#+title:  On the Subject of Being Evil
#+author: Howard X. Abrams
#+date:   2023-12-21
#+filetags: emacs hamacs
#+startup: inlineimages

A literate programming file for configuring Evil mode in Emacs.
#+begin_src emacs-lisp :exports none
  ;;; ha-evil --- configuring Evil mode in Emacs. -*- lexical-binding: t; -*-
  ;;
  ;; © 2023 Howard X. Abrams
  ;;   This work is licensed under a Creative Commons Attribution 4.0 International License.
  ;;   See http://creativecommons.org/licenses/by/4.0/
  ;;
  ;; Author: Howard X. Abrams <http://gitlab.com/howardabrams>
  ;; Maintainer: Howard X. Abrams
  ;; Created: December 21, 2023
  ;;
  ;; While obvious, GNU Emacs does not include this file or project.
  ;;
  ;; *NB:* Do not edit this file. Instead, edit the original literate file at:
  ;;            ~/src/hamacs/ha-evil.org
  ;;       And tangle the file to recreate this one.
  ;;
  ;;; Code:
#+end_src

* Introduction
As a grizzled veteran of the Emacs-VI Wars, I’ve decided to take advantage of both by using VI keybindings on top of Emacs. However, after thirty years of Emacs, my interface follows different goals:

  - Most buffers begin in Evil’s /normal state/, e.g. normal mode for VIers.
  - Pressing ~i~ or ~a~ jumps into a state of total Emacs, with the exception of ~Escape~ going back to Evil. This means, that typing ~C-p~ goes up a line, and doesn’t auto-complete.
  - I don’t use ~:~ and instead use ~M-x~ or better yet, ~SPC SPC~ (typing the space key twice) from [[file:ha-general.org][General project]].
  - The ~Space~ doesn’t advance a letter, but instead displays a tree of highly-customized functions, displayable at the bottom of my screen, e.g.

[[file:screenshots/ha-leader.png]]

Some advice that I followed:
   - [[https://github.com/noctuid/evil-guide][Evil Guide]]
   - [[https://nathantypanski.com/blog/2014-08-03-a-vim-like-emacs-config.html][A Vim-like Emacs Configuration from Nathan Typanski]]
   - [[https://stackoverflow.com/questions/25542097/emacs-evil-mode-how-to-change-insert-state-to-emacs-state-automatically][Evil insert state is really Emacs?]] Real answer to that is to set [[help:evil-disable-insert-state-bindings][evil-disable-insert-state-bindings]]

TODO: Rebind the ~z~ keys
* Evil-Specific Keybindings
I split the configuration of Evil mode into sections. First, global settings:
#+begin_src emacs-lisp
  (use-package evil
    :init
    (setq evil-undo-system 'undo-fu
          evil-auto-indent t
          evil-respect-visual-line-mode t
          evil-want-fine-undo t         ; Be more like Emacs
          evil-disable-insert-state-bindings t
          evil-want-keybinding nil      ; work with evil-collection
          evil-want-integration t
          evil-want-C-u-scroll nil
          evil-want-C-i-jump nil
          evil-escape-key-sequence "jk"
          evil-escape-unordered-key-sequence t))
#+end_src

The Escape key act like ~C-g~ and always go back to normal mode?
#+begin_src emacs-lisp
  (use-package evil
    :config
    ;; (global-set-key (kbd "<escape>") 'keyboard-escape-quit)

    ;; Let's connect my major-mode-hydra to a global keybinding:
    (evil-define-key 'normal 'global "," 'major-mode-hydra)

    (evil-mode))
#+end_src

Even with the [[Evil Collection]], some modes should start in the Emacs state:
#+begin_src emacs-lisp
  (use-package evil
    :config
    (dolist (mode '(custom-mode
                    dired-mode
                    eshell-mode
                    git-rebase-mode
                    erc-mode
                    circe-server-mode
                    circe-chat-mode
                    circe-query-mode
                    vterm-mode))
      (add-to-list 'evil-emacs-state-modes mode)))
#+end_src

I’m not a long term VI user, and I generally like /easy keys/, e.g. ~w~, have larger jumps, and /harder keys/, e.g. ~W~ (shifted), have smaller, fine-grained jumps. So I am switching these around:

#+begin_src emacs-lisp
  (use-package evil
    :config (require 'evil-commands)
    (evil-define-key
      '(normal visual motion operator) 'global
      "w" 'evil-forward-WORD-begin
      "W" 'evil-forward-word-begin
      "e" 'evil-forward-WORD-end
      "E" 'evil-forward-word-end)
#+END_SRC

The ~b~ key bindings seems to need their own call, and I’m not sure why I can’t include it in the ~w~ and ~e~ bindings.

