62 KiB
Introduction
Inspired by other such projects, here is my Emacs initialization file.
Note: This is not my entire Emacs initialization file. It's a work-in-progress.
Setup
I could use org-babel to load this file, but I don't like my initialization file having too many dependencies, especially big dependencies, and org is big. It may be strange for you to read this, as I have placed my entire Emacs configuration in an org-mode file, but here are the make targets I use to tangle and subsequently byte-compile my init file:
%.elc: %.el
emacs -Q -batch -eval "(byte-compile-file \"$<\")"
init.el: init.org
emacs -Q -batch -l "ob-tangle" -eval "(org-babel-tangle-file \"init.org\")
Executing the second target (either through make, or manually) will get you my Emacs initialization file in plain Emacs Lisp.
Note: If you look at this file in its source form you will notice
that I actually tangle init.org
into init2.el
. This is a temporary
measure so that I can gradually move my configuration from my
existing init.el
file into init.org
without much trouble. Once I
have emptied out my init.el
I will instruct babel to tangle into
init.el
, this code already reflects that.
The oni:
prefix
To keep my functions and variables from ever accidentally
interfering with other packages or Emacs internal variables I
prefix all the functions I write and variables I declare with oni:
.
You don't have to copy it if you copy anything from this file, as
long as you do it consistently.
I sometimes forget to do this, though.
The depends-on
snippets
At certain points in this document you will find bits of code that look like:
(depends-on "somepackage")
These calls are from the Cask DSL. They specify that my configuration depends on some package found in one of the ELPA repositories I have setup.
These snippets don't belong in your init.el
, but instead in a
separate Cask
file, and only if you use Cask, not if you just use
the standard package.el
.
Other such projects
As stated, this project is inspired and influenced by other literate configurations. Here is a list of some of the ones I've found:
Preparation
Some things have to be done to make sure that everything works as it should. This includes enabling lexical binding, loading Cask and keeping some comments.
Use lexical binding
For some of my functions, and general coolness, lexical binding is a must. Without it, closures don't work for example. This line needs to appear at the beginning of the file to work.
;; -*- lexical-binding: t -*-
Keep package.el from changing my init
Some time ago my init file was changed by Emacs. It added a single line of code and some explanatory comments. Apparently I need to keep this comment in my initialization file so that Emacs doesn't try to add the code again. I actually use Cask to manage and load my packages so I don't need this.
;; (package-initialize)
Load Cask
I use Cask to manage my installed packages and for that to work I need to load Cask at the start of the initialization file. This should be executed both when byte-compiling and loading at run-time so that other bits and pieces can byte-compile without warnings or errors.
(eval-and-compile
(require 'cask "~/projects/ext/cask/cask.el")
(cask-initialize))
Some helper functions
Certain functions make life a lot easier when configuring Emacs. These don't generally belong to any one specific customization, but make things look a lot prettier.
Optimized with-eval-after-load
First offered here and then later updated here (when
with-eval-after-load
was added). Makes for very nice on-demand
settings loading.
;; http://www.lunaryorn.com/2013/06/25/introducing-with-eval-after-load/
(defmacro stante-after (feature &rest forms)
"After FEATURE is loaded, evaluate FORMS.
FEATURE may be an unquoted feature symbol or a file name, see
`eval-after-load'."
(declare (indent 1) (debug t))
`(,(if (or (not byte-compile-current-file)
(if (symbolp feature)
(require feature nil :noerror)
(load feature :no-message :no-error)))
`progn
(message "stante-after: cannot find %s" feature)
'with-no-warnings)
(with-eval-after-load ',feature ,@forms)))
Turn off minor modes
This macro creates a function that will turn off a minor mode that passed to it.
(defmacro turn-off (func)
"Create a function that will turn off FUNC when called."
`(lambda () (eval (,func -1))))
Change settings
Sometimes a mode just needs to change a certain setting to a specific value.
(defmacro change-settings (&rest settings)
"Create a function that changes the value of NAME to VALUE."
`(lambda () (setq ,@settings)))
Eval after init
Some things can only really work after all other initialization functions have completed. For example, any functions that require any ELPA packages to be loaded, unless you want to load it in your init manually (and have it loaded again later on after your config has run).
(defmacro oni:eval-after-init (&rest body)
"Defer execution of BODY until after Emacs init."
(declare (indent 0))
`(add-hook 'emacs-startup-hook #'(lambda () ,@body)))
Whitespace only with tabs
In some modes I want to see the tabs in a buffer. Though I don't want to be overwhelmed by seeing all spaces and line endings.
(defvar whitespace-style)
(defun oni:whitespace-only-tabs ()
(setq-local whitespace-style '(face tabs))
(whitespace-mode))
Useful functions
During your editing in Emacs you will undoubtedly find the need to define your own editing functions or macros. Here are mine.
Delete the contents of the string at point
First we define the function. It was inspired by Mickey's post on
swapping quote symbols, mostly copied even. First we check if we
are even in a string, and if not we throw an error, after that we
move back to the beginning of the string, store that point, go to
the end of the string (using forward-sexp
) and then delete the
region between the two points (non-inclusive).
(defun oni:delete-string-contents ()
(interactive)
(unless (nth 3 (syntax-ppss))
(error "You must be in a string for this command to work"))
(save-excursion
(while (nth 3 (syntax-ppss)) (forward-char -1))
(let ((bos (point)))
(forward-sexp)
(delete-region (1+ bos) (1- (point))))))
Since for interactive functions it's kind of a pain to have to use a personal "namespace" I prefer naming them regularly as if they're just part of the environment. If ever Emacs comes up with a similar function with the same name, I'd prefer using the built-in version.
(unless (fboundp 'delete-string-contents)
(defalias 'delete-string-contents 'oni:delete-string-contents))
Lastly, any function worth using often should probably be easily accessible with a keybinding. In my case the is inspired by the Vim keybindings like .
