Having Emacs Type for You

We type too much. And with text files, much of the typing is just organizational support. Think of asterisks, spaces and #+ in org-mode, or all the parens, curly braces and do..end in your programming code.

Just because you want those characters in your files doesn’t mean you have to physically type them all.

The following are the major approaches to getting Emacs to do some of the heavy typing for you… like the boilerplate needed when creating a new file of a particular type, or large swatches of text (using templates).

Just to focus this tutorial, I’m not going to cover context completion, like auto-complete or company.

Introduction to YAS

The yasnippet project lets you insert snippets of code. A snippet is a template that can have substitutions replaced either manually or programmatically. The choice of which template to choose and expand is based on the buffer’s mode.

To begin, let’s try to configure YAS and insert static text. If you are using use-package, let’s have it installed and set up a personal directory for our templates (I’ll use snippet and template interchangeably):

(use-package yasnippet
  :ensure t
  (yas-global-mode 1)
  (add-to-list 'yas-snippet-dirs (locate-user-emacs-file "snippets")))

Let’s create a template by typing C-c & C-n … yeah, I never remember that either, so try: M-x yas-new-snippet (and with something like IDO, that can be just as fast).

You should now be presented with a new template buffer, and since it is using YAS, you can fill in the individual fields and use Tab to jump to the next field. Try making it look like:

# -*- mode: snippet -*-
# name: blah
# key: blah
# --
Bling blargh-a bloo bloop!

Hit C-c C-c to install it. It will ask you what mode it should expect to expand it. At this point, hopefully you can just hit Return to accept the default.

Why yes, you normally will want to save it. Notice the default may be exactly where you’ll want to keep it too: ~/.emacs.d/snippets

Now, in a buffer of the same mode, type blah and hit Tab, and the text will be replaced with:

Bling blargh-a bloo bloop!

YAS comes with multiple ways of triggering a template, but let’s next talk about ways to make the results of the template more useful.

Interactive Snippets

Replacing one word with several can be useful, but to make it truly helpful, we could have the result be more context-aware. Let’s make it easy for us to manually change some of the template’s parts.

Open up a code file in some programming language mode, and create a new snippet. I’m going to create an ifelse snippet for JavaScript:

# -*- mode: snippet -*-
# name: ifelse
# key: ife
# --
if ($1) {
else {

Now, I can type ife and have an if.. else template expanded in my JavaScript mode, and the cursor begins between the parens (due to the $1 label), allowing me to type the condition. Hitting Tab jumps between the curly braces allowing me to type the results of a positive condition (due to the $0 label which specifies where the field editing should end).

I’m not finished with YAS yet, but let me take a bit of a detour so that my next YAS example will be more meaningful.

New Files

Remember when your company thought it was a great idea to put their copyright notice at the beginning of every file in the code base? While I’m not sure how much legal weight that carries, Emacs has long had an Auto Insert feature for populating a new file of a particular mode with boilerplate goodness.

A use-package approach to setting it up to dump the contents of a static file into your new file is:

(use-package autoinsert
  ;; Don't want to be prompted before insertion:
  (setq auto-insert-query nil)

  (setq auto-insert-directory (locate-user-emacs-file "templates"))
  (add-hook 'find-file-hook 'auto-insert)
  (auto-insert-mode 1)

  (define-auto-insert "\\.html?$" "default-html.html"))

Now, in this case, creating a file with an extension of .html will insert the contents of ~/.emacs.d/templates/default-html.html. A start, but hardly good enough for the power user.

Combining YAS and Auto Insert

We can use a snippet as the default contents for a new file, which would allow us to tidy up some of that inserted boilerplate.

