Problem:
Stopping a language server and then calling vim.lsp.start() with the same name/root will return the old language server that's in the middle of shutting down. vim.lsp.start() won't return a new server until the old process has terminated.
Solution:
Introducing a client._is_stopping field that tracks the shutdown phase, preventing the client from being reused.
problem: Error notifications from LSP responses were difficult to read due to
inconsistent formatting and missing critical information like client name and error codes.
solution: Implemented a more structured error notification format using pipe separators to
clearly display client name, error code (when available), and error message.
Problem:
enable() routine detaches clients even if they were manually started
and not managed by vim.lsp.config.
Solution:
Skip clients that aren't managed by vim.lsp.config.
Problem: More-window is left visible after leaving it. Cursor may be
hidden behind it and it is hard to re-enter afterwards.
Solution: Close the more-window when another window is entered.
Problem: Spliced conceal_lines marks after changing the buffer text are
left valid, concealing lines that shouldn't be.
Solution: Set the `invalidate` extmark property.
Problem:
`any[]` means nothing, and the return value is not the same as what's
documented in the comment (eg, Lua returns `{ "row", { { "leaf", 1000 },
{ "leaf", 1001 } } }`, not `{ "row", { "leaf", 1000, "leaf", 1001 } }`)
Solution:
Create two classes (vim.fn.winlayout.leaf and vim.fn.winlayout.branch)
and one alias that links the two together.
Also: Due to LuaLS limitations, there is an empty class,
vim.fn.winlayout.empty
Signed-Off-By: VoxelPrismatic <voxelprismatic@pm.me>
Problem:
No way to check if a LSP config is enabled without causing it to
resolve. E.g. `vim.lsp.config['…'] ~= nil` will resolve the config,
which could be an unwanted and somewhat expensive side-effect.
Solution:
Introduce `vim.lsp.is_enabled()`.
Problem:
The check for concealing paths in TOCs in the qf syntax file fails
because the TOC tile has changed.
Solution:
Force the qf syntax file to be reloaded after the qf_toc variable
has been set, so that the it can apply the correct settings.
Using the explicit qf_toc key, already used in the syntax file, instead
of the title is less prone to breaking.
It was also already being set for man pages but it had no effect because
the syntax file had already been loaded when the variable was set.
Fixes#33733
**Problem:** For multiline diagnostics, the end column was improperly
calculated by checking the byte index of the character position on the
*start* line.
**Solution:** Calculate the byte index for end_col using the *end* line.
Problem: The cmdline popupmenu is hidden behind extui windows.
'showmode' message is drawn over the cmdline.
Changing the 'cmdheight' to accommodate space for the text in
the cmdline may change the current cursor position.
Solution: Ensure kZIndexMessages < zindex < kZIndexCmdlinePopupMenu.
Clear the 'showmode' message when the cmdline level is negative.
Temporarily set 'splitkeep' = "screen" when changing the 'cmdheight'.
Problem: Not picking up configured option values when enabling _extui after startup.
Solution: Check option values when enabing _extui.
Fix https://github.com/neovim/neovim/issues/33767
Problem: We have an unmaintained Vimscript parser and cmdline
highlighting mechanism, with which it is hard to leverage the
treesitter highlighter. Long messages result in a hit-enter-prompt.
Solution: Implement a vim.ui_attach() UI, that replaces the message
grid (orphaning some 3000+ LOC core C code). Introduce an experimental
vim._extui module, because removing the message grid at the same time is
too risky. The new UI leverages the bundled treesitter highlighter and
parser for Vimscript, as well as the matchparen plugin, to highlight the
cmdline. Messages are truncated in the cmdline area, or placed in a
floating message box in the bottom right corner. Special ("list_cmd")
messages and the message history are shown in a, "more prompt" (now a
fully interactive regular window). Various default UI elements ('showcmd',
'ruler') are also placed in the cmdline area, as virtual text.
`require('vim._extui').enable({})` enables the experimental UI.
`{ msg.pos = 'box' }` or `:set cmdheight=0` enables the message
box variant.
Followup:
- Come to a consensus for how best to represent messages (by default).
- Start removing message grid when this is deemed a successful replacement.
When that is finished, make this new UI the default and update a lot of tests.
Problem:
virtual_text diagnostics are great when skimming a file, and
virtual_lines are great when "zooming in" on a particular problem.
Having both enabled results in duplicate diagnostics on-screen.
Solution:
This PR expands the behavior of `current_line` for virtual_text and
virtual_lines by making `virtual_text.current_line = false` distinct
from `nil`. If you set:
vim.diagnostic.config({
virtual_text = { current_line = false },
virtual_lines = { current_line = true },
})
With this configuration, virtual_text will be used to display
diagnostics until the cursor reaches the same line, at which point they
will be hidden and virtual_lines will take its place.
Problem:
enable() could be more flexible, so that it works even if called "late".
