performance degradation when working with large files in JS

[redoxahmii]

I didn't make this an issue as it seems that this only occurs for selected few people when working with JS projects.
First I was concerned that It was some misconfiguration on my end but after a post on Reddit it seems that other people are suffering the same issues.

I am using LazyVim and when working in a JS file I have vtsls eslint tailwindcss_language_server and emmet_language_server attached to buffer.

blink.cmp seems to struggle with files with above 2k LOC and part of this problem is also the LSPs but the difference in time for drawing the menu is far more then compared to nvim.cmp.

The most throttling that blink.cmp gets is from emmet_language_server and it cannot seem to handle those completions.

The delay in drawing a menu is large enough that I have to stop typing wait for the menu and then continue and could honestly type faster and I'm not that fast to begin with.

This is the post on Reddit that some other people have similar experiences as well.

While vtsls and emmet_language_server might be the reasons for alot of this throttling in getting completions it should atleast draw the menu and show the buffer sources instead of blocking the whole menu.

This could be an issue in terms of configuration as I'm not that well aware of what LazyVim is doing with blink.cmp as I just recently switched over.

Any help regarding this would be very helpful.

In any other filetype blink.cmp has been miles above nvim.cmp for me so thank you for your amazing work!

[Saghen]

There's an issue about this already, but your post gave me a clue about the source #1115 (comment) so thank you!

[redoxahmii]

emmet-language-server is the biggest culprit i found in my case with tailwindcss-language-server as a close second.
The issue with tailwindcss-language-server is that it just isn't configurable enough to properly attach only when working in an actual tailwind project and just attaches to every file available.

Disabling tailwindcss-language-server just now made it a lot faster than nvim-cmp and it's a very clear difference now.

Keeping eslint and vtsls with passing nvim-lspconfig.opts.servers.eslint with these flags has made it equivalent to the standard experience.

flags = {
            allow_incremental_sync = false,
            debounce_text_changes = 1000,
            exit_timeout = 1500,
          },