(global-set-key (kbd "C-c i s") 'delete-string-contents)
General customization
These customizations don't belong with any specific mode.
Set some personal information
This information is used by some emacs commands and modules to make your life easier.
(setq user-full-name "Tom Willemse"
user-mail-address "tom@ryuslash.org")
Turn-off menu bar
The menu bar is one of the UI elements which work best with mouses. Sure you can change your toolkit's key bindings to allow you to more easily navigate, but 1 are easier if you don't use the mouse. Long story short: It has got to go.
(menu-bar-mode -1)
I also do this in my .Xdefaults
file, as that is applied sooner:
Emacs.menuBar: off
Turn-off tool bar
The toolbar is another such thing, and it takes up quite a bit more space too. Icons can look pretty cool, but in the end if you're not going to click them they don't really server much of a purpose. Again: It has got to go.
(tool-bar-mode -1)
I also do this in my .Xdefaults
as those settings are applied
sooner.
Emacs.toolBar: off
Turn-off blinking cursor
I suppose a blinking cursor doesn't get lost very easily. But on the other hand, it can induce quite a few more headaches.
I've noticed that I don't really lose my cursor position all that much, really, so there doesn't seem to be any point in making it blink. Here we go again: It has got to go.
(blink-cursor-mode -1)
I also do this in my .Xdefaults
, as those settings are applied
sooner.
Emacs.cursorBlink: off
Turn-off line numbers
As I'm currently using svg-mode-line-themes2 for my mode-line
I don't need to show these. Also I didn't really use them much, I
don't often need to know what line I'm on.
(line-number-mode -1)
Turn-off tooltips
Tooltips are another one of those UI elements that aren't quite keyboard-friendly. As usually this information is shown just as well in the echo area, this is not necessary.
(tooltip-mode -1)
Change cursor type
I prefer using a bar as a cursor, as opposed to a box. Using a bar is a better way of representing the location of the cursor, in my opinion.
(setq-default cursor-type 'bar)
I use a bar cursor that's already pretty thin, so having an even thinner one in inactive windows is not very clear. Even when I was using a box cursor I didn't like seeing a hollow box everywhere my focus wasn't.
(setq-default cursor-in-non-selected-windows nil)
Truncate long lines
Sometimes, very long lines can't help but show up in code. The default of wrapping lines around to the next line and showing an image in the fringe is very visually disturbing to me when looking at code3. So I prefer to have it just run off the screen.
(setq-default truncate-lines t)
Change default frame settings
Usually, most of these settings would be done with functions like
set-frame-font
or scroll-bar-mode
, but working with the Emacs
Daemon sometimes creates extra complications, because you might
start a graphical environment at some point (I usually do
immediately), but the startup process isn't.
Use a fantastic font
I've tried several fonts, and all of them were nice, but came up
short on some way. Some didn't have italic variants, some had
dotted zeroes instead of dashed zeroes, some even had entirely
clear zeroes which make them look like O
's, others had
boring-looking parentheses4. Not Fantasque Sans
Mono56, it is awesome. If your web browser supports web
fonts you should also see it used in the code blocks here.
For normal Emacs setups you might use the set-frame-font
function
like so:
(set-frame-font "Fantasque Sans Mono-15" nil t)
However, this doesn't work when running Emacs as a Daemon. You
could (especially with lexical binding) add an
after-make-frame-functions
hook or something similar, but I find
it easier to just specify the font in the default-frame-alist
.
(font . "Fantasque Sans Mono-15")
I also do this in my .Xdefaults
as those settings are applied
sooner.
Emacs.font: Fantasque Sans Mono-15
Hide the scroll bar
The scroll-bar is almost just as informative as the current line number and buffer position information shown in the (my) mode line. As I don't usually need to know where I am, other than the current line number occasionally, and I don't use the mouse, the scroll bar doesn't add anything and only takes up space. Once more: It has got to go.
Normally you would use something along the lines of:
(scroll-bar-mode -1)
However, this doesn't work when running Emacs as a Daemon. So
instead I specify it in the default-frame-alist
.
(vertical-scroll-bars . nil)
Bringing it all together
So I've explained the reasons for each individual setting, but to
get them to work they have to be put in the default-frame-alist
.
Here is the final setting:
(setq default-frame-alist
`(<<default-frame-font>>
<<default-frame-scroll-bar>>))
Set the frame title
Show the buffer name in the frame title to make multiple frames identifiable by the buffer they're showing.
(setq frame-title-format '(:eval (concat "GNU Emacs: " (buffer-name))))
Don't show dialog boxes
Unfortunately this doesn't remove all dialog boxes, but at least it keeps some of them from popping up. According to the docstring it should only change anything for when the mouse is used, which I don't ever do, but I still feel safer keeping this in my configuration.
(setq use-dialog-box nil)
Split windows equally
When splitting windows, give them all an equal amount of space. For those very few time that I have more than two windows in my Emacs frame it is much nicer to have each of them get ~33% space instead of 50%, 25% and 25%. The way I have my desktop set-up I don't ever split my windows horizontally.
(setq window-combination-resize t)
Keep the cursor steady when scrolling
When scrolling through the buffer with, for example, or , I find it much more intuitive to have the cursor stay in the relative screen position from before. Otherwise I get confused every time.
(setq scroll-preserve-screen-position t)
Don't show a mode line for completions
One of the things I dislike about Emacs is its rigidity concerning windows and other UI elements. One very simple way to make it feel less so is to show less of it where possible. A mode line for completion buffers creates more separation than necessary, and I can't recall a single time where I've used the completion's mode line for anything.
(add-hook 'completion-list-mode-hook
(lambda () (setq mode-line-format nil)))
Use pixels, not columns/rows to resize frames.
As of Emacs 24.4 frames can resize to pixel sizes instead of columns and rows. This keeps annoying gaps from forming when using a tiling window manager. The real difference is that when this is done now, the actual text portions of the window get the extra space, not some fringe or margin. This looks much nicer.
