edcfg-lint 0.1.1

A fast EditorConfig linter written in Rust

edcfg-lint

A fast editorconfig linter written in Rust. It is many times faster than editorconfig-checker.

Still a work in progress, in particular the harness is mostly vibe-coded and needs a serious refresh. However, it already approaches the limits of what the hardware and OS can do despite this.

How it works inside

There honestly isn't a lot of magic: we use standard Rust crates.

• ignore is used for walking and parallelism.
• ec4rs is used as our EditorConfig core.
• memchr is used for some fast (substring and character) searches.
• encoding_rs is used to handle charset detection and decoding.

edcfg-lint primarily comprises of:

• The harness (src/main.rs)
• The checker (src/file.rs)
• A caching version of ec4rs::properties_of (src/ec.rs)

These crates are highly optimized (we're using a bunch of stuff that makes ripgrep so fast).

Performance

Benchmarked on an 16-inch MacBook Pro with M1 Max chip with 64 GB of memory and 4 TB SSD. Linting was done over home-assistant/core commit fc281b2faecdc86c0a716dfb4a1f243dde632297.

`.editorconfig` used

root = true

[*]
indent_style = space
indent_size = 4
end_of_line = lf
charset = utf-8
trim_trailing_whitespace = true
insert_final_newline = true

[*.py]
indent_style = space
indent_size = 4

[*.{js,json,yml,yaml}]
indent_style = space
indent_size = 2

[*.md]
trim_trailing_whitespace = false

editorconfig-checker 3.6.0 (without -disable-indent-size, as edcfg-lint doesn't do this check at all):

  Time (mean ± σ):      3.668 s ±  0.021 s    [User: 18.099 s, System: 2.113 s]
  Range (min … max):    3.655 s …  3.692 s    3 runs

edcfg-lint:

  Time (mean ± σ):     432.1 ms ± 133.7 ms    [User: 1394.8 ms, System: 2495.2 ms]
  Range (min … max):   322.0 ms … 586.6 ms    5 runs

edcfg-lint is ~8x faster.