Solution:
- enable(true) calls `doautoall nvim.lsp.enable FileType`.
- enable(false) calls `client:stop()` on matching clients.
This will be useful for e.g. :LspStop/:LspStart also.
Problem:
Directories that are "trusted" by `vim.secure.read()`, are not detectable later
(they will prompt again). https://github.com/neovim/neovim/discussions/33587#discussioncomment-12925887
Solution:
`vim.secure.read()` returns `true` if the user trusts a directory.
Also fix other bugs:
- If `f:read('*a')` returns `nil`, we treat that as a successful read of
the file, and hash it. `f:read` returns `nil` for directories, but
it's also documented as returning `nil` "if it cannot read data with the
specified format". I reworked the implementation so we explicitly
treat directories differently. Rather than hashing `nil` to put in the
trust database, we now put "directory" in there explicitly*.
- `vim.secure.trust` (used by `:trust`) didn't actually work for
directories, as it would blindly read the contents of a netrw buffer
and hash it. Now it uses the same codepath as `vim.secure.read`, and
as a result, works correctly for directories.
Problem: UIs implementing ext_cmdline/message must also implement
ext_popupmenu in order to get cmdline completion with
wildoptions+=pum.
Solution: Allow marking a window as the ext_cmdline window through
nvim_open_win(), including prompt offset. Anchor the cmdline-
completion popupmenu to this window.
clarify complete_match() documentation to better explain its backward
search behavior, argument handling, and return value format and add an
example of isexpand
closes: https://github.com/vim/vim/pull/17212ffc89e47d0
* feat(shada): don't store jumplist if '0 in 'shada'
* fix(shada): don't store search and sub patterns if /0 in 'shada'
* fix(shada): don't store empty replacement string
* fix(shada): don't add '0' mark if f0 in 'shada'
- sort fields alphabetically.
- in the `vim.lsp.Client` docs, reference `vim.lsp.ClientConfig` instead
of duplicating its docs.
- cleanup lots of redundant-yet-drifted field docs.
Problem:
`FileType` event is fired before checkhealth report is finished, so
user can't override report settings or contents.
https://github.com/neovim/neovim/pull/33172#issuecomment-2833513916
Solution:
- Trigger FileType event later.
- Document how to remove emojis.
Problem: Cannot define completion triggers and act upon it
Solution: add the new option 'isexpand' and add the complete_match()
function to return the completion matches according to the
'isexpand' setting (glepnir)
Currently, completion trigger position is determined solely by the
'iskeyword' pattern (\k\+$), which causes issues when users need
different completion behaviors - such as triggering after '/' for
comments or '.' for methods. Modifying 'iskeyword' to include these
characters has undesirable side effects on other Vim functionality that
relies on keyword definitions.
Introduce a new buffer-local option 'isexpand' that allows specifying
different completion triggers and add the complete_match() function that
finds the appropriate start column for completion based on these
triggers, scanning backwards from cursor position.
This separation of concerns allows customized completion behavior
without affecting iskeyword-dependent features. The option's
buffer-local nature enables per-filetype completion triggers.
closes: vim/vim#16716bcd5995b40
Co-authored-by: glepnir <glephunter@gmail.com>
Problem:
Users of the Roslyn (C#) LSP have encountered significant delays when
retrieving pull diagnostics in large documents while using Neovim. For
instance, diagnostics in a 2000-line .cs file can take over 20 seconds
to display after edits in Neovim, whereas in VS Code, diagnostics for
the same file are displayed almost instantly.
As [mparq noted](https://github.com/seblj/roslyn.nvim/issues/93#issuecomment-2508940330)
in https://github.com/seblj/roslyn.nvim/issues/93, VS Code leverages
additional parameters specified in the [LSP documentation for
textDocument/diagnostic](https://microsoft.github.io/language-server-protocol/specifications/lsp/3.17/specification/#documentDiagnosticParams),
specifically:
- previousResultId
- identifier
Solution:
When requesting diagnostics, Neovim should include the
`previousResultId` and `identifier` parameters as part of the request.
These parameters enable the server to utilize caching and return
incremental results.
Support for maintaining state is already present in the
[textDocument/semanticTokens implementation](8f84167c30/runtime/lua/vim/lsp/semantic_tokens.lua (L289)).
A similar mechanism can be implemented in `textDocument/diagnostic` handler.
Problem:
In cases when the (in-process) LSP server responds to the request
immediately and calls `notify_reply_callback` the request will still be
marked as pending, because the code assumes that the response will occur
asynchronously. Then the request will be pending forever, because it was
already set as "completed" before we even set it as "pending".
A workaround is to wrap `notify_replay_callback` in `vim.shedule` ([like
so](https://github.com/neovim/neovim/pull/24338#issuecomment-2809568617)]
but that seems counterintuitive.
Solution:
Handle this case in Client:request().