(setq frame-resize-pixelwise t)
Don't just quit Emacs with in the daemon
When working with Emacs as a daemon, which I do almost all of the time, I prefer using to close the current frame instead of the entire session. Before this change I would occasionally close my session by mistake.
(defun oni:close-client-window ()
"Close a client's frames."
(interactive)
(server-save-buffers-kill-terminal nil))
(when (daemonp)
(global-set-key (kbd "C-x C-c") 'oni:close-client-window))
Don't minimize the frame with
One of the more annoying things that can happen is accidentally minimizing the frame you're working with. This doesn't really matter if you're working on a normal stacking window manager, but with a tiling window manager and no task bar this just causes the Emacs frame to hang until it is refocused or disappear with no way to get it back.
(when (or window-system (daemonp))
(global-unset-key (kbd "C-z")))
Lazily load some buffers
Don't load all buffers right away. Having a lot of buffers and switching between projects a lot can take up quite a bit of time.
Note: See my note on vacuous defvar for this use of defvar
.
(defvar desktop-restore-eager)
(setq desktop-restore-eager 5)
Ask for a y
or n
, not yes
or no
.
Emacs starts out asking for you to type yes
or no
with most
important questions. It is possible that this is used for such
important questions where accidentally saying yes
when you meant no
would be catastrophic (or at least could be). However, I've never
(so far) had this problem, and I find it quite tedious to have to
write out yes
or no
and then press , give me a simple
y
or n
with no required and I'm quite happy.
(defalias 'yes-or-no-p 'y-or-n-p)
Use hippie-expand
, not dabbrev-expand
I've never actually used dabbrev-expand
, and only rarely use
hippie-expand
really, but since hippie-expand
also includes a
dabbrev
expander and a lot more than just that, it seems alright to
do this.
(defalias 'dabbrev-expand 'hippie-expand)
Setting up load-path
First, to help, I create a function that takes a path, adds it to
load-path
and then checks to see if there is a file named
loaddefs.el
in the given path. If there is, it loads it. This
loaddefs.el
file is something that is created from autoload cookies
in the files in some of these paths.
Since the load-path
is also important during byte-compilation, this
function should be defined both at run-time and compile-time.
(eval-and-compile
(defun oni:loadpath-add-and-autoload (path)
"Add PATH to `load-path' and load a `loaddefs.el' if it exists."
(add-to-list 'load-path path)
(let ((loaddefs (concat path "/loaddefs.el")))
(when (file-exists-p loaddefs)
(load loaddefs)))))
After that I add some directories to my load-path
so I can use these
libraries when wanted. One of these is the site-lisp
directory in my
.emacs.d
directory, which is where I keep most of my personal
non-ELPA modules (like module-specific initialization files). There
are also some directories I include in vendor-lisp
, which is where I
keep modules that I didn't write myself and, for some reason, can't
or don't want to use ELPA for. Again it is important to realize that
this information is relevant both at run-time and compile-time, so
we wrap it with an eval-and-compile
.
(eval-and-compile
(mapc #'oni:loadpath-add-and-autoload
'("~/.emacs.d/site-lisp"
"~/.emacs.d/vendor-lisp/mozrepl"
"~/.emacs.d/vendor-lisp/eap" "/usr/share/emacs/site-lisp"
"/usr/lib/node_modules/tern/emacs/")))
Turn off bidirectional text
To speed things up a little, and because I have no contacts at all (so far) who use right-to-left text, there is no reason for me to use bidirectional text. For this reason I tell Emacs to always use left-to-right by default, instead of checking each paragraph.
(setq-default bidi-paragraph-direction 'left-to-right)
Don't use tabs
There is a war going on out there. Actually there are several, but
the one I'm talking about here is the one about tabs vs. spaces. I
have been deeply entrenched in the spaces camp for quite a while.
Indentation always gets screwy for me if I don't use the exact same
tab-width
everyone else uses. I just like the consistency of using
spaces.
(setq-default indent-tabs-mode nil)
Use four spaces for indentation
I once, long ago, started out with using eight spaces for each level of indentation. Although I think, technically, I was using tabs. This means that even just 3 levels of indentation take up a huge amount of space that can't be used for anything else. Since I also try to limit the length of the lines in my code, this can make for very little code per line.
I also tried (and enjoyed) two spaces, but after a while, looking at bigger files, I noticed that it all becomes a blur. Two spaces is not a big-enough visual difference to make code read nicely, in my opinion.
So far, I have quite enjoyed four spaces for most code7.
(setq-default tab-width 4)
Keep the message buffer from growing too large
Keep at most one-thousand messages in the *Messages*
buffer. I can
leave my Emacs session running for quite long times, sometimes
days, at a time. If there are a lot of messages it can get a little
tricky to manage/search.
(setq message-log-max 1000)
Reject SSL/TLS certificates that don't check out
You just can't trust the internet. And I'd prefer to know if something goes wrong.
(defvar gnutls-verify-error)
(setq gnutls-verify-error t)
Use Conkeror to browse URLs
Use the generic browse-url function to open URLs in Conkeror.
(stante-after browse-url
(setq browse-url-browser-function 'browse-url-generic)
(setq browse-url-generic-program "conkeror"))
Major mode customization
Many major modes offer and require some customization. I use the term major mode somewhat loosely here because some of these customizations don't actually deal with major modes, but full-on applications or frameworks like Gnus and Elnode.
Org mode
Org mode offers a lot of customization options. Most of mine are in a separate file.
Store org id locations in my Emacs data directory.
(use-package org-id
:defer t
:config
(setq org-id-locations-file
(concat user-emacs-directory "data/org-id-locations")))
Add org-mode appointments to the diary
Diary offers reminders, which can be useful when scheduling appointments.
(defadvice org-agenda-redo (after ext:org-agenda-redo-add-appts)
"Pressing `r' on the agenda will also add appointments."
