@codincod/codemirror-lang-clojure

This package implements Clojure language support for the CodeMirror code editor.

Written in part for CodinCod, a competitive coding platform, where it colours the editor people solve puzzles in.

This code is released under an MIT license.

Usage

import {EditorView, basicSetup} from "codemirror"
import {clojure} from "@codincod/codemirror-lang-clojure"

const view = new EditorView({
  parent: document.body,
  extensions: [basicSetup, clojure()],
})

Measuring it

The grammar is measured against real Clojure with lezer-survey, which has to sit beside this directory:

npm run measure -- ~/corpora/clojure
npm run gaps -- ~/corpora/clojure

Against the 889 Clojure snippets CodinCod's guessing game holds, 853 parse with no error node. The other 36 are files Clojure's own reader turns down as well, mostly a paren the snippet was cut in half by; every file the reader accepts, the grammar reads clean. The reader turns down 50 in all, so on 14 of them this parser is the more forgiving of the two, which is what an editor wants while somebody is still typing.

What it covers

Clojure, ClojureScript, the .cljc files that are both, and EDN, which are one reader between them. Every reader macro the language has: sets, anonymous functions, regular expressions, metadata, namespaced maps, reader conditionals, tagged literals, symbolic values, and the datum comment that removes the form behind it however long it runs.

The tree says which word in a form is the verb, which words a let or a defn binds, and which of the punctuation a Lisp is made of is a name rather than syntax: / is division, % is an argument, result# is a gensym, and a comma is whitespace.

The oracle

tools/reads asks Clojure's own reader whether a file reads, which is how the corpus is told from the files that only share its extension. It runs babashka out of the code runner's Nix store, and prints one line for every file the reader turned down.