**Problem:** Tree-sitter 0.24.0 introduced a new symbol type to denote
supertype nodes (`TSSymbolTypeSupertype`). Now, `language.inspect()`
(and the query `omnifunc`) return supertype symbols, but with double
quotes around them.
**Solution:** Mark a symbol as "named" based on it *not* being an
anonymous node, rather than checking that it is a regular node (which a
supertype also is not).
**Problems:**
- `vim.treesitter.language.inspect()` returns duplicate
symbol names, sometimes up to 6 of one kind in the case of `markdown`
- The list-like `symbols` table can have holes and is thus not even a
valid msgpack table anyway, mentioned in a test
**Solution:** Return symbols as a map, rather than a list, where field
names are the names of the symbol. The boolean value associated with the
field encodes whether or not the symbol is named.
Note that anonymous nodes are surrounded with double quotes (`"`) to
prevent potential collisions with named counterparts that have the same
identifier.
This commit also marks `child_containing_descendant()` as deprecated
(per upstream's documentation), and uses `child_with_descendant()` in
its place. Minimum required tree-sitter version will now be `0.24`.
Problem:
Using a single RPC request to child server isn't enough to prevent race
between nvim_paste and nvim_input.
Solution:
Ensure both child client and child server have processed pending events
by sending an empty DCS response to the child client and waiting for
TermResponse autocommand on the child server.
Problem: on `CompleteDone` cursor can jump to the end of line instead of
the end of the completed word.
Solution: remove only inserted word for snippet expansion instead of everything
until eol.
Fixes#30656
Co-authored-by: Mathias Fussenegger <f.mathias@zignar.net>
Co-authored-by: Justin M. Keyes <justinkz@gmail.com>
Problem: Currently completion attribute hl_group is combined with
all items, which is redundant and confusing with kind_hlgroup
Solution: Renamed to abbr_hlgroup and combine it only with the abbr item
(glepnir).
closes: vim/vim#158180fe17f8ffb
Co-authored-by: glepnir <glephunter@gmail.com>
Problem:
- Some servers like LuaLS add unwanted blank lines after multiline
`@param` description.
- List items do not wrap nicely.
Solution:
- When rendering the LSP doc hover, remove blank lines in each `@param`
or `@return`.
- But ensure exactly one empty line before each.
- Set 'breakindent'.
Problem: Lua accessors for
- global, local, and special variables (`vim.{g,t,w,b,v}.*`), and
- options (`vim.{o,bo,wo,opt,opt_local,opt_global}.*`),
do not have command-line completion, unlike their vimscript counterparts
(e.g., `g:`, `b:`, `:set`, `:setlocal`, `:call <fn>`, etc.).
Completion for vimscript functions (`vim.fn.*`) is incomplete and does
not list all the available functions.
Solution: Implement completion for vimscript function, variable and
option accessors in `vim._expand_pat` through:
- `getcompletion()` for variable and vimscript function accessors, and
- `nvim_get_all_options_info()` for option accessors.
Note/Remark:
- Short names for options are yet to be implemented.
- Completions for accessors with handles (e.g. `vim.b[0]`, `vim.wo[0]`)
are also yet to be implemented, and are left as future work, which
involves some refactoring of options.
- For performance reasons, we may want to introduce caching for
completing options, but this is not considered at this time since the
number of the available options is not very big (only ~350) and Lua
completion for option accessors appears to be pretty fast.
- Can we have a more "general" framework for customizing completions?
In the future, we may want to improve the implementation by moving the
core logic for generating completion candidates to each accessor (or
its metatable) or through some central interface, rather than writing
all the accessor-specific completion implementations in a single
function: `vim._expand_pat`.
An implication of this current approach is that `NVIM_API_LEVEL` should be
bumped when a new Lua function is added.
TODO(future): add a lint check which requires `@since` on all new functions.
ref #25416
Some composite/compound types even as basic as `(string|number)[]` are
not currently supported by the luacats LPEG grammar used by gen_vimdoc.
It would be parsed & rendered as just `string|number`.
