Move from simrats symbols outline which worked very well but had
specific issues for displaying markdown outlines as soon as any
lsp would attach itself to the same buffer.
Aerial seems to not suffer from those issues so this is the one
we will go with for now.
If tracking stable (versioned) releases, it can get out of sync
with LSPs and display errors on startup - since the stable
version gets updated very rarely (over half a year ago currently)
but LSPs and their configs change more rapidly. Since the master
branch seems stable enough we will simply track this instead
for now.
With the new lazy loaded plugin structure, Magma would not start correctly.
This commit simply removes the lazy designation for Magma, Quarto and the
Jupyter attachments.
It also uses the python environment detection utility to set the correct
python executable on entering a quarto file (if there exists a virtual
environment for the project).
When writing a markdown or similar document, vale will automatically
be enabled if there is a '.vale.ini' or '_vale.ini' file (and a
corresponding styles folder) in the project root directory.
Otherwise, vale stays disabled.
The plugin was not loading since lazyloading was not given an event
to start it with. Now, anytime any zk command is given, or we enter
a prose-like filetype the plugin is sourced.
The zk lsp on the other hand was started twice whenever the plugin
was loaded, since we also loaded it through the lspconfig manually.
This commit fixes both issues by sourcing and plugin and letting the
plugin load the lsp as well.
The utility terminals (lazygit and python repl for now) can now be hidden
even from terminal insert mode (i.e. when interacting with them) with
<C-\>. They can be invoked through their usual chords (<leader>tg and
<leader>tp respectively) again and will pick right up where you left off.
Insert mode in terminals can also be left slightly easier should it
be needed: Instead of the <C-\><C-n> chord you can use j\.
Lastly, the utility terminals can be started in a vertically docked mode
instead of floating. This is done by adding a bang to their commands,
`Lazygit!` and `Pythonterm!`, or using capital versions of their
mappings: <leader>tG and <leader>tP.
Moved plugins into individual component module files which are
automatically required by lazy.nvim. Should make everything a tiny bit
more modular, or at least prepare the way for true modularity if I ever
have the time on my hands to ensure everything works with missing
modules.
Moved core settings into their own directory (`core`), and created a
`personal` folder which contains functions/plugins I wrote that do not
necessarily have to be their own imported plugin yet.
Finally, extended the utility functions a little, so we can detect if a
plugin exists and change e.g. key maps based on that (once again,
extending modularity a little more). Some simple attempts have been made
at that in the `mappings.lua` file, though it is nowhere near extensive
yet - most keymaps are still set regardless of plugin availability.
However, with this slimmer base to work off of, I feel more confident in
changing future things about this setup a little more ad-hoc without
having as many ripple repercussions as before.
# This is the 1st commit message:
nvim: Restructure lua dir
Moved plugins into individual component module files which are
automatically required by lazy.nvim. Should make everything a tiny bit
more modular, or at least prepare the way for true modularity if I ever
have the time on my hands to ensure everything works with missing
modules.
Moved core settings into their own directory (`core`), and created a
`personal` folder which contains functions/plugins I wrote that do not
necessarily have to be their own imported plugin yet.
Finally, extended the utility functions a little, so we can detect if a
plugin exists and change e.g. key maps based on that (once again,
extending modularity a little more). Some simple attempts have been made
at that in the `mappings.lua` file, though it is nowhere near extensive
yet - most keymaps are still set regardless of plugin availability.
However, with this slimmer base to work off of, I feel more confident in
changing future things about this setup a little more ad-hoc without
having as many ripple repercussions as before.
# This is the commit message #2:
Update settings file with 0.9 options
# This is the commit message #3:
Move lazy.nvim setup into core module
# This is the commit message #4:
Rename maps.lua to mappings.lua
Formatting and linting should from now be done with null_ls instead
of formatter.nvim (and nothing for linting so far).
This will still take a little to fully transition, for now we use
null_ls for eslint linting and prettier formatting for a variety
of javascript/typescript and astro files.
null_ls uses Mason installations under the hood and any tool it
uses also gets installed by Mason.
Automatic completion from full-text search was draining battery and generally not too helpful.
Disabled (commented) for now, can be re-enabled more specifically.
Notations will be concealed automatically on entering a textual buffer
and `$...$` style notations are contained.
Concealing can be turned off with <localleader>sV, which will toggle
concealing on or off for all notations in the file.
