NestJS Commander

Have you been building amazing applications with NestJS? Do you want that same structure for absolutely everything you're working with? Have you always wanted to build up some sweet CLI application but don't really know where to start? This is the solution. A package to bring building CLI applications to the Nest world with the same structure that you already know and love :heart: Built on top of the popular Commander package.

Installation

Before you get started, you'll need to install a few packages. First and foremost, this one: nest-commander (name pending). You'll also need to install @nestjs/common and @nestjs/core as this package makes use of them under the hood, but doesn't want to tie you down to a specific version, yay peerDependencies!

1npm i nest-commander @nestjs/common @nestjs/core
2# OR
3yarn add nest-commander @nestjs/common @nestjs/core
4# OR
5pnpm i nest-commander @nestjs/common @nestjs/core

A Command File

nest-commander makes it easy to write new command line applications with decorators via the @Command() decorator for classes and the @Option() decorator for methods of that class. Every command file should implement the CommandRunner abstract class and should be decorated with a @Command() decorator.

CommandRunner

Every command is seen as an @Injectable() by Nest, so your normal Dependency Injection still works as you would expect it to (woohoo!). The only thing to take note of is the abstract class CommandRunner, which should be implemented by each command. The CommandRunner abstract class ensures that all commands have a run method that return a Promise<void> and takes in the parameters string[], Record<string, any>. The run command is where you can kick all of your logic off from, it will take in whatever parameters did not match option flags and pass them in as an array, just in case you are really meaning to work with multiple parameters. As for the options, the Record<string, any>, the names of these properties match the name property given to the @Option() decorators, while their value matches the return of the option handler. If you'd like better type safety, you are welcome to create an interface for your options as well. You can view how the Basic Command test manages that if interested.

@Command()

The @Command() decorator is to define what CLI command the class is going to manage and take care of. The decorator takes in an object to define properties of the command. The options passed here would be the same as the options passed to a new command for Commander

property	type	required	description
name	string	true	the name of the command
arguments	string	false	Named arguments for the command to work with. These can be required <> or optional [], but do not map to an option like a flag does
description	string	false	the description of the command. This will be used by the --help or -h flags to have a formalized way of what to print out
argsDescription	Record<string, string>	false	An object containing the description of each argument. This will be used by -h or --help
Options	CommandOptions	false	Extra options to pass on down to commander

For mor information on the @Command() and @Option() parameters, check out the Commander docs.

@Option()

Often times you're not just running a single command with a single input, but rather you're running a command with multiple options and flags. Think of something like git commit: you can pass a --amend flag, a -m flag, or even -a, all of these change how the command runs. These flags are able to be set for each command using the @Option() decorator on a method for how that flag should be parsed. Do note that every command sent in via the command line is a raw string, so if you need to transform that string to a number or a boolean, or any other type, this handler is where it can be done. See the putting it all together for an example. The @Option() decorator, like the @Command() one, takes in an object of options defined in the table below

property	type	required	description
flags	string	true	a string that represents the option's incoming flag and if the option is required (using <>) or optional (using [])
description	string	false	the description of the option, used if adding a --help flag
defaultValue	string or boolean	false	the default value for the flag

Under the hood, the method that the@Option() is decorating is the custom parser passed to commander for how the value should be parsed. This means if you want to parse a boolean value, the best way to do so would be to use JSON.parse(val) as Boolean('false') actually returns true instead of the expected false.

Running the Command

Similar to how in a NestJS application we can use the NestFactory to create a server for us, and run it using listen, the nest-commander package exposes a simple to use API to run your server. Import the CommandFactory and use the static method run and pass in the root module of your application. This would probably look like below

1import { CommandFactory } from 'nest-commander';
2import { AppModule } from './app.module';
3
4async function bootstrap() {
5  await CommandFactory.run(AppModule);
6}
7
8bootstrap();

By default, Nest's logger is disabled when using the CommandFactory. It's possible to provide it though, as the second argument to the run function. You can either provide a custom NestJS logger, or an array of log levels you want to keep - it might be useful to at least provide ['error'] here, if you only want to print out Nest's error logs.

