23 KiB

Raw Blame History

Contribution guide

We're glad you're interested in contributing Misskey! In this document you will find the information you need to contribute to the project.

[!NOTE] This project uses Japanese as its major language, but you do not need to translate and write the Issues/PRs in Japanese. Also, you might receive comments on your Issue/PR in Japanese, but you do not need to reply to them in Japanese as well.
The accuracy of machine translation into Japanese is not high, so it will be easier for us to understand if you write it in the original language. It will also allow the reader to use the translation tool of their preference if necessary.

Roadmap

See ROADMAP.md

Issues

Before creating an issue, please check the following:

To avoid duplication, please search for similar issues before creating a new issue.
Do not use Issues to ask questions or troubleshooting.
- Issues should only be used to feature requests, suggestions, and bug tracking.
- Please ask questions or troubleshooting in GitHub Discussions or Discord.

[!WARNING] Do not close issues that are about to be resolved. It should remain open until a commit that actually resolves it is merged.

Recommended discussing before implementation

We welcome your proposal.

When you want to add a feature or fix a bug, first have the design and policy reviewed in an Issue (if it is not there, please make one). Without this step, there is a high possibility that the PR will not be merged even if it is implemented.

At this point, you also need to clarify the goals of the PR you will create, and make sure that the other members of the team are aware of them. PRs that do not have a clear set of do's and don'ts tend to be bloated and difficult to review.

Also, when you start implementation, assign yourself to the Issue (if you cannot do it yourself, ask Committer to assign you). By expressing your intention to work on the Issue, you can prevent conflicts in the work.

To the Committers: you should not assign someone on it before the Final Decision.

How issues are triaged

The Committers may:

close an issue that is not reproducible on latest stable release,
merge an issue into another issue,
split an issue into multiple issues,
or re-open that has been closed for some reason which is not applicable anymore.

@syuilo reserves the Final Decision rights including whether the project will implement feature and how to implement, these rights are not always exercised.

Well-known branches

master branch is tracking the latest release and used for production purposes.
develop branch is where we work for the next release.
- When you create a PR, basically target it to this branch.
l10n_develop branch is reserved for localization management.

Creating a PR

Thank you for your PR! Before creating a PR, please check the following:

If possible, prefix the title with a keyword that identifies the type of this PR, as shown below.
- fix / refactor / feat / enhance / perf / chore etc
- Also, make sure that the granularity of this PR is appropriate. Please do not include more than one type of change or interest in a single PR.
If there is an Issue which will be resolved by this PR, please include a reference to the Issue in the text.
Please add the summary of the changes to CHANGELOG.md. However, this is not necessary for changes that do not affect the users, such as refactoring.
Check if there are any documents that need to be created or updated due to this change.
If you have added a feature or fixed a bug, please add a test case if possible.
Please make sure that tests and Lint are passed in advance.
- You can run it with pnpm test and pnpm lint. See more info
If this PR includes UI changes, please attach a screenshot in the text.

Thanks for your cooperation 🤗

Additional things for ActivityPub payload changes

This section is specific to misskey-dev implementation. Other fork or implementation may take different way. A significant difference is that non-"misskey-dev" extension is not described in the misskey-hub's document.

If PR includes changes to ActivityPub payload, please reflect it in misskey-hub's document by sending PR.

The name of purporsed extension property (referred as "extended property" in later) to ActivityPub shall be prefixed by _misskey_. (i.e. _misskey_quote)

The extended property in packages/backend/src/core/activitypub/type.ts must be declared as optional because ActivityPub payloads that comes from older Misskey or other implementation may not contain it.

The extended property must be included in the context definition. Context is defined in packages/backend/src/core/activitypub/misc/contexts.ts. The key shall be same as the name of extended property, and the value shall be same as "short IRI".

"Short IRI" is defined in misskey-hub's document, but usually takes form of misskey:<name of extended property>. (i.e. misskey:_misskey_quote)

One should not add property that has defined before by other implementation, or add custom variant value to "well-known" property.

Reviewers guide

Be willing to comment on the good points and not just the things you want fixed 💯

Review perspective

Scope
- Are the goals of the PR clear?
- Is the granularity of the PR appropriate?
Security
- Does merging this PR create a vulnerability?
Performance
- Will merging this PR cause unexpected performance degradation?
- Is there a more efficient way?
Testing
- Does the test ensure the expected behavior?
- Are there any omissions or gaps?
- Does it check for anomalies?

