Jinx is a fast just-in-time spell-checker for Emacs. Jinx highlights misspelled words in the text of the visible portion of the buffer. For efficiency, Jinx highlights misspellings lazily, recognizes window boundaries and text folding, if any. For example, when unfolding or scrolling, only the newly visible part of the text is checked if it has not been checked before. Each misspelling can be corrected from a list of dictionary words presented as a completion menu.
Installing Jinx is straight-forward and configuring should not need much intervention. Jinx can be used completely on its own, but can also safely co-exist with Emacs’s built-in spell-checker Ispell.
Jinx’s high performance and low resource usage comes from directly calling the
widely-used API of the Enchant library (see libenchant). Jinx automatically
compiles jinx-mod.c
and loads the dynamic module at startup. By binding directly
to the native Enchant API, Jinx avoids the slower backend process communication
with Aspell. Enchant is widely used by other text editors and supports Nuspell,
Hunspell, Aspell and a few language-specific backends.
Jinx supports spell-checking multiple languages in the same buffer. See the
jinx-languages
variable to customize for multiple languages. Jinx can flexibly
ignore misspellings via faces (jinx-exclude-faces
and jinx-include-faces
),
regular expressions (jinx-exclude-regexps
), and programmable predicates. Jinx
comes preconfigured for the most important Emacs major modes. Modes like Java,
Ruby or Rust are listed in jinx-camel-modes
. For these modes composite words in
camelCase
and PascalCase
are accepted.
Jinx can be installed from GNU ELPA or MELPA directly with package-install
.
Jinx requires libenchant
. Enchant library is a required dependency for Jinx to
compile its module at install time. If pkgconf
or pkg-config
is available when
installing Jinx, Jinx will use it to locate libenchant
. Depending on your
operating system and Linux distribution you have to install different packages:
- Debian, Ubuntu:
libenchant-2-dev
,pkgconf
- Arch, Gentoo:
enchant
,pkgconf
- Guix:
emacs-jinx
orenchant
,pkgconf
- Void, Fedora:
enchant2-devel
,pkgconf
- FreeBSD, OpenBSD, Mac:
enchant2
,pkgconf
Jinx has two modes: the command, global-jinx-mode
activates globally; and the
command, jinx-mode
, for activating for specific modes.
;; Alternative 1: Enable Jinx globally
(add-hook 'emacs-startup-hook #'global-jinx-mode)
;; Alternative 2: Enable Jinx per mode
(dolist (hook '(text-mode-hook prog-mode-hook conf-mode-hook))
(add-hook hook #'jinx-mode))
The commands jinx-correct
and jinx-languages
are marked as autoloads. Invoking
jinx-correct
corrects the misspellings. Binding jinx-correct
to M-$
takes over
that key from the default assignment to ispell-word
. Since Jinx is independent
of the Ispell package, M-$
can be re-used.
(keymap-global-set "M-$" #'jinx-correct)
(keymap-global-set "C-M-$" #'jinx-languages)
M-$
triggers correction for the misspelled word before point.C-u M-$
triggers correction for the entire buffer.C-u C-u M-$
forces correction of the word at point, even if it is not misspelled.
A sample configuration with the popular use-package
macro is shown here:
(use-package jinx
:hook (emacs-startup . global-jinx-mode)
:bind (("M-$" . jinx-correct)
("C-M-$" . jinx-languages)))
See also the Jinx Wiki for additional configuration tips. The wiki documents
configurations to save misspellings as global abbreviations and support for
Ispell LocalWords
.
After invoking the command jinx-correct
, suggested corrections are displayed as
a completion menu. You can press the displayed digit keys to quickly select a
suggestion. Furthermore the menu offers options to save the word temporarily for
the current session, in the personal dictionary or in the file-local variables.
Note that you can enter arbitrary input at the correction prompt in order to
make the correction or to store a modified word in the personal dictionary.
For example if you typed alotriomorpc
, the prompt offers you the option
@alotriomorpc
which would add this word to your personal dictionary upon
selection. You can then correct the option to @allotriomorphic
and add it to the
dictionary.
While inside the jinx-correct
prompt, the keys M-n
and M-p
are bound to
jinx-next
and jinx-previous
respectively and allow you to move the next and
previous misspelled word.
The completion menu is compatible with all popular completion UIs: Vertico, Mct,
Icomplete, Ivy, Helm and the default completions buffer. In case you use Vertico
I suggest that you tweak the completion display via vertico-multiform-mode
for
the completion category jinx
. You can for example use the grid display such that
more suggestions fit on the screen and enable annotations.
(add-to-list 'vertico-multiform-categories
'(jinx grid (vertico-grid-annotate . 20)))
(vertico-multiform-mode 1)
As mentioned before, when correcting a word with jinx-correct
, the movement
commands jinx-next
and jinx-previous
are available on the keys M-n
and M-p
to
navigate to the next and previous misspelling respectively. The movement
commands work from within the minibuffer during jinx-correct
and also globally
outside the minibuffer context.
While the commands are not bound globally by default, they are available as M-n
and M-p
if point is placed on top of a misspelled word overlay. If you want you
can add them and other commands to the jinx-mode-map
, such that they are always
available independent of point placement. If repeat-mode
from Emacs 28 is
enabled, the movement can be repeated with the keys n
and p
.
Enchant uses different backends for different languages. The backends are
ordered as specified in the personal configuration file
~/.config/enchant/enchant.ordering
and the system-wide configuration file
/usr/share/enchant-2/enchant.ordering
. Enchant uses Hunspell as default backend
for most languages. There are a few exceptions. For English Enchant prefers
Aspell and for Finnish and Turkish special backends called Voikko and Zemberek
are used. On non-Linux operating systems Enchant may also integrate with the
spell-checker provided by the operating system.
Depending on the backend the personal dictionary will be taken from different
locations, e.g., ~/.aspell.LANG.pws
or ~/.config/enchant/LANG.dic
. It is
possible to symlink different personal dictionaries such that they are shared by
different spell checkers. See the Enchant manual for details.
There exist multiple alternative spell-checking packages for Emacs, most famously the builtin ispell.el and flyspell.el packages. The main advantages of Jinx are its automatic checking of the visible text, its sharp focus on performance and the ability to easily use multiple dictionaries at once. The following three alternative packages come closest to the behavior of Jinx.
- jit-spell: Jinx UI borrows ideas from Augusto Stoffel’s Jit-spell. Jit-spell uses the less efficient Ispell process communication instead of Jinx’s calling a native API. Since Jit-spell does not restrict spell checking to the visible text only, it may enqueue the entire buffer too eagerly for checking. This happens when scrolling around or when stealth font locking is enabled. For this reason, Jit-spell affected load and latency in my tests (issue on Github).
- spell-fu: The idea to check words just in the visible text came from Campbell Barton’s spell-fu package. Spell-fu is fast but incurs high memory overhead on account of its dictionary in a hash table. For languages with compound words and inflected word forms, this overhead magnifies (issue on Codeberg). By accessing the Enchant API directly, Jinx avoids any overhead. Jinx also benefits from the advanced spell-checker algorithms of Enchant (affixation, compound words, etc.).
- flyspell: Flyspell is a built-in package. Flyspell highlights misspellings
while typing. Only the word under the cursor is spell-checked. Jinx, on the
other hand, is more effective because it automatically checks for misspellings
in the entire visible text of the buffer at once. Flyspell can check the
entire buffer but must be instructed to do so via the command
flyspell-buffer
.
Since this package is part of GNU ELPA contributions require a copyright assignment to the FSF.