Problem: completion: cannot use autoloaded funcs in 'complete' F{func}
(Maxim Kim)
Solution: Make it work (Girish Palya)
fixes: vim/vim#17869closes: vim/vim#178851bfe86a7d3
Cherry-pick Test_omni_autoload() from patch 8.2.3223.
Co-authored-by: Girish Palya <girishji@gmail.com>
Problem: The :keep{alt,jumps,marks,patterns} commmands are sometimes
misidentified as :k.
Solution: Make sure one_letter_cmd() only returns true for :k and not
other :keep* commands (Doug Kearns).
This currently manifests as missing completion for :keep* commands and
incorrect results from fullcommand().
E.g., fullcommand("keepmarks") returns "k" rather than "keepmarks".
The correct command, however, is executed as command modifiers are
handled specially in do_one_cmd() rather than using find_ex_command().
Fix exists(':k') so that it returns 2 for a full match.
closes: vim/vim#15742ea84202372
Cherry-pick Test_ex_command_completion() from patch 9.1.0624.
Co-authored-by: Doug Kearns <dougkearns@gmail.com>
Problem: completion: hang (after 9.1.1471) or E684 (after 9.1.1410)
when 'tagfunc' calls complete().
Solution: Check if complete() has been called immediately after getting
matches instead of in the next loop iteration (zeertzjq).
related: vim/vim#1668
related: neovim/neovim#34416
related: neovim/neovim#35163closes: vim/vim#17929982cda6976
Problem: completion: incsearch highlight might be lost after search
completion (Hirohito Higashi)
Solution: Restore incsearch highlight after dismissing pum with Ctrl-E
(Girish Palya)
related: vim/vim#17870closes: vim/vim#1789104c9e78cd3
This change actually isn't needed as Nvim doesn't call update_screen()
to redraw pum, but it doesn't hurt either.
Co-authored-by: Girish Palya <girishji@gmail.com>
When Nvim is started in one terminal emulator,
suspended, and later resumed in a different terminal emulator (as can
happen when using e.g. a multiplexer), the new terminal emulator will
not have all of the same modes enabled that the original terminal
emulator had. This is a problem in particular for in-band resize events
because we were disabling the SIGWINCH signal handler when we determined
that the terminal supported this mode.
However, if the new terminal does not support this mode then the
SIGWINCH handler remains disabled, and Neovim no longer properly
resizes. Instead of disabling the SIGWINCH handler, we track the state
of the resize events mode internally. If the mode is enabled then we
return early from the SIGWINCH handler before performing any ioctls or
other system calls. But if the mode is not enabled we proceed as normal.
Problem: :bnext doesn't go to unlisted help buffers when cycling
through help buffers (after 9.1.0557).
Solution: Don't check if a help buffer is listed (zeertzjq).
From <https://github.com/vim/vim/issues/4478#issuecomment-498831057>:
> I think we should fix that, since once you get to a non-help buffer
> all unlisted buffers are skipped, thus you won't encounter another
> help buffer.
This implies that cycling through help buffers should work even if help
buffers are unlisted. Otherwise this part of :bnext isn't really useful,
as :h makes help buffers unlisted by default.
related: vim/vim#4478
related: vim/vim#15198closes: vim/vim#179139662f33480
Problem: diff: using diff anchors with hidden buffers fails silently
Solution: Give specific error message for diff anchors when using hidden
buffers (Yee Cheng Chin).
Diff anchors currently will fail to parse if a buffer used for diff'ing
is hidden. Previously it would just fail as the code assumes it would
not happen normally, but this is actually possible to do if `closeoff`
and `hideoff` are not set in diffopt. Git's default diff tool "vimdiff3"
also takes advantage of this.
This fix this properly would require the `{address}` parser to be
smarter about whether a particular address relies on window position or
not (e.g. the `'.` address requires an active window, but `'a` or `1234`
do not). Since hidden diff buffers seem relatively niche, just provide a
better error message / documentation for now. This could be improved
later if there's a demand for it.
