(or emacs irrelevant)

Throwing abbrevs into the mix

28 Dec 2014

Currently, I'm using two methods for completing Elisp: company-mode and helm-lisp-completion-at-point. The latter is the cannon, the big gun: it always gets the job done, but I don't want to shoot at sparrows with it. So I only bring it out for hairy cases, like for stuff that starts with LaTeX-. Hence, the company-mode. But too often have I typed region- only to find 7 candidates staring at me, 4 of them useless. Which prompted me to look for an additional completion method.

Enter abbrevs

According to the manual,

A defined "abbrev" is a word which "expands", if you insert it, into some different text

Also,

Abbrevs can have "mode-specific" definitions, active only in one major mode

Sounds like something that could solve my problem with region- commands:

rb -> region-beginning
re -> region-end
ra -> region-active-p

Also, obviously,

Abbrevs expand only when Abbrev mode, a buffer-local minor mode, is enabled

Add it to the mix:

(defun oleh-emacs-lisp-hook ()
  (setq outline-regexp ";; ———")
  (company-mode 1)
  (abbrev-mode 1)
  (set (make-local-variable 'company-backends)
       '((company-elisp :with company-dabbrev-code)))
  (yas-minor-mode-on)
  (lispy-mode 1)
  (auto-compile-mode 1))

Some data acquisition

I had the whole abbrev thing in the back of my mind until I saw a link to the post Abbrevs for the most frequent elisp symbols. That's when I decided to act. That post eventually links to a pastebin, where 1600 abbrevs are defined. With my handy best extension for best browser I've opened the paste in Emacs best editor by just clicking the edit button in the RAW Paste Data section.

I had to M-x emacs-lisp-mode, since the file opened in text-mode. And boy, it's big. In lispy-mode, I usually use 99j to navigate 99 sexps down and therefore to the end of the list. Well, for this file even 999j wasn't enough. I quickly tired of deleting one-by-one the each individual useless abbrev. I mean:

ek -> echo-keystrokes,

when is that ever going to be useful? So I wrote this throw-away code:

(defun foobar ()
  (interactive)
  (lispy-mark-list 2)
  (let ((str (read (lispy--string-dwim)))
        count)
    (other-window 1)
    (goto-char (point-min))
    (setq count (count-matches str))
    (other-window 1)
    (lispy-out-backward 1)
    (deactivate-mark)
    (if (< count 5)
        (lispy-delete 1)
      (message "%d" count))))
(local-set-key (kbd "C-.") 'foobar)

After switching to a two-pane window layout, with point in the pastebin buffer, calling foobar would count the amount of the abbrev matches in my most frequent elisp buffer. If it was less than 5, the abbrev was auto-deleted, otherwise the decision was up to me, as holding C-. would no longer delete. In the end, there were only 56 abbrevs left out of 1600.

The final result

Here's what I have put into my abbrev_defs:

(define-abbrev-table 'emacs-lisp-mode-abbrev-table
    '(("sm" "string-match") ("mm" "major-mode")
      ("rb" "region-beginning") ("ca" "char-after")
      ("smd" "save-match-data") ("mb" "match-beginning")
      ("pm" "point-min") ("ir" "indent-region")
      ("sf" "search-forward") ("ci" "call-interactively")
      ("sn" "symbol-name") ("se" "save-excursion")
      ("scb" "skip-chars-backward") ("fc" "forward-char")
      ("ff" "find-file") ("fs" "forward-sexp")
      ("pa" "prefix-arg") ("re" "region-end")
      ("dc" "delete-char") ("ms" "match-string")
      ("tc" "this-command") ("dd" "default-directory")
      ("bc" "backward-char") ("rsf" "re-search-forward")
      ("snp" "substring-no-properties")
      ("bsnp" "buffer-substring-no-properties")
      ("lep" "line-end-position") ("bs" "buffer-substring")
      ("cc" "condition-case") ("ul" "up-list")
      ("bfn" "buffer-file-name") ("lb" "looking-back")
      ("tap" "thing-at-point") ("rm" "replace-match")
      ("fl" "forward-line") ("df" "declare-function")
      ("ntr" "narrow-to-region") ("dr" "delete-region")
      ("rsb" "re-search-backward") ("scf" "skip-chars-forward")
      ("wcb" "with-current-buffer") ("ie" "ignore-errors")
      ("gc" "goto-char") ("jos" "just-one-space")
      ("la" "looking-at") ("ow" "other-window")
      ("dk" "define-key") ("dm" "deactivate-mark")
      ("bod" "beginning-of-defun") ("sic" "self-insert-command")
      ("eol" "end-of-line") ("me" "match-end")
      ("nai" "newline-and-indent") ("cb" "current-buffer")
      ("atl" "add-to-list") ("rris" "replace-regexp-in-string")))