LSPelling

A language server to spellcheck code using different stategies

Trying out lspelling

With the Helix editor

• Build the project using cargo build

• Add the built binary to your PATH

  • fish: set -gax PATH (realpath ./target/debug)
  • bash: PATH=$PATH:$(realpath target/debug)

• Change directory to ./examples/helix and play with the two files in there

Status

• Only Rust files have a custom tree-sitter fragmentizer, other use a dumb(-ish) fragmentizer (works fine though).
• Only supports ruspell as a spellchecking backend
• ruspell doesn't implement custom dictionaries so the "Saving to a custom dictionary" code action is a no-op.

How it works

ls space

• waits for on_change events and passes the whole source to wordc's Checker

• send diagnostics to the client and listens to code actions events

wordc space

• checker: combines a fragment processor with a spellchecking provider, uses the tokens spit out of the processor to feed the spellchecker and provide diagnostics.

• fragment processor: parses source code into text fragments of different kind, uses a fragmentizer.

  • Rust fragmentizer is implemented with tree-sitter and a query to select idents and comments

  • a Dumb fallback fragmentizer skips non-alphanumeric characters

• fragmentizer: depends on the language. uses domain knowledge to fragment source code and avoid keyword.

• spellchecking provider: takes tokens and spellcheck them

  • only tokens as of today are Words

  • is though to be be extended to support full Sentences to pass to tools like LanguageTool.

---

This work is licensed under CeCILL-2.1, a strong copyleft French OSS license. This license allows modification and distribution of the software while requiring the same license for derived works.