|
From: | Ergus |
Subject: | Re: [SPAM UNSURE] Maybe we're taking a wrong approach towards tree-sitter |
Date: | Wed, 28 Jul 2021 14:36:12 +0200 |
On Wed, Jul 28, 2021 at 07:43:03PM +0800, Andrei Kuznetsov wrote:
Stephen Leake <stephen_leake@stephe-leake.org> writes:The tree-sitter runtime, that Emacs would link with, it implemented in C, partly for this reason. It would be compiled with whatever Emacs is compiled with, or the system compiler.Interesting. I was not aware of that.Some of the tree-sitter development tools are implemented in Rust; you only need Rust if you are developing/fixing a grammar for a language.If I understand this correctly, it means one would require the Rust toolchain to support new languages in tree-sitter, or to improve existing support. Would that really fit Emacs? I think many people might not be comfortable learning such a large language and toolchain to develop editing tools for Emacs. Furthermore, is there any concrete reason this could not be done in Lisp?
I will say: 1) Performance (discussed in the previous thread): 2) Not reinvent the wheel. Tree-sitter is very well maintained, optimized and with very specialized algorithms; and we lack manpower to duplicate all that effort; and implementing it in lisp won't really worth the efforts and may be unmaintainable and slow. Tree-sitter hopefully won't get abandoned in the future because many editors use it right now (including neovim) and the community is very dynamic. Another advantage is that with tree-sitter as a back-end we could officially (almost for free) support many languages that are currently unsupported officially and may require a lot of effort to support them in a minimal way (or currently supported in some inconsistent way, with incoherent bindings/colors/indentations. Ex: Typescripts, Rust, Julia)
Note: Somehow I sent a reply earlier, and not a follow-up. I apologize for the duplicate.
[Prev in Thread] | Current Thread | [Next in Thread] |