Re: Eglot cannot work with default completion-in-region?

[JD Smith <jdtsmith@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1538 bytes --]

> That's because it's a fundamentally different problem.

I agree with this.  Fancy completion styles work by sending all possible completions, and the completion-at-point systems suggests you do absolutely no filtering based on what the user has typed, leaving the style functions to do that.  This only works when all that data is in memory or close at hand.  A completion style simply has no relevance to the LSP server; it only knows what's in the buffer at point when completion was requested. 

> On Jan 29, 2024, at 5:21 PM, João Távora <joaotavora@gmail.com> wrote:
> 
> It may look like there is, but it's an illusion.
> Some people are using "orderless" with LSP (via Eglot or others), but
> they're guaranteed to be missing completions here and there except in
> perhaps in the most trivial cases.

Unless such people configure orderless to pass every single LSP completion result from the initial orderless term, then turn on all the bells and whistles for winnowing that set down for subsequent terms, which is how I use orderless with LSP.  Definitely not an illusion, but does require a "context" switch as you go.  This switch is aided by the corfu in-buffer completion system's mnemonic of M-SPACE to insert a space (the default orderless separator) and keep completion active.  Before the space?  LSP rules.  After the space?  Orderless takes over.  

But yes, it's a tricky impedance mismatch to manage (and there are many others in the completion-of-data-which-lives-outside-the-emacs-process space). 

[-- Attachment #2: Type: text/html, Size: 4508 bytes --]