Updated the `rpc.connect` function to support connecting to LSP servers
using hostnames, not just IP addresses. This change includes updates to
the documentation and additional test cases to verify the new
functionality.
- Modified `connect` function to resolve hostnames.
- Updated documentation to reflect the change.
- Added test case for connecting using hostname.
Added a TCP echo server utility function to the LSP test suite. This
server echoes the first message it receives and is used in tests to
verify LSP server connections via both IP address and hostname.
Refactored existing tests to use the new utility function.
Problem: prefix can be a symbol like period, the fuzzy matching can't
handle it correctly.
Solution: when prefix is empty or a symbol add all lsp completion
result into matches.
Problem: Some items of completion results include function signatures that can
cause the pum to be very long when a function has many params, because pum
scales with the longest word/abbr.
Solution: add custom covert function that can customise abbr to remove params.
Problem: the autotrigger mechanism could fire completion requests despite
completion already being active from another completion mechanism or manual
trigger
Solution: add a condition to avoid an additional request.
Problem: When preparing the parameters for a code actions LSP request,
the code set `context.diagnostics` when processing the first LSP client,
and then reused those `context.diagnostics` for subsequent LSP clients.
This meant that the second and next LSP clients got diagnostics that
did not originate from them, and they did not get the diagnostics that
they sent.
Solution: Avoid setting `context.diagnostics` (which is referenced by
all clients). Instead, set `params.context.diagnostics` directly, which
is specific to a single client.
Fixes#30001
Caused by #29501
`buf.code_action` always included diagnostics on a given line from all
clients. Servers should only receive diagnostics they published, and in
the exact same format they sent it.
Should fix https://github.com/neovim/neovim/issues/29500
Problem:
Some language servers (e.g., rust-analyzer, texlab) are desynced when
the user deletes the entire contents of the buffer. This is due to the
discrepancy between how nvim computes diff and how nvim treats empty
buffer.
* diff: If the buffer became empty, then the diff includes the last
line's eol.
* empty buffer: Even if the buffer is empty, nvim regards it as having
a single empty line with eol.
Solution:
Add special case for diff computation when the buffer becomes empty so
that it does not include the eol of the last line.
Problem: CompletionItem in lsp spec mentioned the deprecated attribute
Solution: when item has deprecated attribute set hl_group to DiagnosticDeprecated
in complete function
Reverts https://github.com/neovim/neovim/pull/29212 and adds a few
additional test cases
From the spec
> All text edits ranges refer to positions in the document they are
> computed on. They therefore move a document from state S1 to S2 without
> describing any intermediate state. Text edits ranges must never overlap,
> that means no part of the original document must be manipulated by more
> than one edit. However, it is possible that multiple edits have the same
> start position: multiple inserts, or any number of inserts followed by a
> single remove or replace edit. If multiple inserts have the same
> position, the order in the array defines the order in which the inserted
> strings appear in the resulting text.
The previous fix seems wrong. The important part:
> If multiple inserts have the same position, the order in the array
> defines the order in which the inserted strings appear in the
> resulting text.
Emphasis on _appear in the resulting text_
Which means that in:
local edits1 = {
make_edit(0, 3, 0, 3, { 'World' }),
make_edit(0, 3, 0, 3, { 'Hello' }),
}
`World` must appear before `Hello` in the final text. That means the old
logic was correct, and the fix was wrong.
Problem:
When there are multiple inlay hints present at the same position, they
should be rendered in the order they are received in the response from
LSP as per the LSP spec. Currently, this is not respected.
Solution:
Gather all hints for a given position, and then set it in a single
extmark call instead of multiple set_extmark calls. This leads to fewer
extmark calls and correct inlay hints being rendered.
This patch replaces fswatch with inotifywait from inotify-toools:
https://github.com/inotify-tools/inotify-tools
fswatch takes ~1min to set up recursively for the Samba source code
directory. inotifywait needs less than a second to do the same thing.
https://github.com/emcrisostomo/fswatch/issues/321
Also it fswatch seems to be unmaintained in the meantime.
Signed-off-by: Andreas Schneider <asn@cryptomilk.org>
Although the built-in pum completion mechanism will filter anyway on the
next input it is odd if the initial popup shows entries which don't
match the current prefix.
Using fuzzy match on the label/prefix is compatible with
`completeopt+=fuzzy` and also doesn't seem to break postfix snippet
cases
Closes https://github.com/neovim/neovim/issues/29287
Co-authored-by: Ilia Choly <ilia.choly@gmail.com>
Co-authored-by: Jose Pedro Oliveira <jose.p.oliveira.oss@gmail.com>
Co-authored-by: Maria José Solano <majosolano99@gmail.com>
Co-authored-by: zeertzjq <zeertzjq@outlook.com>
Problem:
For snippets lsp.completion prefers the label if it is shorter than the
insertText or textEdit to support postfix completion cases but clangd
adds decoration characters to labels. E.g.: `•INT16_C(c)`
Solution:
Use parse_snippet on insertText/textEdit before checking if it is
shorter than the label.
Fixes https://github.com/neovim/neovim/issues/29301
This reduces the number of nil checks around buf_versions usage
Test changes were lifted from 5c33815
Co-authored-by: Mathias Fussenegger <f.mathias@zignar.net>
this only changes the error message, so that it is clear that
the error is with the LSP server, rather than being a crash inside
nvim runtime scripts. We are already doing a lot of validation,
it's just that nil was being overlooked here.
This fixes issue #27395
Problem:
Text edits with the same position (both line and character) were being
reverse sorted prior to being applied which differs from the lsp spec
Solution:
Change the sort order for just the same position edits
* Revert "fix(lsp): account for changedtick version gap on modified reset (#29170)"
This reverts commit 2e6d295f79.
* Revert "refactor(lsp): replace util.buf_versions with changedtick (#28943)"
This reverts commit 5c33815448.
LspDetach is now triggered by the main on_detach callback that is added
when an LSP client is attached to a buffer. The semantic_tokens module
already includes a LspDetach handler that does the right thing. When the
LspDetach trigger was added to the main LSP on_detach, it created a race
condition in semantic tokens when a buffer was deleted that would
trigger both its own on_detach and the LspDetach handlers. If the former
came last, an error was thrown trying to delete a non-existent augroup
(destroy() was being called twice).
Problem: Completion side effects not working randomly.
Solution: When creating the table of LSP responses, the table index
was used, but this is not the same as the actual client_id, so it was changed
to use the client_id directly.
`lsp.util.buf_versions` was already derived from changedtick (`on_lines`
from `buf_attach` synced the version)
As far as I can tell there is no need to keep track of the state in a
separate table.
The `complete()` mechanism matches completion candidates against
the typed text, so strict pre-filtering isn't necessary.
This is a first step towards supporting postfix snippets (like
`items@insert` in luals)
