rubyfast

A Ruby performance linter rewritten in Rust. Detects 19 common performance anti-patterns in Ruby code.

Rust rewrite of fasterer — same detection rules, but faster execution, parallel file scanning, and zero Ruby runtime dependency.

Installation

cargo install --path .

Usage

# Scan current directory
rubyfast

# Scan a specific file
rubyfast path/to/file.rb

# Scan a specific directory
rubyfast path/to/ruby/project

# Auto-fix safe offenses in-place
rubyfast --fix path/to/ruby/project

# Choose output format
rubyfast --format file path/to/project   # group by file (default)
rubyfast --format rule path/to/project   # group by rule
rubyfast --format plain path/to/project  # one per line (for CI/grep/reviewdog)

Exit code 0 if no offenses, 1 if any offenses found.

Output Formats

`--format file` (default)

Groups offenses by file path — compact and easy to scan:

app/controllers/concerns/lottery_common.rb
  L13  Hash#fetch with second argument is slower than Hash#fetch with block
  L94  Hash#fetch with second argument is slower than Hash#fetch with block

app/controllers/api/v1/health_articles_controller.rb
  L11  Hash#fetch with second argument is slower than Hash#fetch with block

Offenses that support --fix are tagged with (fixable), and the summary line shows how many can be auto-fixed:

tests/fixtures/19_for_loop.rb
  L1  For loop is slower than using each (fixable)

22 files inspected, 41 offenses detected, 21 offenses autocorrectable (run rubyfast --fix)

`--format rule`

Groups offenses by rule — useful for understanding which patterns are most common:

Hash#fetch with second argument is slower than Hash#fetch with block. (3 offenses)
  app/controllers/api/v1/health_articles_controller.rb:11
  app/controllers/concerns/lottery_common.rb:13
  app/controllers/concerns/lottery_common.rb:94

`--format plain`

One offense per line — suitable for grep, reviewdog, and CI pipelines:

app/controllers/api/v1/health_articles_controller.rb:11 Hash#fetch with second argument is slower than Hash#fetch with block.

Auto-fix

rubyfast --fix automatically corrects 8 safe offenses in-place:

#	Pattern	Fix
1	`.shuffle.first`	`.sample`
2	`.select{}.first`	`.detect{}`
4	`.reverse.each`	`.reverse_each`
5	`.keys.each`	`.each_key`
6	`.map{}.flatten(1)`	`.flat_map{}`
7	`.gsub("x","y")`	`.tr("x","y")`
13	`(1..10).include?`	`.cover?`
19	`for x in arr`	`arr.each do \|x\|`

Fixes are applied in reverse byte order with syntax verification — if a fix would produce invalid Ruby, the file is left unchanged.

Detected Offenses

#	Pattern	Suggestion	Auto-fix
1	`.shuffle.first`	`.sample`	Yes
2	`.select{}.first`	`.detect{}`	Yes
3	`.select{}.last`	`.reverse.detect{}`	No
4	`.reverse.each`	`.reverse_each`	Yes
5	`.keys.each`	`.each_key`	Yes
6	`.map{}.flatten(1)`	`.flat_map{}`	Yes
7	`.gsub("x","y")` (single chars)	`.tr("x","y")`	Yes
8	`.sort { \|a,b\| ... }`	`.sort_by`	No
9	`.fetch(k, v)`	`.fetch(k) { v }`	No
10	`.merge!({k: v})`	`h[k] = v`	No
11	`.map { \|x\| x.foo }`	`.map(&:foo)`	No
12	`.each_with_index`	`while` loop	No
13	`(1..10).include?`	`.cover?`	Yes
14	`.module_eval("def ...")`	`define_method`	No
15	`rescue NoMethodError`	`respond_to?`	No
16	`def foo(&block); block.call; end`	`yield`	No
17	`def x; @x; end`	`attr_reader`	No
18	`def x=(v); @x=v; end`	`attr_writer`	No
19	`for x in arr`	`arr.each`	Yes

Inline Disable

Suppress specific offenses with inline comments (similar to RuboCop):

# Disable on the same line
x = [].shuffle.first # rubyfast:disable shuffle_first_vs_sample

# Disable the next line
# rubyfast:disable-next-line shuffle_first_vs_sample
x = [].shuffle.first

# Block disable/enable
# rubyfast:disable for_loop_vs_each
for i in arr
  puts i
end
# rubyfast:enable for_loop_vs_each

# Disable all rules
code # rubyfast:disable all

# Multiple rules
code # rubyfast:disable shuffle_first_vs_sample, for_loop_vs_each

# Backwards compatible with fasterer
code # fasterer:disable shuffle_first_vs_sample

Configuration

Create a .rubyfast.yml (or .fasterer.yml for backwards compatibility) in your project root to disable specific checks or exclude files:

speedups:
  shuffle_first_vs_sample: true
  for_loop_vs_each: false          # disable this check

exclude_paths:
  - vendor/**/*.rb
  - tmp/**/*.rb

The config file is auto-discovered by walking up from the scan directory. .rubyfast.yml takes precedence if both exist.

CI/CD Integration

GitHub Action

- uses: 7a6163/rubyfast-action@v1

With reviewdog inline PR comments (uses --format plain internally):

rubyfast:
  runs-on: ubuntu-latest
  permissions:
    contents: read
    checks: write
    pull-requests: write
  steps:
    - uses: actions/checkout@v4
    - uses: 7a6163/rubyfast-action@v1
      with:
        reviewdog: "true"
        github-token: ${{ secrets.GITHUB_TOKEN }}
        reviewdog-reporter: github-pr-review

See rubyfast-action for all options.

GitLab CI

rubyfast:
  image:
    name: ghcr.io/7a6163/rubyfast:latest
    entrypoint: [""]
  script:
    - rubyfast .

Docker

docker run --rm -v $(pwd):/workspace ghcr.io/7a6163/rubyfast:latest .

Benchmark

Compared against fasterer (v0.11.0, Ruby + ruby_parser) and a prism-based fork (Ruby + prism, unreleased). Measured on Apple Silicon, macOS.

2,235 Ruby files

Tool	Parser	Time	Relative
rubyfast v1.3.1	Rust + ruby-prism	0.21s	1x
fasterer (prism fork)	Ruby + prism	2.09s	10x slower
fasterer v0.11.0	Ruby + ruby_parser	34.1s	162x slower

rubyfast is 162x faster than the original fasterer and 10x faster than the prism-based Ruby fork.

Development

cargo build
cargo test
cargo clippy -- -D warnings

License

MIT

rubyfast 1.3.2