« first day (5193 days earlier)      last day (27 days later) » 

03:20
Finally got a palindrome:
user image
5
 
1 hour later…
04:40
@mickep Barely able to handle the heartbreak of all my favourites losing at this tournament: Williams, Selby, Ding, and Robertson. The quarterfinals was devastating.
 
2 hours later…
06:41
Quick question: since NewCM is more or less metrically compatible with LM/CM, what's the best way to use NewCM in pdflatex? Can we keep everything same, and just replace the .pfb files with the ones from NewCM, after renaming them? Would that work?
@Skillmoni I see. Thanks
 
2 hours later…
cfr
cfr
08:26
@ApoorvPotnis NewCM doesn't provide .pfbs. it is only available in opentype format as far as I know.
@ApoorvPotnis there is no 'best way' to use it in pdflatex. if you want to use it, use lualatex. if you want to use pdflatex, use cfr-lm.
cfr
cfr
09:18
github question: how can I change github.com/cfr42/nfssext/actions/runs/12842756496/workflow so that it reports failure, but still runs all three jobs?
[example is silly.]
Got to edit proofs in skylatex. Never heard of it before.
10:08
Thanks to @Skillmon, expl3 is working again with LuaMetaTeX :)
10:32
@JosephWright :) l3bigint will need the same treatment as well.
@Skillmon You know what, I was thinking exactly that
@JosephWright also dipped my toes in it a while back to try to understand parts of it so that maybe I can implement the divisions, but I can't comfortably say that I feel ready tackling that, yet.
@Skillmon Yeah, I've looked but as Bruno says, picking the right approach is hard
@Skillmon I'm going to do a dev release in a second so we can have expl3 working again in ConTeXt ASAP
@Skillmon I had thought about taking the approach from bigintcalc, but I have a feeling it's not the most efficient
@JosephWright is ; only special in \numexpr or also in \dimexpr et al.?
@Skillmon All of them, I think, based on a quick test
It's not actually in the LuaMetaTeX manual ...
10:38
@JosephWright and are we sure we found all places in which ; is used as the delimiter of a \numexpr or similar? Is there any such usage in l3draw, l3coffin or l3box?
@JosephWright what is? :D
@Skillmon I think we have them all - certainly enough to load in ConTeXt - the ; stuff is all Bruno, and I wrote most of the code in l3draw and l3coffins , while l3box is all low-level
@Skillmon :)
\startbuffer
\the\dimexpr .4999pt : 2 \relax =.24994pt
\the\dimexpr .4999pt / 2 \relax =.24995pt
\the\dimexpr .4999pt ; 2 \relax =.00002pt
\scratchdimen.4999pt \divide \scratchdimen 2 \the\scratchdimen =.24994pt
\scratchdimen.4999pt \edivide\scratchdimen 2 \the\scratchdimen =.24995pt
\scratchdimen 4999pt \rdivide\scratchdimen 2 \the\scratchdimen =2500.0pt
\scratchdimen 5000pt \rdivide\scratchdimen 2 \the\scratchdimen =2500.0pt
@Skillmon @JosephWright From the just bashed manual...
@mickep so we correctly guessed as much :)
@mickep thanks :) Aside: I would've preferred a Python-esque // integer division (though it adds parsing overhead and would slow down division, maybe it's better that an entirely different character got used)
10:55
@Skillmon That means that // is free to use as the end-thingy in fp... ;)
11:24
@mickep Er, no, as that would lead to an issue at the first /
@JosephWright That was a joke.
11:56
@ApoorvPotnis no that sounds like a disaster "more or less metrically compatible with LM/CM" translates to "uses different metrics to either lm or cm" if you really want to do this use one of the standard convertors to generate type1 fonts and matching tfm metrics,
I see tblr is getting faster:-)
@yo' it might even become usable on overleaf again github.com/lvjr/tabularray/issues/553
4
12:16
@DavidCarlisle wow, I wasn't expecting this.
@dedded welcome to the game :)
Drat, now I lost the game.
@mickep :)
12:44
@DavidCarlisle oh, I meant more or less the same in the following manner. NewCM glyphs are perfectly compatible with LM/CM. The metric changes have been done afterwards, in order to improve the appearance. They can work with the original metrics, and give an output exactly similar to LM/CM, just with slightly thicker glyphs designed to imitate the ink spread
@ApoorvPotnis still sounds wrong to me, I'd generate new tfms especially if you want any document to be portable. but otherwise you probably don't need to rename any files, just set up map files to use them. But I wouldn't.
@ApoorvPotnis but lm and cm are not exactly the same. So one cannot be compatible with both.
@mickep that would be like agreeing with me and @egreg
@DavidCarlisle Even a blind hen...
@ApoorvPotnis But instead of converting fonts back to type1 again it must be easier to get things to work in luatex.
13:07
@DavidCarlisle We only use regex functions once in the whole of expl3 and the kernel - there's a reason
@mickep But in seriousness, it will be good to have things working - and we could think about adding ; to the other engines now
yo'
yo'
@DavidCarlisle Cool!
@DavidCarlisle Unthinkable!
@JosephWright wait, where?
@Skillmon lttemplates, checking for \AssignTemplateKeys - used only at point-of-definition for template code, so once per template implementation
I could code another way, but this is not a major issue and I think the balance of effort here makes snse
