Skip to content

Latest commit

 

History

History
469 lines (372 loc) · 13.7 KB

README.md

File metadata and controls

469 lines (372 loc) · 13.7 KB

Nest-Typed-Config

Never write strings to read config again.

NPM Version Package License NPM Downloads build Coverage

Features

  • Provide a type-safe and intuitive way to config your Nest projects, just as smooth as request DTO.
  • Load your configuration with environment variables, json/yaml/toml configuration files or remote endpoints.
  • Validate your configuration with class-validator and class-transformer
  • Provide easy to use options by default, meanwhile everything is customizable.

Installation

Yarn

yarn add nest-typed-config

NPM

$ npm i --save nest-typed-config

Inspiration

There are various popular configuration modules for Nest framework, such as the official configuration module, nestjs-config and nestjs-easyconfig. These modules can help to manage configurations, validate them, and load them through the ConfigService. But that's when type-safety is gone. For example:

// @nestjs/config, with type-casting
const dbUser = this.configService.get<string>('DATABASE_USER');
// nestjs-config, returns `any` type
const env = this.config.get('app.environment');
// nestjs-easyconfig, only string is supported
const value = this.config.get('key');

Writing type casting is a pain and hard to maintain, and it's common to use non-string configurations in real-world projects. This module aims to provide an intuitive and type-safe way to load, validate and use configurations. Just import any config model, and inject it with full TypeScript support. In a nutshell:

// config.ts
export class Config {
    @IsString()
    public readonly host!: string;

    @IsNumber()
    public readonly port!: number;
}

// app.service.ts
import { Config } from './config';

@Injectable()
export class AppService {
    constructor(private readonly config: Config) {}

    show() {
        console.log(`http://${this.config.host}:${this.config.port}`)
    }
}

Quick Start

Let's define the configuration model first. It can be nested at arbitrary depth.

// config.ts
import { Allow } from 'class-validator';

// validator is omitted for simplicity
export class TableConfig {
    @Allow()
    public readonly name!: string;
}

export class DatabaseConfig {
    @Type(() => TableConfig)
    @Allow()
    public readonly table!: TableConfig;
}

export class RootConfig {
    @Type(() => DatabaseConfig)
    @Allow()
    public readonly database!: DatabaseConfig;
}

Then, add a configuration file such as .env.yaml under project root directory:

database:
    table:
        name: example

After creating the configuration file, import TypedConfigModule and fileLoader to load configuration from file system.

// app.module.ts
import { Module } from '@nestjs/common';
import { TypedConfigModule, fileLoader } from 'nest-typed-config';
import { AppController } from './app.controller';
import { AppService } from './app.service';
import { RootConfig } from './config';

// Register TypedConfigModule
@Module({
    imports: [
        TypedConfigModule.forRoot({
            schema: RootConfig,
            load: fileLoader(),
        }),
    ],
    providers: [AppService],
    controllers: [AppController],
})
export class AppModule {}

That's it! You can use any config or sub-config you defined as injectable services now!

// app.service.ts
import { Injectable } from '@nestjs/common';
import { RootConfig, DatabaseConfig, TableConfig } from './config';

@Injectable()
export class AppService {
    // inject any config or sub-config you like
    constructor(
        private config: RootConfig,
        private databaseConfig: DatabaseConfig,
        private tableConfig: TableConfig,
    ) {}

    // enjoy type safety!
    public show(): any {
        const out = [
            `root.name: ${this.config.name}`,
            `root.database.name: ${this.databaseConfig.name}`,
            `root.database.table.name: ${this.tableConfig.name}`,
        ].join('\n');

        return `${out}\n`;
    }
}

For a full example, please visit our examples folder

Using loaders

Using dotenv loader

The dotenvLoader function allows you to load configuration with dotenv, which is similar to the official configuration module. You can use this loader to load configuration from .env files or environment variables.

Example

TypedConfigModule.forRoot({
    schema: RootConfig,
    load: dotenvLoader({ /* options */ }),
})

Passing options

The dotenvLoader function optionally expects a DotenvLoaderOptions object as a first parameter:

export interface DotenvLoaderOptions {
    /**
     * If set, use the separator to parse environment variables to objects.
     *
     * @example
     *
     * ```bash
     * app__port=8080
     * db__host=127.0.0.1
     * db__port=3000
     * ```
     *
     * if `separator` is set to `__`, environment variables above will be parsed as:
     *
     * ```json
     * {
     *     "app": {
     *         "port": 8080
     *     },
     *     "db": {
     *         "host": "127.0.0.1",
     *         "port": 3000
     *     }
     * }
     * ```
     */
    separator?: string;

    /**
     * If "true", environment files (`.env`) will be ignored.
     */
    ignoreEnvFile?: boolean;

    /**
     * If "true", predefined environment variables will not be validated.
     */
    ignoreEnvVars?: boolean;

    /**
     * Path to the environment file(s) to be loaded.
     */
    envFilePath?: string | string[];

    /**
     * A boolean value indicating the use of expanded variables.
     * If .env contains expanded variables, they'll only be parsed if
     * this property is set to true.
     * 
     * Internally, dotenv-expand is used to expand variables.
     */
    expandVariables?: boolean;
}

Using file loader

The fileLoader function allows you to load configuration with cosmiconfig. You can use this loader to load configuration from files with various extensions, such as .json, .yaml, .toml or .js.

