Hacker News Reader

Comment by konschubert | Hacker News Reader

I strongly dislike LaTeX mostly because of its terrible error messages, because of the mess of incompatible packages and because of its inflexibility in layout adjustments.

I admit that I never sat down to understand the latex design principles and learn it the hard way. So maybe I'm just uninformed.

But seriously, I just want to write some text with formulas.

The error messages are so useless, usually I'm forced to trace a bug by commenting out sections of text. For even the most trivial features, packages have to be included. Finding the right set of compatible packages is a science on its own. Usually you start with somebody else's document header and try to tweak it to your needs.

And good luck placing an image on the page where you want it.

I assume that if you spend some time understanding the language, it becomes clearer and less of a mess.

But I feel like there is room for a declarative text editor that's a little bit more intuitive.

View on HN

To some extent, ShareLaTeX has actually been building a wrapper to help you decode obscure error messages [0]. In addition, their online editor provides some basic linting [1].

[0]: https://www.sharelatex.com/blog/2016/06/29/latex-log-hints.h...

[1]: https://www.sharelatex.com/blog/2016/11/01/introducting-late...

Yea, I'm now using sharelatex whenever I am teXing something. Its features ease the pain.

I agree. You mention two problems that I both agree with.

First, LaTeX wants to be so much, that it can be difficult to get simple things done.

Second, LaTeX's tooling is abhorrent. The quite useless error messages are just one example. Why, for the sake of whoever, would a user-facing tool require me to compile the source twice to get some references right? There is no valid excuse for that.

The real problem here is that LaTeX is not what a document should be described in. LaTeX is a middleware that should be hidden from the user wherever possible.

There is no valid excuse for that.

There was when TeX was originally written to run on memory constrained mini computers. I agree that there isn’t an excuse today.

What you are describing sort of goes against the tenets of LaTeX. You aren't supposed to worry about formatting, just content. You have to explicitly tell it when you want to deviate from the template. In terms of placing figures, you just have to write \begin{figure}[h!].

I've never really had a problem with the error messages or packages personally.

I think it's a legitimate wish to change the footer height in a document.

\setlength{\footskip}{length} should do the job. You might have to reduce the \textheight to make room. What are you trying to achieve?

I'd probably try the /setlength command, but not consider that I have to adjust the invisible /textheight. Then I would wonder why it did not work.

With HTML, at least I have the inspector to see where bounding boxes are. With latex I'm blind.

    \documentclass[a4paper]{article}
    \usepackage{layout}
    \begin{document}
    \layout{}
    \end{document}

Ought to sort you out. There’s also the showframe package.

No, you're supposed to worry separately about formatting and content. I know that math-y people (I guess that means math, CS and physics people) think that just using the default settings is fine and those who are hit particularly hard with Stockholm syndrome sometimes even call it 'beautiful', but outside of that world, there is a stigma against papers formatted using just the default Latex (Tex?) settings. So wanting to use other fonts, more narrow margins and different header styles is something that should be if not easy, then at least within reach of even casual users. Which it isn't, right now. (in my experience, having tried to set up Latex and VCS-based authoring workflows several times with people who only took programming classes because they had to).

HTML is infinitely easier to memorize than LaTeX, and Github-Flavored-Markdown is easier still. If they had math notation (Mathjax integration with GFM is still not there), I think they would eliminate 95% of the need for LaTeX.

Doesn't latex offer a lot more control over layout (page breaks, columns)? For html you have to use css for that and markdown (thankfully) gives you nothing.

Latex the document language? No.

Latex the design template system? Yes, just about.

They are to each other as HTML is to CSS - completely complementary.

The main downfall of Latex is confusing the two spaces. The intent is good though.

HTML + CSS is not as cleanly abstracted for pure document markup and layout (as it's a more general solution), but it's getting closer every year.

On this, I think we can just agree to disagree. For one, TeX did not try and make an arbitrary distinction between style language and content language. So, it is both. Period. LaTeX did try and layer the style to a separate file, but still the same language.

Contrast this with the supposed HTML and CSS divide which has to later in a scripting language and all the joy they all enrich on each other. Is it any wonder that JSX is taking off?

Now, tooling is behind. But that is because it isn't cool to build up tooling on something that doesn't have the user base of a browser.

> So, it is both. Period.

Yes, I know. That was my point. I concider that the problem and a major reason why it's still so hard to use and one reason why tooling sucks for it.

> JSX

... has nothing to do with HTML & CSS which are both quite happy if you don't use scripting at all. If you think JSX applies in any way here I posit you don't understand the problem space.

Apologies for the misdirection. I posit that JSX exists because having a separate scripting language (JS) from the display language (HTML) is actually a pain from the scripting point of view. So, yes, nothing to do directly with HTML & CSS, but everything to do with the mess that is all of them.

