Suggestion: start LSP client in BufEnter

[lbiaggi]

Feature description

Moving the LSP start to BufEnter event reduces ~ 20-30% of the startup, which is really helpful when opening in an RPI.

IMO, moving to BufEnterfelt more smoothly

I achieve it through by disabling auto-attach and running the original command.

...
auto_attach = function(bufnr)
            local augroup = _G.custom.commands.augroup
            augroup("RUST_LSP", {
                {
                    event = "BufEnter",
                    pattern = "*.rs",
                    command = function()
                        require("rustaceanvim.lsp").start()
                    end,
                    once = true,
                },
            })

            return false
        end,
...

I wonder if this would be beneficial for other users / more powerful environment

[mrcjkb]

Sounds reasonable 😄

I've pushed the change to master, but I'll wait a bit before publishing a release, to see if someone complains.

[mrcjkb]

@lbiaggi I've had to revert this because it led to problems with the LSP client starting at all, and I don't know if it has any other side effects.

[lbiaggi]

@lbiaggi I've had to revert this because it led to problems with the LSP client starting at all, and I don't know if it has any other side effects.

Yeek, seriously? Could you give me an example?

[mrcjkb]

Yeek, seriously? Could you give me an example?

I don't have a repro myself.
#408 (comment)

In any case, the uncertainty is enough reason to revert it. You seem to have a workaround that works for you, and this is an edge case.

[lbiaggi]

Yeah you are right about my use being an edge case. I'm just curious, because I would like to help/debug it.

…

On Sun, 26 May 2024, 09:01 Marc Jakobi, ***@***.***> wrote: Yeek, seriously? Could you give me an example? I don't have a repro myself. #408 (comment) <#408 (comment)> In any case, the uncertainty is enough reason to revert it. You seem to have a workaround that works for you, and this is an edge case. — Reply to this email directly, view it on GitHub <#407 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABJGX4ARYWVQSMLPW2US6RTZEGJEZAVCNFSM6AAAAABIG5UMG6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMZSGEZDAOBUHE> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[mrcjkb]

The official recommendation is for vim.lsp.start to be called on a FileType event.
I fear that deferring to BufEnter could have unexpected side effects. So even if we were to fix this particular case, I won't include this in rustaceanvim.

[lbiaggi]

Look, I'm not gonna argue with you about this ("the official statement") which is just an example. But as I've said before it was just a suggestion if it doesn't work for the others, I keep doing it on my own for my use case, as I have been doing anyway. I was really curious about what caused it, it wasn't my intention to try to force it into mainline, sorry for the confusion and noise. Btw, you did a good work with the plugin

…

On Sun, 26 May 2024, 10:04 Marc Jakobi, ***@***.***> wrote: The official recommendation <https://neovim.io/doc/user/lsp.html#lsp-quickstart> is for vim.lsp.start to be called on a FileType event. I fear that deferring to BufEnter *could have unexpected side effects*. So even if we were to fix this particular case, I won't include this in rustaceanvim. — Reply to this email directly, view it on GitHub <#407 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABJGX4B3FW546JU3IVI5JBTZEGQTDAVCNFSM6AAAAABIG5UMG6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMZSGE2DIOBYGU> . You are receiving this because you were mentioned.Message ID: ***@***.***>