related: vim/vim#17615closes: vim/vim#17904cad3b2421d
Co-authored-by: Yee Cheng Chin <ychin.git@gmail.com>
vim-patch:8.1.0710: when using timers may wait for job exit quite long
vim-patch:8.1.0969: message written during startup is truncated
vim-patch:8.1.1043: Lua interface does not support Blob
vim-patch:8.1.1389: changes are not flushed when end and start overlap
vim-patch:8.1.1395: saving for undo may access invalid memory
vim-patch:8.1.1534: modeless selection in popup window selects too much
vim-patch:8.1.1571: textprop highlight starts too early if just after a tab
vim-patch:8.1.1573: textprop test fails if screenhots do not work
vim-patch:8.1.1663: compiler warning for using size_t
vim-patch:8.1.1866: modeless selection in GUI does not work properly
vim-patch:8.1.1871: modeless selection in GUI still not correct
vim-patch:8.1.2085: MS-Windows: draw error moving cursor over double-cell char
vim-patch:8.1.2107: various memory leaks reported by asan
vim-patch:8.1.2146: build failure
vim-patch:8.1.2296: text properties are not combined with syntax by default
vim-patch:8.1.2299: ConPTY in MS-Windows 1909 is still wrong
vim-patch:8.1.2337: double-click time sometimes miscomputed
vim-patch:8.2.0300: Vim9: expression test fails without channel support
vim-patch:8.2.0414: channel connect_waittime() test is flaky
vim-patch:8.2.0501: Vim9: script test fails when channel feature is missing
vim-patch:8.2.0508: Vim9: func and partial types not done yet
vim-patch:8.2.0527: Vim9: function types insufficiently tested
vim-patch:8.2.0565: Vim9: tests contain superfluous line continuation
vim-patch:8.2.0640: Vim9: expanding does not work
vim-patch:8.2.0700: Vim9: converting error message to exception not tested
vim-patch:8.2.0701: Vim9 test fails without job feature
vim-patch:8.2.1481: Vim9: line number reported with error may be wrong
vim-patch:8.2.1609: Vim9: test fails when build without +channel
vim-patch:8.2.1724: Vim9: assignment tests spread out
vim-patch:8.2.1759: Vim9: Some tests are still using :let
vim-patch:8.2.1865: Vim9: add() does not check type of argument
vim-patch:8.2.1867: Vim9: argument to add() not checked for blob
vim-patch:8.2.1965: Vim9: tests fail without the channel feature
vim-patch:8.2.2301: Vim9: cannot unlet a dict or list item
vim-patch:8.2.2304: Vim9: no test for unletting an imported variable
vim-patch:8.2.3006: crash when echoing a value very early
vim-patch:8.2.3007: Vim9: test for void value fails
vim-patch:8.2.3008: startup test may hang
vim-patch:8.2.3009: startup test may hang
vim-patch:8.2.3181: Vim9: builtin function test fails without channel feature
vim-patch:8.2.3242: Vim9: valgrind reports leaks in builtin function test
vim-patch:8.2.3382: crash when getting the type of a NULL partial
vim-patch:8.2.3845: Vim9: test fails when the channel feature is missing
vim-patch:8.2.4407: Vim9: some code not covered by tests
vim-patch:8.2.4410: Vim9: some code not covered by tests
vim-patch:9.0.0563: timer_info() test fails
vim-patch:9.0.2080: vim9_script test too large
vim-patch:9.1.1541: Vim9: error when last enum value ends with a comma
The '?' needs to be escaped, because the autocommand is using
file-patterns (glob like) and not a regex). See :h file-pattern
closes: vim/vim#17890f7deb815b0
Co-authored-by: Girish Palya <girishji@gmail.com>
Using the preprocessor before generating prototypes provides some
"niceties" but the places that rely on these are pretty few.
Vastly simplifying the BUILD SYSTEM is a better trade-off.
Unbalancing { } blocks due to the preprocessor is cringe anyway (think
of the tree-sitter trees!), write these in a different way.
Add some workarounds for plattform specific features.
INCLUDE_GENERATED_DECLARATIONS flag is now technically redundant,
but will be cleaned up in a follow-up PR as it is a big mess.
Problem: Vim9: no optional arguments in func type.
Solution: Check for question mark after type. Find function reference
without function().
5deeb3f1f9
Co-authored-by: Bram Moolenaar <Bram@vim.org>
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.
Platform specific options are cringe and should either be fixed or
deleted.
In this case, the platform difference can trivially be implemented
using a conditional default value. If the user overrides the value,
then it is because the user wants that value, regardless of the
corporation who manufactured the OS that the user is running.
Possible alternative: delete the option by making it always on.
Problem:
Not easy for a user to tell ":restart" to "run this command(s) after restarting".
Solution:
All ":restart" args following the optional +cmd arg are treated as a big cmdline that is passed as a "-c" CLI arg when restarting nvim.
