Use sign_getplaced instead of :sign place when possible.

[aktau]

There's a new signs API (available in Vim 8.1.0772 and in Neovim), it doesn't need text parsing to get at the juicy bits, unlike :signs place list.

Of course, wholesale moving to this easier-to-use API does mean the plugin wouldn't be compatible with older versions o (Neo)Vim anymore. Another option is conditional usage, though that likely only has the disadvantage of quasi-duplicating code.

I'm unsure whether it has performance benefits.

Usage of it in vim-lsp:

$ rg 'sign_[a-z_]*\('
autoload/lsp/ui/vim/signs.vim
55:    call sign_define(a:sign_name, l:options)
78:    call sign_undefine('LspError')
79:    call sign_undefine('LspWarning')
80:    call sign_undefine('LspInformation')
81:    call sign_undefine('LspHint')
104:    let l:sign_group = s:get_sign_group(a:server_name)
105:    call sign_unplace(l:sign_group, { 'buffer': a:path })
108:function! s:get_sign_group(server_name) abort
115:    let l:sign_group = s:get_sign_group(a:server_name)
125:                let l:sign_id = sign_place(0, l:sign_group, l:sign_name, a:path, { 'lnum': l:line })
   vim

[ZeroKnight]

Wonderful, I was hoping they'd add to the signs API.

I would like to continue supporting versions without this new API, so I wonder what the best way would be to go about that. Using has v:version is what immediately comes to mind, but I wonder if there's a better way?

[aktau]

I think the best option is to check for the existence of this API:

if exists('*sign_define') ...

(See prabirshrestha/vim-lsp#322, which has a discussion on this topic.)

[aktau]

Thinking about this, I think it's best to emulate vim-lsp check before using the new API:

if exists('*sign_define') && (has('nvim') || has('patch-8.1.0772'))
   " Both group=* and the new API are available.
endif