huacnlee / autocorrect

A linter and formatter to help you to improve copywriting, correct spaces, words, and punctuations between CJK (Chinese, Japanese, Korean).
https://huacnlee.github.io/autocorrect
MIT License
929 stars 34 forks source link
autocorrect copywriting formatter linter rust spellcheck webassembly

AutoCorrect Icon

AutoCorrect

Go GitHub release (latest by date) Docker Image Version (latest server) Crates.io NPM PyPI version Gem Version Maven Central

🎯 AutoCorrect 的愿景是提供一套标准化的文案较正方案。以便于在各类场景(例如:撰写书籍、文档、内容发布、项目源代码...)里面应用,让使用者轻松实现标准化、专业化的文案输出 / 校正。

AutoCorrect is a linter and formatter to help you to improve copywriting, correct spaces, words, and punctuations between CJK (Chinese, Japanese, Korean).

Like Eslint, Rubocop and Gofmt ..., AutoCorrect allows us to check source code, and output as colorized diff with corrected suggestions. You can integrate to CI (GitLab CI, GitHub Action, Travis CI....) for use to check the contents in source code. Recognize the file name, and find out the strings and the comment part.

AutoCorrect 是一个基于 Rust 编写的工具,用于「自动纠正」或「检查并建议」文案,给 CJK(中文、日语、韩语)与英文混写的场景,补充正确的空格,纠正单词,同时尝试以安全的方式自动纠正标点符号等等。

类似 ESlint、Rubocop、Gofmt 等工具,AutoCorrect 可以用于 CI 环境,它提供 Lint 功能,能便捷的检测出项目中有问题的文案,起到统一规范的作用。

支持各种类型源代码文件,能自动识别文件名,并准确找到字符串、注释做自动纠正。

此方案最早于 2013 年 出现于 Ruby China 的项目,并逐步完善规则细节,当前准确率较高(极少数异常情况),你可以放心用来辅助你完成自动纠正动作。

autocorrect lint output

Features

典型应用场景

Installation

Install on macOS You can install it via [Homebrew](https://brew.sh): ```bash $ brew install autocorrect ```
Install on Windows You can install it via [Scoop](https://scoop.sh): ```bash $ scoop install autocorrect ```

Or you can just install it via this on Unix-like system:

$ curl -sSL https://git.io/JcGER | sh

After that, you will get autocorrect command.

$ autocorrect -V
AutoCorrect 2.4.0

Or install NPM:

$ yarn add autocorrect-node
$ yarn autocorrect -V

Upgrade

Since: 1.9.0

AutoCorrect allows you to upgrade itself by autocorrect update command.

$ autocorrect update

NOTE: This command need you input your password, because it will install bin into /usr/local/bin directory.

Usage

Use in CLI

$ autocorrect text.txt
你好 Hello 世界

$ echo "hello世界" | autocorrect --stdin
hello 世界

$ autocorrect --fix text.txt
$ autocorrect --fix zh-CN.yml
$ autocorrect --fix

Lint

$ autocorrect --lint --format json text.txt

$ autocorrect --lint text.txt
Error: 1, Warning: 0

text.txt:1:3
-你好Hello世界
+你好 Hello 世界

You also can lint multiple files:

$ autocorrect --lint

How to lint all changed files in Git:

$ git diff --name-only | xargs autocorrect --lint

Use in NPM

since: 2.7.0

AutoCorrect has been published in NPM with CLI command support. If you want to use it in Frontend or Node.js project, you can just install autocorrect-node package for without install AutoCorrect bin.

cd your-project
yarn add autocorrect-node

Now you can run yarn autocorrect command in your project. This command is same as autocorrect command.

$ yarn autocorrect -h

More docs: autocorrect-node/README.md

Configuration

Default config: .autocorrect.default

$ autocorrect init
AutoCorrect init config: .autocorrectrc

NOTE: If you download fail, try to use autocorrect init --local command again.

Now the .autocorrectrc file has been created.

.autocorrectrc is allows use YAML, JSON format.

Config file example:

# yaml-language-server: $schema=https://huacnlee.github.io/autocorrect/schema.json
# Config rules
rules:
  # Auto add spacing between CJK (Chinese, Japanese, Korean) and English words.
  # 0 - off, 1 - error, 2 - warning
  space-word: 1
  # Add space between some punctuations.
  space-punctuation: 1
  # Add space between brackets (), [] when near the CJK.
  space-bracket: 1
  # Add space between ``, when near the CJK.
  space-backticks: 1
  # Add space between dash `-`
  space-dash: 0
  # Convert to fullwidth.
  fullwidth: 1
  # To remove space near the fullwidth.
  no-space-fullwidth: 1
  # Fullwidth alphanumeric characters to halfwidth.
  halfwidth-word: 1
  # Fullwidth punctuations to halfwidth in english.
  halfwidth-punctuation: 1
  # Spellcheck
  spellcheck: 2
# Enable or disable in a specific context
context:
  # Enable or disable to format codeblock in Markdown or AsciiDoc etc.
  codeblock: 1
textRules:
  # Config special rules for some texts
  # For example, if we wants to let "Hello你好" just warning, and "Hi你好" to ignore
  # "Hello你好": 2
  # "Hi你好": 0
fileTypes:
  # Config the files associations, you config is higher priority than default.
  # "rb": ruby
  # "Rakefile": ruby
  # "*.js": javascript
  # ".mdx": markdown
spellcheck:
  # Correct Words (Case insensitive) for by Spellcheck
  words:
    - GitHub
    - App Store
    # This means "appstore" into "App Store"
    - AppStore = App Store
    - Git
    - Node.js
    - nodejs = Node.js
    - VIM
    - DNS
    - HTTP
    - SSL

Ignore option

Since: 2.2.0

When you want to config some special words or texts to ignore on format or lint.

