Question: does \ref fully expand its argument by default? I think that's likely, but I don't know the definition. Though I gather that \show or similar would give it to me.

@FaheemMitha Try latexdef ref from the command line... It's a very useful little tool! (and no, I do not know if it fully expand its argument...) ;-)

@Rmano Thanks for pointing that out. I get

    \ref:
macro:#1->\expandafter \@setref \csname r@#1\endcsname \@firstoftwo {#1}

As usual, this doesn't tell me much.

@FaheemMitha source2e chapter 62 about ltxref

@FaheemMitha you can continue also... latexdef @setref ;-) and so on--- but @daleif advise is better, I reckon.

Does anybody know when there will be the "freeze" for TeXLive 2020? I would like to target it with some small changes for circuitikz and schedule bigger ones after that...

@daleif Thank you for the reference.

That still leaves me unsure what's happening. I see that

\@setref has three arguments, and there is some expansion going on. For example, #1 gets expanded once, it looks like.

@Rmano That's a good suggestion.

> "Everyone knows that debugging is twice as hard as writing a program in the first place. So if you're as clever as you can be when you write it, how will you ever debug it?" – Brian Kernighan

Hmmmm

@PauloCereda use dark mode, so there are no bugs!

@Skillmonlikestopanswers.xyz ooh

@FaheemMitha It's a \csname construct, so has to expand to text

@JosephWright Does it substitute the argument value into r@#1?

@FaheemMitha It's the standard situation: #1 here is the argument to \ref

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

@FaheemMitha if even ignores \protected while expanding.

@FaheemMitha #1 is the argument value.

Right, I mean that it's expanded to r@ followed by whatever the argument value is.

@FaheemMitha Yes, some text that reads r@<expansion-of-#1>

@JosephWright Yes, that's what I meant to say.

@MarcelKrüger currently luacolor and the color font feature clash - as soon as one loads luacolor the font colors are gone. Did you already thought about this at some time?

@JosephWright it is always a bit surprising how many places something like "perhaps textcomp is not there" can affect.

@FaheemMitha well it answers your question as #1 goes in \csname so will fully expand and give errors if that expansion is not just character tokens.

@DavidCarlisle quack

@PauloCereda breakfast

@DavidCarlisle ooh Frühstück

I mean

@DavidCarlisle oh no

@PauloCereda Learning German for your next trip?

Es ist gut, eine Ente zum Abendessen einzuladen.

@egreg ooh

@egreg I know Hallo, entschuldigen sie! Ich spreche kein Deutsch! Sprechen Sie Englisch? Danke! Ein bier, bitte! I will probably survive. :)

@DavidCarlisle I see. I guess it's time to go back and read about \csname again.

@egreg @PauloCereda @egreg's German is no better than his TeX, what he meant to say was Es ist gut, eine Ente zum Abendessen zu essen.

@DavidCarlisle oh nein

@DavidCarlisle why is there an hyperref update? Did I miss something?

@UlrikeFischer er ...

@UlrikeFischer ctan shows this ctan.org/ctan-ann/id/… which is as expected, did you see a later one?

@UlrikeFischer oh I just got an update via tlmgr, I see. odd..

@UlrikeFischer oh is it because @JosephWright flagged some new hyperref dependencies so the metadata changed...

@DavidCarlisle @UlrikeFischer @yo' (sorry to repeat the question, I feel it was lost in the noise) does anybody know when there will be the "freeze" for TeXLive 2020? I would like to target it with some small changes for circuitikz and schedule bigger ones after that...

@Rmano See tug.org/texlive (at the bottom).

@Rmano timetable is always on this page tug.org/texlive

@DavidCarlisle @TeXnician thanks!

Ah, ok, so it's already gone. Well, no big problem. v1.0 is a nice one to have frozen...

Or the final date for CTAN packages is March, 22?

@Rmano no 28feb: tlnet (and TL'19) frozen, tlpretest starts, CTAN updates continue there.

