Jinx provides just-in-time spell-checking via libenchant. The package aims to achieve high performance and low resource usage, without impacting your editing experience. Overall Jinx should just work out of the box without much intervention.

Jinx highlights misspellings lazily only in the visible part of the text. The window boundaries and text folding are taken into account. Jinx binds directly to the native libenchant API, such that process communication with a backend Aspell process can be avoided. Libenchant is widely used as spell-checking API by text editors and supports Nuspell, Hunspell, Aspell and a few lesser known backends. Jinx automatically compiles jinx-mod.c and loads the dynamic module at startup.

Jinx supports multiple languages in a buffer at the same time via the jinx-languages customization variable. It offers flexible settings to 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 major modes.

The package is available on GNU ELPA and MELPA and can be installed with package-install. Libenchant must be installed on your system for compilation. If pkg-config is available it will be used to locate libenchant. On Debian or Ubuntu, install the packages libenchant-2-2, libenchant-2-dev and pkg-config.

Jinx offers two modes global-jinx-mode and jinx-mode. You can either enable global-jinx-mode or add jinx-mode to the hooks of the modes.

(add-hook 'emacs-startup-hook #'global-jinx-mode)

(dolist (hook '(text-mode-hook prog-mode-hook conf-mode-hook))
  (add-hook hook #'jinx-mode))

Furthermore Jinx brings two auto-loaded commands jinx-correct and jinx-languages. In order to correct misspellings bind jinx-correct to a convenient key in your configuration. Jinx is independent of the Ispell package, so you can reuse the binding M-$ which is bound to ispell-word by default. When pressing M-$, Jinx offers correction suggestions for the misspelling next to point. If the prefix key C-u is pressed, the entire buffer is spell-checked.

(keymap-global-set "<remap> <ispell-word>" #'jinx-correct)

Enchant uses different backends depending on the language. The ordering of the backends is configured by the file ~/.config/enchant/enchant.ordering. For most languages Hunspell is used by default. Depending on the backend the personal dictionary will be taken from different locations, e.g., ~/.aspell.LANG.pws or ~/.config/enchant/. It is possible to symlink different personal dictionaries such that they are shared by different spell checkers. See the Enchant manual for more details.

• jit-spell: Jinx offers a similar UI as Augusto Stoffel’s jit-spell package and borrows ideas from it. Jit-spell uses Ispell process communication instead of a native API. It does not restrict the highlighting to the visible text. In my setup I observed an increase in load and latency as a consequence, in particular in combination with stealth locking and commands which trigger fontification eagerly like consult-line from my Consult package.
• spell-fu: The technique to spell-check only the visible text was inspired by Campbell Barton’s spell-fu package. Spell-fu maintains the dictionary itself via a hash table, which results in high memory usage for languages with compound words or inflected word forms. In Jinx we avoid the complexity of managing the dictionary and access the advanced spell-checker algorithms directly via libenchant (affixation, compound words, etc.).
• flyspell: Flyspell is a builtin package which highlight misspellings while typing and when you move the cursor to a word. Jinx uses a different mode of operation, where the entire visible text of the buffer is checked always.

Since this package is part of GNU ELPA contributions require a copyright assignment to the FSF.