Problem:
Generated docs sections are ordered randomly. This matters when showing
an outline or table of contents (e.g. `gO`).
Solution:
Specify which sections have an intentional ordering; sort the rest by
name.
Problem: cannot perform autocompletion
Solution: Add the 'autocomplete' option value
(Girish Palya)
This change introduces the 'autocomplete' ('ac') boolean option to
enable automatic popup menu completion during insert mode. When enabled,
Vim shows a completion menu as you type, similar to pressing |i\_CTRL-N|
manually. The items are collected from sources defined in the
'complete' option.
To ensure responsiveness, this feature uses a time-sliced strategy:
- Sources earlier in the 'complete' list are given more time.
- If a source exceeds its allocated timeout, it is interrupted.
- The next source is then started with a reduced timeout (exponentially
decayed).
- A small minimum ensures every source still gets a brief chance to
contribute.
The feature is fully compatible with other |i_CTRL-X| completion modes,
which can temporarily suspend automatic completion when triggered.
See :help 'autocomplete' and :help ins-autocompletion for more details.
To try it out, use :set ac
You should see a popup menu appear automatically with suggestions. This
works seamlessly across:
- Large files (multi-gigabyte size)
- Massive codebases (:argadd thousands of .c or .h files)
- Large dictionaries via the `k` option
- Slow or blocking LSP servers or user-defined 'completefunc'
Despite potential slowness in sources, the menu remains fast,
responsive, and useful.
Compatibility: This mode is fully compatible with existing completion
methods. You can still invoke any CTRL-X based completion (e.g.,
CTRL-X CTRL-F for filenames) at any time (CTRL-X temporarily
suspends 'autocomplete'). To specifically use i_CTRL-N, dismiss the
current popup by pressing CTRL-E first.
---
How it works
To keep completion snappy under all conditions, autocompletion uses a
decaying time-sliced algorithm:
- Starts with an initial timeout (80ms).
- If a source does not complete within the timeout, it's interrupted and
the timeout is halved for the next source.
- This continues recursively until a minimum timeout (5ms) is reached.
- All sources are given a chance, but slower ones are de-prioritized
quickly.
Most of the time, matches are computed well within the initial window.
---
Implementation details
- Completion logic is mostly triggered in `edit.c` and handled in
insexpand.c.
- Uses existing inc_compl_check_keys() mechanism, so no new polling
hooks are needed.
- The completion system already checks for user input periodically; it
now also checks for timer expiry.
---
Design notes
- The menu doesn't continuously update after it's shown to prevent
visual distraction (due to resizing) and ensure the internal list
stays synchronized with the displayed menu.
- The 'complete' option determines priority—sources listed earlier get
more time.
- The exponential time-decay mechanism prevents indefinite collection,
contributing to low CPU usage and a minimal memory footprint.
- Timeout values are intentionally not configurable—this system is
optimized to "just work" out of the box. If autocompletion feels slow,
it typically indicates a deeper performance bottleneck (e.g., a slow
custom function not using `complete_check()`) rather than a
configuration issue.
---
Performance
Based on testing, the total roundtrip time for completion is generally
under 200ms. For common usage, it often responds in under 50ms on an
average laptop, which falls within the "feels instantaneous" category
(sub-100ms) for perceived user experience.
| Upper Bound (ms) | Perceived UX
|----------------- |-------------
| <100 ms | Excellent; instantaneous
| <200 ms | Good; snappy
| >300 ms | Noticeable lag
| >500 ms | Sluggish/Broken
---
Why this belongs in core:
- Minimal and focused implementation, tightly integrated with existing
Insert-mode completion logic.
- Zero reliance on autocommands and external scripting.
- Makes full use of Vim’s highly composable 'complete' infrastructure
while avoiding the complexity of plugin-based solutions.
- Gives users C native autocompletion with excellent responsiveness and
no configuration overhead.
- Adds a key UX functionality in a simple, performant, and Vim-like way.
closes: vim/vim#17812af9a7a04f1
Co-authored-by: Girish Palya <girishji@gmail.com>
Problem: Using `addr` without `range` in nvim_create_user_command gives
"No range allowed" error, inconsistent with `:command -addr` behavior.
Solution: Set EX_RANGE flag when `addr` option is specified to match
`:command` behavior.
