This also fixes the following warning in tests with ASAN or TSAN:
-------- Running tests from test/functional/plugin/lsp/inline_completion_spec.lua
RUN T4604 vim.lsp.inline_completion enable() requests or abort when entered/left insert mode: 225.00 ms OK
RUN T4605 vim.lsp.inline_completion get() applies the current candidate: 212.00 ms OK
nvim took 2013 milliseconds to exit after last test
This indicates a likely problem with the test even if it passed!
RUN T4606 vim.lsp.inline_completion get() accepts on_accept callback: 212.00 ms OK
RUN T4607 vim.lsp.inline_completion select() selects the next candidate: 220.00 ms OK
-------- 4 tests from test/functional/plugin/lsp/inline_completion_spec.lua (3437.00 ms total)
-------- Running tests from test/functional/plugin/lsp/linked_editing_range_spec.lua
nvim took 2011 milliseconds to exit after last test
This indicates a likely problem with the test even if it passed!
The flakiness happens because get() uses vim.schedule(), and a following
key may be processed before the scheduled event. Use poke_eventloop() to
ensure that the scheduled event is processed.
Problem: make_floating_popup_options only shows when opts.border is explicitly set, ignoring global winborder setting
Solution: check both opts.border and vim.o.winborder when determining whether to show title
The cursor movement autocommand can not detect when the final tabstop $0
is directly adjacent to another tabstop, which prevents ending the
snippet session. The fix is an early return when jumping.
Problem:
Previously, 'null' value in LSP responses were decoded as 'nil'.
This caused ambiguity for fields typed as '? | null' and led to
loss of explicit 'null' values, particularly in 'data' parameters.
Solution:
Decode all JSON 'null' values as 'vim.NIL' and adjust handling
where needed. This better aligns with the LSP specification,
where 'null' and absent fields are distinct, and 'null' should
not be used to represent missing values.
This also enables proper validation of response messages to
ensure that exactly one of 'result' or 'error' is present, as
required by the JSON-RPC specification.
**Problem:** For unchanged document diagnostic reports, the `resultId`
is ignored completely, even though it should still be saved for the
request (in fact, the spec marks it as mandatory for unchanged reports,
so it should be extra important).
**Solution:** Always store the `resultId`.
Problem:
Closes#31453
Solution:
Introduce `vim.lsp.Capability`, which may serve as the base class for
all LSP features that require caching data. it
- was created if there is at least one client that supports the specific method;
- was destroyed if all clients that support the method were detached.
- Apply the refactor for `folding_range.lua` and `semantic_tokens.lua`.
- Show active features in :checkhealth.
Future:
I found that these features that are expected to be refactored by
`vim.lsp.Capability` have one characteristic in common: they all send
LSP requests once the document is modified. The following code is
different, but they are all for this purpose.
- semantic tokens:
fb8dba413f/runtime/lua/vim/lsp/semantic_tokens.lua (L192-L198)
- inlay hints, folding ranges, document color
fb8dba413f/runtime/lua/vim/lsp/inlay_hint.lua (L250-L266)
I think I can sum up this characteristic as the need to keep certain
data synchronized with the latest version computed by the server.
I believe we can handle this at the `vim.lsp.Capability` level, and
I think it will be very useful.
Therefore, my next step is to implement LSP request sending and data
synchronization on `vim.lsp.Capability`, rather than limiting it to the
current create/destroy data approach.
Problem:
In LSP configs, the function form of `cmd()` cannot easily get the
resolved root dir (workspace). One of the main use-cases of a dynamic
`cmd()` is to be able to start a new server whose binary may be located
*in the workspace* ([example](https://github.com/neovim/nvim-lspconfig/pull/3912)).
Compare `reuse_client()`, which also receives the resolved config.
Solution:
Pass the resolved config to `cmd()`.
Co-authored-by: Justin M. Keyes <justinkz@gmail.com>
Problem:
Some LSPs cause the following completion error (reformatted slightly):
Error executing vim.schedule lua callback:
.../runtime/lua/vim/lsp/completion.lua:373
attempt to index field 'range' (a nil value)
This is because an internal function assumes edits are either missing
or of type `TextEdit`, but there's a third [possibility][0] that's not
handled: the `InsertReplaceEdit`.
This was previously reported in at least two issues:
- https://github.com/neovim/neovim/issues/33142
- https://github.com/neovim/neovim/issues/33224
Solution:
Don't assume the edit is a `TextEdit`. This implicitly handles
`InsertReplaceEdit`s.
Also, add a test case for this, which previously caused an error.
[0]: 2c07428966/runtime/lua/vim/lsp/_meta/protocol.lua (L1099)
**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:
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:
Nvim needlessly requests inlay_hints even if they are disabled for a given buffer.
Solution:
Add the missing `enabled` check in `on_refresh`.
Rest of the code has this check already so that's the only needed one to fix this.
Problem: `trigger` is a custom word not yet used in APIs.
