Problem:
nvim_parse_cmd() in pcall() may show an error message (side-effect):
:lua pcall(vim.api.nvim_parse_cmd, vim.fn.getcmdline(), {})
E16: Invalid range
Solution:
Avoid emsg() in the nvim_parse_cmd() codepath.
- refactor(api): add error message output parameter to get_address()
- fix: null check emsg() parameter
- refactor: remove emsg_off workaround from do_incsearch_highlighting()
- refactor: remove emsg_off workaround from cmdpreview_may_show()
- refactor: remove remaining calls to emsg() from parse_cmd_address() and get_address()
- (refactor): lint set_cmd_dflall_range()
- refactor: addr_error() - move output parameter to return value
Fix#20339
TODO:
These are the functions called by `get_address()`:
```
nvim_parse_cmd() -> parse_cmdline() -> parse_cmd_address() -> get_address()
skipwhite()
addr_error()
qf_get_cur_idx()
qf_get_cur_valid_idx()
qf_get_size()
qf_get_valid_size()
mark_get()
mark_check()
assert()
skip_regexp()
magic_isset()
> do_search()
> searchit()
ascii_isdigit()
getdigits()
getdigits_int32()
compute_buffer_local_count()
hasFolding()
```
From these functions, I found at least two that call emsg directly:
- do_search()
- seems to be simple to refactor
- searchit()
- will be more challenging because it may generate multiple error messages,
which can't be handled by the current `errormsg` out-parameter.
For example, it makes multiple calls to `vim_regexec_multi()` in a loop that
possibly generate error messages, and later `searchit()` itself may generate
another one:
- c194acbfc4/src/nvim/search.c (L631-L647)
- c194acbfc4/src/nvim/search.c (L939-L954)
---------
Co-authored-by: Justin M. Keyes <justinkz@gmail.com>
Problem:
`:lua vim.cmd.win_getid(30,10)` is interpreted as `:win[size] 30 10`.
User intention was to call `vim.fn.win_getid(30,10)`.
Solution:
Check that the `cmd` actually matches the resolved command.
Enforce consistent terminology (defined in
`gen_help_html.lua:spell_dict`) for common misspellings.
This does not spellcheck English in general (perhaps a future TODO,
though it may be noisy).
Adds new API helper macros `CSTR_AS_OBJ()`, `STATIC_CSTR_AS_OBJ()`, and `STATIC_CSTR_TO_OBJ()`, which cleans up a lot of the current code. These macros will also be used extensively in the upcoming option refactor PRs because then API Objects will be used to get/set options. This PR also modifies pre-existing code to use old API helper macros like `CSTR_TO_OBJ()` to make them cleaner.
problem: can we have Serde?
solution: we have Serde at home
This by itself is just a change of notation, that could be quickly
merged to avoid messy merge conflicts, but upcoming changes are planned:
- keysets no longer need to be defined in one single file. `keysets.h` is
just the initial automatic conversion of the previous `keysets.lua`.
keysets just used in a single api/{scope}.h can be moved to that file, later on.
- Typed dicts will have more specific types than Object. this will
enable most of the existing manual typechecking boilerplate to be eliminated.
We will need some annotation for missing value, i e a boolean will
need to be represented as a TriState (none/false/true) in some cases.
- Eventually: optional parameters in form of a `Dict opts` final
parameter will get added in some form to metadata. this will require
a discussion/desicion about type forward compatibility.
libnvim couldn't be easily used in C++ due to the use of reserved keywords.
Additionally, add explicit casts to *alloc function calls used in inline
functions, as C++ doesn't allow implicit casts from void pointers.
Problem:
The Lua-API bridge allows Dict params to be empty Lua (list) tables at
the function-signature level. But not for _nested_ Dicts, because they
are not modeled:
fae7540732/src/nvim/api/keysets.lua (L184)
Some API functions like nvim_cmd check for kObjectTypeDictionary and
don't handle the case of empty Lua tables (treated as "Array").
Solution:
Introduce VALIDATE_T_DICT and use it in places where
kObjectTypeDictionary was being checked directly.
fixes#21005
Problem:
Validation messages are not consistently formatted.
- Parameter names sometimes are NOT quoted.
- Descriptive names (non-parameters) sometimes ARE quoted.
Solution:
Always quote the `name` value passed to a VALIDATE macro _unless_ the
value has whitespace.
Problem:
- API validation involves too much boilerplate.
- API validation errors are not consistently worded.
Solution:
Introduce some macros. Currently these are clumsy, but they at least
help with consistency and avoid some nesting.
This is cherry-picked from these Vim patches:
Only applicable change outside vi_diff.txt in patch 8.1.1226:
6c60f47fb9
Most changes outside starting.txt and vi_diff.txt in patch 8.1.1280:
25c9c680ec
Missing docs for 'mousemoveevent':
cbaff5e06e
Allow Include What You Use to remove unnecessary includes and only
include what is necessary. This helps with reducing compilation times
and makes it easier to visualise which dependencies are actually
required.
Work on https://github.com/neovim/neovim/issues/549, but doesn't close
it since this only works fully for .c files and not headers.
Adds a `name` key to the opts dict passed to Lua command callbacks
created using `nvim_create_user_command()`. This is useful for when
multiple commands use the same callback.
Note that this kind of behavior is not as strange as one might think,
even some internal Neovim commands reuse the same internal C function,
differing their behavior by checking the command name. `substitute`,
`smagic` and `snomagic` are examples of that.
This will also be useful for generalized Lua command preview functions
that can preview a wide range of commands, in which case knowing the
command name is necessary for the preview function to actually be able
to execute the command that it's supposed to preview.
I don't think using an integer as a NUL-terminated string can work on
big-endian systems, at least.
This is also not tested. Add a test.
Also fix a mistake in the docs of nvim_parse_cmd.
Now nvim_parse_cmd and nvim_create_user_command use a "tab" value which
is the same as the number passed before :tab modifier instead of the
number plus 1, and "tab" value is -1 if :tab modifier is not used.
vim-patch:8.2.0559: clearing a struct is verbose
Problem: Clearing a struct is verbose.
Solution: Define and use CLEAR_FIELD() and CLEAR_POINTER().
a80faa8930
Problem: Support for user commands is spread out. No good reason to make
user commands optional.
Solution: Move user command support to usercmd.c. Always enable the
user_commands feature.
ac9fb18020
Reduce the amount of string allocations and length calculations.
With the following benchmark:
```lua
total = 0
for _ = 1, loops do
local start = now()
vim.api.nvim_cmd({cmd = 'let', args = {'a', '=', '1'}}, {})
total = total + (now() - start)
end
print('nvim_cmd', total / loops)
```
```
hyperfine 'nvim --clean test.lua +source +q'
```
Before: 234.5ms
After: 173.8ms
Problem: Value of cmod_verbose is a bit complicated to use.
Solution: Use zero for not set, value + 1 when set. (closesvim/vim#10564)
cd7496382e
Omit has_cmdmod(): only used for Vim9 script
Problem: Command modifiers are saved and set inconsistently.
Solution: Separate parsing and applying command modifiers. Save values in
cmdmod_T.
5661ed6c83
Cherry-pick: :0verbose fix from patch 8.2.4741
