Plugin support ?

[BastienVigneron]

Hi,

Nice to see this kind of initiative. What about steel /Lua plugins support ?
Is it something you plan to integrate, or did you want to keep the all-in-one philosophy ?

[warmwaffles]

Lua is going to be the better option to pick because these LLMs are not great with Lisp like languages. I do hope something is chosen, whether it be a quickjs runtime or something else that LLMs can be better leveraged with.

[giomf]

Maybe webassmebly is a good candidate?

[trojkat]

There is a draft PR already #130.

[warmwaffles]

nice it's in wasm

[413x1nkp]

helix had a draft PR for steel-based plugin system for years now, why not just use that?

helix-editor/helix#8675

[matoous]

To be honest, I don't want to go with scheme, Lua or WASM seem like more user-friendly options to me. There were good arguments in the original PR and discussions in Helix as to why one versus the other, the main argument being efficiency. While that's a reasonable argument and I would very much love this editor to be blazing fast the core design principle is that core features come out of the box (and are fast out of the box). With that it mind I think it is justifiable to take slight performance hit with Lua or WASM and plugins authoring simpler while giving users the option to skip plugins if they want to squeeze out every bit of performance.

For full original discussion see: helix-editor/helix#3806, there are many good arguments on all sides and I still haven't fully made my mind yet so I am open for more arguments and changing my mind.

[413x1nkp]

i think there's good arguments for both WASM/Lua and Scheme. there's a few plugins already made for the Scheme-based system, thus they could work out of the box. if another helix fork pops out with the Scheme-based plugin system, the compatibility of plugins between mitos and a hypothetical fork wouldn't be too hard to maintain.

on the other hand, WASM-based approach allows to use virtually any language for plugins, be it C/C++/rust/etc, and performance would probably be better. iirc, WASM modules are explicitly designed to be streamable from the disk, thus RAM usage should be pretty low overall. WASM module files can also be compiled AoT for native performance, though logistics for how that would work in terms of a plugin system is kind of shaky (plus you lose the benefit of streaming the module from the disk).

[v-Nyo]

My dream is to eventually replace Emacs with Mitos (Helix), which would be faster and simpler. The problem is that I'm not really qualified to judge what the best plugin system for that would be, especially if you'd want to support things like Org mode, Org-roam, and Org Babel.