Emacs: The Plan to Move Away from Evil Mode

It would make me very happy if Viper (the built-in Vi emulation in Emacs) was a viable option, but it messes with Emacs’ default keybindings and behaviour for some things (notably minibuffer completion with Vertico), is largely unmaintained, and is a bit too loyal to Vi’s behaviour (e.g. it doesn’t immediately remove text deleted with c from the display), lacking a lot of what I’ve come to expect from Vim.

These are both creations of the infamous Xah Lee. At a glance they seem to have similar keybindings, aiming to prioritise the most frequently used editing operations. Xah-Fly-Keys is modal, so it seems to be catered more towards existing Vim keys users, but I think the similarity would just make me question the effort taken to switch. The downside is that it would just be another large framework slapped on top of Emacs, replacing Evil.

The default Emacs keybindings are somewhat notorious for causing hand pain and RSI. I would need to be extremely careful with how I used them, although this would be the ideal solution from a minimalism perspective. A lot of modes aren’t supported by Evil, so primarily using Emacs’ default keys should theoretically lead to a more consistent experience. Since the default keybindings are available through C-z in evil-mode, I could probably try transitioning to these gradually instead of going “cold turkey”, with the safety net of evil-mode still there if I need to get something done and find the keybindings are getting in the way.

God Mode seems to just hold down a modifier key for you (specifically Ctrl). I could see this being beneficial, but it seems like a somewhat ad-hoc hacky solution that “goes outside the system” of Emacs’ keybindings.

The idea here would be to create a large set of custom bindings on top of Emacs according to my own preference. This would be quite the time commitment, but would have a high chance of producing a valuable result. I have considered how I could approach creating an efficient set of keybindings, and I reckon repeat maps would be valuable for this.

I’ve created a package called repeat-map-define, which makes it easier to use the repeat-mode added in Emacs 28 to define “repeat maps”. Repeat maps function in almost exactly the same way as Hydra as far as I can tell, although my package is only 72 lines of code as of writing (repeat-mode does 99% of the work), compared to the 1500+ of Hydra. The idea is that, after pressing a key chord, the chord’s prefix is “pinned”, so you only need to press the last key of that and any other bindings in the map. This allows for dynamic, pseudo-modal keymaps to be seamlessly activated and deactivated automatically, which seems much nicer than the God Mode approach. The obvious downside is that repeat maps must be manually defined, but with my package or Hydra, this is hardly any more effort than defining key bindings normally.

My idea would be to define repeat maps for just about everything that isn’t a minor mode and doesn’t create a separate buffer with its own special bindings. The primary of repeat maps is that they don’t diminish the pain in the (common!) case of pressing one chord in a map then doing something else. In addition to this, it can be easy to overlook the fact that a repeat map was activated, or it may be unwanted, forcing the user to press C-g if the normal binding of a key in the repeat map is required.

In the end, the goal should be writing code in the most efficient, healthy way possible. If Vim keys work for me, then I should probably stick with them. I also have a lot of evil-mode-specific key bindings, which would feel somewhat depressing to give up. In the depths of my despair, I even considered switching back to NeoVim, but that would definitely not be worth it.