(setq appt-time-msg-list nil)
(org-agenda-to-appt))
Ansi term
I don't use ansi term much. Though I probably should do it more. Eshell is much nicer for many reasons but it can't handle everything.
Close ansi-term buffer after exit
After the ansi-term process ends it leaves a buffer. I don't use ansi term in such a way that this has ever been useful, so just kill the ansi-term buffer after the process quits, no matter the exit status. Usually this comes about when I press at the command prompt.
(defadvice term-handle-exit (after oni:kill-buffer-after-exit activate)
"Kill the term buffer if the process finished."
(kill-buffer (current-buffer)))
Fix some term keybindings
ansi-term
passes along a lot of characters correctly, but things
like forward-delete-word
are not, by default. This is confusing when
you see one thing and another is sent. Passing the correct keys
directly to the terminal fixes this problem.
Note: See my note on vacuous defvar for this use of defvar
.
Note: See my note on function declarations about the use of
declare-function
.
(defvar term-raw-map)
(declare-function term-send-raw-string "term")
(defun oni:set-term-keys ()
(cl-flet ((zcommand (key)
(lambda ()
(interactive) (term-send-raw-string key))))
(define-key term-raw-map
(kbd "C-<backspace>") (zcommand "\C-H"))))
(add-hook 'term-mode-hook #'oni:set-term-keys)
Gnus
Gnus is a very powerful news reader that also handles Email quite well. I've been using it for a while now, though not to its full extent I'm sure. This section contains customization for multiple major modes that are all closely related to Gnus.
Use the right dictionary
One of the caveats of using two (or more) languages in a single installation of Gnus is that ispell sometimes gets confused. Having come across a stackoverflow question8 about just this subject it was easy to modify the source code posted there to come up with this.
Note: See my note on function declarations about the use of
declare-function
.
(declare-function message-narrow-to-headers-or-head "message")
(declare-function message-fetch-field "message")
(defun oni:switch-ispell-dictionary ()
(save-excursion
(message-narrow-to-headers-or-head)
(when (string-match (rx (and "@" (or "aethon" "picturefix") ".nl>") eol)
(message-fetch-field "From"))
(ispell-change-dictionary "nl_NL"))))
(add-hook 'message-setup-hook 'oni:switch-ispell-dictionary)
Don't let shr use background color
Reading mail in Gnus is very nice, but shr has become a little too good at its job. Add to this the many occasions when a background is specified without specifying a foreground, plus a color theme that is the inverse of what is usually expected, and you can get hard-to-read HTML messages, gray foreground and gray background.
I've looked at the other possible renderers, but they don't look very nice compared to shr. So just remove its ability to add background colors.
Note: See my note on function declarations about the use of
declare-function
.
(declare-function shr-colorize-region "shr")
(defun oni:shr-colorize-remove-last-arg (args)
"If ARGS has more than 3 items, remove the last one."
(if (> (length args) 3)
(butlast args)
args))
(with-eval-after-load 'shr
(advice-add #'shr-colorize-region :filter-args
#'oni:shr-colorize-remove-last-arg))
Init file
I put my gnus initialization file right where I put all my
module-specific initialization files. Gnus is special, though: It
loads the file every time you start it. That keeps it from using a
simple (eval-after-load 'gnus '(load "gnus-init"))
.
Note: See my note on vacuous defvar for this use of defvar
.
(defvar gnus-init-file)
(setq gnus-init-file "~/.emacs.d/site-lisp/gnus-init")
SQL
This is a generic comint mode for multiple SQL implementations such as PostgreSQL and MariaDB (MySQL).
Remember SQL input
Remembering input between sessions is a good thing.
(stante-after sql
(setf sql-input-ring-file-name
(expand-file-name "~/.emacs.d/etc/sqliinput")))
Python
I used to code Python for a living, that's not so much the case anymore. And for almost everything I will prefer Lisp over Python. So these customizations might be a little old or badly tested.
Show package name instead of file name
When working with python, knowing which package I'm in usually tells
me more about what I'm doing than the file name, especially when
working with django where almost every app will have a tests.py
and
a models.py
. Of course uniquify
fixes this pretty well too, though
in this case, it's less cool (imo).
First we define two functions that help us determine the package name of the current file and the parent package name of the current file. Finally we define a third function which determines the full package name of the current buffer.
(defun oni:python--get-current-module-name ()
"Get the name of the current python module.
This is very simply just the filename of the current buffer with
the extension and pyath removed."
(file-name-sans-extension
(file-name-nondirectory (buffer-file-name))))
(defun oni:python--get-parent-module-name (&optional dir)
"This gets the currend file's parent module.
This function recursively gathers the parent package name of
either DIR or the current buffer's file name. Any directory where
an `__init__.py' file is found is considered to be a package.
This function returns either the parent package, with its
parents, or nil if the current directory isn't a python
package.."
(let* ((base (directory-file-name
(file-name-directory (or dir (buffer-file-name)))))
(package (file-name-nondirectory base)))
(if (file-exists-p (concat base "/__init__.py"))
(let ((parent (oni:python--get-parent-module-name base)))
(if parent
(concat parent "." package)
package))
nil)))
(defun oni:python-package-name ()
(let ((current-module (oni:python--get-current-module-name)))
(if (file-exists-p "__init__.py")
(concat (oni:python--get-parent-module-name)
"." current-module)
current-module)))
After all this we make Emacs show the package name rather than the file name in the mode-line.
(defun oni:python-package-buffer-identification ()
"Have `mode-line-buffer-identification' show the python package name."
(setq mode-line-buffer-identification
'(:eval (oni:python-package-name))))
(add-hook 'python-mode-hook #'oni:python-package-buffer-identification)
Elnode
Elnode is an awesome project and I'm still looking for a moment where I have the inspiration and time to actually do something with it. I started at some point, but then I couldn't get the cookies to work and I switched over to using Common Lisp, only to eventually stop developing the project because there was not chance of it seeing any use in the foreseeable future.