Deploy

The /deploy command by issue comment can be used to deploy the contents of a PR to the preview environment.

/deploy sha=<commit hash>

An actual domain will be assigned so you can test the federation.

Merge

Release

Release Instructions

Commit version changes in the develop branch (package.json)
Create a release PR.
- Into master from develop branch.
- The title must be in the format Release: x.y.z.
  - x.y.z is the new version you are trying to release.
Deploy and perform a simple QA check. Also verify that the tests passed.
Merge it. (Do not squash commit)
Create a release of GitHub
- The target branch must be master
- The tag name must be the version

[!NOTE] Why this instruction is necessary:

To perform final QA checks

To distribute responsibility

To check direct commits to develop

To celebrate the release together 🎉

Localization (l10n)

Misskey uses Crowdin for localization management. You can improve our translations with your Crowdin account. Your changes in Crowdin are automatically submitted as a PR (with the title "New Crowdin translations") to the repository. The owner @syuilo merges the PR into the develop branch before the next release.

If your language is not listed in Crowdin, please open an issue.

Development

Setup

Before developing, you have to set up environment. Misskey requires Redis, PostgreSQL, and FFmpeg.

You would want to install Meilisearch to experiment related features. Technically, meilisearch is not strict requirement, but some features and tests require it.

There are a few ways to proceed.

Use system-wide software

You could install them in system-wide (such as from package manager).

Use `docker compose`

You could obtain middleware container by typing docker compose -f $PROJECT_ROOT/compose.local-db.yml up -d.

Use Devcontainer

Devcontainer also has necessary setting. This method can be done by connecting from VSCode.

Instead of running pnpm locally, you can use Dev Container to set up your development environment. To use Dev Container, open the project directory on VSCode with Dev Containers installed. Note: If you are using Windows, please clone the repository with WSL. Using Git for Windows will result in broken files due to the difference in how newlines are handled.

It will run the following command automatically inside the container.

git submodule update --init
pnpm install --frozen-lockfile
cp .devcontainer/devcontainer.yml .config/default.yml
pnpm build
pnpm migrate

After finishing the migration, you can proceed.

Start developing

During development, it is useful to use the

pnpm dev

command.

Server-side source files and automatically builds them if they are modified. Automatically start the server process(es).
Vite HMR (just the vite command) is available. The behavior may be different from production.
Service Worker is watched by esbuild.
The front end can be viewed by accessing http://localhost:5173.
The backend listens on the port configured with port in .config/default.yml. If you have not changed it from the default, it will be "http://localhost:3000". If "port" in .config/default.yml is set to something other than 3000, you need to change the proxy settings in packages/frontend/vite.config.local-dev.ts.

`MK_DEV_PREFER=backend pnpm dev`

pnpm dev has another mode with MK_DEV_PREFER=backend.

MK_DEV_PREFER=backend pnpm dev

This mode is closer to the production environment than the default mode.
Vite runs behind the backend (the backend will proxy Vite at /vite).
You can see Misskey by accessing http://localhost:3000 (Replace 3000 with the port configured with port in .config/default.yml).
To change the port of Vite, specify with VITE_PORT environment variable.
HMR may not work in some environments such as Windows.

Testing

You can run non-backend tests by executing following commands:

pnpm --filter frontend test
pnpm --filter misskey-js test

Backend tests require manual preparation of servers. See the next section for more on this.

Backend

There are three types of test codes for the backend:

Unit tests: /packages/backend/test/unit
Single-server E2E tests: /packages/backend/test/e2e
Multiple-server E2E tests: /packages/backend/test-federation

Running Unit Tests or Single-server E2E Tests

Create a config file:

cp .github/misskey/test.yml .config/

Start DB and Redis servers for testing:

docker compose -f packages/backend/test/compose.yml up

Instead, you can prepare an empty (data can be erased) DB and edit .config/test.yml appropriately.

Run all tests:

pnpm --filter backend test     # unit tests
pnpm --filter backend test:e2e # single-server E2E tests

If you want to run a specific test, run as a following command:

pnpm --filter backend test -- packages/backend/test/unit/activitypub.ts
pnpm --filter backend test:e2e -- packages/backend/test/e2e/nodeinfo.ts

Running Multiple-server E2E Tests

See /packages/backend/test-federation/README.md.

