eslint-plugin-codegen

An eslint plugin for inline codegen. Auto-fixes out of sync code, with presets for barrels, jsdoc to markdown and more.

Motivation

Sometimes the same information is useful in multiple places - for example, jsdoc comments in code can double as markdown-formatted documentation for a library.

This allows generating code in a project using eslint, without having to incorporate any extra build tools, either for the codegen itself, or to validate that the generated code is up to date. So references to other parts of the project will always stay up to date - and your existing CI tools can enforce this just by running eslint.

Here's an example of it being used along with VSCode's eslint plugin, with auto-fix-on-save:

Contents

• Motivation
• Contents
• How to use
  • Setup
    • Usage with eslint-plugin-markdown
  • Presets
    • custom
    • barrel
    • markdownFromJsdoc
    • monorepoTOC
    • markdownFromJsdoc
    • markdownTOC
    • markdownFromTests
    • labeler
  • Customisation

How to use

Caveat

Before you use this, note that it's still in v0. That means:

1. Breaking changes might happen. Presets might be renamed, or have their options changed. The documentation should stay up to date though, since that's partly the point of the project.
2. There are missing features, or incompletely-implemented ones. For example, markdownFromJsdoc only works with export const ... style exports. Currently most of the features implemented are ones that are specifically needed for this git repo.
3. There might be bugs. The project is in active development - raise an issue if you find one!

Setup

In an eslint-enabled project, install with

1npm install --save-dev eslint-plugin-codegen

or

1yarn add --dev eslint-plugin-codegen

Then add the plugin and rule to your eslint config, for example in eslintrc.js:

1module.exports = {
2  //...
3  plugins: [
4    // ...
5    'codegen',
6  ],
7  rules: {
8    // ...
9    'codegen/codegen': 'error',
10  },
11}

You can use the rule by running eslint in a standard way, with something like this in an npm script: eslint --ext .ts,.js,.md .

In vscode, if using the eslint plugin, you may need to tell it to validate markdown files in your repo's .vscode/settings.json file (see this repo for an example):

1{
2  "eslint.validate": ["markdown", "javascript", "typescript"],
3  "editor.codeActionsOnSave": {
4    "source.fixAll.eslint": true
5  }
6}

To trigger the rule, add a comment line to a source file.

In markdown:

<!-- codegen:start {{ OPTIONS }} -->

In typescript/javascript:

// codegen:start {{ OPTIONS }}

Where {{ OPTIONS }} are an inline object in the format:

{preset: presetName, key1: value1, key2: value2}

Where key1 and key2 are options passed to the codegen preset. yaml is used to parse the object, So any valid yaml that fits on one line can be passed as options. In practise, the one-line restriction means using yaml's "flow style" for collections.

See below for documentation. This repo also has lots of usage examples.

Usage with eslint-plugin-markdown

This plugin uses an ESLint processor to handle markdown and YAML files. ESLint only allows one processor per file type, so the processor from this plugin is designed to be compatible with eslint-plugin-markdown. But to use both plugins, you need to use the process for eslint-plugin-codegen, not eslint-plugin-markdown. You can do this by adding the recommended config for eslint-plugin-codegen second, e.g.

1module.exports = {
2  plugins: ['markdown', 'codegen'],
3  extends: ['plugin:markdown/recommended', 'plugin:codegen/recommended'],
4}

Or specify the processor explicitly - when you switch to flat config this will be required:

1module.exports = {
2  // 1. Add the plugin.
3  plugins: ['markdown'],
4  overrides: [
5    {
6      // 2. Enable the Markdown processor for all .md files.
7      files: ['**/*.md'],
8      processor: 'codegen/processor', // NOT 'markdown/markdown'
9    },
10  ],
11}

Presets

custom

Define your own codegen function, which will receive all options specified. Import the Preset type from this library to define a strongly-typed preset function:

Example

1export const jsonPrinter: import('eslint-plugin-codegen').Preset<{myCustomProp: string}> = ({meta, options}) => {
2  const components = meta.glob('**\/*.tsx') // uses 'globSync' from glob package
3  const json = JSON.stringify({filename: meta.filename, customProp: options.myCustomProp, components}, null, 2)
4  return `export default ${json}`
5}
6
7// codegen:start {export: jsonPrinter}

This can be used in other files by specifying the source option like:

<!-- codegen:start {source: ./lib/my-custom-preset.js, export: jsonPrinter, myCustomProp: hello}

Note that some helpers passed via dependencies, such as glob, fs, path, child_process, lodash, jsYaml, dedent, and readPkgUp, corresponding to those node modules respectively. These can be useful to allow access to those libraries without them being production dependencies. This also allows your lint process to use these node-only dependencies, even in a file that is not run in node - only the calls would be included in any bundled output, not the dependencies themselves.