YAS’s work-horse for expanding snippets is yas-expand-snippet, which takes, as input, the contents of the snippet to insert. Shove this in your *scratch* buffer, and evaluate it (with C-x C-e):

(yas-expand-snippet ";; Bah-da $1 Bing")

You can probably see where I’m going here, eh? Let’s create a helper function that takes the static text that auto-insert puts into a new file, but treat it as if it were a snippet:

(defun autoinsert-yas-expand()
  "Replace text in yasnippet template."
  (yas-expand-snippet (buffer-string) (point-min) (point-max)))

Here, the (buffer-string) is the contents of the buffer, and yas-expand-snippet takes two more parameters to specify the contents in the current buffer to replace with the results… for us, this will be the entire buffer: (point-min) and (point-max).

The define-auto-insert function can take a vector, which it either includes (if it is a string) or executes (if it is a function name), so:

(define-auto-insert "\\.el$" [ "defaults-elisp.el" autoinsert-yas-expand ])

Will take new files ending in .el, insert the defaults-elisp.el file into it, and then execute our autoinsert-yas-expand function which will replace that template as if it were a YAS snippet.

This allows you punctuate it with $1 and $2 and other field placeholders.

I wrap up my templates with use-package, like:

(use-package autoinsert
  (define-auto-insert "\\.el$" ["default-lisp.el" ha/autoinsert-yas-expand])
  (define-auto-insert "\\.sh$" ["default-sh.sh" ha/autoinsert-yas-expand])
  (define-auto-insert "/bin/"  ["default-sh.sh" ha/autoinsert-yas-expand])
  (define-auto-insert "\\.html?$" ["default-html.html" ha/autoinsert-yas-expand]))

Programmatic Snippets

Entering the field details by hand is pretty good, but what if a snippet could programmatically enter some of the information?

For instance, we traditionally start our Emacs Lisp files like:

;;; demo-it --- Utility functions for creating demonstrations
;; Copyright (C) 2014  Howard Abrams
;; Author: Howard Abrams <howard.abrams@gmail.com>
;; Keywords: demonstration presentation
;; This program is free software; you can redistribute it and/or modify
;; it under the terms of the GNU General Public License as published by
;; the Free Software Foundation, either version 3 of the License, or
;; ...

Where the first line contains the file’s name and a description. YAS will execute code within back-ticks as Emacs Lisp, so:

(yas-expand-snippet "`(buffer-file-name)`")

Will insert the complete file name of the buffer, and:

(yas-expand-snippet "`user-full-name`")

Will insert the contents of the variable, user-file-name (which should be your nom de plume). Our Emacs Lisp template could be this snippet:

;;; `(upcase (file-name-nondirectory (file-name-sans-extension (buffer-file-name))))` --- $1
;; Author: `user-full-name` <`user-mail-address`>
;; Copyright © `(format-time-string "%Y")`, `user-full-name`, all rights reserved.
;; Created: `(format-time-string "%e %B %Y")`
;;; Commentary:
;;  $2
;;; Code:


;;; `(file-name-nondirectory (buffer-file-name))` ends here

Full Programmatic Inserts

For me, a journal is a file in the ~/journal directory that has the simple file format of YYYYMMDD. We might be tempted to create a YAS snippet to automatically include a title, like:

#+TITLE: Journal Entry for `(format-time-string "%e %B %Y")`

But that only works if I am really good and write in my journal each day. Perhaps I could create a snazzy title with the date based on the file name. Let’s define that format:

(setq org-journal-date-format "#+TITLE: Journal Entry- %e %B %Y")

And a function that will parse a buffer-file-name to the appropriate values for that format1:

(defun journal-title ()
  "The journal heading based on the file's name."
  (let* ((year  (string-to-number (substring (buffer-name) 0 4)))
         (month (string-to-number (substring (buffer-name) 4 6)))
         (day   (string-to-number (substring (buffer-name) 6 8)))
         (datim (encode-time 0 0 0 day month year)))
    (format-time-string org-journal-date-format datim)))

Now, our template would be:

#+TITLE: Journal Entry for `(journal-title)`

Nice, but let’s level up on this…

While I’m quite intrigued with Habitica, and am looking forward to better Emacs integration, what I really like, is the idea of Dailies… tasks checked-off each day come back the next day.

I already have some good task capturing code, but nothing that keeps coming back. Perhaps, I could have my daily journal entry track that.