Environment Variable

MISSKEY_CONFIG_YML: Specify the file path of config.yml instead of default.yml (e.g. 2nd.yml).
MISSKEY_WEBFINGER_USE_HTTP: If it's set true, WebFinger requests will be http instead of https, useful for testing federation between servers in localhost. NEVER USE IN PRODUCTION.

Continuous integration

Misskey uses GitHub Actions for executing automated tests. Configuration files are located in /.github/workflows.

Vue

Misskey uses Vue(v3) as its front-end framework.

Use TypeScript.
When creating a new component, please use the Composition API (with setup sugar and ref sugar) instead of the Options API.
- Some of the existing components are implemented in the Options API, but it is an old implementation. Refactors that migrate those components to the Composition API are also welcome.

nirax

niraxは、Misskeyで使用しているオリジナルのフロントエンドルーティングシステムです。 vue-routerから影響を多大に受けているので、まずはvue-routerについて学ぶことをお勧めします。

ルート定義

ルート定義は、以下の形式のオブジェクトの配列です。

{
	name?: string;
	path: string;
	component: Component;
	query?: Record<string, string>;
	loginRequired?: boolean;
	hash?: string;
	globalCacheKey?: string;
	children?: RouteDef[];
}

[!WARNING] 現状、ルートは定義された順に評価されます。たとえば、/foo/:idルート定義の次に/foo/barルート定義がされていた場合、後者がマッチすることはありません。

複数のルーター

vue-routerとの最大の違いは、niraxは複数のルーターが存在することを許可している点です。これにより、アプリ内ウィンドウでブラウザとは個別にルーティングすることなどが可能になります。

Storybook

Misskey uses Storybook for UI development.

Setup & Run

Setup

pnpm --filter misskey-js build

Run

pnpm --filter frontend storybook-dev

Usage

When you create a new component (in this example, MyComponent.vue), the story file (MyComponent.stories.ts) will be automatically generated by the .storybook/generate.js script. You can override the default story by creating a impl story file (MyComponent.stories.impl.ts).

/* eslint-disable @typescript-eslint/explicit-function-return-type */
import { StoryObj } from '@storybook/vue3';
import MyComponent from './MyComponent.vue';
export const Default = {
	render(args) {
		return {
			components: {
				MyComponent,
			},
			setup() {
				return {
					args,
				};
			},
			computed: {
				props() {
					return {
						...this.args,
					};
				},
			},
			template: '<MyComponent v-bind="props" />',
		};
	},
	args: {
		foo: 'bar',
	},
	parameters: {
		layout: 'centered',
	},
} satisfies StoryObj<typeof MyComponent>;

If you want to opt-out from the automatic generation, create a MyComponent.stories.impl.ts file and add the following line to the file.

import MyComponent from './MyComponent.vue';
void MyComponent;

You can override the component meta by creating a meta story file (MyComponent.stories.meta.ts).

export const argTypes = {
	scale: {
		control: {
			type: 'range',
			min: 1,
			max: 4,
		},
	},
};

Also, you can use msw to mock API requests in the storybook. Creating a MyComponent.stories.msw.ts file to define the mock handlers.

import { HttpResponse, http } from 'msw';
export const handlers = [
	http.post('/api/notes/timeline', ({ request }) => {
		return HttpResponse.json([]);
	}),
];

Don't forget to re-run the .storybook/generate.js script after adding, editing, or removing the above files.

Nest

Nest Service Circular dependency / Nestでサービスの循環参照でエラーが起きた場合

forwardRef

まずは簡単にforwardRefを試してみる

export class FooService {
	constructor(
		@Inject(forwardRef(() => BarService))
		private barService: BarService
	) {
	}
}

OnModuleInit

できなければOnModuleInitを使う

import { Injectable, OnModuleInit } from '@nestjs/common';
import { ModuleRef } from '@nestjs/core';
import { BarService } from '@/core/BarService';

@Injectable()
export class FooService implements OnModuleInit {
	private barService: BarService // constructorから移動してくる

	constructor(
		private moduleRef: ModuleRef,
	) {
	}

	async onModuleInit() {
		this.barService = this.moduleRef.get(BarService.name);
	}

	public async niceMethod() {
		return await this.barService.incredibleMethod({ hoge: 'fuga' });
	}
}

Service Unit Test

テストでonModuleInitを呼び出す必要がある

// import ...

