Experiment with Tauri and Solid

[martpie]

I've been playing a little bit with Tauri and it seems an early, but very nice solution. It has less APIs but look much more frameworky than Electron (binaries compilation included, etc).

I don't think we can re-use 1:1 compared with Electron, and we may lose some features along the way, but consuming 5 times less RAM and dividing binary size by 15 seem like awesome things to have.

This is very much experimental, and may not be a successful port.

• Isolate all calls to Electron and centralize them, so it's easier to rewrite this layer in Rust
• As much as possible, do not rely on Node.js modules for the frontend (example: scan files from Rust rather than the frontend?)
• Can the DB (linvodb3) survive this? Can we use a common solution between back-end and front-end?
• ...

I am also thinking on switching to something lighter than React. The main reason is I'm going to have to rewrite a good part of the React codebase, so if there's a way to simplify this part as well, why not?

[binyamin]

@martpie There's also Preact, which tries to expose the same React API in around 3kb.

[martpie]

So the main advantage I see with Solid.js is the absence of Virtual DOM. That means the tracks list would be heavily optimized (we currently have to rely on a lot of small tricks to keep the thing performant enough).

I tested with 10'000 tracks on Tauri, and it didn't need to use any kind of optimizations to perform well, so that's a big plus. The downside is the ecosystem is VERY limited.

The UI stack is not set in stone anyway, I am currently mostly experimenting with the Rust back-end, and see if I have all the tools to re-create the library scan, the ID3 tags extraction, the DB management, etc.