@samcarter I worry that they'll run into the same issues that have blighted other attempts: 90% of the ideas are easy enough, it's that last 10% that's hard; and critical mass more than anything is why LaTeX succeeds (there are alternatives, some pretty good, but they've got to actually 'win')
@samcarter Also, they don't really say much about their typographic approach: have they for example looked at the things that TeX can't do or are 'LuaMetaTeX + Lua' only ... (rivers, global page break optimisation, other things Frank has listed in the past)
@samcarter Oh indeed: I think many people here can see that a 'modern LaTeX alternative' is attractive, but the challenge is covering all the tricky stuff - I am interested to see if they can manage it with their team size and if they can get enough buy-in from 'community' coders to get the kind package/feature cover that LaTeX has
@samcarter I'm a bit surprised by their approach to tables - doesn't feel that smooth
@JosephWright Same here, I don't like their table syntax at all. I mean, yes, I like the table generation interface (but I like ConTeXt's one even more so). But for manual input, I think even Markdown tables are more readable.
@JosephWright For simple things even HTML tables look clear to my eyes compared to the typst thing (at least when laid out as in the reference/documentation, maybe the community will develop some more readable formatting). And I have never considered HTML tables to be any point of reference for readability :D
@samcarter, @TeXnician (@barbarabeeton) I just tried out numbering equations: you can do it, but again feels tricky and I don't think at the moment it can do selected lines in a set of aligned ones
@samcarter Did you see the comment that suggests that all of that input is the 'right' way to do it? That seems ... odd
@JosephWright yes, I noticed. Maybe it will be better suited for machines than humans? I've seen a post that the upcoming pandoc version will support converting to typst - let's see how well it deals with such things.
@JosephWright so does MRTtable, depends on which styling you need.
(which reminds me of two things I wanted to create: A general purpose well-documented variant of MRTtable that I can release; and a box-based tabular implementation, the latter might very well never happen)
@JosephWright you don't have to, it's an unpublished package of mine that does auto-styling of tables in a particular style (works with LaTeX's tabular and array, and longtable, also key=value based table-wrapper that allows to globally change the position of the caption, supports \label, etc.)
(unpublished is maybe not the correct word, it's just not on CTAN, it is however published and was used by at least 3 other people than myself...)
@JosephWright If you want to skim through its documentation: github.com/Skillmon/MRTbundle (documentation pdf is in the Doc folder)
@youthdoo I haven't looked recently but texmacs has been available for decades and not really attracted any use away from tex. this is not a good forum for an objective comparison from users of both systems though.
@JosephWright -- Thanks for the pointer. I'm following Typst in a desultory manner, not testing, but interested in learning reactions from serious *TeX users. A presentation has been proposed for TUG'23. My questions are whether this is sufficiently easier to use than LaTeX while producing output of equal or better quality to be acceptable for use by math publishers. That's a hard nut to crack.
@barbarabeeton I think 'easier' is going to depend on what people want to achieve. Certainly they've go the abstraction of 'design properties' clean and built-in, so that's a plus. But I'm a bit worried about their 'whitespace important' approach to math mode, and there are a lot of obvious places at the moment where things won't work smoothly
@barbarabeeton I've tested maybe an hour this morning, then tried to work out how to do a table and ... decided to do some day-job work
@JosephWright -- Yeah. I'm trying to work on a TUGboat issue and a TUG'23 presentation. "Free" time is sparse at the moment. But please keep reporting.
@barbarabeeton I plan to keep an eye on the project: it's interesting and like I said earlier, it's not like people don't see there are real issues with TeX's architecture, it's just that a true replacement has to do 'all the stuff', not the easy 90%
@JosephWright There's also a lot of specialized packages that LaTeX so it will depend on how easy package development is and how motivated people are in their respective fields to develop them. And reproducing TikZ and its myriad associated packages is a tall order.