#+BEGIN_SRC emacs-lisp
  (evil-define-key
    '(normal visual motion operator) 'global
    "b" 'evil-backward-WORD-begin)
  (evil-define-key
    '(normal visual motion operator) 'global
    "B" 'evil-backward-word-begin)
#+end_src

In other words, with the above settings in place, ~w~ and ~e~ should jump from front to back of the entire line, but ~W~ and ~E~ should stop as /subword/:
  - =word-subword-subword=
  - =word_subword_subword=

Note I don’t bind =evil-backward-word-end= with a single key, but bind it to ~g e~ below.

While an absolute heresy to most VI users, I'm Evil and use ~M-x~ and ~SPC~ instead of ~:~ for running commands. I bind the ~:~ as a reverse of the ~;~ which continues the search from ~f~ and ~t~:
#+BEGIN_SRC emacs-lisp
  (evil-define-key
    '(normal visual motion operator) 'global
    ":" 'evil-repeat-find-char-reverse))
#+END_SRC

This clever hack from [[https://manueluberti.eu//emacs/2022/10/16/back-last-edit/][Manuel Uberti]] got me finding these useful bindings:
  - ~g ;~ ::  [[help:goto-last-change][goto-last-change]]
  - ~g ,~ :: [[help:goto-last-change-reverse][goto-last-change-reverse]]

Keybindings I would like to use more:
  - ~*~ :: jumps to the next instance of the word under point
  - ~#~ :: jumps to the previous instance of the word under point

  While I’m pretty good with the VIM keybindings, I would like to play around with the [[https://evil.readthedocs.io/en/latest/extension.html#text-objects][text objects]] and how it compares to others (including the surround).

  - ~diw~ :: deletes a word, but can be anywhere in it, while ~de~ deletes to the end of the word.
  - ~daw~ :: deletes a word, plus the surrounding space, but not punctuation.
  - ~xis~ :: changes a /sentence,/ and if ~i~ is ~a~, it gets rid of the surrounding whitespace as well. For instance, I mainly use ~das~ and ~cis~.
  - ~xip~ :: changes a /paragraph/.
  - ~xio~ :: changes a /symbol/, which can change for each mode, but works with =snake_case= and other larger-than-word variables.
  - Surrounding punctuation, like quotes, parenthesis, brackets, etc. also work, so ~ci)~ changes all the parameters to a function call, for instance
     - ~xa”~ :: a double quoted string
     - ~xi”~ :: inner double quoted string
     - ~xa'~ :: a single quoted string
     - ~xi'~ :: inner single quoted string
     - ~xa`~ :: a back quoted string
     - ~xi`~ :: inner back quoted string

*NOTE:* The ~x~ in the above examples are /operations/, e.g. ~d~ for /delete,/ ~v~ for /select,/ ~y~ for /copy/ and ~c~ for /change/.

What text objects are known?
  - ~w~ :: word
  - ~s~ :: sentence
  - ~p~ :: paragraph
  - ~l~ :: lines, with the [[Evil Text Object Line][Text Object Line]] package, configured below.
  - ~o~ :: symbol, like a variable, but also words, so ~vio~ is an easy sequence for selecting a word.
  - ~’~ :: a string, surround by quotes, also ~`~ for backticks
  - ~)~ :: parenthesis, also ~}~ and ~]~, see ~x~
  - ~x~ :: within a brace, paren, etc., with the [[Better Parenthesis with Text Object][my extensions below]], see ~b~ and ~f~ offer similar functionality.
  - ~d~ / ~f~ :: a /defun/, or code block, see Tree-Sitter approach [[file:ha-programming.org::*Evil Text Object from Tree Sitter][defined here]], or the old Emacs approach defined below.
  - ~i~ :: indention area, for YAML and Python, with the [[Text Objects based on Indentation][evil-indent-plus]] package, configured below.
  - ~t~ :: an HTML tag
  - ~c~ :: for comments
  - ~u~ :: for URLs, really? Useful much?
  - ~a~ :: function arguments (probably a lot like symbol, ~o~), but the ~a~ can include commas. This comes from [[https://github.com/wcsmith/evil-args][evil-args]] extension (see below).

