.emacs.d

This is my personal emacs configuration.

Everything here is borrowed or stolen. Help yourself.

Goals:

• Deterministic
  • Git subtrees for all dependencies
  • no network access required
• Start as fast as possible
  • use-package defers as much as possible

Usage

git clone https://github.com/thumphries/.emacs.d
ln -s "$(pwd)/.emacs.d" "${HOME}/.emacs.d"
./compile

Subtrees

Adding new subtrees:

# ./subtree-add <remote> [<ref>]
./subtree-add https://github.com/abo-abo/hydra master

Updating subtrees:

# ./subtree-pull <remote> [<ref>]
./subtree-pull https://github.com/abo-abo/hydra master

Profiling

./profile

Recommended

A few things that have worked really well:

Subtrees

Everything is checked into site-lisp using git subtrees.

I haven't had to think about emacs packages in anger since I started doing this.

Sometimes a package will break when I upgrade emacs itself, almost always resolved with a subtree pull or a new dependency.

Lazy loading

See use-package.

With byte compilation and occasional startup profiling everything starts up in the order of 200ms.

This is not a common thing to need. I use emacs completely wrong - kinda like it's vi with lots of sessions and without the GUI. Fast startup is pretty useful for this workflow.

Garbage collector thresholds

The default emacs garbage collection threshold is/was rather low. Doesn't matter much for text editing, but at session init time we're interpreting webapp quantities of random bytecode.

Bumping the collection threshold during init and dropping it afterward led to a substantial reduction in init time. The actual numbers are not magic, I probably pasted them from somewhere. Their magnitude is what matters.

;; init.el prologue
(setq gc-cons-threshold (* 50 1000 1000))

;; init.el
; ...

;; init.el epilogue
(setq gc-cons-threshold (* 2 1000 1000))