Params

name	description
source	Relative path to the module containing the custom preset. Default: the file being linted.
export	The name of the export. If omitted, the module's default export should be a preset function.
require	A module to load before source. If not set, defaults to tsx/cjs or ts-node/register/transpile-only for typescript sources.
dev	Set to true to clear the require cache for source before loading. Allows editing the function without requiring an IDE reload. Default false if the CI enviornment variable is set, true otherwise.

Caching

Sometimes, you may want to write a custom function that takes a long time to run. To avoid having to wait for it to run every time you lint, you can set the dev option to true. This will clear the require cache for the custom function, so it will be run fresh each time.

1export const longRunningFunction: import('eslint-plugin-codegen').Preset = ({cache}) => {
2  const result = cache({maxAge: '1 year'}, () => {
3    // do something that takes a long time to run, but doesn't need to run too often
4  })
5  return result
6}
7// codegen:start {preset: custom, export: longRunningFunction}

You can use some helpers that are passed to the preset function:

1export const secretaryGeneralLogStatement: import('eslint-plugin-codegen').Preset = ({cache, dependencies}) => {
2  const result = cache({maxAge: '1 year'}, () => {
3    const res = dependencies.fetchSync('https://www.un.org/sg')
4    const secretaryGeneral = (/Secretary-General ([A-Z].+?) [a-z]/.exec(res.text))?.[1]
5    return `console.log('The UN Secretary-General is ${secretaryGeneral}')`
6  })
7  return result
8}
9
10// codegen:start {preset: custom, export: secretaryGeneralLogStatement}

This will transform to something like:

1export const secretaryGeneralLogStatement: import('eslint-plugin-codegen').Preset = ({cache, dependencies}) => {
2  return cache({maxAge: '4 weeks'}, () => {
3    const res = dependencies.fetchSync('https://en.wikipedia.org/wiki/Secretary-General_of_the_United_Nations')
4    const $ = dependencies.cheerio.load(res.text)
5    const incumbent = $('.infobox div:contains("Incumbent")')
6    const secretaryGeneral = incumbent.find('a').text()
7    return `console.log('The UN Secretary-General is ${secretaryGeneral}')`
8  })
9  return result
10}
11
12// codegen:start {preset: custom, export: secretaryGeneralLogStatement}
13// codegen:hash {input: 4119892f2e6eaf56ae5c346de91be718, output: eed0d07c81b82bff1d3e4751073b0112, timestamp: 2025-03-05T18:58:13.921Z}
14console.log('The UN Secretary-General is António Guterres')
15// codegen:end

It will not re-run unless the input has changed, the output hash doesn't match, or until 1 year after the recorded timestamp. "The input" is a hash of the following:

• the filename the directive appears in
• the source code excluding any existing content between the codegen:start and codegen:end directives
• the options passed to the preset

The output (i.e. the generated code between the start and end directives) is also hashed and written to the hash directive.

If the the generated code doesn't match the output hash, the generator function will re-run.

This means that if you change the filename, or any of the source code, it will re-run. You can control this behaviour when defining your generator function:

1export const secretaryGeneralLogStatement: import('eslint-plugin-codegen').Preset =
2  ({cache, dependencies}) => {
3    return cache({maxAge: '4 weeks'}, () => {
4      const res = dependencies.fetchSync(
5        'https://en.wikipedia.org/wiki/Secretary-General_of_the_United_Nations',
6      )
7      const $ = dependencies.cheerio.load(res.text)
8      const incumbent = $('.infobox div:contains("Incumbent")')
9      const secretaryGeneral = incumbent.find('a').text()
10      return `console.log('The UN Secretary-General is ${secretaryGeneral}')`
11    })
12  }
13
14// codegen:start {preset: custom, export: secretaryGeneralLogStatement}
15// codegen:hash {input: 4119892f2e6eaf56ae5c346de91be718, output: eed0d07c81b82bff1d3e4751073b0112, timestamp: 2025-03-05T18:58:13.921Z}
16console.log('The UN Secretary-General is António Guterres')
17// codegen:end

The helpers that are provided to the generator function via the dependencies prop are listed below. You can use all of them in a type-safe way in your generator function, without having to add them as dependencies or even devDependencies:

• fs: https://nodejs.org/api/fs.html
• path: https://nodejs.org/api/path.html
• child_process: https://nodejs.org/api/child_process.html
• lodash: https://npmjs.com/package/lodash
• jsYaml: https://npmjs.com/package/js-yaml
• dedent: https://npmjs.com/package/dedent
• glob: https://npmjs.com/package/glob
• readPkgUp: https://npmjs.com/package/read-pkg-up
• cheerio: https://npmjs.com/package/cheerio
• makeSynchronous: A function for making functions synchronous by running them in a subprocess. See the code for more details. It's a simplified version of this. Note: it's strongly recommended to use this with the cache feature to avoid slowing down your lint process.
• fetchSync: A simplified fetch wrapper that runs synchronously via makeSynchronous. See the code for more details. Useful for fetching data from the internet without adding a production dependency. Note: it's strongly recommended to use this with the cache feature to avoid slowing down your lint process.

TypeScript

The plugin will attempt to run generator functions written in typescript. If tsx or ts-node are available, they will be used to register their respective loaders before requiring the file containing the generator function. If you have trouble, try installing tsx even if you don't otherwise use it - or just write the generator function in a separate javascript file.

Note: to get type safety for the helpers in javascript files, use the {@type ...} syntax:

1/** @type {import('eslint-plugin-codegen').Preset} */
2module.exports.myGenerator = ({dependencies}) => {
3  const subpackages = dependencies.glob.sync('subpackages/**/*.package.json')
4  return `const subpackages = ${JSON.stringify(subpackages)}`
5}

Demo

barrel

Bundle several modules into a single convenient one.

Example

1// codegen:start {preset: barrel, include: some/path/*.ts, exclude: some/path/*util.ts}
2export * from './some/path/module-a'
3export * from './some/path/module-b'
4export * from './some/path/module-c'
5// codegen:end

Params

name	description
include	[optional] If specified, the barrel will only include file paths that match this glob pattern
exclude	[optional] If specified, the barrel will exclude file paths that match these glob patterns
import	[optional] If specified, matching files will be imported and re-exported rather than directly exported with export * from './xyz'. Use import: star for import * as xyz from './xyz' style imports. Use import: default for import xyz from './xyz' style imports.
export	[optional] Only valid if the import style has been specified (either import: star or import: default). If specified, matching modules will be bundled into a const or default export based on this name. If set to {name: someName, keys: path} the relative file paths will be used as keys. Otherwise the file paths will be camel-cased to make them valid js identifiers.
extension	[optional] Useful for ESM modules. If set to true files will be imported with the file extension. If set to an object, extensions will be converted using this object.

Demo

markdownFromJsdoc

Convert jsdoc for an es export from a javascript/typescript file to markdown.

Example

<!-- codegen:start {preset: markdownFromJsdoc, source: src/foo.ts, export: bar} -->

Params

name	description
source	{string} relative file path containing the export with jsdoc that should be copied to markdown
export	{string} the name of the export
headerLevel	{1

monorepoTOC

Generate a table of contents for a monorepo.

Example (basic)

<!-- codegen:start {preset: monorepoTOC} -->

Example (using config options)

<!-- codegen:start {preset: monorepoTOC, repoRoot: .., workspaces: lerna, filter: {package.name: foo}, sort: -readme.length} -->

Params

name	description
repoRoot	[optional] the relative path to the root of the git repository. By default, searches parent directories for a package.json to find the "root".
filter	[optional] a dictionary of filter rules to whitelist packages. Filters can be applied based on package.json keys, examples: - filter: '@myorg/.*-lib' (match packages with names matching this regex) - filter: { package.name: '@myorg/.*-lib' } (equivalent to the above) - filter: { package.version: '^[1-9].*' } (match packages with versions starting with a non-zero digit, i.e. 1.0.0+) - filter: '^(?!.*(internal$))' (match packages that do not contain "internal" anywhere (using negative lookahead)) - filter: { package.name: '@myorg', path: 'libraries' } (match packages whose name contains "@myorg" and whose path matches "libraries") - filter: { readme: 'This is production-ready' } (match packages whose readme contains the string "This is production-ready")
sort	[optional] sort based on package properties (see filter), or readme length. Use - as a prefix to sort descending. examples: - sort: package.name (sort by package name) - sort: -readme.length (sort by readme length, descending) - sort: toplogical (sort by toplogical dependencies, starting with the most depended-on packages)

Demo

markdownFromJsdoc

Convert jsdoc to an es export from a javascript/typescript file to markdown.