@JosephWright I once thought using l3regex at definition time once was fine. The result was that it took multiple seconds just to load ducksay...
@JosephWright depends. How many templates do you reckon will be used per document?
13:16
@Skillmon Perhaps 20
@Skillmon If you look at how long the template test suite takes, almost all of the time is in the data structures (so gets worse as they get longer), not the preamble
@Skillmon We can look to optimise, of course, but I think there are other higher priorities
It's also a simple regex - \c{AssignTemplateKeys}
@JosephWright hm, I must admit I didn't spend even a second in the code of lttemplates/xtemplates
@Skillmon Its when you define templates, so look at the tagging exames
Examples
They are all template based
The lost ones use the most templates
@JosephWright now I'm at 10 seconds (opened and searched for \\regex). Looks like something that \str_if_in could do with a very low error rate...
I just realised, \regex_match_case:nnTF has a wrong argument order compared to the rest of the case-functions.
@Skillmon I’m surprised it’s that much: if so, yes the saying is worth it
@Skillmon log an issue :)
The regrex stuff was originally a stop gap
@mickep are the glyphs, not the metrics, different for CM and LM?
13:30
@ApoorvPotnis Some of them are.
@DavidCarlisle I see. I just want the easiest experimental way of using newcm in pdflatex
@mickep oh I see. I thought that the glyphs are same. Is there any documentation regarding the differences?
@JosephWright Oh, I just was about to suggest to Hans to remove the ; and : from expressions.
@ApoorvPotnis Not that I know of.
Ok. I'll see the fonts themselves
@ApoorvPotnis why?
@ApoorvPotnis some may be similar or even the same, but as far as I understand it is (as latin modern was) essentially a new font just inspired by the original cm design.
@DavidCarlisle because I find CM/LM too thin. NewCM has a slightly thicker book weight. There's mlmodern, but it seems a bit too thick.
In any case, I'm sure the differences between individual glyphs would be very few. I guess most of the stuff would look just fine
13:44
@ApoorvPotnis but why use pdflatex if you want to use an opentype font. As @mickep says it would be far easier to process the document with lualatex than to convert the fonts, especially if you need the document ever to be processed by anyonee els
@ApoorvPotnis why do you say it will look fine? you are worried about often sub-pixel differences in stroke widths, yet you are unconcerned if the metrics are out and the character advance widths are arbitrarily wrong?
@DavidCarlisle oh, that's just a curiosity.
@DavidCarlisle newcm is designed to be metrically compatible with lm, that's what the maintainer told me. there have been recent changes to the metrics, but they are supposed to improve the appearance
@mickep Would be my preference - I'd say the e-TeX primitives really should work unchanged - Hans has after all added new extended expressions
@DavidCarlisle the effect of thickening the glyphs won't distort the look. the effect would be equivalent to printing a document with LM fonts with an ink that spreads a little
@DavidCarlisle i agree that using lualatex would be easy. not disagreeing with that
@JosephWright He will not remove : and ;. But he added the e-TeX mode, so I guess you can just use that. I have no opinions about the other engines.
14:00
@ApoorvPotnis yes but I don't think the font is cm with strokes mechanically thickened, as far as I know it's a new font so (without checking) I wouldn't expect the advance widths or inter-letter kerns to be the same, so if you place newcm glyphs at points calculated for cm metrics the inter-letter spacing can not be guaranteed to be at all reasonable. But it seems a lot of work to make a non portable document so I wouldn't even consider it, but it's a free world you are free to try.
@DavidCarlisle i will try and see if the output is acceptable
@ApoorvPotnis why don't you want to use tfms that match the fonts? (how are you making tex-encoded type1 fonts anyway)?
15:05
latex's default input encoding has changed a lot since that package was written, I may look again at the weekend and see if I can steal a tick from @egreg, he's always so happy when that happens. — David Carlisle 23 hours ago
@egreg ^
 
1 hour later…
16:17
@DavidCarlisle 😛
 
2 hours later…
18:42
@JosephWright I think it's too late to change it, so maybe not?
19:33
@the_guruji I don't care too much about who wins to be honest. And I don't watch it all. Impressed by Shaun Murphy.
@Skillmon isn't it a feature that regex commands have their own argument orders which is why renamed versions exist like \seq_set_regex_extract_once:Nnn == \regex_extract_once:nnN ?
@DavidCarlisle correct, but there's no l3case so you can't provide another name for \regex_case:nn that has the correct argument order.
19:49
@Skillmon depends how you view it, you could say order is correct for regex module conventions and another version isn't needed as their is no suitable module (perhaps)?
 
4 hours later…
23:32
@mickep yes, I just saw his 147 on youtube. i dont have much time anymore to watch full matches either. highlights are thankfully easy to find.

« first day (5193 days earlier)      last day (27 days later) »