Don't start elnode
when Emacs starts
There is one little annoyance and that is the fact that elnode will start itself up when Emacs starts. I don't want that.
This bit of code can't be put in an eval-after-load
or anything
like that because by the time it would be evaluated, elnode would
already have started.
Note: See my note on vacuous defvar for this use of defvar
.
(defvar elnode-do-init)
(setq elnode-do-init nil)
js2
js2-mode
isn't just an "improved" JavaScript mode, it's a full-on
JavaScript parser. It's easy to notice typo's and such when Emacs
can show you that a certain variable is or isn't declared. It does
come with some baggage, as it's not immediately clear how I can
specify which global names exist. For simple files it is fine to do
something like:
/*global $ Routes jQuery */
Which will tell js2-mode
that $
, Routes
and jQuery
are known to be
define elsewhere. This doesn't work well for big lists of globals.
Add some known symbols for .conkerorrc/init.js to js2-mode
Conkeror has a lot of functions, and I don't like seeing them all as unknowns. So add them to known symbols.
Note: See my note on vacuous defvar for this use of defvar
.
(defvar js2-additional-externs)
(defun oni:js2-add-conkeror-symbols ()
"Add known/used conkeror symbols to additional externs."
(when (string-suffix-p ".conkerorrc/init.js" (buffer-file-name))
(setq js2-additional-externs
'( ;; Functions
"add_hook" "check_buffer" "co_return" "content_buffer"
"define_browser_object_class" "define_key" "define_webjump"
"dumpln" "get_current_profile" "get_home_directory"
"get_recent_conkeror_window"
"hints_minibuffer_annotation_mode" "interactive" "load_spec"
"load_spec_uri_string" "load_url_in_new_buffer" "make_file"
"make_uri" "mode_line_adder"
"open_download_buffer_automatically" "prefix_completer"
"read_browser_object" "register_user_stylesheet"
"remove_hook" "require" "send_http_request" "session_pref"
"shell_command_blind" "theme_load"
;; Variables
"Cc" "Ci" "browser_object_history_url" "browser_object_links"
"buffer_count_widget" "buffer_icon_widget" "content_buffer"
"content_buffer_form_keymap" "content_buffer_normal_keymap"
"content_buffer_text_keymap" "content_policy_accept"
"content_policy_bytype" "content_policy_reject" "cwd"
"default_base_keymap" "default_global_keymap"
"downloads_status_widget" "external_content_handlers"
"hint_digits" "load_paths" "read_buffer_show_icons"
"read_url_handler_list" "session_auto_save_auto_load"
"theme_load_paths" "title_format_fn" "url_remoting_fn"
;; Keyword argument
"$alternative" "$browser_object" "$completer" "$completions"
"$initial_value" "$options" "$prompt" "$sort_order"
"$use_bookmarks" "$use_history" "$use_webjumps"))))
(add-hook 'js2-init-hook #'oni:js2-add-conkeror-symbols)
Eww
I've been excited about the "Emacs Web Wowser" since I first read about it on the mailing list. Previously I'd used some integration with w3 which was ok, but it certainly didn't match up to a fully-integrated Emacs application.
This application uses shr
just like Gnus, so there is some overlap
here with reading Emails.
Teach eww about <code> tags
Strangely enough, eww
doesn't seem to be aware of <code>
HTML tags.
Luckily it's trivial to teach it. It does know about <pre>
HTML
tags, and basically I just want <code>
tags to be treated almost as
<pre>
tags, so to do that we just have to define a shr-tag-code
function. I've copied the shr-tag-pre
function and removed the calls
to ensure-newline
, because <code>
tags are inline tags.
In order to remain a little future-proof, it should only be done if it doesn't already exist.
Note: See my note on vacuous defvar for this use of defvar
.
Note: See my note on function declarations about the use of
declare-function
.
(defvar shr-folding-mode)
(declare-function shr-indent "shr")
(declare-function shr-generic "shr")
(with-eval-after-load 'shr
(unless (fboundp 'shr-tag-code)
(defun shr-tag-code (cont)
(let ((shr-folding-mode 'none))
(shr-indent)
(shr-generic cont)))))
Setup eww-lnum
(depends-on "eww-lnum")
As recommended in the README, set the keys in the eww-mode-map
.
(defvar eww-mode-map)
(with-eval-after-load 'eww
(define-key eww-mode-map "f" 'eww-lnum-follow)
(define-key eww-mode-map "F" 'eww-lnum-universal))
Scheme
I really like programming in Lisp. One of the more comfortable Lisps is Scheme because most of the implementations I've worked with are more like other compiled or interpreted languages, whereas Common Lisp usually re-compiles on every load. Aside from that, there are some neat programs written in some scheme dialects and of course scsh is the most awesome shell scripting language ever conceived.
Use scheme-mode for scsh interpreted files
Set the major mode for files interpreted by scsh (for example, by
having #!/usr/local/bin/scsh
at the top) to use scheme-mode
.
(add-to-list 'interpreter-mode-alist '("scsh" . scheme-mode))
Set default scheme implementation
Set the default implementation for geiser to guile so it doesn't ask which implementation to use every time.
(stante-after geiser-impl
(setq geiser-default-implementation 'guile))
SCSS
SCSS is a CSS preprocessor that makes writing CSS files much more fun. Add autocompletion and some custom imenu function.
(add-hook 'scss-mode-hook 'auto-complete-mode)
(add-hook 'scss-mode-hook 'scss-imenu-setup)
Dired
Dired is an excellent file manager.
Change listings in dired
The number of bytes a file is doesn't usually tell me much when it's
something like 292837
. I prefer seeing just how many Kb or Mb a
certain file is. I also don't need to see the .
and ..
directories
when I insert directories into the current dired buffer, as there is
a great chance that the current and parent directory are already
shown in the buffer.