This idea of inserted daily information and check-lists should only happen if I am creating today’s journal (not catching up with the past), and I might have special dailies inserted based on the day of the week.

I could use YAS snippets, but then the amount of `(...)` code may over-shadow the text, so I’ll create a small collection of templates:

  • journal-dailies.org to contain the real dailies
  • journal-dailies-end.org to contain any follow-up notes
  • journal-mon.org for additional text to be inserted on Monday journals
  • journal-tue.org for additional text to be inserted on Monday journals
  • And a journal-XYZ.org for each additional weekday

With these files, editing my daily lists should be clear and straight-forward.

Now, I just need to update my snippet-based template. Since I will need to create a fair amount of Emacs Lisp functions to support that, I might as well make the entire thing programmatic:

(define-auto-insert "/[0-9]\\{8\\}$" [journal-file-insert])

Now when I load up a file with a name that contains only 8 digits, it will run the function journal-file-insert:

(defun journal-file-insert ()
  "Insert's the journal heading based on the file's name."
  (insert (journal-title))
  (insert "\n\n") ; Start with a blank separating the title

  ;; If the journal entry I'm creating matches today's date:
  (when (equal (file-name-base (buffer-file-name))
               (format-time-string "%Y%m%d"))

    ;; Note: The `insert-file-contents' leaves the cursor at the
    ;; beginning, so the easiest approach is to insert these files
    ;; in reverse order:
    (insert-file-contents "journal-dailies-end.org")
    (insert "\n")

    ;; Insert dailies that only happen once a week:
    (let ((weekday-template (downcase
                             (format-time-string "journal-%a.org"))))

      (when (file-exists-p weekday-template)
        (insert-file-contents weekday-template)))

    (insert-file-contents "journal-dailies.org")
    (previous-line 2)))

That’s all I know about Auto Insert and the yasnippet project. Do you have any questions or tips I could use?



I generally don’t like using functions like substring for manipulating strings since such functions generally lack precision. Since my jounal-title function will not be called without a (buffer-name) that matches an 8 digit pattern, I felt like I could get away.

However, a version that uses regular expressions could be:

(defun journal-title ()
  "The journal heading based on the file's name."
  (when (string-match "\\([[:digit:]]\\{4\\}\\)\\([[:digit:]]\\{2\\}\\)\\([[:digit:]]\\{2\\}\\)" (buffer-name))
    (let* ((year  (string-to-number (match-string 1 (buffer-name))))
           (month (string-to-number (match-string 2 (buffer-name))))
           (day   (string-to-number (match-string 3 (buffer-name))))
           (datim (encode-time 0 0 0 day month year)))

      (format-time-string org-journal-date-format datim))))

Hopefully, that regular expression isn’t too scary. Emacs’s version of regular expressions look a tad worse due to all the escaping. Might be better if I used concat and broke into three parts:

(concat "\\([[:digit:]]\\{4\\}\\)"   ; year
        "\\([[:digit:]]\\{2\\}\\)"   ; month
        "\\([[:digit:]]\\{2\\}\\)")  ; day
  • \\( \\) creates a group that match-string let’s me extract a part
  • [:digit:] refers to any number, and putting it in another pair of brackets is… well, just a Emacsism.
  • \\{ \\} specifies a repeating value, so \\{4\\} says, look for the previous pattern 4 times… in other words, 4 numbers.

However, as Jason Milkins pointed out, using rx would make that function much more readable:

(require 'rx)

(defun journal-title ()
  "The journal heading based on the file's name."
  (when (string-match (rx (group (= 4 digit))
                          (group (= 2 digit))
                          (group (= 2 digit))) (buffer-name))
    (let* ((year  (string-to-number (match-string 1 (buffer-name))))
           (month (string-to-number (match-string 2 (buffer-name))))
           (day   (string-to-number (match-string 3 (buffer-name))))
           (datim (encode-time 0 0 0 day month year)))

      (format-time-string org-journal-date-format datim))))