And this is highlighting where we agree to disagree. I consider it somewhat of a strength of TeX that it is a single language for what it does. Makes scripting it much easier to consider. (Far from easy, I grant.)

As for tooling. Tooling is almost always strictly a matter of labor. More people working with it would have produced tooling that was what people wanted. Consider how terrible the tooling for early web was.

But again, I do not think you are wrong, per se. It would take an empirical argument without appeal to beliefs to convince me that one is truly better than the other. I just happen to like the TeX way in absence of these experiments.

I'm not saying LaTex is bad just that putting it all into one system is why it is hard to use, pretty much only the domain of programmers and those who have to use it, and has awful tooling.

None of these things are impossible to solve for LaTex, but they do make it conceptually harder to detangle.

So we aren't disagreeing heavily. Sounds like the main difference is that i disagree that it is any technical achievement of the newer languages that makes them "better." Rather, my position is that it was more from resources. LaTeX just didn't have the resources to build around it.

I don't know how to test this hypothesis. I can present my reasoning for feeling it, though. And that is simply that the technical makeup of the current technologies is built on sand.

Javascript is a mess technical capabilities that is constantly getting frameworks rewritten to address previously solved by other framework's problems.

CSS is a beast when you get in the woods and is basically dominated by whoever has produced the most recent resetting sheet. More, it seems that you can pseudo script CSS things now, what with duration and other items making their way into the standard. (Or, just the :content meta where you can now author crap in the css....)

And HTML, the cornerstone, is a mess of different ways to nest things into pseudo xml. There was a push to make it legitimate XML, but that failed due to the fact that it wouldn't work.

Contrast with LaTeX which just doesn't have a gui. The equivalent of the "Inspector" tab. You seem to be claiming that this is not fixable, but I have some serious doubts on that claim. Especially since I didn't have it at the beginning of my web career, either.

Right, I'm not extolling the virtues of CSS/HTML from a technical standpoint (though they actually do quite well, given the problem space - they are unfairly maligned).

It's more about how they separate out the ideas conceptually - document structual markup + document layout styling.

LaTeX kind of does this in how most people use it but it bundles the ideas together which makes it very hard to reason about and complicates the space.

> You seem to be claiming that this is not fixable

Not at all! I'm saying it's not there yet and progress on that GUI/Inspector/software has been hampered by the above conceptual problems due to the design of the language.

They are solvable, of course, but LaTeX seems to go out of its way to make it harder to solve.

This is the reason I write pandoc flavored markdown these days and generate latex with it. That way I can still use full latex power when needed inline, but for simple things like headers, italic and monospace I can just use markdown syntax.

SILE could be a new-and-improved LaTeX, someday.

http://sile-typesetter.org/what-is/

LaTeX isn't used much outside academia, so I don't think there's hope of it getting any better.

Yea, academia doesn't exactly have a track record of taking good care of its software...

At least for LaTeX there's also a ton of "cargo cult" involved. You either build a template, or get one from a professor, when you start as a student, and you just use that template for the rest of your time at the university. It's not uncommon to see a template that was initially created in the 90's and people just sort of added to it over the years.

Sure, there is room for a better latex, it's not like it's the tool to end all tools. But nobody has managed to produce an alternative so far...

By the way, try the option [H] on floats, it forces it to stay in the position it is in the code, instead of letting TeX rearrange it to minimise wasted space.

I wonder if the problem is that nobody will write something else because of the huge effort it will be, and there will be no users of the alternative until it achieves some level of functionality. I wonder if there's an economics term for this conundrum.

I wonder if there's an economics term for this conundrum

It doesn't come from the field of economics but there is a very good term for why it's so hard to develop an alternative to complex systems like LaTeX:

The Anna Karenina Principle [0]

The reason nobody will jump on a LaTeX alternative is because everybody who uses LaTeX needs something different from it. In a way, you could say that LaTeX has a 'niche' in the long tail [1] of the distribution. The 'happy families' in this analogy would be everybody who gets by with Microsoft Word.

[0] https://en.wikipedia.org/wiki/Anna_Karenina_principle

[1] https://en.wikipedia.org/wiki/Long_tail

Maybe something like TeXmacs? It is actually unrelated to TeX (AFAIK) and is basically a WYSIWYG-ish editor for structured documents. The only drawback is that it is kinda crash friendly (especially the latest version, try one of the 1.0.7.xxx instead of 1.99 version).

Indeed, I think Latex is one of this things you like or dislikes forever the first time you tried.

Indeed, you seem uninformed. You probably should look at ConTeXt, rather than LaTeX.

Reply on news.ycombinator.com