This workarounds a bug likely in nvim__get_runtime, and fixes#35124
Though I'd argue it is more correct anyway as the point of
vim.SUBMODULE lazy loading is "only pay for what you use". If no one
has require'vim.diagnostic' yet in LSP or otherwise, there cannot
be any diagostics available and loading the lua module is wasteful.
Problem: Visual block insert on a single line incorrectly triggers two
on_lines callbacks - one for the correct line (0-indexed) and another
for a non-existent additional line.
Solution: Only call changed_lines() in block_insert() when additional
lines beyond the first were actually modified (start.lnum < end.lnum).
Problem:
":restart" always executes ":qall" to exit the server.
Solution:
Support ":restart +cmd" so the user can control the command
used to exit the server.
Co-authored-by: Justin M. Keyes <justinkz@gmail.com>
Co-authored-by: zeertzjq <zeertzjq@outlook.com>
Problem: cannot easily trigger wildcard expansion
Solution: Introduce wildtrigger() function
(Girish Palya)
This PR introduces a new `wildtrigger()` function.
See `:h wildtrigger()`
`wildtrigger()` behaves like pressing the `wildchar,` but provides a
more refined and controlled completion experience:
- Suppresses beeps when no matches are found.
- Avoids displaying irrelevant completions (like full command lists)
when the prefix is insufficient or doesn't match.
- Skips completion if the typeahead buffer has pending input or if a
wildmenu is already active.
- Does not print "..." before completion.
This is an improvement on the `feedkeys()` based autocompletion script
given in vim/vim#16759.
closes: vim/vim#17806b486ed8266
While at it, also make Ctrl-Z trigger search completion.
Co-authored-by: Girish Palya <girishji@gmail.com>
Problem: unlike win_close, win_close_othertab could be used to close the
autocommand window from a different tabpage. This causes aucmd_restbuf to close
the wrong window, potentially causing a crash.
Solution: disallow closing it. Also replace a deprecated use of exc_exec in the
test file.
Fixes#21409.
Problem: no check for nvim_open_win opening a new floating window into a closing
buffer, which can lead to crashes.
Solution: call check_split_disallowed; opening a new float semantically splits
from a window, so the same problems as regular splits apply. Also restore the
error if switch_win_noblock in win_set_buf fails (may not be possible to hit
this, but win_set_buf can silently succeed there since #31595).
As the lock check applies to curbuf (not the target buffer) this may feel a bit
restrictive, though this isn't different to how things like ":split" or
nvim_open_win with "split = true" works when targeting a different buffer. Only
checking the target buffer's lock will cause issues if win_set_buf doesn't end
up in the target buffer for whatever reason.
Maybe we could consider checking the lock of the buffer after win_set_buf and
close the window if it's locked (maybe a bit fiddly, especially as closing a
window can fail...), or make the open + switch operation more atomic, like how
Vim does for its popup windows..?
It also used to be the case that win_set_buf would set an error if autocommands
sent us to a different buffer than what was requested, but #31595 appears to
have also changed that... I haven't touched that here.
Problem: can't accurately know if close_buffer directly (e.g: not via autocmds)
decremented b_nwindows. This can cause crashes if win_close_othertab decides to
keep the window after calling close_buffer (if it did not free the buffer), as
b_nwindows may remain out-of-sync.
Solution: change the return value of close_buffer to accurately depict whether
it decremented b_nwindows. Check it in win_close_othertab to avoid a crash.
Similar issues may exist in other places that call close_buffer, but I've not
addressed those here (not to mention only one other place even checks its return
value...)
Problem: TabClosed is fired after close_buffer is called (after b_nwindows is
decremented) and after the tab page is removed from the list, but before it's
freed. This causes inconsistencies such as the removed tabpage having a valid
handle and functions like nvim_tabpage_get_number returning nonsense.
Solution: fire it after free_tabpage. Try to maintain the Nvim-specific
behaviour of setting `<amatch>` to the old tab page number, and the
(undocumented) behaviour of setting `<abuf>` to the buffer it was showing
(close_buffer sets w_buffer to NULL if it was freed, so it should be OK pass it
to apply_autocmds_group, similar to before).
Problem: No check for closing the only non-floating window in a non-current
tabpage that contains floats. This can lead to a tabpage that contains only
floats, causing crashes.
Solution: Copy the relevant check from win_close to win_close_othertab. Fix some
uncovered issues.
Closes#34943Fixes#31236
Co-authored-by: glepnir <glephunter@gmail.com>