The textRules config may help you.

For example, we want:

Use can config:

textRules:
  Hello世界: 2
  Hi你好: 0

After that, AutoCorrect will follow your textRules to process.

Ignore files

Use .autocorrectignore to ignore files

Sometimes, you may want to ignore some special files that not want to check.

By default, the file matched .gitignore rule will be ignored.

You can also use .autocorrectignore to ignore other files, format like .gitignore.

Disable by inline comment

If you just want to disable some special lines in a file, you can write a comment autocorrect-disable, when AutoCorrect matched the comment include that, it will disable temporarily.

And then, you can use autocorrect-enable to reopen it again.

For example, in JavaScript:

function hello() {
  // autocorrect-disable
  console.log("现在这行开始autocorrect会暂时禁用");
  console.log("这行也是disable的状态");
  // autocorrect-enable
  let a = "现在起autocorrect回到了启用的状态";
}

The output will:

function hello() {
  // autocorrect-disable
  console.log("现在这行开始autocorrect会暂时禁用");
  console.log("这行也是disable的状态");
  // autocorrect-enable
  let a = "现在起 autocorrect 回到了启用的状态";
}

Disable some rules

Since: 2.0

You can use autocorrect-disable <rule> in a comment to disable some rules.

Rule names please see: Configuration

function hello() {
  // autocorrect-disable space-word
  console.log("现在这行开始autocorrect会暂时禁用.");
  // autocorrect-disable fullwidth
  console.log("这行也是disable的状态.");
  // autocorrect-enable
  let a = "现在起autocorrect回到了启用的状态.";
}

Will get:

function hello() {
  // autocorrect-disable space-word
  console.log("现在这行开始autocorrect会暂时禁用。");
  // autocorrect-disable fullwidth, space-word
  console.log("这行也是disable的状态.");
  // autocorrect-enable
  let a = "现在起 autocorrect 回到了启用的状态。";
}

VS Code Extension

Install Extension

https://marketplace.visualstudio.com/items?itemName=huacnlee.autocorrect

内置 Visual Studio Code 插件,安装后会将 AutoCorrect 和 Visual Studio Code 完整集成,可以达到「保存自动格式化」或「纠正提示」。

如下图:

AutoCorrect for VS Code Extension

Intellij Platform Plugin

AutoCorrect for Intellij Platform Plugin

https://github.com/huacnlee/autocorrect-idea-plugin

GitHub Action

https://github.com/huacnlee/autocorrect-action

Add to your .github/workflows/ci.yml

steps:
  - name: Check source code
    uses: actions/checkout@v4

  - name: AutoCorrect
    uses: huacnlee/autocorrect-action@main

GitLab CI

Add to your .gitlab-ci.yml, to use huacnlee/autocorrect Docker image to check.

autocorrect:
  stage: build
  image: huacnlee/autocorrect:latest
  script:
    - autocorrect --lint
  # Enable allow_failure if you wants.
  # allow_failure: true

Work with ReviewDog

Since: 2.8.0

AutoCorrect can work with reviewdog, so you can use it in CI/CD. ReviewDog will post a comment to your PR with the AutoCorrect change suggestions. Then the PR committer can easy to accept the suggestions.

Use --format rdjson option to output the lint results as the reviewdog supported format.

autocorrect --lint --format rdjson | reviewdog -f=rdjson -reporter=github-pr-review

Use huacnlee/autocorrect-action can help you setup GitHub Action.

Use for programming

AutoCorrect makes for support use in many programming languages.

Benchmark

MacBook Pro (13-inch, M1, 2020)

Use make bench to run benchmark tests.

See autocorrect/src/benches/example.rs for details.

format_050              time:   [8.2420 µs 8.2657 µs 8.2937 µs]
format_100              time:   [14.199 µs 14.246 µs 14.298 µs]
format_400              time:   [40.511 µs 41.923 µs 43.798 µs]
format_html             time:   [204.94 µs 208.61 µs 214.07 µs]
halfwidth_english       time:   [2.4983 µs 2.5541 µs 2.6293 µs]
format_json             time:   [54.037 µs 57.023 µs 61.821 µs]
format_javascript       time:   [102.81 µs 104.41 µs 106.92 µs]
format_json_2k          time:   [8.7609 ms 8.9099 ms 9.1201 ms]
format_jupyter          time:   [81.765 µs 83.038 µs 85.321 µs]
format_markdown         time:   [879.27 µs 894.86 µs 918.30 µs]

spellcheck_50           time:   [1.6012 µs 1.6122 µs 1.6306 µs]
spellcheck_100          time:   [3.0968 µs 3.1696 µs 3.2653 µs]
spellcheck_400          time:   [10.136 µs 10.478 µs 10.898 µs]

lint_markdown           time:   [937.57 µs 942.59 µs 949.15 µs]
lint_json               time:   [59.174 µs 60.302 µs 61.763 µs]
lint_html               time:   [238.03 µs 241.38 µs 245.77 µs]
lint_javascript         time:   [111.64 µs 113.05 µs 114.82 µs]
lint_yaml               time:   [348.56 µs 350.11 µs 352.80 µs]
lint_to_json            time:   [941.25 µs 948.95 µs 958.26 µs]
lint_to_diff            time:   [1.0573 ms 1.0823 ms 1.1134 ms]

Real world benchmark

With MDN Translated Content project, it has about 30K files.

~/work/translated-content $ autocorrect --fix
AutoCorrect spend time: 8402.538ms

Other Extensions

The other implementations from the community.

User cases

License

This project under MIT license.