describe('test', () => {
	let app: TestingModule;
	let fooService: FooService; // for test case
	let barService: BarService; // for test case

	beforeEach(async () => {
		app = await Test.createTestingModule({
			imports: ...,
			providers: [
				FooService,
				{ // mockする (mockは必須ではないかもしれない)
					provide: BarService,
					useFactory: () => ({
						incredibleMethod: jest.fn(),
					}),
				},
				{ // Provideにする
					provide: BarService.name,
					useExisting: BarService,
				},
			],
		})
			.useMocker(...
			.compile();

		fooService = app.get<FooService>(FooService);
		barService = app.get<BarService>(BarService) as jest.Mocked<BarService>;

		// onModuleInitを実行する
		await fooService.onModuleInit();
	});

	test('nice', () => {
		await fooService.niceMethod();

		expect(barService.incredibleMethod).toHaveBeenCalled();
		expect(barService.incredibleMethod.mock.lastCall![0])
			.toEqual({ hoge: 'fuga' });
	});
})

Notes

Misskeyのドメイン固有の概念は`Mi`をprefixする

例えばGoogleが自社サービスをMap、Earth、DriveではなくGoogle Map、Google Earth、Google Driveのように命名するのと同じコード上でMisskeyのドメイン固有の概念にはMiをprefixすることで、他のドメインの同様の概念と区別できるほか、名前の衝突を防ぐ。ただし、文脈上Misskeyのものを指すことが明らかであり、名前の衝突の恐れがない場合は、一時的なローカル変数に限ってMiを省略してもよい。

How to resolve conflictions occurred at pnpm-lock.yaml?

Just execute pnpm to fix it.

INSERTするときにはsaveではなくinsertを使用する

#6441

placeholder

SQLをクエリビルダで組み立てる際、使用するプレースホルダは重複してはならない例えば

query.andWhere(new Brackets(qb => {
	for (const type of ps.fileType) {
		qb.orWhere(`:type = ANY(note.attachedFileTypes)`, { type: type });
	}
}));

と書くと、ループ中でtypeというプレースホルダが複数回使われてしまいおかしくなるだから次のようにする必要がある

query.andWhere(new Brackets(qb => {
	for (const type of ps.fileType) {
		const i = ps.fileType.indexOf(type);
		qb.orWhere(`:type${i} = ANY(note.attachedFileTypes)`, { [`type${i}`]: type });
	}
}));

Not `null` in TypeORM

const foo = await Foos.findOne({
	bar: Not(null)
});

のようなクエリ(barがnullではない)は期待通りに動作しない。次のようにします:

const foo = await Foos.findOne({
	bar: Not(IsNull())
});

`null` in SQL

SQLを発行する際、パラメータがnullになる可能性のある場合はSQL文を出し分けなければならない例えば

query.where('file.folderId = :folderId', { folderId: ps.folderId });

という処理で、ps.folderIdがnullだと結果的にfile.folderId = nullのようなクエリが発行されてしまい、これは正しいSQLではないので期待した結果が得られないだから次のようにする必要がある

if (ps.folderId) {
	query.where('file.folderId = :folderId', { folderId: ps.folderId });
} else {
	query.where('file.folderId IS NULL');
}

`[]` in SQL

SQLを発行する際、INのパラメータが[](空の配列)になる可能性のある場合はSQL文を出し分けなければならない例えば

const users = await Users.find({
	id: In(userIds)
});

という処理で、userIdsが[]だと結果的にuser.id IN ()のようなクエリが発行されてしまい、これは正しいSQLではないので期待した結果が得られないだから次のようにする必要がある

const users = userIds.length > 0 ? await Users.find({
	id: In(userIds)
}) : [];

配列のインデックス in SQL

SQLでは配列のインデックスは1始まり。 [a, b, c]の aにアクセスしたいなら[0]ではなく[1]と書く

null IN

nullが含まれる可能性のあるカラムにINするときは、そのままだとおかしくなるのでORなどでnullのハンドリングをしよう。

enumの削除は気をつける

enumの列挙の内容の削除は、その値をもつレコードを全て削除しないといけない

削除が重たかったり不可能だったりする場合は、削除しないでおく

Migration作成方法

packages/backendで:

pnpm dlx typeorm migration:generate -d ormconfig.js -o <migration name>

生成後、ファイルをmigration下に移してください
作成されたスクリプトは不必要な変更を含むため除去してください

JSON SchemaのobjectでanyOfを使うとき

JSON Schemaで、objectに対してanyOfを使う場合、anyOfの中でpropertiesを定義しないこと。バリデーションが効かないため。（SchemaTypeもそのように作られており、objectのanyOf内のpropertiesは捨てられます） https://github.com/misskey-dev/misskey/pull/10082