(defvar dired-subdir-switches)
(with-eval-after-load 'dired
(setq dired-listing-switches "-alh"
dired-subdir-switches "-Alh"))
PHP
I occasionally write PHP for work. Not usually in my free time, but there are some open source PHP projects that I sometimes tinker with.
Show tabs and spaces in indent
I'm working with some WordPress plugins nowadays and their style guide insists on using tabs, not spaces… I'd like to know that I'm following this rule.
(add-hook 'php-mode-hook #'oni:whitespace-only-tabs)
Web
Web mode is good for files that contain a lot of HTML, CSS and JavaScript together. Most other major modes or multi-major modes don't quite do it.
Turn off the fill column indicator
web-mode
has some quirks, such as not being able to handle the
fact that fci-mode
puts a red line at the 80-column margin. This is
annoying to say the least.
(declare-function fci-mode "fci-mode")
(add-hook 'web-mode-hook (turn-off fci-mode))
Show tabs in indentation
Just like in php-mode
I want to see the tabs.
(add-hook 'web-mode-hook #'oni:whitespace-only-tabs)
Use tabs for indentation
Set indent-tabs-mode
for web-mode
as well.
(defvar web-mode-code-indent-offset)
(defvar web-mode-markup-indent-offset)
(add-hook 'web-mode-hook
(change-settings indent-tabs-mode t
web-mode-code-indent-offset 4
web-mode-markup-indent-offset 4))
Use it for Embedded Ruby HTML files
Use it for .html.erb
files.
(oni:eval-after-init
(add-to-list 'auto-mode-alist '("\\.html\\.erb$" . web-mode)))
Use it for HTML-heavy PHP files
I have to work with a lot of PHP and HTML interspersed. This makes
a difficult case since php-mode
very deliberately doesn't support
that very well. On the other hand I really don't like web-mode
for
PHP without any HTML in it. So I decided to name the files that
contain mostly HTML with some PHP .html.php
and have them load
web-mode
instead of php-mode
, whilst keeping the association for
plain .php
files as it is.
Something tricky about doing this is that if this setting gets
evaluated before php-mode
is loaded it'll be further down the list
from php-mode
's definition. This would cause the php-mode
auto
mode definition from being accepted first (since .html.php
also
matches .php
) and consequently render this definition useless.
(oni:eval-after-init
(add-to-list 'auto-mode-alist '("\\.html\\.php$" . web-mode)))
PO mode
Autoload po-mode
, because it didn't come with an autloads file or
cookie.
(autoload 'po-mode "po-mode" nil t)
Automatically enable po-mode
for files that end in .po
or that have
a .po
extension followed by others (such as .po.erb
).
(add-to-list 'auto-mode-alist '("\\.po\\'\\|\\.po\\." . po-mode))
Magit
Recently Magit gained the annoying habit of producing a huge warning message whenever you don't tell it that you've already seen it. To tell it you've already seen the message you need to specify the following.
(eval-and-compile
(defvar magit-last-seen-setup-instructions "1.4.0"))
I use a defvar
here in order to keep the byte-compiler from
complaining about an undefined variable. It needs to be specified
before magit is loaded otherwise magit will keep complaining.
Project directory
I keep all my projects in ~/projects/
, so Magit shouldn't have to
look anywhere else.
(setq magit-repository-directories '("~/projects/"))
Show fine differences
I like to see all the little differences in diffs that I can. They really help reading diffs. I also just want to see them on all diffs and not the selected one, which would make an unnecessary amount of navigation required to properly read the diffs.
(setq magit-diff-refine-hunk 'all)
Delay setting
The settings in the previous sections should only be set after Magit has loaded.
(stante-after magit
<<magit-repo-dirs>>
<<magit-diff-refine-hunk>>)
Lui
These settings will be tangled to site-lisp/lui-init.el
and loaded
when lui.el is loaded.
(with-eval-after-load 'lui
(require 'lui-init))
Circe
These settings will be tangled to site-lisp/circe-init.el
and
loaded when Circe is loaded.
(use-package circe
:ensure t
:commands (circe)
:load-path "vendor-lisp/circe"
:config (require 'circe-init))
Twig
Since twig is essentially a HTML template language, many tags must be edited. Tagedit is good for restructuring existing tag structures and adding single tags, whereas emmet mode is good when you know more about the HTML hierarchy you're going to add.
(add-hook 'twig-mode-hook 'tagedit-mode)
(add-hook 'twig-mode-hook 'emmet-mode)
Since Twig is a template language, and thus not just HTML, electric pairing is a blessing.
(add-hook 'twig-mode-hook 'electric-pair-local-mode)
Both HTML and Twig mode aren't traditional text modes, so
auto-fill-mode
doesn't make sense.
(add-hook 'twig-mode-hook (turn-off auto-fill-mode))
Ruby
Turn on abbrev-mode
in Ruby buffers.
(add-hook 'ruby-mode-hook 'abbrev-mode)
Turn on /ryuslash/legacy-dotfiles/src/branch/drd/emacs/.emacs.d/Electric%20pair mode.
(add-hook 'ruby-mode-hook #'electric-pair-local-mode)
Turn on /ryuslash/legacy-dotfiles/src/branch/drd/emacs/.emacs.d/Electric%20indent mode.
(add-hook 'ruby-mode-hook #'electric-indent-local-mode)
Turn on Flycheck mode.
(add-hook 'ruby-mode-hook 'flycheck-mode)
Turn on Yard mode.
(add-hook 'ruby-mode-hook 'yard-mode)
Turn on /ryuslash/legacy-dotfiles/src/branch/drd/emacs/.emacs.d/Eldoc mode.
(add-hook 'ruby-mode-hook 'eldoc-mode)
Turn on Ruby tools mode.
(add-hook 'ruby-mode-hook 'ruby-tools-mode)
Turn on Ruby refactor mode.
