The Emacs startup process: Init files

Published on: 21 July 2014 In categories:

I started thinking about this when I arranged a short remote pairing session to help a friend work through some pain he was having with his Emacs configuration. The session hasn’t happened yet but it got me interested enough in what Emacs does under the hood when it starts up, particularly when it loads your config and I though it might make an interesting blog post.

The .emacs file.

When Emacs starts up it loads your configuration from a file. It looks for this file in a number of places in your home directory, in order, these are: ~/.emacs, ~/.emacs.el and ~/.emacs.d/init.el. The last form is the most common as it allows you to split your configuration up into multiple files. You can then control when these are loaded by using require, load or autoload in your configuration. This keeps your initial Emacs memory footprint quite small and ensures Emacs is quick to load, and just pulls in extra libraries and configuration when you need it.

The mode system helps a lot. You’ll often see people with files like ruby.el, python.el or scheme.el in their emacs directories. They will then put all their specific language related config in there and with a couple of lines in their main init.el they can only require those files the first time they open a specific file, or start a specific mode in Emacs

Emacs does not load these its config files sequentially. If it finds a ~/.emacs file, it will interpret that file and then it will stop searching, so you cannot use this mechanism to provide default overrideable configuration. But it’s OK, Emacs has you covered!

The site-start and default.el files

If you want to provide default Emacs configuration; for instance, if you’re a system administrator and want to set up some friendly overrideable default configuration for your users; Or you use Emacs on a bunch of different machines and you want some system specific variables defined for each machine, and want to keep them out of your main, shared Emacs config; you’d put that shared config in either site-start.el or default.el

These two files need to live somewhere in your Emacs library load path. They are loaded on startup around your own custom configuration and which one you use depends on how easy you want to make it for people to ignore these settings.

The site-start file gets loaded before your Emacs file, so any configuration set in here will get overridden by conflicting information in your config files. For example, you can define some keyboard shortcuts in here that will be made available to every Emacs user on the system, but won’t clobber anything they’ve configured to use the same keys. You can prevent site-start from being required, but because it runs before your config file you have to do it using by passing the --no-site-file argument when you start Emacs.

The default file gets loaded after your Emacs configuration, so it provides a way of clobbering users’ custom settings or providing extra config that they may have forgotten. This is not something you should rely on however, as it’s possible by bypass loading this file.

If you set the variable inhibit-default-init to something non-nil in your Emacs config it will prevent default.el from being run.

So that’s the basics of how Emacs starts and what it loads. You can do a lot more than just that however and Emacs offers massive levels of control. So let’s take a look at some other things that you might need to know.

Customize mode

I’ve spelt customise with the American spelling in the title for a reason. At some point in your Emacs journey you’ll end up looking at the Emacs ‘Customize mode’. You can get to this by running M-x customize and it provides an excellent interface for configuring Emacs. Most things that you’ll want to do commonly have entries in Customize mode, and Emacs provides an API for adding Customize mode compatability to your own Emacs lisp libraries.

By default when you change a setting using the Customize interface it adds an entry into your init.el that looks like the following:

;; custom-set-variables was added by Custom.
;; If you edit it by hand, you could mess it up, so be careful.
;; Your init file should contain only one such instance.
;; If there is more than one, they won't work right.
'(ansi-color-faces-vector [default bold shadow italic underline bold bold-italic bold]))

As the warning says, you should avoid touching this block by hand and make sure there is only one such block in your Emacs config. Because this section feels a little more subject to churn than my standard init.el file and because there are more serious consequences to breaking it, I like to split it out into its own file, that I can deal with separately. Emacs provides a variable that you can set called custom-file just for this purpose. By default this value is set to nil which tells Emacs to just use your init.el.

;; Store emacs customisation system stuff in a seperate file
(setq custom-file "~/.emacs.d/customisations.el")
(load custom-file)

Note that both of the above lines are necessary. The first tells Emacs where your Customize options should be saved, and the second tells Emacs to load those customisations.

After this change, our current load order looks like this.

  1. site-start.el
  2. ~/.emacs || ~/.emacs.el || ~/.emacs.d/init.el
  3. ~/.emacs.d/customisations.el
  4. default.el

Other variables of note.

  1. user-emacs-directory - Tell Emacs that you want your config directory to live somewhere specific. By Default this is set to ~/.emacs.d/

  2. package-user-dir - Tell Emacs where you want ELPA packages placed. These are packages that you install using Emacs 24’s built in package manager (M-x package-install)

I’m sure there are plenty more, but these are some important ones that I’ve used recently. user-emacs-directory particularly can be helpful when you want to test or debug changes to your config as you can load one then the other to spot changes between them.

Don’t forget as well that Emacs has an exceptionally in-depth help system built in. You can search it with M-x apropros or find information about any variable, function of keybinding using C-h v, C-h f and C-h k respectively. This even works for things defined in your own Emacs config!

Byte compiling

Byte compiling your configuration can speed up the loading of your config, and in fact, any Emacs lisp that you write. This is the process of turning the source code you write into an intermediary form that can be read directly by the Lisp interpreter built into Emacs.

After byte compilation has been carried out you will have a .elc file corresponding to every .el file that you have compiled. Their contents will contain stuff that looks like the following:

(byte-code "\303\304\305\"\211\203)\306\307!\203\307\303\207"

When emacs looks for a file to load, it will automatically prefer any file ending with elc. That is, when you (require 'foo) If Emacs finds a file called foo.elc before it find foo.el while searching the load path, it will always load it first, even if the compiled version appears later in the load path. ### Aside: Emacs & load-path

When we said that Emacs looks in your library load path for the site specific files, we neglected to mention what that is. By default it contains only 2 entries: /usr/local/share/emacs/version/site-lisp and /usr/local/share/emacs/site-lisp in that order. The order matters because Emacs will only load the first file that matches what it’s looking for.

You can override the initial value of your load path by setting the environment variable EMACSLOADPATH.

When you are within Emacs you can treat the load path as any other list, pushing and popping variables off it at will. There are a few other ways that directories can end up on the load path, but they are less comonnly used. You can read more about them here:

I hope this has been helpful. If you want to dig into this further there are some excellent resources online, particularly the Emacs manual, which you can access using C-h r from within Emacs or on the web at the GNU Emacs homepage