テキストhogeおよびfugaについて、片方を必須としつつ両方の指定もありうる場合:

export const paramDef = {
	type: 'object',
	properties: {
		hoge: { type: 'string', minLength: 1 },
		fuga: { type: 'string', minLength: 1 },
	},
	anyOf: [
		{ required: ['hoge'] },
		{ required: ['fuga'] },
	],
} as const;

コネクションには`markRaw`せよ

Vueのコンポーネントのdataオプションとしてmisskey.jsのコネクションを設定するとき、必ずmarkRawでラップしてください。インスタンスが不必要にリアクティブ化されることで、misskey.js内の処理で不具合が発生するとともに、パフォーマンス上の問題にも繋がる。なお、Composition APIを使う場合はこの限りではない(リアクティブ化はマニュアルなため)。

JSONのimportに気を付けよう

TypeScriptでjsonをimportすると、tscでコンパイルするときにそのjsonファイルも一緒にdistディレクトリに吐き出されてしまう。この挙動により、意図せずファイルの書き換えが発生することがあるので、jsonをimportするときは書き換えられても良いものかどうか確認すること。書き換えされて欲しくない場合は、importで読み込むのではなく、fs.readFileSyncなどの関数を使って読み込むようにすればよい。

コンポーネントのスタイル定義でmarginを持たせない

コンポーネント自身がmarginを設定するのは問題の元となることはよく知られている marginはそのコンポーネントを使う側が設定する

その他

HTMLのクラス名で follow という単語は使わない

広告ブロッカーで誤ってブロックされる

indexというファイル名を使うな

ESMではディレクトリインポートは廃止されているのと、ディレクトリインポートせずともファイル名が index だと何故か一部のライブラリ？でディレクトリインポートだと見做されてエラーになる

CSS Recipe

Lighten CSS vars

color: hsl(from var(--MI_THEME-accent) h s calc(l + 10));

Darken CSS vars

color: hsl(from var(--MI_THEME-accent) h s calc(l - 10));

Add alpha to CSS vars

color: color(from var(--MI_THEME-accent) srgb r g b / 0.5);

23 KiB Raw Blame History Unescape Escape

Contribution guide

Roadmap

Issues

Recommended discussing before implementation

How issues are triaged

Well-known branches

Creating a PR

Additional things for ActivityPub payload changes

Reviewers guide

Review perspective

Deploy

Merge

Release

Release Instructions

Localization (l10n)

Development

Setup

Use system-wide software

Use docker compose

Use Devcontainer

Start developing

MK_DEV_PREFER=backend pnpm dev

Testing

Backend

Running Unit Tests or Single-server E2E Tests

Running Multiple-server E2E Tests

Environment Variable

Continuous integration

Vue

nirax

ルート定義

複数のルーター

Storybook

Setup & Run

Setup

Run

Usage

Nest

Nest Service Circular dependency / Nestでサービスの循環参照でエラーが起きた場合

forwardRef

OnModuleInit

Service Unit Test

Notes

Misskeyのドメイン固有の概念はMiをprefixする

How to resolve conflictions occurred at pnpm-lock.yaml?

INSERTするときにはsaveではなくinsertを使用する

placeholder

Not null in TypeORM

null in SQL

[] in SQL

配列のインデックス in SQL

null IN

enumの削除は気をつける

Migration作成方法

JSON SchemaのobjectでanyOfを使うとき

コネクションにはmarkRawせよ

JSONのimportに気を付けよう

コンポーネントのスタイル定義でmarginを持たせない

その他

HTMLのクラス名で follow という単語は使わない

indexというファイル名を使うな

CSS Recipe

Lighten CSS vars

Darken CSS vars

Add alpha to CSS vars

23 KiB

Raw Blame History

Use `docker compose`

`MK_DEV_PREFER=backend pnpm dev`

Misskeyのドメイン固有の概念は`Mi`をprefixする

Not `null` in TypeORM

`null` in SQL

`[]` in SQL

コネクションには`markRaw`せよ