This minor mode does only one thing: it breaks the evil undo sequence when the buffer is changed over a line boundary (i.e. when at least one line is created, or deleted in/from the buffer). It is inspired from the following Vim mapping:
inoremap <silent> <cr> <c-g>u<cr>… which it also generalizes so that:
- it doesn’t depend on the actual keys pressed
- it also handles text removal
Instead of trying to map every possible newline-inserting key that exists in the
large world of Emacs modes and users, evil-nl-break-undo follows the buffer
changes directly; in this way, it doesn’t matter what key was pressed, or how
the user changed the buffer: it could have been with RET, or maybe with
M-RET, mapped to one utility command or another (insert a new list item; or
continue a comment on the next line; indent; add a paren etc.). The same thing
happens when text is removed in insert state: if the text deleted includes a
newline character, the current evil-based undo sequence is split.
It means that, for example, after you write an entire paragraph in insert state,
and then you hit u in normal state to undo, changes are undone line by line,
instead of the whole paragraph disappearing with one swift stroke.
(add-hook 'text-mode-hook #'evil-nl-break-undo-mode)
(add-hook 'prog-mode-hook #'evil-nl-break-undo-mode)Or, the equivalent, with use-package, and since the package is now on MELPA:
(use-package evil-nl-break-undo
:ensure t
:hook ((text-mode prog-mode) . evil-nl-break-undo-mode))The mode depends on, and assumes that evil-mode is enabled in the buffer, and
that the variable evil-want-fine-undo is nil (this is the default in evil-mode;
IIRC, not so in Spacemacs).
This is still experimental (but working well for quite some time now!), and more
of an exercise in emacs lisp: there might be evil-enabled buffers that change
their text automatically, or in special ways, without explicit user action. This
could therefore trigger evil undo breaks by themselves in that buffer, which
may, or may not be meaningful there. It would be best then to have such buffers
configured for evil emacs state by default (or any other state different than
evil insert state). Not sure though if having this minor mode enabled, and with
such a buffer in insert state, would do any harm.
It probably doesn’t deserve to be a whole minor mode for such a small thing; but it’s my first! and anyway, Emacs has too many minor modes already.