(add-hook 'ruby-mode-hook 'ruby-refactor-mode-launch)
Turn on /ryuslash/legacy-dotfiles/src/branch/drd/emacs/.emacs.d/Robe mode.
(add-hook 'ruby-mode-hook 'robe-mode)
Set Fill Column Indicator column to the community Ruby Style Guide recommended value.
(add-hook 'ruby-mode-hook 'oni:ruby-set-rsg-margin)
Minor mode customization
Many minor modes also offer a bit of customization possibilities.
Robe
(depends-on "robe")
Robe is a Ruby completion and documentation lookup library.
Eldoc
Seeing the arguments to a function whilst typing its name is excellent.
Show eldoc when evaluating expressions
Thanks to this post it was brought to my attention that eldoc mode can be enabled when evaluating expressions using .
I vaguely remember having had this before, I just don't know how or why it stopped working.
(add-hook 'eval-expression-minibuffer-setup-hook #'eldoc-mode)
Electric pair
Electric pairing of delimiters is one of those features that is just so essential to my feeling comfortable with an editor. Most of the time I don't even use it, really. It's just that I'm so used to having it and when I do expect it to be there it is so frustrating when it's not, or when it doesn't work properly.
This functionality, much like /ryuslash/legacy-dotfiles/src/branch/drd/emacs/.emacs.d/Electric%20indent isn't something I
want enabled in all modes, though for different reasons, and for a
time there was only the global electric-pair-mode
. Again I'm very
happy that a local version was added.
The reason that I don't want it enabled for all modes is that some modes (mostly Lisp-like language modes) have better alternatives. But most non-Lisp-like language modes I really do need to have it.
(add-hook 'c-mode-hook #'electric-pair-local-mode)
(add-hook 'coffee-mode-hook #'electric-pair-local-mode)
(add-hook 'css-mode-hook #'electric-pair-local-mode)
(add-hook 'haml-mode-hook #'electric-pair-local-mode)
(add-hook 'java-mode-hook #'electric-pair-local-mode)
(add-hook 'lua-mode-hook #'electric-pair-local-mode)
(add-hook 'python-mode-hook #'electric-pair-local-mode)
(add-hook 'scss-mode-hook #'electric-pair-local-mode)
(add-hook 'sh-mode-hook #'electric-pair-local-mode)
Remove whitespace when closing delimiters
In electric-pair-mode
, skip over and delete white space if it stands
between the cursor and the closing delimiter.
(setq electric-pair-skip-whitespace 'chomp)
Electric indent
Automatically indenting code upon typing certain characters can be very useful for certain modes where the indentation level can easily be determined. One of the first things I liked about Emacs was the way the key worked: It indents to the "proper" level of indentation, instead of adding a tab character. It quickly grew into a habit to press tab several times when editing a line or a block of code. Electric indent is just an extension of this that, for the most part, allows me to forget about pressing tab.
It doesn't fit all modes though. When I worked in Python a lot I was fighting the electric indent a lot more than it was helping me. This is because instead of scope influencing indentation as in most languages I've worked with, indentation determines scope in Python, as anyone who's looked at it for more than a minute or two will know. This means that any line can usually have several "proper" indentation levels, depending on the meaning of meaning of that line.
So, almost all modes use electric-indent-local-mode
, but a few
don't. So I'm also very happy that recently this mode was added,
because electric-indent-mode
is a global minor mode and I only want
to use it in some 99% of the available modes.
(add-hook 'css-mode-hook #'electric-indent-local-mode)
(add-hook 'php-mode-hook #'electric-indent-local-mode)
(add-hook 'scss-mode-hook #'electric-indent-local-mode)
(add-hook 'sh-mode-hook #'electric-indent-local-mode)
Switch keys back
When electric-indent-mode
is enabled the default function bound to
(electric-newline-and-maybe-indent
) stops indenting
after adding a newline, whilst starts doing it.
Since I use almost exclusively and don't use
at all, it's really not useful to me. So I want to
switch the two when electric-indent-mode
is enabled.
This is very simple. First I define a simple function that checks
if the electric-indent-mode
variable is set (which it should be if
the mode is turned on) and if so I set the proper keys locally. If
electric-indent-mode
is not set, which happens when the mode is
turned off, I remove the local keybindings.
(defun oni:switch-newline-keys ()
"Switch the `C-j' and `RET' keys in the local buffer."
(if electric-indent-mode
(progn
(local-set-key (kbd "C-j") 'newline)
(local-set-key (kbd "RET") 'electric-newline-and-maybe-indent))
(local-unset-key (kbd "C-j"))
(local-unset-key (kbd "RET"))))
And then I add it to the electric indent mode's hook.
(add-hook 'electric-indent-local-mode-hook #'oni:switch-newline-keys)
Auto completion
(depends-on "auto-complete")
I'm not a great fan of any type of auto completion functionality in programming. I think it's basically only good for getting really long names in your file faster. I started programming in C#, in Visual Studio, and they have their very powerful Intellisense mechanism, which I used a lot. What I noticed though was that it was keeping me from learning the APIs. Whenever I wrote some code, all but the most basic and most-used calls I wrote from memory. Many others were done after a few seconds of browsing through the Intellisense pop-up.
Because of this, I disable the normal auto-complete pop-up.
(setq ac-auto-show-menu nil)
On the rare occasion that I do call up the pop-up to see what the completions I have available to me are, I don't want it to show the quick help. The quick help usually shows up as a big extra pop-up next to the completion list and it gets very messy when combined with some other modes that add overlays to the buffer.
(setq ac-use-quick-help nil)
To keep the byte-compiler from complaining about undefined
variables, I only want this to be done after auto-complete
is
loaded, not sooner.
(stante-after auto-complete
<<auto-complete-dont-show>>
<<auto-complete-no-quickhelp>>)
Highlight indentation
(depends-on "hl-indent")
Some languages base their ideas of scope on levels of indentation.