Changeset adds better support for these types.
Ensure the screen is fully updated before blocking for input. This did
not always happen before, for example when setting `cursorline
scrolloff=9999`, which lead to jerky movement when using some GUI
applications.
Because of the duality of redraw_later, this can't be done in
command-line or when waiting for "Press ENTER". In many of those cases
the redraw is expected AFTER the key press, while normally it should
update the screen immediately. So, those special cases are excluded.
Problem: fixed order of items in insert-mode completion menu
Solution: Introduce the 'completeitemalign' option with default
value "abbr,kind,menu" (glepnir).
Adding an new option `completeitemalign` abbr is `cia` to custom
the complete-item order in popupmenu.
closes: vim/vim#14006closes: vim/vim#157606a89c94a9e
Problem:
The `_watch.watch()` strategy may fail if the given path does not exist:
…/vim/_watch.lua:101: ENOENT: no such file or directory
stack traceback:
[C]: in function 'assert'
…/vim/_watch.lua:101: in function <…/vim/_watch.lua:61>
[string "<nvim>"]:5: in main chunk
- `_watch.watch()` actively asserts any error returned by `handle:start()`.
- whereas `_watch.watchdirs()` just ignores the result of `root_handle:start()`.
Servers may send "client/registerCapability" with "workspace/didChangeWatchedFiles"
item(s) (`baseUri`) which do not actually exist on the filesystem:
https://github.com/neovim/neovim/issues/28058#issuecomment-2189929424
{
method = "client/registerCapability",
params = {
registrations = { {
method = "workspace/didChangeWatchedFiles",
registerOptions = {
watchers = { {
globPattern = {
baseUri = "file:///Users/does/not/exist",
pattern = "**/*.{ts,js,mts,mjs,cjs,cts,json,svelte}"
}
},
...
}
Solution:
- Remove the assert in `_watch.watch()`.
- Show a once-only message for both cases.
- More detailed logging is blocked until we have `nvim_log` / `vim.log`.
fix#28058
Problem: Wrong display when typing in diff mode with 'smoothscroll'.
Solution: Use adjust_plines_for_skipcol() (zeertzjq).
closes: vim/vim#1577647f8584a80
Problem:
Linematch used to use strchr to navigate a string, however strchr does
not supoprt embedded NULs.
Solution:
Use `mmfile_t` instead of `char *` in linematch and introduce `strnchr()`.
Also remove heap allocations from `matching_char_iwhite()`
Fixes: #30505
Problem: No clear way to check whether parsers are available for a given
language.
Solution: Make `language.add()` return `true` if a parser was
successfully added and `nil` otherwise. Use explicit `assert` instead of
relying on thrown errors.
Problem:
Headings in :help do not stand out visually.
Solution:
Define a non-standard `@markup.heading.1.delimiter` group and
special-case it in `highlight_group.c`.
FUTURE:
This is a cheap workaround until we have #25718 which will enable:
- fully driven by `vimdoc/highlights.scm` instead of using highlight
tricks (`guibg=bg guifg=bg guisp=fg`)
- better support of "cterm" ('notermguicolors')
Problem: diff mode does not handle overlapping diffs correctly
Solution: correct the logic to handle overlapping blocks
(Yukihiro Nakadaira)
Vim merges overlapped diff blocks and it doesn't work expectedly
in some situation.
closes: vim/vim#1573506fe70c183
Co-authored-by: Yukihiro Nakadaira <yukihiro.nakadaira@gmail.com>
Problem:
checkhealth report sections are not visually separated.
Solution:
Highlight with "reverse".
TODO: migrate checkhealth filetype to use treesitter.
TODO: default :help should also highlight headings more boldy!
**Problem:** `is_ancestor()` uses a slow, bottom-up parent lookup which
has performance pitfalls detailed in #28512.
**Solution:** Take `is_ancestor()` from $O(n^2)$ to $O(n)$ by
incorporating the use of the `child_containing_descendant()` function
Problem: When using nvim_paste in a mapping during a macro recording,
both the mapping and the paste are recorded, causing the paste
to be performed twice when replaying the macro.