@Rmano so you have a week to get things into the final version of tl2019

@DavidCarlisle Ok, thanks! Will do a 1.0.1 I think.

I made the decision: I'll see you all in Lübeck!

@yo' YAAAAAAAAAAAAAAAAAAAAAAY

@PauloCereda :)

@yo' oh no, that means I need to practice my Czech!

/quacks in linguistic despair

@PauloCereda pivo, prosím suffices

@yo' ooh that one <3

@Skillmonlikestopanswers.xyz ^^ the bunny police

@JosephWright I implemented a more expkv-esque \keyval_parse:NNn, which should still be equivalent to \keyval_parse:NNn (I hope you have good regression tests :). New timing results below. "Patched 1" is my first patch, "Patched 2" is the new one. Argument used for benchmark: {a,b=c}.

=== expkv: ===
1.07e-5 seconds (33.4 ops)
1.07e-5 seconds (33.6 ops)
1.07e-5 seconds (33.7 ops)
=== Original: ===
1.65e-5 seconds (52.1 ops)
1.67e-5 seconds (52.8 ops)
1.65e-5 seconds (52.2 ops)
=== Patched 1: ===
1.68e-5 seconds (53 ops)
1.67e-5 seconds (52.5 ops)
1.68e-5 seconds (52.8 ops)
=== Patched 2: ===
1.27e-5 seconds (40 ops)
1.27e-5 seconds (40.1 ops)
1.27e-5 seconds (40 ops)

@Skillmonlikestopanswers.xyz Looks pretty good

@Skillmonlikestopanswers.xyz Looking forward to a PR

@JosephWright @DavidCarlisle Btw, I'll be in London (and around) April 20 (Mon) -- 29 (Wed), and I'll be free from Friday onwards. Any chance for sharing a beer? I certainly plan to do some trips around as well

@JosephWright well, currently it just exists in a pretty messy file on my pc, will take some time to get it kernel-ready :)

@JosephWright are there any \__keyval_... functions used outside of \keyval_parse:NNn?

@Skillmonlikestopanswers.xyz No

@Skillmonlikestopanswers.xyz (AFAIK your changes shouldn't affect this, but) have you tried benchmarking some very long key-vals with like 20 items in the clist?

@Skillmonlikestopanswers.xyz The team have thought about a rename as it's an oddity

@yo' my expkv is O(n) up to at least 100 key=val pairs, since this is pretty close to the new patch I don't see why it should be worse. But I'll run a test and show you the results :)

=== expkv: ===
7.34e-4 seconds (2.33e3 ops)
7.39e-4 seconds (2.32e3 ops)
7.33e-4 seconds (2.31e3 ops)
=== Original: ===
0.00133 seconds (4.21e3 ops)
0.00133 seconds (4.21e3 ops)
0.00134 seconds (4.2e3 ops)
=== Patched 1: ===
0.00126 seconds (3.94e3 ops)
0.00127 seconds (3.98e3 ops)
0.00126 seconds (4e3 ops)
=== Patched 2: ===
0.00104 seconds (3.29e3 ops)
0.00103 seconds (3.25e3 ops)
0.00103 seconds (3.24e3 ops)

@yo' ^^^ 100 keys and 100 key=vals. (a,b=c, 100 times, without the last trailing comma)

The new code is somewhat atrocious though. The worst macro definition it has is:

  \cs_gset:Npn \__keyval_end_loop_other:w
      \q_mark \q_recursion_tail
      \__keyval_if_has_equal_other:w ##1 = \q_stop
      \__keyval_has_false:w \q_mark \q_stop \use_i:nn
      ##2
      \__keyval_loop_other:NNw ##3 \q_mark
    {}

And all branching tests are some kind of \gobble_until_q_mark_q_stop:w #1 \ugly_true:w \q_mark \q_stop \use_ii:nn

@Skillmonlikestopanswers.xyz I'm glad I don't have to code many :w macros (this is rreally a word of appreciation!)