TODO: Search for a plugin, like [[https://github.com/coderifous/textobj-word-column.vim][textobj-word-column]] for text objects based on “columns”.

I am not a long term VI user, and don’t have much need for any of its control sequences (well, not all), so I made the following more Emacsy. I’ll admit, I like ~C-v~ (and use that all the time), so I need to futz around with the scrolling:

#+begin_src emacs-lisp
  (use-package evil
    :config
    (evil-define-key '(normal visual motion operator) 'global
      (kbd "C-a") 'ha-beginning-of-line
      (kbd "C-e") 'move-end-of-line

      ;; Since C-y scrolls the window down, Shifted Y goes up:
      (kbd "C-y") 'evil-scroll-line-down
      (kbd "C-b") 'evil-scroll-line-up
      (kbd "C-S-y") 'evil-scroll-line-up

      (kbd "C-d") 'scroll-down-command
      (kbd "C-S-d") 'scroll-other-window-down
      (kbd "C-f") 'scroll-up-command
      (kbd "C-S-f") 'scroll-other-window

      (kbd "C-o") 'open-line ; matches evil's o
      (kbd "C-p") 'previous-line
      (kbd "C-n") 'next-line

      ;; I have better window control:
      (kbd "C-w") 'sp-kill-region))
#+end_src
** Evil Text Object Line
Delete a line, ~d d~ is in basic VI. Since some commands use text objects, and the basic text object doesn’t include lines,  the [[https://github.com/emacsorphanage/evil-textobj-line][evil-textobj-line]] project adds that:
#+begin_src emacs-lisp
  (use-package evil-textobj-line)
#+end_src
Now ~v i l~ and ~v a l~ works as you’d expect, but does this improve on ~S-v~?
** Text Objects based on Indentation
The [[https://github.com/TheBB/evil-indent-plus][evil-indent-plus]] project creates text objects based on the indentation level, similar to how the ~b~ works with “blocks” of code.
#+begin_src emacs-lisp
  (use-package evil-indent-plus)
#+end_src
This can be handy for Python, YAML, and lists in org files. Note that ~i~ works for the current indent, but ~k~ includes one line above and ~j~ includes one line above and below.
** Arguments as Text Objects
The [[https://github.com/wcsmith/evil-args][evil-args]] projects creates text objects for symbols, but with trailing ~,~ or other syntax.
#+begin_src emacs-lisp
  (use-package evil-args
    :config
    ;; bind evil-args text objects
    (define-key evil-inner-text-objects-map "a" 'evil-inner-arg)
    (define-key evil-outer-text-objects-map "a" 'evil-outer-arg)

    ;; bind evil-forward/backward-args
    (define-key evil-normal-state-map "L" 'evil-forward-arg)
    (define-key evil-normal-state-map "H" 'evil-backward-arg)
    (define-key evil-motion-state-map "L" 'evil-forward-arg)
    (define-key evil-motion-state-map "H" 'evil-backward-arg)

    ;; bind evil-jump-out-args
    (define-key evil-normal-state-map "K" 'evil-jump-out-args))
#+end_src
For a function, like this Python example, with the cursor on =b=:
#+begin_src python :tangle no
  def foobar(a, b, c):
    return a + b + c
#+end_src
Typing ~d a a~ will delete the argument leaving:
#+begin_src python :tangle no
  def foobar(a, c):
    return a + b + c
#+end_src
** Better Parenthesis with Text Object
I took the following clever idea and code from [[http://blog.binchen.org/posts/code-faster-by-extending-emacs-evil-text-object/][this essay]] from Chen Bin for creating a ~xix~ to grab code within any grouping characters, like parens, braces and brackets. For instance, ~dix~ cuts the content inside brackets, etc. First, we need a function to do the work (I changed the original from =my-= to =ha-= so that it is easier for me to distinguish functions from my configuration):
#+begin_src emacs-lisp
  (defun ha-evil-paren-range (count beg end type inclusive)
    "Get minimum range of paren text object.
  COUNT, BEG, END, TYPE follow Evil interface, passed to
  the `evil-select-paren' function.