Solution: Only record a paste when it is from RPC.
Unfortunately this means there is no way for a script to make a recorded
paste. A way to enable that can be discussed later if there is need.
Before this PR, the behavior of nvim_paste is:
- When vim.paste() returns false, return false to the client, but treat
following chunks normally (i.e. rely on the client cancelling the
paste as expected).
- When vim.paste() throws an error, still return true to the client, but
drain the following chunks in the stream without calling vim.paste().
There are two problems with such behavior:
- When vim.paste() errors, the client is still supposed to send the
remaining chunks of the stream, even though they do nothing.
- Having different code paths for two uncommon but similar situations
complicates maintenance.
This PR makes both the cancel case and the error case return false to
the client and drain the remaining chunks of the stream, which, apart
from sharing the same code path, is beneficial whether the client checks
the return value of nvim_paste or not:
- If the client checks the return value, it can avoid sending the
following chunks needlessly after an error.
- If the client doesn't check the return value, chunks following a
cancelled chunk won't be pasted on the server regardless, which leads
to less confusing behavior.
In the api_info() output:
:new|put =map(filter(api_info().functions, '!has_key(v:val,''deprecated_since'')'), 'v:val')
...
{'return_type': 'ArrayOf(Integer, 2)', 'name': 'nvim_win_get_position', 'method': v:true, 'parameters': [['Window', 'window']], 'since': 1}
The `ArrayOf(Integer, 2)` return type didn't break clients when we added
it, which is evidence that clients don't use the `return_type` field,
thus renaming Dictionary => Dict in api_info() is not a breaking change.
In the api_info() output:
:new|put =map(filter(api_info().functions, '!has_key(v:val,''deprecated_since'')'), 'v:val')
...
{'return_type': 'ArrayOf(Integer, 2)', 'name': 'nvim_win_get_position', 'method': v:true, 'parameters': [['Window', 'window']], 'since': 1}
The `ArrayOf(Integer, 2)` return type didn't break clients when we added
it, which is evidence that clients don't use the `return_type` field,
thus renaming Dictionary => Dict in api_info() is not (in practice)
a breaking change.
Problem: When input immediately follows end of bracketed paste, the
nvim_input may be processed before the nvim_paste.
Solution: Ensure some waiting after the end of a bracketed paste.
Problem: incorrect internal diff with an empty file
Solution: Set pointer to NULL, instead of using an empty line file
(Yukihiro Nakadaira)
When using internal diff, empty file is read as one empty line file.
So result differs from external diff.
closes: vim/vim#15719f1694b439b
Co-authored-by: Yukihiro Nakadaira <yukihiro.nakadaira@gmail.com>
- Fixes 'autoindent' being applied during redo.
- Makes redoing a large paste significantly faster.
- Stores pasted text in the register being recorded.
Fix#28561
- Allow function command modifiers.
- Match function bodies starting with empty lines.
Command modifiers reported by @Konfekt.
fixesvim/vim#15671closes: vim/vim#1567435699f1749
Co-authored-by: Doug Kearns <dougkearns@gmail.com>
Problem:
The default builtin UI client does not declare its client info. This
reduces discoverability and makes it difficult for plugins to identify
the UI.
Solution:
- Call nvim_set_client_info after attaching, as recommended by `:help dev-ui`.
- Also set the "pid" field.
- Also change `ui_active()` to return a count. Not directly relevant to
this commit, but will be useful later.
This module is generally used by any tests that need the full Nvim TUI
instead of `screen.lua`. Thus it should live in `functional/` instead of
in `functional/terminal/`.
Problem: Wrong cursor-screenline when resizing window
Solution: Invalidate saved left_col and right_col when width1 or width2
change.
closes: vim/vim#1567986dc4f8b43
**Problem:** `vim.treesitter.get_parser` will throw an error if no parser
can be found.
- This means the caller is responsible for wrapping it in a `pcall`,
which is easy to forget
- It also makes it slightly harder to potentially memoize `get_parser`
in the future
- It's a bit unintuitive since many other `get_*` style functions
conventionally return `nil` if no object is found (e.g. `get_node`,
`get_lang`, `query.get`, etc.)