1import { CommandFactory } from 'nest-commander';
2import { AppModule } from './app.module';
3import { LogService } './log.service';
4
5async function bootstrap() {
6  await CommandFactory.run(AppModule, new LogService());
7
8  // or, if you only want to print Nest's warnings and errors
9  await CommandFactory.run(AppModule, ['warn', 'error']);
10}
11
12bootstrap();

And that's it. Under the hood, CommandFactory will worry about calling NestFactory for you and calling app.close() when necessary, so you shouldn't need to worry about memory leaks there. If you need to add in some error handling, there's always try/catch wrapping the run command, or you can chain on some .catch() method to the bootstrap() call.

Testing

So what's the use of writing a super awesome command line script if you can't test it super easily, right? Fortunately, nest-commander has some utilities you can make use of that fits in perfectly with the NestJS ecosystem, it'll feel right at home to any Nestlings out there. Instead of using the CommandFactory for building the command in test mode, you can use CommandTestFactory and pass in your metadata, very similarly to how Test.createTestingModule from @nestjs/testing works. In fact, it uses this package under the hood. You're also still able to chain on the overrideProvider methods before calling compile() so you can swap out DI pieces right in the test. A nice example of this can be seen in the basic.command.factory.spec.ts file.

Putting it All Together

The following class would equate to having a CLI command that can take in the subcommand basic or be called directly, with -n, -s, and -b (along with their long flags) all being supported and with custom parsers for each option. The --help flag is also supported, as is customary with commander.

1import { Command, CommandRunner, Option } from 'nest-commander';
2import { LogService } from './log.service';
3
4interface BasicCommandOptions {
5  string?: string;
6  boolean?: boolean;
7  number?: number;
8}
9
10@Command({ name: 'basic', description: 'A parameter parse' })
11export class BasicCommand extends CommandRunner {
12  constructor(private readonly logService: LogService) {
13    super();
14  }
15
16  async run(passedParam: string[], options?: BasicCommandOptions): Promise<void> {
17    if (options?.boolean !== undefined && options?.boolean !== null) {
18      this.runWithBoolean(passedParam, options.boolean);
19    } else if (options?.number) {
20      this.runWithNumber(passedParam, options.number);
21    } else if (options?.string) {
22      this.runWithString(passedParam, options.string);
23    } else {
24      this.runWithNone(passedParam);
25    }
26  }
27
28  @Option({
29    flags: '-n, --number [number]',
30    description: 'A basic number parser'
31  })
32  parseNumber(val: string): number {
33    return Number(val);
34  }
35
36  @Option({
37    flags: '-s, --string [string]',
38    description: 'A string return'
39  })
40  parseString(val: string): string {
41    return val;
42  }
43
44  @Option({
45    flags: '-b, --boolean [boolean]',
46    description: 'A boolean parser'
47  })
48  parseBoolean(val: string): boolean {
49    return JSON.parse(val);
50  }
51
52  runWithString(param: string[], option: string): void {
53    this.logService.log({ param, string: option });
54  }
55
56  runWithNumber(param: string[], option: number): void {
57    this.logService.log({ param, number: option });
58  }
59
60  runWithBoolean(param: string[], option: boolean): void {
61    this.logService.log({ param, boolean: option });
62  }
63
64  runWithNone(param: string[]): void {
65    this.logService.log({ param });
66  }
67}

Make sure the command class is added to a module

1@Module({
2  providers: [LogService, BasicCommand]
3})
4export class AppModule {}

And now to be able to run the CLI in your main.ts you can do the following

1async function bootstrap() {
2  await CommandFactory.run(AppModule);
3}
4
5bootstrap();

And just like that, you've got a command line application.

License

This project is licensed under the MIT License - see the LICENSE file for details.