Additionally, the notation under curser can be viewed in a popup with
<localleader>sv.
In addition to my standard file manager, vifm, being integrated into
neovim, I have now also added a side-pane file tree (akin to nerdtree)
that is easily reachable to get a quick overview of a file layout.
For now, I do not intend to do much more with the plugin, only keep it
for those rare cases I want to have a view on my file layout at the same
time as working in a buffer. For all other things (file operations
especially) I still have vifm.
Added fidget plugin which shows the current loading status of LSPs. Only
works for a few LSP (so far), including lua and python. Should
automatically pick up new implementations on update. Will display a
small loading notification in the lower right corner, useful to display
status for those situations where LSP loading takes a long time (e.g.
rust compilation requirements or a large python environment).
FIXME Does not work for each python environment startup yet, and I am
not sure why - sometimes just does not display its loading startup.
When completing in command line I want the completion canditates to be
displayed, but *not* selected on confirmation. By doing so, it makes it
almost impossible to quickly quit with ':q' or write with ':w' for
example, as those always try to expand themselves automatically into
completion items.
This commit changes the default behavior for command mode to show
completions but not auto accept any on confirmation, instead simply
invoking whatever is currently on the command line (as if we had no
completion plugin running).
Whereas previously we had lsp-related mappings both on <localleader>l...
and g... mappings, they are now all unified under the <localleader>l
prefix group. Some mnemonics unfortunately had to give way to a weaker
version of themselves (definition becomes de[f]inition, implementation
becomes i[m]plementation) but overall I believe this to be much more
cohesive for my future lsp usage. With which-key enabled and everything
under the +l group we should be able to easily adapt to the new
mappings.
Additionally, some mappings will invoke the telescope version of their
lsp command if telescope is indeed installed, otherwise fall back to the
native neovim lsp implementation.
Switched the configuration of lsp-zero to its less integrated v2
version. Switched back to manually configuring most of nvim-cmp.
Addded some manual formatting to cmp which displays completion kind as
icons not as text.
Manually add luasnip integration.
Neovim will source the `colorscheme.lua` file in its state directory on
startup, as well as whenever the file contents are changed.
This allows any colorscheme definition to be put into the file and vim
will apply it as soon as the file contents change.
Using the lazy option 'version' we default to updating only to the
latest stable (semver) version of plugins. This should make it a little
more stable in the long run to keep up with plugin updates.
Not all plugins support this versioning scheme and for those that do not
it just keeps tracking the main branch.
Currently from the plugins that support it, only `nvim-lspconfig` needs
to be manually kept on the main branch since it is missing the correct
lua language server otherwise. This should be a problem of the past with
the release of the next version of the plugin.
cmp-pandoc.nvim did not work sufficiently for my use case so far
(sometimes it did, most of the times it did not at all, every now and
again it sputtered some references to the list).
cmp-pandoc-references seems like a plugin kept relatively 'simple'
requiring no setup, a single `bibliography: ` line in the pandoc
meta-data header and it works flawlessly from there. I might delve
deeper at some point, especially with the papis.nvim integration, but
for now this is perfectly adequate.
Neorg is fine but not for me right now (especially with its own syntax
spec). zk seems to fit my workflow much better, this is the beginning of
trying it out.
This commit is a change to some of the foundational mappings I use
through my leader key: opening telescope searches and toggling several
additional display riders.
First of all, all telescope functionality has been subsumed under the
<leader>f prefix:
f (<leader>ff, that is) for finding files
F for grepping
h for finding hidden files
o for finding old files
b for finding buffers
l for searching git log
s for searching git status
c for finding git commits (for current file)
Aside from these (which are pretty massive changes for me), we have a
new leader prefix group for the (now freed) <leader>s:
o toggle symbols outline
m toggle minimap
s open the (mini.nvim) startup screen
We are not making use of vsnip anymore (replaced by luasnip in zero-lsp
package) but still had a couple of vsnip mappings that lead to erratic
<tab> key behavior.
Mini changed the configuration for its indentline module to use a table
instead of a function for its animation generation configuration.
This commit updates the config accordingly.
lsp-zero offers a simple to set-up lsp configuration which is based
around the plugins I use anyway (lsp, cmp, luasnip) but integrates them
very well. I add a few small quality of life things on top (cmp
completion for / search and for vim command mode) as well as a boatload
of custom completion sources but otherwise make use of the nice presets.
