Added euporia term to split it from other python variations. Fixed
python command selection. Improved mapping and which key display.
Removed indentline settings since that is responsibility of indentline.
Removed the plugin since all its functionality can be replicated with an
equal amount of lines without it nowadays. Should be even simpler in the
future when we can just have individual lsp server config files in the
config dir.
Creates a hidden kernel file with the current filename and
'.kernel.json' suffix by default (in the file directory).
But 'JupyterStart' command can also be given an argument to use as the
filename / file path for the kernel file. That way we can also attach to
already running kernels for longer running projects.
When formatting femaco buffers it would sometimes remove the newline
necessary at the end of code fences to have the (```) remain on its own
line. This attempts to fix it.
Switched completion engines. Blink is supposedly faster and so on but
what I really love is the batteries-included style.
I don't have to set up the basic lsp, path, buffer and cmdline
completions, nor styling for basic kind displays and more.
Hundreds of lines of config shrink down to 60, very nice.
While we are still hacking our way around a specific pipx venv (which
has to change), we are at least not hardcoding my own home directory
name anymore.
While I never disliked tmux I have not been using it for absolute ages
now, ever since starting to multiplex with wezterm. Wezterm can (at
least with my current setup) *not* replace all tmux functionality -
especially running multiple sessions in a detachable way on a remote
server - but I have never needed those in a long time now.
Detachable sessions I can create instead with `abduco`.
Instead of completely detaching marksman as soon as we are in a
zettelkasten directory, this disables displaying the 'non-existent link'
diagnostic instead. Only disables signs and virtual_text for now (still
displayed as diagnostic in e.g. Trouble windows) but should provide a
good first step into more fine-grained control.
Oh if only marksman just provided a flag to toggle this feature..
By default disable the line length for any markdown file. I never use it
and I always turn it off. If I need it for a specific project I can
still figure out how to make the configuration override this.
Extract lualine into its own 'statusline.lua' plugin file. This makes
the ui plugin file a little more lean but also prepares the way of
tinkering with statusline alternatives, like heirline.
Since sometimes it is necessary to have correct access to the files
functionality earlier than was possible with the old lazy loading this
instead loads it at program start as well.
Check that the trouble plugin exists before adding its maps to the
telescope buffer. Should probably go into trouble setup instead but
works as a quick safeguard for now.
Remove `<localleader>s` 'set' group of mappings (setting buffer wrapping
mode, setting colorizing) and replace it with explicit on/off mappings
via [o<setting> and ]o<setting>.
Wrapping can be set to soft with `[ow`, hard with `]ow`.
Colorizer can be disabled with `[oc`, enabled with `]oc`.
Colorizer can highlight virtual text with `[oC`, highlight background
with `]oC`.
This should make it hopefully still clear that they are expected plugins
for the configuration while also delineating it from the 'core' module
which bootstraps lazy and sets fundamental options and so on.
Before we load the lazy spec from the 'lua/plugins' directory, we check
that the directory exists and contains at least one .lua file. We do not
check that the file returns a valid spec but that is just assumed in
this case.
Begin a larger refactor with nvim-cmp, which separates out the different
nvim-cmp modules to be loaded at different times and for different files
(such as beancount completion on loaded for beancount files and so on).
This is by no means complete but the start of a larger overhaul process.
Except for mini.starter which we need right at the - well - start, we
can lazy load most other modules. It's a little awkward since we lazy
load them *within* the setup function with a manual autocommand but it
should work fine for the time being.