Solution: Use `get` instead because the main effect is that the
completion candidates will be collected (and shown by default,
but follow-up commits are planned to add an `on_result` callback
that allows more general handling).
---------
Co-authored-by: Christian Clason <c.clason@uni-graz.at>
Problem:
Given that `vim.snippet.expand()` sets temporary `<tab>`/`<s-tab>`
keymaps there is no way to build "smart-tab" functionality where `<tab>`
chooses the next completion candidate if the popup menu is visible.
Solution:
Set the keymap permanent in `_defaults`.
The downside of this approach is that users of multiple snippet engine's
need to adapt their keymaps to handle all their engines that are in use.
For example:
vim.keymap.set({ 'i', 's' }, "<Tab>", function()
if foreign_snippet.active() then
return "<Cmd>lua require('foreign_snippet').jump()<CR>"
elseif vim.snippet.active({ direction = 1 }) then
return "<Cmd>lua vim.snippet.jump(1)<CR>"
else
return key
end
end, { expr = true })
Upside is that using `vim.keymap.set` to override keymaps is a well
established pattern and `vim.snippet.expand` calls made by nvim itself
or plugins have working keymaps out of the box.
Co-authored-by: Maria José Solano <majosolano99@gmail.com>
Problem: The current implementation creates a unique autocommand group for each floating preview window, which is inefficient and can lead to numerous autocommand groups.
Solution: Use a single shared autocommand group with improved window validation to properly clean up LSP floating preview windows.
Problem:
vim.lsp.completion with "autotrigger" enabled, does not send
completion context, even though it has all the necessary info.
Solution:
Include the context for "autotrigger".
trigger() also optionally accepts context when manually invoked.
Problem: Last line in a window does not store correct `wl_lastlnum` if
lines below it are concealed (resulting in e.g. incorrect
cursor row).
Solution: Increment `wl_lastlnum` while it points to a line above a
concealed line.
Problem: After 47aaddfa the max_height option is no longer respected.
Hover documentation and Signature help windows take up the
entire text height.
Solution: Compare to window's current height and only modify the height
if it would reduce the height, not enlarge it.
Problem: _on_conceal_line callbacks are not invoked if callback has not
let Nvim know it wants to receive them. But this may change on
factors other than what is currently checked (changed buffer).
Solution: Forego this optimization, callback is still guarded behind
'conceallevel'.
Problem:
Indenting text is a common task in plugins/scripts for
presentation/formatting, yet vim has no way of doing it (especially
"dedent", and especially non-buffer text).
Solution:
Introduce `vim.text.indent()`. It sets the *exact* indentation because
that's a more difficult (and thus more useful) task than merely
"increasing the current indent" (which is somewhat easy with a `gsub()`
one-liner).
Problem: Height of a (markdown) `vim.lsp.util.open_floating_preview()`
window can be reduced to account for concealed lines (after #31324).
Solution: Set the window height to the text height of the preview window.
Set 'concealcursor' to avoid unconcealing the cursorline when
entering the hover window.
Problem:
After https://github.com/neovim/neovim/pull/32377 selecting snippets
provided by luals inserted the multi-line text before accepting the
candidates. That's inconsistent with servers who provide `textEdit`
instead of `insertText` and having lines shift up/down while cycling
through the completion candidates is a bit irritating.
Solution:
Use the logic used for `textEdit` snippets also for `insertText`
`command` was already resolved via a `completionItem/resolve` request
but only if `additionalTextEdits` were also present, and the
`resolveSupport` capability wasn't listed.
Closes https://github.com/neovim/neovim/issues/32406
Problem: When expanding a completion item that contains a multi-line word, the word is not deleted correctly.
Solution: If the word contains a line break, delete the text from Context.cursor to the current cursor position.
Problem: autotrigger option of vim.lsp.completion.enable() would trigger
all clients, as long as it matched at least one client's
triggerCharacters.
Solution: trigger only the clients with triggerCharacters matching the
character. overtriggering still happens if any client returns
isIncomplete=true (this case is more involved).
Co-authored-by: Mathias Fussenegger <f.mathias@zignar.net>
Follow up to https://github.com/neovim/neovim/pull/32072
If there is no prefix (e.g. at the start of word boundary or a line), it
always used the `filterText` because the `match` function always
returned false.
Problem:
With language servers like lemminx, completing xml tags like `<mo` first
shows the right candidates (`modules`) but after typing `d` the
candidates disappear.
This is because the server returns:
[...]
filterText = "<module",
label = "module",
textEdit = {
newText = "<module>$1</module>$0",
Which resulted in `module` being used as `word`, and `module` doesn't
match the prefix `<mo`. Typing `d` causes the `complete()` filtering
mechanism to kick in and remove the entry.
Solution:
Use `<module` from the `filterText` as `word` if the textEdit/label
heuristic doesn't match.
Problem: open_floating_preview() may be hidden behind current window if
that is floating and has a higher zindex.
Solution: Open floating preview with zindex higher than current window.
