deno_lint VS deno_sdl2

Fast compilation seems very appealing. It is one of the main reason why I am interested into Go and Zig.

I recently started working with Rust for contributing to projects like Rome/tools [1] and deno_lint [2]. The compilation and IDE experience is frustrating. Compilation is slow. I am afraid that this is rooted to the inherent complexity of Rust.

[1] https://github.com/rome/tools

[2] https://github.com/denoland/deno_lint

Though, for large scale projects, I’d wait until https://github.com/denoland/deno_lint/issues/303 is done; if they tackle that, they tackled types, and that’s the single big thing yet to tackle.

There isn't any uniformity that could be reasonably achieved once plugins are added to a linter. Someone will always want an edge case for their project covered even if it's not in the uniform configuration; most large projects either use plugins or custom rules outside of what ESLint provides. Deno understands this too given that plugin support is being considered.

Deno's lint config is under deno.json or deno.jsonc and is limited to the following rules: https://lint.deno.land/

$ deno lint (prefer-const) `order` is never reassigned let order = new Order() ^^^^^ at /Users/dina/try-deno/design-patterns/state.ts:106:4 hint: Use `const` instead help: for further information visit https://lint.deno.land/#prefer-const Found 24 problems Checked 25 files

I built one of the tools mentioned in the article, Deno's linter. Its binary is over 30 MiB:

After a long time, I started contributing to open-source software. For a long time, I had my eye on Deno. This week, I merged my first PR in Deno Lint. Following is the detailed post about the issue.

sdl2 gets around the first problem by downloading a CLI service & starting it, that speaks JSON over a TCP socket. It's a limited surface of the API, and missing some pretty basic stuff, but I could see it working ok. It's not canvas, so I'd have to translate those sort of calls to it. I discussed that a bit here and it sounds like I could blit images of the canvas to sdl2 surface, but since it doesn't support in-memory images, I'd have to implement that in sdl2, or save a "frame" image to a file, for every blit. Either way, I have the overhead of redrawing every frame as a complete image. This seems incredibly jenky, and I assume will not perform very well, but I haven't tested.