:

%lualatex/xelatex
 \documentclass{article}
 \usepackage{expl3}
 \usepackage[]{qrcode}
 \begin{document}
 \ExplSyntaxOn
  \str_set_convert:Nnnn \l_tmpa_str {Grüße~❤🦆}{}{utf8/bytes}
  \exp_args:No\qrcode{\l_tmpa_str}
 \ExplSyntaxOff
 \end{document}

bp2017

11:01 PM

@DavidCarlisle, so if I enclose the chardef within {} inside the macro (so that its inside a group) and then expand the macro with \ii@SaveEachDigit2019 and \ii@SaveEachDigit2020, the \i@SavedDigit0 (would have value of 2 in both cases) gets put on the stack twice? I don't get it, why put it on the stack twice if there is no way to access it unless \global is used? Do you mean if you put one chardef in one group and another chardef with the same name in another group they would both be on stack?

Otherwise I don't see the point of tracking chardefs (or anything) with the same name instead of overwriting them, unless they are in DIFFERENT groups (so that tex keeps track of each by putting them on stack, so that correct name can be accessed within respective group).

And it doesn't really matter whether I see the point in it or not, I just wanted to know whether additional resources are allocated unnecessarily (like when \newdimen allocates new registers even for a dimen of the same name in the same group) in the case I showed you.

David Carlisle

11:16 PM

@bp2017 I'm clearly not explaining this well,and it's late so I'll let Knuth explain it, see if that's clearer

\ddanger The special case of ``^|save size|'' capacity exceeded is one
of the most troublesome errors to correct, especially if you run into
the error only on long jobs. \TeX\ generally uses up two words of save
size whenever it performs a non-global assignment to some quantity
whose previous value was not assigned at the same level of ^{grouping}.
When macros are written properly, there will rarely be a need for
more than 100 or~so things on the ``^{save stack}''; but it's possible to
make save stack usage grow without limit if you make both local and

@bp2017 if you do \def\z{} zzz {\chardef\z=1} zzz {\chardef\z=1} zzz it gets put on the stack twice but only stack position is used as it is popped twice as well at the end of each of the groups

@bp2017 what are you doing with the chardef, it why \chardef\foo=\numexpr`\0+#1\relax rather than just \let\foo=#1 ?

@bp2017 \newdimen isn't really comparable to \chardef, the former is a tex macro, and the latter is one of the tex primitive definition commands.

Transcript for

$TeX - LaTeX$ TeX, LaTeX and Friends