  If INCLUSIVE is t, the text object is inclusive."
    (let* ((open-rx  (rx (any "(" "[" "{" "<")))
           (close-rx (rx (any ")" "]" "}" ">")))
           (range    (condition-case nil
                         (evil-select-paren
                          open-rx close-rx
                          beg end type count inclusive)
                       (error nil)))
           found-range)

      (when range
        (cond
         (found-range
          (when (< (- (nth 1 range) (nth 0 range))
                   (- (nth 1 found-range) (nth 0 found-range)))
            (setf (nth 0 found-range) (nth 0 range))
            (setf (nth 1 found-range) (nth 1 range))))
         (t
          (setq found-range range))))
      found-range))
#+end_src

Extend the text object to call this function for both /inner/ and /outer/:
#+begin_src emacs-lisp
  (evil-define-text-object ha-evil-a-paren (count &optional beg end type)
    "Select a paren."
    :extend-selection t
    (ha-evil-paren-range count beg end type t))

  (evil-define-text-object ha-evil-inner-paren (count &optional beg end type)
    "Select 'inner' paren."
    :extend-selection nil
    (ha-evil-paren-range count beg end type nil))
#+end_src

And the keybindings:
#+begin_src emacs-lisp
  (define-key evil-inner-text-objects-map "x" #'ha-evil-inner-paren)
  (define-key evil-outer-text-objects-map "x" #'ha-evil-a-paren)
#+end_src
** Text Object for Functions
While Emacs has the ability to recognize functions, the Evil text object does not. But text objects have both an /inner/ and /outer/ form, and what does that mean for a function? The /inner/ will be the /function itself/ and the /outer/ (like words) would be the surrounding /non-function/ stuff … in other words, the distance between the next functions.

#+begin_src emacs-lisp
  (defun ha-evil-defun-range (count beg end type inclusive)
    "Get minimum range of `defun` as a text object.
  COUNT, is the number of _following_ defuns to count. BEG, END,
  TYPE are not used. If INCLUSIVE is t, the text object is
  inclusive acquiring the areas between the surrounding defuns."
    (let ((start (save-excursion
                   (beginning-of-defun)
                   (point)))
          (end (save-excursion
                 (end-of-defun count)
                 (point))))
      (when inclusive
        ;; Let's see if we can grab more text ...
        (save-excursion
          ;; Don't bother if we are at the start of buffer:
          (when (> start (point-min))
            (goto-char start)
            ;; go to the end of the previous function:
            (beginning-of-defun)
            (end-of-defun count)
            ;; if we found some more text to grab, reset start:
            (if (< (point) start)
                (setq start (point))))
          ;; Same approach with the end:
          (when (< end (point-max))
            (goto-char end)
            (end-of-defun)
            (beginning-of-defun)
            (if (> (point) end)
                (setq end (point))))))

      (list start end)))
#+end_src

Extend the text object to call this function for both /inner/ and /outer/:
#+begin_src emacs-lisp
  (evil-define-text-object ha-evil-a-defun (count &optional beg end type)
    "Select a defun and surrounding non-defun content."
    :extend-selection t
    (ha-evil-defun-range count beg end type t))

  (evil-define-text-object ha-evil-inner-defun (count &optional beg end type)
    "Select 'inner' (actual) defun."
    :extend-selection nil
    (ha-evil-defun-range count beg end type nil))
#+end_src

And the keybindings:
#+begin_src emacs-lisp
  (define-key evil-inner-text-objects-map "d" #'ha-evil-inner-defun)
  (define-key evil-outer-text-objects-map "d" #'ha-evil-a-defun)
#+end_src

Why not use ~f~? I’m reserving the ~f~ for a tree-sitter version that is not always available for all modes… yet.
* Evil Extensions
** Evil Exchange
I often use the Emacs commands, ~M-t~ and whatnot to exchange words and whatnot, but this requires a drop out of normal state mode. The [[https://github.com/Dewdrops/evil-exchange][evil-exchange]] project attempts to do something similar, but in a VI-way, and the /objects/ do not need to be adjacent.

