Feature Request: Inline LaTeX for Ghostty

[cheater]

It would be cool to be able to have inline LaTeX working.

Send an escape code, then just latex source code, then send the matching closing escape code. The terminal would render the image with LaTeX (external tool call) and display it inline, just like all those inline LaTeX JS libraries for the web.

Bonus point if somehow:

• colored text is supported
• colors follow terminal preferences
• the text gets antialiased even while colored
• sub-pixel rendering and hinting happens
• reflow can happen when window is resized
• reflow can break up inline formulas in known places, like e.g. on binary operators, or in a place somehow indicated by the user
• text size is reflected correctly
• large formulas are centered on the line (implies additional line spacing)

I know it's a big ask. But maybe... maybe it's possible one day :)

I won't tell anyone this one's easy. But maybe some simple version of this is within reach?

[pluiedev]

This is pretty much a textbook example of something we won't implement because it's way too niche, but you're more than welcome to maintain a fork for it if you wish :) Maybe in the far off future there will be a plugin system to realize this, but it's frankly too invasive for anything outside of a fork

[00-kat]

Why not render it yourself, to an image, then display that with the Kitty image protocol? I've seen some Neovim plugins which do that for example.

[chy-chiu]

Someone just built it:
https://github.com/mikewang817/TFormula

[cheater]

Oh I love this

[cheater]

OK, so it sort of illustrates why I requested for support to be explicitly added to Ghostty, especially if you read my FR here: mikewang817/TFormula#2

I think the final version of this sort of feature would be a capability that programs should be able to discover via termcap, and no wrapper programs would need to be used. That would ultimately mean its place is in the terminal. Wonder what you think about this, @pluiedev and @00-kat ?

You could argue with "but what if someone wants some other thing that also gets rendered to graphics" and then my argument would be that:

1. TeX is special, useful, and desired enough that it warrants special case support
2. If many things start appearing with the same scenario, maybe some sort of plugin system would be appropriate rather than using wrappers. However, until then, it's not worth even talking about, because we don't even have two things that need some sort of text-to-graphics capability, we just have one so far.

I would be interested to hear everyone's thoughts on this. Thank you.

[cheater]

PS to address the "too niche" argument, I think this might be a combination of point of view and survivorship bias.

Maybe the person making the argument isn't part of a community that uses TeX a lot (POV). And if you look at e.g. emoji support, that was a "niche requirement" too for terminals for a long time, until someone added it (I think MacOS was first?) and people saw it's actually pretty useful and now everyone's using it (you can't run a test suite without seeing some emoji).

[pluiedev]

From the perspective of maintainers, each new feature we add is additional code we have to maintain, and for TeX specifically that also means more dependencies that can get quite large (TeX Live is already massive even in its minimal form), and that just means more burdens for us and for e.g. distro maintainers who have to make sure Ghostty continues to build and function correctly. This is a massive cost that, in the face of alternatives like simply using a third-party tool that renders TeX into an image and takes care of sizing, positioning, scale, etc. with the Kitty Images Protocol, simply doesn't justify itself.

I'm decidedly biased in this regard, but I don't think any of us are inclined to accept this feature any time soon.

[cheater]

How can this be structured so that

1. The burden on maintainers is minimal to zero long-term
2. Programs can discover the capability using termcap
3. Wrappers don't need to be used?

I'm sure there is some solution that might become apparent but not immediately.