Instead, we now presume that if neovim is installed, the user will use
it over the default vim.
We also slightly refactor the entrypoint for neovim, using the default
paths for MYVIMRC, over setting VIMINIT and MYVIMRC explicitly.
these are a collection of functions that we have built in house, those
their insides replaced to work better with the new plugin manager
namely
- Format
- OpenConfig
- ToggleGutter
We remove the old bind.lua functionality for creating recursive auto
initialized groups of bindings, replacing it with a simpler data
structure that only provides a common, simple mechanism for storing all
the information needed to make a new key bind in the future.
This change is good for two reasons:
1. the luals LSP can now provide autocompletion for keybinds
2. the module is much simpler, and can be extended easily to map the
existing data structure into other formats, like the one used by
lazy.nvim's keys map, or Neovim's own vim.keymap.set
We also update, refactor and rename the old map.lua file to binds.lua,
so that we remain consistent in our terminology with Neovim itself
Lastly, we touch up key/init.lua to handle these changes, and make it
easier for others to potentially create their own keybinds
These are simple "prebaked" functions that can modify the PsoxizshConfig
object before it is utilized further.
It allows the primary users of this repo to have seperate "profiles"
that they can maintain in the main code base, while also allowing others
to just pass a similiar function to PsoxizshConfig.profile.
This commit rewrites the psoxizsh init to conform to lazy.nvim plugin
schema.
This is a port of the old early/pre/post/late stages we used to use to
functions we can call during our toplevel module's eventual setup()
call.
We should probably add optional callbacks to each so that users can
retain the older method of initialization...
This is our inhouse overrides over LazyVim's key maps for the LSP on_attach
function.
Unfortunately there's no good way to do this, other than directly
mutating the key binds LazyVim creates
Adds back some of the functionality of the late lsp.server, such that we
can paper over the exact implementation details of _how_ we're enabling
the language server.
In practice, this will allow us to create a bunch of preset
PsoxizshLspServer types, which users can then opt into enabling in their
plugin/* configs
Largely, this is simply bumping versions and fixing a few configurations
that were changed in newer versions of the plugins.
I did however notice our package manager (packer.nvim) is orphaned now,
and I need to switch to a different one in the near future...
These gives us back the ability to use cspell, which was lost in the
transition from coc-nvim to our current lua based setup.
However, it requires the following to take effect:
1. Have nodejs installed (nvm is fine)
2. Run ':MasonInstall cspell'
3. Run ':PackerCompile'
4. Restart
All goes well, when you next open nvim, you should have cspell powered
spell checking and :LspInfo should show null-ls running.
Word substitutes triggered via:
@Buffer.Lsp.N.Leader.CodeAction
Next/Prev word (or other diagnostic) via:
@Global.N.DiagnosticNext
@Global.N.DiagnosticPrev
Where @ = nvim/lua/psoxizsh/key/bind.lua
neovim moved buf.formatting to buf.format, a while back, leaving this as
effectively setting `action = null`, which throws warnings on start; so
we fix that.
This commit enables nvim-cmp's cmdline completion support, overriding
the builtin neovim's.
We gain a nicer menu, more sources (paths, history), and a slightly more
intelligent listing with some support for file specific commands (e, r,
w).
We also switch out the 'path' completion source with 'async_path',
preventing delays / freezes when traversing slow or large fs trees
This is a plugin which overrides vim.prompt and vim.input handlers,
providing Telescope integration, and a much more pleasant experience
over the default menus provided by nvim.
The primary reason for this plugin is vim.lsp.buf.code_action/1, which
uses vim.input.
This allows callers of the struct to perform the more normal '='
assignment and have sane behavior.
Previously, this would not do what one expects when used with bind.Bind
or other bind.Group(s): it merely adds them to the Group object
directly, not adding it to the ._children table or downward propagating
the Group's settings.
This is fixed here, by adding a __newindex metamethod to the bind.Group
object, which checks to see if the addition is a known Bind or Group
type, and does the right if it is, falling back to the previous behavior
if it isn't
Paul Stemmet
psox st8l