@AlanMunn I think people who find TeX to cumbersome to use will not miss TikZ because they already use Inkscape or similar. So maybe they will more or less go for first-class SVG support instead of getting TikZ/pstricks/metapost equivalents. Given how many programs are able to export SVG today and you “only” have to be able to adjust text/fonts/… to the document, this should cover a lot for the potential user base of unhappy LaTeX users.
@AlanMunn Sure. I use TikZ for many things as well. I just believe that typst's target group (according to all discussions I have read up to now on various sites) is a set of frustrated LaTeX users at the moment. And they will probably not miss what they found too complex anyway.
Especially as with first-class SVG support they could even get both (import from mainstream programs and through means of dvisvgm and similar even simply running TeX documents for the graphics and including the resulting SVG). So for me it's more about how they choose to lay foundations for vector graphics in their ecosystem (or even in the engine itself), rather than about the actual frontend/input language.
@TeXnician I don't doubt that. Anyone who's using R for example is unlikely to care about the existence of pgfplots (I certainly don't) so I can definitely see the attraction. It might also attract LyX users who are already used to instant preview but often find the LaTeX underpinnings difficult.
@mickep but at least they give a name that does not suggest it is related, unlike TeXmacs which is unrelated to TeX and emacs, so has a name formed by combining the names....
@AlanMunn I always find that really odd: R is a processing program, pgfplots is for plotting - the two are separate tasks (no-one would after all say Excel does a good job of producing production-quality plots, but it's fine for a good amount of analysis)
@TeXnician What I think is odd is they've got no EPS or PDF import, only SVG. Most programs I use only do EPS output for vectors, so SVG-only would be a real issue
@JosephWright It's probably really a matter of what the authors are confronted with. I have not used a tool that did EPS export for quite some years. Probably they do hat they have test cases for and rely on “there are converters” for the rest of it until there is more demand by users.
@JosephWright Well technically R is a language... :) But there is a very tight connection between manipulating data and visualizing data, and in R provides extremely sophisticated packages for the latter, so why would I ever need to transform my data from R into something that a different plotting system is expecting.
@AlanMunn I guess I'd always expect to export the manipulated data in a format for re-plotting, but that likely reflects the fact that analysis for me is processing experimental data which uses specialist programs that can do that part, but are awful at plots
@TeXnician Really? I think lots of specialist programs only do EPS ... again, probably shows my focus on experimental data
@JosephWright Ah, well maybe. R certainly processes all of the experimental data we produce just fine, and so it's an extra unnecessary step to put it in a form that e.g. pgfplots would use. And ggplot2 in R is really, really good.
@AlanMunn Likely in part reflects that I've never learnt R properly (similar issue with Python for plots), and partly that the processing I need at a statistical level is trivial (regression fits, basically)
@JosephWright Well, I have a rather theoretical CS background. Most graphics I do are created for visualization of concepts, no exports needed. I only had to deal with a program doing EPS export in an applied CS course at university. And that was about all my exposure to specialist programs doing EPS. So yes, I guess it's really field-dependent and therefore I suspect that the typst devs had no user demand for EPS yet.
@TeXnician For example, I use a piece of software written in Fortran (currently Fortran90 but the code dates from 1969 so ... not originally), which only exports as EPS. It's the market leader in my opinion for the specific graphics it does, so I'm not changing. I suspect SVG output is ... unlikely!
@JosephWright I think the key question for developers of modern tools like typst is how well EPS→SVG converters work for the outputs of such specialist programs. If their output converts well, there's probably less motivation to add a rather arcane language like postscript into the maintenance chain (there are no good postscript parsers I know of in the Rust ecosystem which is their language of choice).
@PabloGonzálezL TikZ' catcode regime is known to be broken under ConTeXt LMTX. ConTeXt MkIV (LuaTeX not LMTX) works fine, so if you need TikZ in ConTeXt, especially involving funny catcodes, go for MkIV.