By default, fileLoader searches for .env.{ext} (ext = json, yaml, toml, js) configuration file starting at process.cwd(), and continues to search up the directory tree until it finds some acceptable configuration (or hits the home directory). Moreover, configuration of current environment takes precedence over general configuration (.env.development.toml is loaded instead of .env.toml when NODE_ENV=development)

Example

TypedConfigModule.forRoot({
    schema: RootConfig,
    load: fileLoader({ /* options */ }),
})

Passing options

The fileLoader function optionally expects a FileLoaderOptions object as a first parameter:

import { Options } from 'cosmiconfig';

export interface FileLoaderOptions extends Partial<Options> {
  /**
   * basename of config file, defaults to `.env`.
   *
   * In other words, `.env.yaml`, `.env.yml`, `.env.json`, `.env.toml`, `.env.js`
   * will be searched by default.
   */
  basename?: string;
  /**
   * Use given file directly, instead of recursively searching in directory tree.
   */
  absolutePath?: string;
  /**
   * The directory to search from, defaults to `process.cwd()`. See: https://github.com/davidtheclark/cosmiconfig#explorersearch
   */
  searchFrom?: string;
}

If you want to add support for other extensions, you can use loaders property provided by cosmiconfig:

TypedConfigModule.forRoot({
    schema: RootConfig,
    load: fileLoader({
        // .env.ini has the highest priority now
        loaders: {
          '.ini': iniLoader
        }
    }),
})

Using remote loader

The remoteLoader function allows you to load configuration from a remote endpoint, such as configuration center. Internally axios is used to perform http requests.

Example

TypedConfigModule.forRoot({
    schema: RootConfig,
    load: remoteLoader('http://localhost:8080', { /* options */ }),
})

Passing options

The remoteLoader function optionally expects a RemoteLoaderOptions object as a second parameter, which accepts all axios request configuration except url.

export interface RemoteLoaderOptions extends AxiosRequestConfigWithoutUrl {
  /**
   * Config file type
   */
  type?: ((response: any) => RemoteLoaderConfigType) | RemoteLoaderConfigType;

  /**
   * A function that maps http response body to corresponding config object
   */
  mapResponse?: (config: any) => Promise<any> | any;

  /**
   * A function that determines if the request should be retried
   */
  shouldRetry?: (response: AxiosResponse) => boolean;

  /**
   * Number of retries to perform, defaults to 3
   */
  retries?: number;

  /**
   * Interval in milliseconds between each retry
   */
  retryInterval?: number;
}

You can use the mapResponse function to preprocess the server response before parsing with type, and use shouldRetry function to determine whether server response is valid or not. When server response is not valid, you can use retries and retryInterval to adjust retry strategies. For example:

/*
  Example server response: 
  {
    "code": 0,
    "fileName": ".env.yaml",
    "fileType": "yaml",
    "fileContent": "database:\n    table:\n        name: example"
  }
*/

TypedConfigModule.forRoot({
    schema: RootConfig,
    load: remoteLoader('http://localhost:8080', {
        type: response => response.fileType,
        mapResponse: response => response.fileContent
        // retry when http status is not 200, or response code is not zero
        shouldRetry: response => response.data.code !== 0
        retries: 3,
        retryInterval: 3000
    }),
})

Using multiple loaders

Loading configuration from file system is convenient for development, but when it comes to deployment, you may need to load configuration from environment variables, especially in a dockerized environment. This can be easily achieved by providing multiple loaders. For example:

TypedConfigModule.forRoot({
    schema: RootConfig,
    // Loaders having larger index take precedence over smaller ones,
    // make sure dotenvLoader comes after fileLoader ensures that
    // environment variables always have the highest priority
    load: [
        fileLoader({ /* options */ }),
        dotenvLoader({ /* options */ }),
    ]
})

Using custom loader

If native loaders provided by nest-typed-config can't meet your needs, you can implement a custom loader. This can be achieved by providing a function which returns the configuration object synchronously or asynchronously through the load option. For example:

TypedConfigModule.forRoot({
    schema: RootConfig,
    load: async () => {
        return {
            host: '127.0.0.1',
            port: 3000
        }
    },
})

Default values

Just define your default values in config schema, and you are ready to go:

// config.ts
export class Config {
    @IsString()
    public readonly host: string = '127.0.0.1';

    @IsNumber()
    public readonly port: number = 3000;
}

API

TypedConfigModule.forRoot

export type ConfigLoader = () =>
  | Promise<Record<string, any>>
  | Record<string, any>;

export interface TypedConfigModuleOptions {
  /**
   * The root object for application configuration.
   */
  schema: ClassConstructor<any>;

  /**
   * Function to load configurations, can be sync or async.
   */
  load: ConfigLoader | ConfigLoader[];

  /**
   * If "true", registers `ConfigModule` as a global module.
   * See: https://docs.nestjs.com/modules#global-modules
   */
  isGlobal?: boolean;

  /**
   * Custom function to normalize configurations. It takes an object containing environment
   * variables as input and outputs normalized configurations.
   *
   * This function is executed before validation, and can be used to do type casting,
   * variable expanding, etc.
   */
  normalize?: (config: Record<string, any>) => Record<string, any>;

  /**
   * Custom function to validate configurations. It takes an object containing environment
   * variables as input and outputs validated configurations.
   * If exception is thrown in the function it would prevent the application from bootstrapping.
   */
  validate?: (config: Record<string, any>) => Record<string, any>;

  /**
   * Options passed to validator during validation.
   * @see https://github.com/typestack/class-validator
   */
  validationOptions?: ValidatorOptions;
}

changelog

Please refer to CHANGELOD.md

License

MIT.