2.2 KiB

Raw Blame History

Tree-sitter Grammar for Emacs Lisp

A simple tree-sitter grammar for elisp.

Syntax supported:

Atoms (integers, floats, strings, characters, symbols)
Lists (normal syntax (a b) and dotted (a . b))
Vectors
Quoting and unquoting (', #', `, ,, ,@)
Some special read syntax ($#, ##, #("foo" 1 2 x))
Bytecode literals (#[1 2 3 4])
Comments

Currently unsupported:

Autoload cookies
Special forms (e.g. let, currently treated as symbols)
Definitions (e.g. defun, defvar, defmacro)

Limitations

Elisp is a lisp-2 with user-defined macros. A simple parser cannot detect if e.g. (foo (let ...)) is a function call with a let expression argument, or a macro call where let means something else.

Currently tree-sitter-elisp treats everything as an s-expression. This is accurate, but makes this package less useful for generating a summary of file contents, or for syntax highlighting.

Developing

Check out the repo, then use npm to install dependencies.

$ npm install

You can then parse your favourite elisp files.

$ npm run parse ~/.emacs.d/init.el

The grammar itself is in grammar.js. You'll need to regenerate the code after editing the grammar.

$ npm run generate

This project also contains a few tests.

$ npm test

You can also run this parser against your .emacs.d to confirm it can parse everything.

$ npm run parse -- '/home/wilfred/.emacs.d/**/*.el' --quiet --stat

Why?

The best place to read and write elisp is of course Emacs.

However, there is a growing ecosystem of tools built on top of tree-sitter, such as GitHub. This project should allow them to support emacs lisp too.

References

tree-sitter-clojure is another tree-sitter package for the lisp family. It's a useful project to compare with, and has notes discussing lisp-specific challenges.

language-emacs-lisp is a textmate grammar for elisp that's used for Atom and GitHub.

2.2 KiB Raw Blame History