**Solution:** Return `nil` if no parser can be found or created
- This requires a function signature change, and some new assertions in
places where the parser will always (or should always) be found.
- This commit starts by making this change internally, since it is
breaking. Eventually it will be rolled out to the public API.
Problem:
floating windows did not correctly inherit the NormalFloat highlight
group from the global namespace when it was not defined in the window-specific
namespace. This led to floating windows losing their background highlight when
switching between namespaces.
Solution:
Updated the window highlight logic in update_window_hl() to handle the fallback.
This fix resolves issues with floating window backgrounds not displaying as expected
in certain namespace configurations.
$NVIM_LOG_FILE: /Users/runner/work/neovim/neovim/build/.nvimlog
WRN 2024-09-08T21:48:13.279 ?.21134 vim_mktempdir:3281: $TMPDIR tempdir not a directory (or does not exist): TMPDIR-should-be-ignored
WRN 2024-09-08T21:48:13.312 ?.21137 vim_mktempdir:3281: $TMPDIR tempdir not a directory (or does not exist): TMPDIR-should-be-ignored
- `alter_slashes` belongs in `testutil.lua`, not `testnvim.lua`.
- `alter_slashes` is an unusual name. Rename it to `fix_slashes`.
- invert its behavior, to emphasize that `/` slashes are the preferred,
pervasive convention, not `\` slashes.
Problem:
If $NVIM_APPNAME is a relative dir path, Nvim fails to start its
primary/default server, and `v:servername` is empty.
Root cause is d34c64e342, but this wasn't
noticed until 96128a5076 started reporting the error more loudly.
Solution:
- `server_address_new`: replace slashes "/" in the appname before using
it as a servername.
- `vim_mktempdir`: always prefer the system-wide top-level "nvim.user/"
directory. That isn't intended to be specific to NVIM_APPNAME; rather,
each *subdirectory* ("nvim.user/xxx") is owned by each Nvim instance.
Nvim "apps" can be identified by the server socket(s) stored in those
per-Nvim subdirs.
fix#30256
Problem: vim.tbl_deep_extend had an undocumented feature where arrays
(integer-indexed tables) were not merged but compared literally (used
for merging default and user config, where one list should overwrite the
other completely). Turns out this behavior was relied on in quite a
number of plugins (even though it wasn't a robust solution even for that
use case, since lists of tables (e.g., plugin specs) can be array-like
as well).
Solution: Revert the removal of this special feature. Check for
list-like (contiguous integer indices) instead, as this is closer to the
intent. Document this behavior.
Problem:
$XDG_RUNTIME_DIR may be broken on WSL, which prevents starting (and even
building) Nvim. #30282
Solution:
- When startup fails, mention the servername in the error message.
- If an autogenerated server address fails, log an error and continue
with an empty `v:servername`. It's only fatal if a user provides a bad
`--listen` or `$NVIM_LISTEN_ADDRESS` address.
Before:
$ nvim --headless --listen ./hello.sock
nvim: Failed to --listen: "address already in use"
$ NVIM_LISTEN_ADDRESS='./hello.sock' ./build/bin/nvim --headless
nvim: Failed to --listen: "address already in use"
After:
$ nvim --headless --listen ./hello.sock
nvim: Failed to --listen: address already in use: "./hello.sock"
$ NVIM_LISTEN_ADDRESS='./hello.sock' ./build/bin/nvim --headless
nvim: Failed $NVIM_LISTEN_ADDRESS: address already in use: "./hello.sock"
Problem:
str_byteindex_enc could return an error if the index was longer than the
lline length. This was handled in each of the calls to it individually
Solution:
* Fix the call at the source level so that if the index is higher than
the line length, line length is returned as per LSP specification
* Remove pcalls on str_byteindex_enc calls. No longer needed now that
str_byteindex_enc has a bounds check.
Font-family names must be enclosed in quotation marks to ensure that
fonts are applied correctly when there are spaces in the name.