#+begin_src emacs-lisp
  (use-package evil-exchange
    :init
    (setq evil-exchange-key (kbd "gx")
          evil-exchange-cancel-key (kbd "gX"))

    :general (:states 'normal
                      "g x" '("exchange" . 'evil-exchange)
                      "g X" '("cancel exchange" . 'evil-exchange-cancel)

                      ;; What about a "normal mode" binding to regular emacs transpose?
                      "z w" '("transpose words" . transpose-words)
                      "z x" '("transpose sexps" . transpose-sexps)
                      "z k" '("transpose lines" . transpose-lines))

    :config (evil-exchange-install))
#+end_src

Let’s explain how this works as the documentation assumes some previous knowledge. If you had a sentence:

       The ball was blue and the boy was red.

Move the point to the word, /red/, and type ~g x i w~ (anywhere since we are using the inner text object). Next, jump to the word /blue/, and type the sequence, ~g x i w~ again, and you have:

       The ball was blue and the boy was red.

The idea is that you can exchange anything. The ~g x~ marks something (like what we would normally do in /visual mode/), and then by marking something else with a ~g x~ sequence, it swaps them.

Notice that you can swap:
  - ~gx i w~ :: words, ~W~ words with dashes, or ~o~ for programming symbols (like variables)
  - ~gx i s~ :: sentences
  - ~gx i p~ :: paragraphs
  - ~gx i x~ :: programming s-expressions between parens, braces, etc.
  - ~gx i l~ :: lines, with the [[Evil Text Object Line][line-based text object]] project installed
** Evil Lion
The [[https://github.com/edkolev/evil-lion][evil-lion]] package is a wrapper around Emacs’ [[help:align][align]] function. Just a little easier to use. Primary sequence is ~g a i p =~  to align along all the equal characters in the paragraph (block), or ~g a i b RET~ to use a built in rule to align (see below), or ~g a i b /~ to specify a regular expression, similar to [[help:align-regexp][align-regexp]].

#+begin_src emacs-lisp
  (use-package evil-lion
    :after evil
    :general
    (:states '(normal visual)
             "g a" '("lion ←" . evil-lion-left)
             "g A" '("lion →" . evil-lion-right)))
#+end_src
Lion sounds like /align/ … get it?

Where I like to align, is on variable assignments, e.g.
#+begin_src emacs-lisp :tangle no
  (let ((foobar        "Something something")
        (a             42)
        (very-long-var "odd string"))
    ;;
    )
#+end_src

If you press ~RETURN~ for the /character/ to align, =evil-lion= package simply calls the built-in [[help:align][align]] function. This function chooses a regular expression based on a list of /rules/, and aligning Lisp variables requires a complicated regular expression. Extend [[elisp:(describe-variable 'align-rules-list)][align-rules-list]]:
#+begin_src emacs-lisp
  (use-package align
    :straight (:type built-in)
    :config
    (add-to-list 'align-rules-list
                 `("lisp-assignments"
                   (regexp . ,(rx (group (one-or-more space))
                                  (or
                                   (seq "\"" (zero-or-more any) "\"")
                                   (one-or-more (not space)))
                                  (one-or-more ")") (zero-or-more space) eol))
                   (group . 1)
                   (modes . align-lisp-modes))))
#+end_src
** Evil Commentary
The [[https://github.com/linktohack/evil-commentary][evil-commentary]] is a VI-like way of commenting text. Yeah, I typically type ~M-;~ to call Emacs’ originally functionality, but in this case, ~g c c~ comments out a line(s), and ~g c~ comments text objects and whatnot. For instance, ~g c $~ comments to the end of the line.

#+begin_src emacs-lisp
  (use-package evil-commentary
    :config (evil-commentary-mode)

    :general
    (:states '(normal visual motion operator)
             "g c" '("comments" . evil-commentary)
             "g y" '("yank comment" . evil-commentary-yank)))
#+end_src
** Evil Collection
Dropping into Emacs state is better than pure Evil state for applications, however, [[https://github.com/emacs-evil/evil-collection][the evil-collection package]] creates a hybrid between the two, that I like.