When these languages get long functions/branches it can get tricky
to see where these blocks end. Of course if you have such large
functions you might have other problems, but this is sometimes out
of your hands. To help, I use hl-indent-mode
.
For other languages it might look fun for a little while as well, but since they don't care about indentation as much it can sometimes get messy.
First, hl-indent-mode
doesn't have an ;;;###autoload
cookie for its
main entry-point, so I add it manually:
(autoload 'hl-indent-mode "hl-indent" nil t)
Then I enable it for the languages I want.
(add-hook 'python-mode-hook #'hl-indent-mode)
(add-hook 'yaml-mode-hook #'hl-indent-mode)
(add-hook 'haml-mode-hook #'hl-indent-mode)
Django helper
(depends-on "pony-mode")
Autoload pony-mode
because it doesn't come with an autoload cookie.
(autoload 'pony-mode "pony-mode" nil t)
Turn on compilation-shell-minor-mode
when pony-mode
is enabled.
(defun oni:turn-on-compilation-shell-for-pony ()
"Turn on option `compilation-shell-minor-mode' for `pony-minor-mode'."
(add-hook 'pony-minor-mode-hook 'compilation-shell-minor-mode nil t))
(add-hook 'comint-mode-hook #'oni:turn-on-compilation-shell-for-pony)
Automatic syntax checking on-the-fly
(depends-on "flycheck")
(depends-on "flycheck-cask")
(depends-on "flycheck-commit-check" :git "git://github.com/ryuslash/flycheck-commit-check.git")
Flycheck provides awesome syntax checkers for many languages. Enable it for the languages I use.
(add-hook 'haml-mode-hook 'flycheck-mode)
(add-hook 'emacs-lisp-mode-hook 'flycheck-mode)
(add-hook 'git-commit-mode-hook 'flycheck-mode)
(add-hook 'go-mode-hook 'flycheck-mode)
(add-hook 'html-mode-hook 'flycheck-mode)
(add-hook 'lua-mode-hook 'flycheck-mode)
(add-hook 'perl-mode-hook 'flycheck-mode)
(add-hook 'php-mode-hook 'flycheck-mode)
(add-hook 'python-mode-hook 'flycheck-mode)
(add-hook 'rst-mode-hook 'flycheck-mode)
(add-hook 'rust-mode-hook 'flycheck-mode)
(add-hook 'sh-mode-hook 'flycheck-mode)
(add-hook 'texinfo-mode-hook 'flycheck-mode)
Whenever flycheck is started, try using it with Cask.
(add-hook 'flycheck-mode-hook 'flycheck-cask-setup)
Load my own git commit checker.
(require 'flycheck-commit-check)
Disable certain checkers.
(mapc (lambda (c) (delq c flycheck-checkers))
'(python-pylint python-pyflakes))
Set the highlighting mode to columns so I can see (if possible) where errors/warnings belong.
(setq flycheck-highlighting-mode 'columns)
Try not to display the errors. A jumpy echo area makes me nervous and is a nicer way to look at it usually. It still doesn't keep it from being displayed, only when nothing else is displayed though.
(setq flycheck-display-errors-function (lambda (_) nil))
Do most of these things after flycheck is loaded.
(stante-after flycheck
<<flycheck-commit-checker>>
<<flycheck-delete-checkers>>
<<flycheck-columns>>
<<flycheck-display>>)
Hightlight numbers mode
I find this regular expression to work better at identifying general numbers as the default one doesn't consider numbers starting with a sign or a decimal point.
(stante-after highlight-numbers
(setq highlight-numbers-generic-regexp
"\\_<[-+]?[[:digit:]]+\\(?:\.[[:digit:]]+\\)?.*?\\_>"))
Emmet mode
I'm one of the strange people who use instead of to insert a new line (and indent), so Emmet mode's default binding of using to expand a line is very disruptive to me.
(stante-after emmet-mode
(define-key emmet-mode-keymap (kbd "C-j") nil)
(define-key emmet-mode-keymap (kbd "RET") #'emmet-expand-line))
Applications
Emacs has a number of packages that are basically applications within Emacs. These are the ones I use or have used in the past, and may again use in the future.
Elfeed
(use-package elfeed
:ensure t
:defer t)
Final touches
These options and calls need to come last so they don't interfere with the rest of the initialization process, or get interfered with.
Load custom file
I don't really use the Emacs customization interface much, but I
have used it as a kind-of persistent datastore, specifically for
desktop-registry9. I do very much like the idea of it, it's a
very cool thing to have. I also use custom.el
for storing some
things that I really can't store in a public viewing location like
this file or the git repository it lives in.
(setq custom-file "~/.emacs.d/custom.el")
(load custom-file)
Notes
Here are some random or somewhat general notes about things you may run into when looking through my Emacs init.
Vacuous defvar
A defvar
without a value like (defvar some-variable)
tells the
byte-compiler that the variable will appear, but doesn't give it a
value. It should only count for the file where it is used and once
the file with the actual defvar
is loaded it will be populated with
its value, contrary to what would happen if you'd given it a value
before loading its original file.
Function declarations
The function declare-function
tells the byte-compiler where to find
a certain function. This keeps the byte-compiler from complaining
about certain functions possibly not being defined at run-time.
Footnotes
This runs tmm-menubar
, which lets you navigate the menubar in a
text-driven way. I don't ever use it because I know what all my
favorite functions are called, but it seems a great deal more
efficient than having to click on everything.
It works fine for me with something like jabber chats and the like.
I like to program in Lisp, parentheses are important to me! Parentheses should be nice and round, not almost like bars!
Used to be Cosmic Sans Neue Mono, the name changed because people misread it as "Comic" (me included, which was the original reason I checked it out, for laughs) and hate Comic Sans, and also because there was already a Cosmic Sans font as well, which could cause confusion.
I still use 2 spaces for some languages, like HTML.