Fix an issue where multiple fonts specified in `vim.o.guifont` are
inserted as a single element, treating them as a single font.
Support for escaping commas with backslash and ignoring spaces
after a comma.
ref `:help 'guifont'`
Problem:
Since 96128a5076 the test logs have noise from tests that *expect*
failures:
$NVIM_LOG_FILE: /tmp/cirrus-ci-build/build/.nvimlog
(last 100 lines)
ERR 2024-09-04T13:38:45.181 T949.28335.0/c terminfo_start:486: uv_pipe_open failed: no such device or address
ERR 2024-09-04T13:38:45.181 T949.28335.0/c flush_buf:2527: uv_write failed: bad file descriptor
ERR 2024-09-04T13:38:45.181 T949.28335.0/c flush_buf:2527: uv_write failed: bad file descriptor
WRN 2024-09-04T13:43:43.294 ?.35904 server_start:173: Failed to start server: address already in use: /…/Xtest_tmpdir/…/T7159.35895.0
WRN 2024-09-04T13:43:43.314 ?.35907 server_start:173: Failed to start server: illegal operation on a directory: /
ERR 2024-09-04T13:43:43.332 ?.35909 socket_watcher_init:60: Host lookup failed: https://example.com
Solution:
Rewrite the test to use `vim.system()`. Set NVIM_LOG_FILE in the child
process to a "throwaway" logfile.
Problem: Marktree meta count still includes invalidated marks, making
guards that check the meta total ineffective.
Solution: Revise marktree metadata when in/revalidating a mark.
- The exclusion of lists was never justified in the commit history and is
the wrong thing to do for a function that deals with tables.
- Move the error checks out of the recursive path.
Fixes#23654
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:
crash when calling nvim_buf_get_text() with a large negative start_col:
call nvim_buf_get_text(0, 0, -123456789, 0, 0, {})
Solution:
clamp start_col after subtracting it from the line length.
Problem:
137f98cf64 added the `create` parameter to `tmpname()` but didn't
fully implement it.
Solution:
- Update impl for the `os.tmpname()` codepath.
- Inspect all usages of `tmpname()`, update various tests.
Problem:
`nvim --listen` does not error on EADDRINUSE. #30123
Solution:
Now that `$NVIM_LISTEN_ADDRESS` is deprecated and input *only* (instead
of the old, ambiguous situation where it was both an input *and* an
output), we can be fail fast instead of trying to "recover". This
reverts the "recovery" behavior of
704ba4151e, but that was basically
a workaround for the fragility of `$NVIM_LISTEN_ADDRESS`.
For context, see https://github.com/neovim/neovim/pull/24738. Before
that PR, Nvim did not correctly handle captures with quantifiers. That
PR made the correct behavior opt-in to minimize breaking changes, with
the intention that the correct behavior would eventually become the
default. Users can still opt-in to the old (incorrect) behavior for now,
but this option will eventually be removed completely.
BREAKING CHANGE: Any plugin which uses `Query:iter_matches()` must
update their call sites to expect an array of nodes in the `match`
table, rather than a single node.
Problem: Window is updated with potentially invalid skipcol in recursive
window update path. I.e. cursor outside of visible range in
large line that does not fit.
Solution: Make sure it is valid (Luuk van Baal).
3d5065fc75
Problem: inserting with a count is inefficient
Solution: Disable calculation of the cursor position and topline, if a
count has been used (Ken Takata)
Optimize insertion when using :normal 10000ix.
This patch optimizes the insertion with a large count (e.g. `:normal
10000ix`).
It seems that calculation of the cursor position for a long line is slow
and it takes O(n^2). Disable the calculation if not needed.