#+begin_src emacs-lisp
  (use-package evil-collection
    :after evil
    :config
    (evil-collection-init))
#+end_src

Do I want to specify the list of modes to change for =evil-collection-init=, e.g.
#+begin_src emacs-lisp :tangle no :eval no
  '(eww magit dired notmuch term wdired)
#+end_src
** Evil Owl
Not sure what is in a register? Have it show you when you hit ~”~ or ~@~ with [[https://github.com/mamapanda/evil-owl][evil-owl]]:
#+begin_src emacs-lisp
  (use-package posframe)

  (use-package evil-owl
    :after posframe
    :config
    (setq evil-owl-display-method 'posframe
          evil-owl-extra-posframe-args
          '(:width 50 :height 20 :background-color "#444")
          evil-owl-max-string-length 50)
    (evil-owl-mode))
#+end_src
** Evil Surround
I like both [[https://github.com/emacs-evil/evil-surround][evil-surround]] and Henrik's [[https://github.com/hlissner/evil-snipe][evil-snipe]], but they both start with ~s~, and conflict, and getting them to work together means I have to remember when does ~s~ call sniper and when it calls surround. As an original Emacs person, I am not bound by that key history, but I do need them consistent, so I’m choosing the ~s~ to be /surround/.

#+begin_src emacs-lisp
  (use-package evil-surround
    :config
    (defun evil-surround-elisp ()
      (push '(?\` . ("`" . "'")) evil-surround-pairs-alist))
    (defun evil-surround-org ()
      (push '(?\" . ("“" . "”")) evil-surround-pairs-alist)
      (push '(?\' . ("‘" . "’")) evil-surround-pairs-alist)
      (push '(?b . ("*" . "*")) evil-surround-pairs-alist)
      (push '(?* . ("*" . "*")) evil-surround-pairs-alist)
      (push '(?i . ("/" . "/")) evil-surround-pairs-alist)
      (push '(?/ . ("/" . "/")) evil-surround-pairs-alist)
      (push '(?= . ("=" . "=")) evil-surround-pairs-alist)
      (push '(?~ . ("~" . "~")) evil-surround-pairs-alist))

    (global-evil-surround-mode 1)

    :hook
    (org-mode . evil-surround-org)
    (emacs-lisp-mode . evil-surround-elisp))
#+end_src
Notes:
  - ~cs'"~ :: to convert surrounding single quote string to double quotes.
  - ~ds"~ :: to delete the surrounding double quotes.
  - ~yse"~ :: puts single quotes around the next word.
  - ~ysiw'~ :: puts single quotes around the word, no matter the points position.
  - ~yS$<p>~ :: surrouds the line with HTML =<p>= tag (with extra carriage returns).
  - ~ysiw'~ :: puts single quotes around the word, no matter the points position.
  - ~(~ :: puts spaces /inside/ the surrounding parens, but ~)~ doesn't. Same with ~[~ and ~]~.
** Evil Jump, er Better Jump
The [[https//github.com/gilbertw1/better-jumper][better-jumper project]] replaces the [[https://github.com/bling/evil-jumper][evil-jumper project]], essentially allowing you jump back to various movements. While I already use ~g ;~ to jump to the last change, this jumps /to the jumps/ … kinda. I’m having a difficult time determining /what jumps/ are remembered.

#+begin_src emacs-lisp
  (use-package better-jumper
    :config
    (better-jumper-mode +1)

    (with-eval-after-load 'evil-maps
      (define-key evil-motion-state-map (kbd "M-[") 'better-jumper-jump-backward)
      (define-key evil-motion-state-map (kbd "M-]") 'better-jumper-jump-forward)))
#+end_src

* Technical Artifacts                                :noexport:

Let's =provide= a name so we can =require= this file:

#+begin_src emacs-lisp :exports none
  (provide 'ha-evil)
  ;;; ha-evil.el ends here
#+end_src

#+description: configuring Evil mode in Emacs.

#+property:    header-args:sh :tangle no
#+property:    header-args:emacs-lisp  :tangle yes
#+property:    header-args    :results none :eval no-export :comments no mkdirp yes

#+options:     num:nil toc:t todo:nil tasks:nil tags:nil date:nil
#+options:     skip:nil author:nil email:nil creator:nil timestamp:nil
#+infojs_opt:  view:nil toc:t ltoc:t mouse:underline buttons:0 path:http://orgmode.org/org-info.js