Editing the C Standard
thephd.github.io
thephd.github.io
Do you remember what gcc errors were like just as clang got popular? They were pretty bad before the c compilers started competing to have nice ones.
Those are wonderful in comparison to the errors in LaTeX.
Fucking extra newline breaking equation* and causing nonsense.
In general complaining in such terms about a project you've just taken over doesn't seem very professional.[2] We've all been there, done that, etc. LaTeX diagnostics are atrocious. All received large projects are filled with decisions one wouldn't have made oneself, and consequences one would have avoided. But it pays to remember that any senior engineer has presumably left in their wake several large projects that befuddled and frustrated subsequent maintainers, both because of regrettable reasons, yet also for reasons that one would readily defend.
But to suggest Microsoft Word as being a superior option... I can't even begin to fathom such a comparison. I can imagine, however, how insanely frustrated and pissed I would be if I had to take over the task of managing a huge, complex Microsoft Word document. I'm also prepared to admit that there may exist a universe (in which I might be living) where Word is a defensible choice over LaTeX, and that the choice might be more akin to common religious preferences--vi vs emacs, CLI vs GUI, etc--and not something that can be strictly judged in objective terms. Hopefully that admission would give me pause before I let loose such a tirade.
[1] AFAIU, Jens Gustedt was the previous editor, both initiating and executing the switch to LaTeX. But I've never participated in the committee. I may have had a mistaken belief about who was doing the editing over the past several years.
[2] That's a comment on the article, not on the author himself. Professionals do unprofessional things all the time.
I realize there was a subheading saying that, but the article has other subheadings, and cherry-picking is the essence of editorializing. Titles have by far the biggest influence on discussion, so this is a big deal.
I used to use pandoc (with pandoc-flavored Markdown) for this, but it just couldn't handle layout properly, even with mixed LaTeX and Markdown in the same files.
I'm not overly fond of LaTeX, but it is the only tool that does what I want.
What's fascinating is that this situation has existed for decades. Alternatives have risen and [mostly] fallen, but LaTeX persists. For that reason an alternative would not only have to be better, but significantly better, for me to question someone's choice of LaTeX.
Disclaimer: I have been using LaTeX (mostly XeLaTeX) for over a decade for most of my writing, mathematical and non-mathematical, although nowadays I often do so via Markdown+pandoc. Also, at my last company we had to parse mathematical formulas typeset in LaTeX and understand their structure... it was a nightmare.
As a community (programming, engineering, science - whatever) we should probably be louder that the one thing any man or woman with the raw cash can do to help the scientific community is funding LaTeX-2030. The problem is oh-so-difficult but having a modern statically typed (for reflection mainly) well-designed programming language for typesetting could save huge amounts of time for researchers and the like.
That and the real agenda which is to kill word.
The degrees of separation from the HN readers stereotype of a HN reader to a billionaire can't be that high, right?
Irony, on the other hand, is a single message for multiple audiences. It requires a degree of mutual incomprehension.
The vast majority just don't have a use case where "LaTeX-2030" is going to help, and this wouldn't make much difference to them. I'm a mathematician and I write papers in LaTeX every day. Sure I could benefit from programability occasionally, but it's just not a serious problem. If I need to automatically generate a table or something, I have a Python script output LaTeX, it takes me 30 seconds.
Obviously there are some uses cases where it would make a big difference, but there just aren't very many in the grand scheme. We already have LaTeX3 which lets you do scripting, and lualatex. The fact that basically nobody uses either gives you a sense of how much the general community wants this.
! Undefined control sequence.
<write> ...\protect \numberline {\plist@algorithm
1}{\ignorespaces $\protect...
l.350 \end{theorem}
None of these commands (except \end{theorem}) were in my paper, they were all in an included package. I later found out the issue was nowhere near line 350 or inside of the theorem environment. There just doesn't seem to be a good way to debug LaTeX errors. The best method I found was to use some sort of binary search, removing parts of the document to see if the error persists — which is ridiculously time consuming.https://tex.stackexchange.com/questions/309010/teach-me-to-d...
https://news.ycombinator.com/item?id=22865357
https://news.ycombinator.com/item?id=22865357&p=2
For me great tooling means Word/Framemaker like, with ability to generate whatever document format I feel like.
Yeah there are plenty of relatively easy to use lightweight markup languages which can generate LaTeX[0]. If you don't need the specific properties of LaTeX but want the option of LaTeX-typeset output, use something else and compile to PDF through LaTeX.
[0] because Pandoc generates LaTeX even if the reference implementation doesn't.
https://journals.plos.org/plosone/article?id=10.1371/journal...
Big quote from the abstract :
"On most measures, expert LaTeX users performed even worse than novice Word users."
However, it is worth noting that anyone who has ever dealt with a Word document that is a collation from other people usually winds up with similar to hassles to what the writer finds in the article.
But in reality, most people just accept how LaTeX handles the figure placement by default (which is usually quite good), and doesn't waste time fighting it. So it's not really a real-life comparison of the two.
If the challenge had been "Write a document formatted according to the norms of your field", which is what people are actually doing most of the time, LaTeX would win easily. For scientific papers the journal is going to redo all the formatting anyway (using their own custom LaTeX set-ups), so there's no reason to waste time on this.
In my experience, LaTeX's default figure placement is actually quite bad, and you have to spend some time trying to paste it in different places to get it to place the figure in the place you want it.
IIRC, TeX only becomes defined bug-free on Knuth's death. However, it is feature-complete, in the sense that there are no more features to be added, even if such functionality is desperately needed. (Most notably, TeX does have issues with i18n and especially right-to-left text support)
LaTeX is certainly the best LaTeX, but it doesn't really have that many competitors.
It's been a few years since you used CSS :) Both of these things are trivial to do now
But to be clear: while I think my CSS experience would give me a grasp of what needs to be expressed and some ideas for how to express it, there is a whole other layer of actually applying that, and reconciling it with the content, which would be a challenge (and is done for you by the browser in CSS' case)
> As far as LaTeX's abilities as a programming language go, I have an interpreter for a reasonable subset of C written in TeX (it was an academic exercise but this is just to show what is possible). So it is much more than a mere markup language.
I believe you, but I also question whether that level of power is actually being used by people making documents with it, or whether it just adds complexity. I genuinely don't know. But those are the kinds of places one might find improvements to be made.
I would love to work on this, but definitely can't commit the time.
There's a lot of competitors to the first step, in the vein of having multiple tools that are plain-text files but convert to rich-text output. However, most of these tools lack good, or even any, support for the fine-tuning capabilities. Where there is competition in the fine-tuning space, the relevant tools are largely based on WYSIWYG editors--which makes sense when you consider that it's easier to fine-tune properties when you get immediate visual feedback on what the fine-tuning you're doing does to the result. But these editors don't provide an easy way to integrate in the workflows of the other tools, since the underlying formats don't really support the fine-tuning capabilities.
I never realised how awful LaTeX was until I started working with a co-author who is blind, if you just "write LaTeX" you end up with equations, algorithms and tables constantly half-off the page.
Where it fails is in reporting the problem in a meaningful way. This is, of course, very difficult to accomplish, since there is no nice mapping from line-of-TeX to visual error. What's more, any tool which could provide such an error would be so constrained as to be useless for typesetting.