Lewis Russell

> I don't know; my impression was also that the fold queries were pretty solid, but someone (you?) really didn't like the foldexpr implementation... That's purely from an efficiency POV....

This makes the following work: ```lua local parser_config = require "nvim-treesitter.parsers".get_parser_configs() parser_config.c = { install_info = { url = 'https://github.com/nvim-treesitter/tree-sitter-c', files = {"src/parser.c"}, revision = "nvimc", } } ```

> Why do you want to change the behavior regarding git in that situation. It will just be slower and less reliable since git can only retrieve tree-revisions. I thought...

> Wouldn't a revision field that takes precedence over the lockfile be sufficient? This would just be a two word change. I don't really want to specify the revision, just...

The last change supports `revision=false` which does what I want.

Happens on my Macbook M1 Pro so I don't think it is limited to slow systems.

Not from our experience. We are successfully able to apply this PR via a patch. See https://github.com/neovim/neovim/pull/19632

https://github.com/kabouzeid/nvim-lspinstall/pull/173 should alleviate the problem

Here's my startup time currently using [impatients](https://github.com/lewis6991/impatient.nvim) profiler. ``` ───────────────────────────┬────────────────┬────────────┬────────────┬────────────┬────────────┐ │ Loader │ Resolve │ Load │ Exec │ Total │ ───────────────────────────┼────────────────┼────────────┼────────────┼────────────┼────────────┤ Total │ │ 2.2017ms │ 4.4617ms │ 93.9100ms...

No problem, we can just treat the server differently depending on whether it is a string or a table. If it's a string, treat it as a module name. If...