Example

<!-- codegen:start {preset: markdownFromJsdoc, source: src/foo.ts, export: bar} -->

Params

name	description
source	{string} relative file path containing the export with jsdoc that should be copied to markdown
export	{string} the name of the export

Demo

markdownTOC

Generate a table of contents from the current markdown file, based on markdown headers (e.g. ### My section title)

Example

<!-- codegen:start {preset: markdownTOC, minDepth: 2, maxDepth: 5} -->

Params

name	description
minDepth	exclude headers with lower "depth". e.g. if set to 2, # H1 would be excluded but ## H2 would be included. @default 2
maxDepth	exclude headers with higher "depth". e.g. if set to 3, #### H4 would be excluded but ### H3 would be included. @default Infinity

Demo

markdownFromTests

Use a test file to generate library usage documentation. Note: this has been tested with vitest and jest. It might also work fine with mocha, and maybe ava, but those haven't been tested. JSDoc/inline comments above tests will be added as a "preamble", making this a decent way to quickly document API usage of a library, and to be sure that the usage is real and accurate.

Example

<!-- codegen:start {preset: markdownFromTests, source: test/foo.test.ts, headerLevel: 3} -->

Params

name	description
source	the test file
include	if defined, only tests with titles matching one of these regexes will be included
exclude	if defined, tests with titles matching one of these regexes will be excluded
headerLevel	The number of # characters to prefix each title with
includeEslintDisableDirectives	If true, // eslint-disable ... type comments will be included in the preamble. @default false

Demo

labeler

Generates a yaml config for the GitHub Pull Request Labeler Action. Creates a label per package name, which will be applied to any file modified under the leaf package path. When packages are added or removed from the repo, or renamed, the yaml config will stay in sync with them. Additional labels can be added outside of the generated code block. See https://github.com/mmkal/ts/tree/main/.github/labeler.yml for an example.

Example

1# codegen:start {preset: labeler}

Note: eslint and related tools make it quite difficult to lint github action yaml files. To get it working, you'll need to:

• add '!.github' to your .eslintignore file, or the ignorePatterns property in your lint config.
• {vscode} add "yaml" to the "eslint.validate" list in vscode/settings.json.
• {@typescript/eslint} add '.yml' (and/or '.yaml') to the parserOptions.extraFileExtensions list in your lint config.
• {@typescript/eslint} explicitly include 'hidden' files (with paths starting with .) in your tsconfig. See https://github.com/mmkal/ts/tree/main/tsconfig.eslint.json for an example.

Params

name	description
repoRoot	[optional] path to the repository root. If not specified, the rule will recursively search parent directories for package.json files

Demo

Customisation

In addition to the custom preset, you can also define your own presets in eslint configuration, e.g.:

1module.exports = {
2  // ...
3  plugins: [
4    // ...
5    'codegen',
6  ],
7  rules: {
8    // ...
9    'codegen/codegen': ['error', {presets: require('./-presets')}],
10  },
11}

presets should be a record of preset functions, conforming to the Preset interface from this package. This can be used to extend the in-built ones. For example, you could make generated markdown collapsible:

Before:

 <!-- codegen:start {preset: markdownTOC}-->
 - [Section1](#section1)
 - [Section2](#section2)
 <!-- codeg```

`my-custom-presets.js`:

<!-- eslint-disable @typescript-eslint/no-var-requires -->
<!-- eslint-disable import/no-extraneous-dependencies -->
```js
const {presets} = require('eslint-plugin-codegen')

module.exports.markdownTOC = params => {
  const toc = presets.markdownTOC(params)
  return [
    '<details>',
    '<summary>click to expand</summary>',
    '',
    toc,
    '</details>',
  ].join('\n')
}

.eslintrc.js:

1module.exports = {
2  // ...
3  plugins: [
4    // ...
5    'codegen',
6  ],
7  rules: {
8    // ...
9    'codegen/codegen': ['error', {presets: require('./my-custom-presets')}],
10  },
11}

After:

readme.md:

 <!-- codegen:start {preset: markdownTOC}-->
 <details>
  <summary>click to expand</summary>

 - [Section1](#section1)
 - [Section2](#section2)
 </details>
 <!-- codegen:end -->

Rendered:

click to expand

• Section1
• Section2

---

_{The code in this repository was moved from https://github.com/mmkal/ts}