Before:
```
$ time ./vim --clean -c 'normal 10000ix' -cq!
real 0m1.879s
user 0m1.328s
sys 0m0.139s
$ time ./vim --clean -c 'normal 20000ix' -cq!
real 0m5.574s
user 0m5.421s
sys 0m0.093s
$ time ./vim --clean -c 'normal 40000ix' -cq!
real 0m23.588s
user 0m23.187s
sys 0m0.140s
```
After:
```
$ time ./vim --clean -c 'normal 10000ix' -cq!
real 0m0.187s
user 0m0.046s
sys 0m0.093s
$ time ./vim --clean -c 'normal 20000ix' -cq!
real 0m0.217s
user 0m0.046s
sys 0m0.108s
$ time ./vim --clean -c 'normal 40000ix' -cq!
real 0m0.278s
user 0m0.093s
sys 0m0.140s
$ time ./vim --clean -c 'normal 80000ix' -cq!
real 0m0.494s
user 0m0.311s
sys 0m0.140s
$ time ./vim --clean -c 'normal 160000ix' -cq!
real 0m1.302s
user 0m1.140s
sys 0m0.094s
```
closes: vim/vim#1558809b80d23cf
Co-authored-by: Ken Takata <kentkt@csc.jp>
Problem: Behavior of Enter in completion depends on typing speed.
Solution: Don't make whether Enter selects original text depend on
whether completion has been interrupted, which can happen
interactively with a slow completion 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.
Use the grapheme break algorithm from utf8proc to support grapheme
clusters from recent unicode versions.
Handle variant selector VS16 turning some codepoints into double-width
emoji. This means we need to use ptr2cells rather than char2cells when
possible.
Problem:
Things like underlines are always given a default foreground highlight
regardless of the value of `sp`.
Solution:
Check for `sp` first, and apply that color to the text decoration color if it
exists.
Limitations:
If there is no value of `sp`, vim applies a text decoration color that matches
the foreground of the text. This is still not implemented (and seems like a much
more complex problem): in TOhtml, the underline will still be given a default
foreground highlight.
Problem: Installing treesitter parser is hard (harder than
climbing to heaven).
Solution: Add optional support for wasm parsers with `wasmtime`.
Notes:
* Needs to be enabled by setting `ENABLE_WASMTIME` for tree-sitter and
Neovim. Build with
`make CMAKE_EXTRA_FLAGS=-DENABLE_WASMTIME=ON
DEPS_CMAKE_FLAGS=-DENABLE_WASMTIME=ON`
* Adds optional Rust (obviously) and C11 dependencies.
* Wasmtime comes with a lot of features that can negatively affect
Neovim performance due to library and symbol table size. Make sure to
build with minimal features and full LTO.
* To reduce re-compilation times, install `sccache` and build with
`RUSTC_WRAPPER=<path/to/sccache> make ...`
Problem: The matchparen plugin is slow on a long line.
Solution: Don't use a regexp to get char at and before cursor.
(zeertzjq)
Example:
```vim
call setline(1, repeat(' foobar', 100000))
runtime plugin/matchparen.vim
normal! $hhhhhhhh
```
closes: vim/vim#1556881e7513c86
Problem: cannot set special highlight kind in popupmenu
Solution: add kind_hlgroup item to complete function
(glepnir)
closes: vim/vim#1556138f99a1f0d
Co-authored-by: glepnir <glephunter@gmail.com>
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.
Upon `terminal_enter`, `mapped_ctrl_c` is set in order to avoid `CTRL-C`
interrupts (which is proxied to the terminal process instead), `os_inchar`
will then test `mapped_ctrl_c` against `State` and set `ctrl_c_interrupts=false`
which prevents `process_ctrl_c` from setting `got_int=true` in a terminal
state.
However, if `got_int` is set outside of `process_ctrl_c`, e.g. via
`interrupt()`, this will hang the neovim process as `terminal_execute` will
enter an endless loop as `got_int` will never be cleared causing `safe_vgetc`
to always return `Ctrl_C`.
A minimal example reproducing this bug:
```vim
:autocmd TermEnter * call timer_start(500, {-> interrupt()})
:terminal
:startinsert
```
To fix, we make sure `got_int` is cleared inside `terminal_execute` when
it detects `Ctrl_C`.
Closes#20726
Co-authored-by: zeertzjq <zeertzjq@outlook.com>
