Skip to content

corgi-emacs/corkey

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

71 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Corkey

Corkey is Corgi's key binding system.

The main ideas behind Corkey:

  • Define key bindings as "just data", in dedicated key binding files
  • Never define keys implicitly just by loading a package
  • Unify evil and plain emacs bindings
  • Provide which-key style discoverability, with docstrings being first class
  • Decouple conceptual operations (e.g. "eval current form") from concrete and mode-specific implementations

The last point is achieved to a Corkey-specific concept: signals, which we'll elaborate on further down.

Getting started

Create a file named user-keys.el and place it in your Emacs user directory (user-emacs-directory), or anywhere on Emacs's load-path.

(bindings
 ;; "global" bindings are always active regardless of Evil's "state" (= vim mode)
 ;; If you don't provide this the default is `normal'.
 (global
  ("M-RET" projectile-switch-project))

 ;; Bindings for commands are usually only active in normal and visual state.
 (normal|visual
  ("SPC"
   ("o" "Open"
    ("u" "Open URL at point" browse-url-at-point)
    ("s" "Edit string at point" string-edit-at-point)))))

Now start Corkey:

(corkey-mode)
(corkey/load-and-watch '(user-keys))

corkey-mode is a globalized minor mode which will enable Corkey in all buffers.

Bindings are nested, e.g. ("SPC" ("b" ("k" kill-buffer))) means that "space" followed by "b" and then "k" will invoke M-x kill-buffer. Instead of a prefix key (string) you can use an evil state name (what vim calls a "mode", should be a symbol). global is a special case, it means "these bindings should work regardless of the evil state". To apply bindings to multiple states, separate them with a |.

You can add a descriptions before the command, this will show up in a pop-up when you press the prefix key and wait a bit. (This uses which-key)

Now M-RET should work anywhere for switching to another project, and SPC o u / SPC o s will work when you are in evil's normal or visual state. If you press SPC o and wait you will get a popup showing your options, with human-readable descriptions.

To add or change bindings, simply update and save user-keys.el, and Corkey will reload it.

Signals

Instead of a concrete command like kill-buffer or projectile-switch-project you can put a keyword like :eval/buffer (notice the leading colon). Corkey calls this a "signal". It's a way of saying "conceptually what this key does is evaluate the current buffer, but what that does in a given instance is context-specific".

;; user-keys.el
(bindings
 ("," ("e" ("b" "Evaluate buffer" :eval/buffer))))

At this point this doesn't do anything yet, Corkey will not create a , e e binding. For that you need to define what this signal does.

;; user-signals.el
((emacs-lisp-mode (:eval/buffer eval-buffer))
 (clojure-mode    (:eval/buffer cider-eval-buffer)))

Now install this signals file, the optional second argument to corkey/load-and-watch is a list of signal file names:

(corkey/load-and-watch '(user-keys) '(user-signals))

Now it will stitch the two together, in Emacs Lisp buffers , e e will call eval-buffer, whereas in Clojure buffers it will call cider-eval-buffer.

You can bind signals per major or minor mode. Use default to provide a fallback value, if no specific mode applies.

Having this indirection promotes a degree of consistency that in other configs is achieved by manually ensuring analogous bindings across modes, which is difficult to enforce and maintain, and where it's often left to contributors to try to deduce the conventions used by other modes.

Layering

Note that the arguments to corkey/load-and-watch are both lists, it is possible to provide multiple key files, and multiple signal files, for instance you can load both the corgi-keys and user-keys keys files, and both the corgi-signals and user-signals signal files. In fact this is the default when calling corkey/load-and-watch without arguments.

(corkey/load-and-watch '(corgi-keys user-keys) '(corgi-signals user-signals))

This will cause Corkey to layer later files over earlier ones, in other words: any definitions in user-keys and user-signals will take precedence over corgi-keys and corgi-signals.

Typically the corgi-* files are provided by the corgi-bindings package, installed via straight. These provide the base set of bindings for Corgi. The user-* files are files you can place in your emacs-user-directory, to add your own customizations.

For any file name emacs-user-directory will always be searched first, followed by scanning the emacs load-path. This means you can also copy the corgi-* files to your emacs-user-directory to customize them directly.

File names that Corkey can't find are currently silently ignored. This means you can use the default values even if you don't have a user-keys.el or user-signals.el. If you decide to create them later they will be picked up then.

Signal and third party packages

We've already explained the main purpose of signals, to allow mode-specific bindings. But signals become really interesting when considering the implications for third party packages.

Providing signal bindings

Say someone wants to create a Corgi integration for a programming language that isn't covered yet by the base setup. That means they just need to create a package that contains a my-lang-signals.el

((my-lang-mode (:eval/last-sexp my-lang-eval-last-sexp
                :eval/buffer my-lang-eval-buffer
                :eval/region my-lang-eval-region
                :repl/connect my-lang-connect-repl
                :repl/jack-in my-lang-start-repl)))

Now a Corkey user can load this file

(corkey/load-and-watch '(corgi-keys user-keys) '(corgi-signals my-lang-signals user-signals))

And get all their familiar Corgi bindings available for said language. But what's more, if they have user-specific bindings (maybe they have a different preference for how to eval a form), then this will also "just work" for the new language.

Providing alternate key maps

Say someone has particularly fond memories of using Notepad, and wants to create a keys file, so that others can enjoy it too.

;; notepad-keys.el
(bindings 
 (global
  ("C-o" :file/open)
  ("C-s" :file/save)
  ...))

Now a Corkey user can load this file

(corkey/load-and-watch '(notepad-keys) '(corgi-signals user-signals))

Now currently (this is might change) Corgi uses counsel-find-file to open a file, but maybe the user has decided to use something else than Counsel, so they can declare that in their user-signals, and the C-o binding will now honor that.

Undefined commands

If Corkey can not find a listed command, i.e. if the package that provides it is not loaded when your key bindings are loaded, then this will simply be ignored, and no binding for the given keys will be made.

This is intentional, this allows us to define bindings in Corgi-bindings for various packages, even though not everyone may load all of those packages.

keys-file grammar

BINDINGS  := '(bindings' <def>+ ')'
<def>     := '(' <key> <doc> <target> ')' | '(' <prefix> <def> + ')'
<target>  := <signal> | <command>
<prefix>  := <state> | <key>
<state>   := 'normal' | 'insert' | 'visual' | 'emacs' | 'motion' | 'global'
<key>     := stringp
<doc>     := stringp
<signal>  := keywordp
<command> := symbolp

License

Copyright © 2020-2022 Arne Brasseur and Contributors

Licensed under the term of the GNU General Public License, version 3.

About

Corgi's keybinding system

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published