@yo' I only use them when I'm after maximum performance.

btw, do you go to Lübeck?

I wish there would be a package PartialFractionDecomposition.sty like polynom.sty .....

@yo' unfortunately the meeting is in the last week before the deadline for my master's thesis, so unfortunately not (see the starred /envious rabbit sounds message)

@Skillmonlikestopanswers.xyz ah! :-( Well, good luck!

@cis now that you see there is demand, go for it!

@yo' thanks. Maybe next year. I'll miss out on TUG in the US as well.

@Skillmonlikestopanswers.xyz too bad. I shall be there as well

@yo' I can't afford it currently, and I don't like the US's border policies.

@Skillmonlikestopanswers.xyz yeah, that's a PITA thing

@cis well, you don't need to call it expl3-something, you can simply just code it in expl3 :)

@yo' I do not think, that "I" can do that.... :()

@cis Well, it would be a nice thing to learn expl3 at; it's an isolated, reasonably well-defined task.

@cis well, you'll want to not have floating point numbers but rationals as long as possible (iirc, there is a L2 package implementing rational number arithmetic, but I could be wrong here)

@MarcelKrüger how do you call verapdf to check the pdf syntax? I just tried with a faulty pdf and get lots of reports that it is not standard compliant but not that it is broken.

@Skillmonlikestopanswers.xyz -- There is the bursary possibility. Unfortunately, we can't do anything about the border policies -- I hate them too; they-re totally unreasonable and unjustifiable.

@barbarabeeton guess I'd still pass, I don't want to cost the TUG money just because I want to visit, flying to the US is quite expensive.

@JosephWright just realized that there will be at least one false-positive regression: I'll have to use \__kernel_msg_expandable_error instead of normal errors.

@Skillmonlikestopanswers.xyz That's OK

@UlrikeFischer Normally verapdf should give a warning while parsing the file, but I normally used it with already syntax-checked PDF files, so I'm not sure.

@UlrikeFischer For pure syntax checking I normally prefer qpdf check <name>.pdf, especially since I use qpdf a lot anyway.

@JosephWright I found a minor inconvenience in the current keyval: an empty key will throw the misplaced-equals-sign error, for which case the help text will be wrong.

@MarcelKrüger not very reliable. I created a clearly faulty pdf, and preflight complains about the wrong operator, but qpdf (and verapdf) let it pass:

\documentclass{article}
\usepackage{l3pdf}
\ExplSyntaxOn
\pdf_uncompress:
\ExplSyntaxOff
\begin{document}
some text
\pdfliteral page {/Span /BMC}
\end{document}

@UlrikeFischer I think especially qpdf does mostly check the outer object structure, at least that's what I used it for. But this certainly needs like we need a different checking system.

@MarcelKrüger Hi, I will meet you in the Dante meeting too, yay!

@PauloCereda YAY! Looking forward to it!

@JosephWright youtube.com/watch?v=1GKsK2e1zcE

@JosephWright ok, I got the following after running l3build check:

  Check failed with difference files
  - ../build/test/m3deprecation002.luatex.diff
  - ../build/test/m3show002.ptex.diff
  - ../build/test/m3show003.ptex.diff
  - ../build/test/m3keyval001.pdftex.diff
  - ../build/test/d3dvipdfmx.uptex.diff
  - ../build/test/m3deprecation002.ptex.diff
  - ../build/test/m3expl001.ptex.diff
  - ../build/test/d3dvisvgm.uptex.diff
  - ../build/test/m3expl003.luatex.diff
  - ../build/test/m3deprecation001.xetex.diff
  - ../build/test/m3deprecation002.pdftex.diff
  - ../build/test/m3keys002.ptex.diff

Haven't checked the diffs yet.

Found one issue. Forgot that \tl_trim_spaces_apply:nN will not trim outermost braces.