This has de-bloated my _cmp.lua settings somewhat and allows me to have
a simple treesitter and a simple(r) lsp setup!
Additionally, switched to mason.nvim to manage my lsp sources (and
linters), which is both the new hotness and actually really useful.
Allows portable installation of most lsp servers into neovim data
directory so they do not clutter any other directories or global
namespaces.
Added the previously important 'site' directory to the runtime path of
neovim so it stops complaining about not finding spell dictionaries on
starting up with a spell-enabled file.
Removed the notation velocity plugin for vim. I have not been making use
of it for ages and should I need a vim-internal way of digging through
my wiki I can probably find easier (and more clear) ways of doing so.
Nice plugin, but of no more use to me.
Switch to use the nice new nvim-internal keymapping functions.
Cartographer was a really useful plugin when they did not yet exist but
now everything I wanted to do can be done without a mapping plugin, so
it is time to remove it.
Switched to the neovim variant of quarto plugins since it may contain
improved integration with lsp and cmp - and offers completion and
diagnostics for code snippets *within* the overall markdown sources of
quarto, which is very exciting.
Migrated from packer.nvim to lazy.nvim - it gives really fast loading
times, easy lazy loading for any package and a snazzy package manager
overview page (invoked through `:Lazy`) which, importantly, all do not
rely on a compiled plugin list.
In essence it feels like an easier to manage and robust packer
replacement (even though packer is a really good manager as well) that
is well worth to switch to. Migration was pretty painless as well.
One thing I might do in the future is split up the plugin structure
itself from a single plugin list (which invokes plugin settings files)
to several files containing the plugins and their settings since it
might give more structure.
So far, my repl integration consisted of vim-slime and vim-ipython-cell.
I am basically purely working in the python repl for now and while the
two plugins (together with tmux) *worked* they didn't work well enough
that I used them in an extensive way. Just as often as I could make use
of their productivity enhacements, I had to wrangle with the plugins
themselves to get what I wanted out of them.
Hence the switch to magma. It integrates more deeply into python, and
neovim both - showing cell outputs in floating windows and so on.
It does not, on its own, recognize e.g. jupyter cells or quarto code
cells, so I have set up some simple key binds to improve
inter-operability between those. Especially [r and ]r to jump between
quarto cells and <localleader>R to execute the contents of the cell the
cursor is currently in may be of interest. They are fairly crude and
probably have edge cases where they will not work but until then they
provide an easy quality of life improvement.
Lastly, I silently added a plugin for markdown previews (that also
scrolls with the editing cursor). I don't foresee myself using it *too*
often but it might come handy for smaller articles and writing. It can
be invoked with <localleader>mp - a binding which I will no doubt also
have to overhaul at some point.
Added simple bibtex source display: Will show the bibtex entries
existing in the bib file set through `bibliography: path/to/mine.bib`.
Shows author, title, date on hovering.
Never used the compile script for years now, and I neither see that
changing nor the files being of any use anymore. They were something I
whipped up when I first started getting into Linux and bash scripting
(and as such a little nostalgic I suppose) but nothing that has actual
purpose anymore.
Added a quick way to fix the last spelling mistake: Use <c-s> while
writing (in insert mode) or localleader-s while in normal mode.
It will fix the mistake and keep your cursor at the current position.
Neovim sometimes errors out when enabling spelling without having all
the necessary spellfiles available. This ensures on setting up the
environment that a spellfile is downloaded.
Had to disable automatic formatting since it was messing with my
contributions to other git projects, if they did either not have a
formatter enabled (most of the time) or had different formatters or
those set up differently than this setup (fixable, but I don't think
it's worth the time).
Instead, formatting can be invoked with `<localleader>f`/`F` to format
or format and save respectively.
Updated buffer navigation to make use of the <c-w>hjkl paradigm instead
of the <c-hjkl> since it adheres closer to vanilla nvim and works better
with the new navigation features of Navigator.nvim.
Switched out the old distraction-free writing plugins for neovim
variants zen-mode and twilight by folke - they are simple (one command
to invoke zen-mode, mapped to `F11`, and that's it), they are written in
lua and they work together beautifully (invoking zen-mode also invokes
twilight paragraph highlighting).
Marty Oehme