@Skillmonlikestopanswers.xyz but that can be solved easily with a quark, no?

@yo' of course, currently fixing it.

@yo' still an annoyance :) Especially since \tl_trim_spaces_apply:nN waste time in the last step to preserve braces and that after keyval wastes time to use it, stripping a \q_mark which is immediately added by the internals of the trimming macro. I'm currently considering writing a \__keyval_trim_spaces:nn routine.

@Skillmonlikestopanswers.xyz oh \tl_trim_spaces_braces:nN #l3poetry

could anyone point me in the right direction for how to render pseudocode? This effort fails badly bpaste.net/VKWA

@UlrikeFischer @PauloCereda @barbarabeeton I got the tickets, I'll arrive on Tuesday at 15:52 by train via Büchen. Unfortunately, I'll have to leave on Friday early morning (at 08:09). I'm staying in Motel One.

ah.. the \to should be to!

@yo' cool!

@PauloCereda it is!

@yo' mind if I ask for a music advice? :)

@JosephWright New benchmarking results of my keyval implementation for expl3 (without using \tl_trim_spaces_apply:nN but instead a adapted copy of it):

=== expkv: ===
1.09e-5 seconds (34.4 ops)
1.09e-5 seconds (34.5 ops)
1.09e-5 seconds (34.4 ops)
=== Original: ===
1.66e-5 seconds (52.2 ops)
1.66e-5 seconds (52.3 ops)
1.67e-5 seconds (52.6 ops)
=== Patched 1: ===
1.65e-5 seconds (51.8 ops)
1.66e-5 seconds (52.3 ops)
1.64e-5 seconds (51.9 ops)
=== Patched 2: ===
1.03e-5 seconds (32.4 ops)
1.05e-5 seconds (33.1 ops)
1.04e-5 seconds (32.4 ops)

@Skillmonlikestopanswers.xyz Pretty impressive

It's beating expkv currently (next release of expkv will most likely be faster again -- and downward incompatible)

@JosephWright wait till you see the code :)

@JosephWright will have to build the new changes into the dtx, then run the checks. We'll see.

@PauloCereda go for it!

@Skillmonlikestopanswers.xyz Looking forward to it

@JosephWright it's passing all the l3keys tests except for m3keyval001 :)

@Skillmonlikestopanswers.xyz Er, good

@JosephWright and that's just the expandable error message.

@JosephWright github.com/latex3/latex3/pull/678

You will now be preparing Japanese omelets in a Japanese omelette pan!

I prepared myself well for this task and thought everything through from A to Z in various thought experiments.

@Skillmonlikestopanswers.xyz I'll read over your code myself and ask @DavidCarlisle to too

@JosephWright thanks.

@cis are there any demons involved in your thought experiments?

@Skillmonlikestopanswers.xyz Oh, I don't know. I have e.g. forget why I ordered these silicone pliers (3rd picture). I only remember that they occurred in my thought experiments.

@cis I guess as chop-stick replacement.

@cis -- 'Oh, these are yummy, and very pretty as well.

@JosephWright are quarks \begin{macro} or \begin{variable}?

@Skillmonlikestopanswers.xyz Variable

@barbarabeeton I never eat Japanese Omelette, so I am curious.

@cis -- So that's how they get such nicely squared edges. These omelets look rather like eggy rectangular cousins of croissants. They're slightly sweet, and are the non-rice component of tamago sushi. (We consider tamago as almost the dessert course.)

@JosephWright I tried to replace \quark_new:N with \scan_new:N, but now m3quark001 is failing, is that expected (duplicate scan marks error changed to include keyval, I guess it is expected...)?

@Skillmonlikestopanswers.xyz Did you give them a private name?

@JosephWright of course. \s__keyval_nil, \s__keyval_mark, \s__keyval_stop and \s__keyval_tail

@Skillmonlikestopanswers.xyz Ah, right, that is expected yes