NestJS Commander

Have you been building amazing REST and RPC 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!

npm i nest-commander @nestjs/common @nestjs/core
# OR
yarn add nest-commander @nestjs/common @nestjs/core
# OR
pnpm 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 interface 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 interface CommandRunner, which should be implemented by each command. The CommandRunner interface 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.

Inquirer Integration

nest-commander also can integrate with inquirer to allow for user input during your CLI run. I tried to keep this integration as smooth as possible, but there are some caveats to watch for:

1. Whatever inputs you want to handle via inquirer, must be omitted from commander if you don't want them passed in at all, or they must be optional if you want them to be passable from the command line. If you use a required option and it is not passed from the command line, commander will fail the CLI call.
2. Inquirer plugins are not yet supported. I do have an idea for this, but they are out of scope for the initial integration.
3. You, as the developer, have options on how to set up the inquirer integration. The details will come later, but know that you are given power here. Use it wisely.

QuestionSet

A class decorated with @QuestionSet() is a class that represents a related set of questions. Looking at inquirer's own examples, this could be like the pizza example. There's nothing too special about this decorator, all it does is allow the underlying engine to find the appropriate question set when it is needed. The @QuestionSet() decorator takes an object of options defined below

| property | type | required | description | | -------- | ------ | -------- | --------------------------------------------------------------------------------- | | name | string | true | The name that will be used by the InquirerService when getting a prompt to run. |

Question

Here's where the options start to open up. Each @Question() should decorate a class method. This method will essentially become the filter property for inquirer. If you don't need any filtering done, simply return the value that comes into the method. All of the other properties come from, and adhere to the types of, Inquirer and their documentation can better illustrate what values are needed when and where.

Question Functional Properties

With Inquirer, several of the properties can have functions instead of simple types. For these properties, you can do one of two things: 1) pass the function to the decorator or 2) use the @*For()^ decorator. Each @*For() decorator takes in an object similar to the @Question() decorator as described below

| property | type | required | description | | -------- | ------ | -------- | -------------------------------------------------------------------------------------- | | name | string | true | The name that will be used to determine which @Question() this decorator belongs to. |

Passing to the @Question() decorator

Below is an example of using the validate method in the @Question() decorator

@Question({
  type: 'input',
  name: 'phone',
  message: "What's your phone number?",
  validate: function(value: string) {
    const pass = value.match(
      /^([01]{1})?[-.\s]?\(?(\d{3})\)?[-.\s]?(\d{3})[-.\s]?(\d{4})\s?((?:#|ext\.?\s?|x\.?\s?){1}(?:\d+)?)?$/i,
    );
    if (pass) {
      return true;
    }
    return 'Please enter a valid phone number';
  }
})
parsePhone(val: string) {
  return val;
}

Using the @*For() decorator

Below is an example of a @Question() and @ValidateFor() decorator in use

@Question({
  type: 'input',
  name: 'phone',
  message: "What's your phone number?",
})
parsePhone(val: string) {
  return val;
}

@ValidateFor({ name: 'phone' })
validatePhone(value: string) {
  const pass = value.match(
    /^([01]{1})?[-.\s]?\(?(\d{3})\)?[-.\s]?(\d{3})[-.\s]?(\d{4})\s?((?:#|ext\.?\s?|x\.?\s?){1}(?:\d+)?)?$/i,
  );
  if (pass) {
    return true;
  }

  return 'Please enter a valid phone number';
}

As you can see, the name of both @Question() and @ValidateFor() align, allowing the underlying engine to properly map the validatePhone method to the phone's property set.

^ Please note that @*For() is shorthand for @ValidateFor(), @ChoicesFor(), @MessageFor(), @DefaultFor(), and @WhenFor().

InquirerService

The InquirerService is an injectable provider that allows you to call inquirer for a specific set of questions (named with @QuestionSet()). When calling the question set, you can pass in the already obtained options as well, and inquirer will skip over the options that are already answered, unless the askAnswered property is set to true as mentioned in their docs. You can use either InquirerService#ask or InquirerService#prompt, as they are aliases for each other. The return from the InquirerService#prompt method is the non-partial variant of the options passed in; in other words, the return is the answers that the user provided, mapping appropriately in the cases where necessary, such as lists. For an example usage, please check the pizza integration test.

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

import { CommandFactory } from 'nest-commander';
import { AppModule } from './app.module';

async function bootstrap() {
  await CommandFactory.run(AppModule);
}

bootstrap();

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.

Error Handling

By default, nest-commander does not add in any error handling, other that the default that commander itself does. If you would like to use commander's exitOverride you can pass an errorHandler property to the options object of the CommandFactory.run method. This error handler should take in an error object, and return void.

import { CommandFactory } from 'nest-commander';
import { AppModule } from './app.module';

async function bootstrap() {
  await CommandFactory.run(AppModule, {
    errorHandler: (err) => {
      console.error(err);
      process.exit(1); // this could also be a 0 depending on how you want to handle the exit code
    }
  });
}

bootstrap();

Testing

There is a testing helper package called nest-commander-testing that works very similarly to @nestjs/testing. Check out it's documentation and examples for help.

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.

import { Command, CommandRunner, Option } from 'nest-commander';
import { LogService } from './log.service';

interface BasicCommandOptions {
  string?: string;
  boolean?: boolean;
  number?: number;
}

@Command({ name: 'basic', description: 'A parameter parse' })
export class BasicCommand implements CommandRunner {
  constructor(private readonly logService: LogService) {}

  async run(passedParam: string[], options?: BasicCommandOptions): Promise<void> {
    if (options?.boolean !== undefined && options?.boolean !== null) {
      this.runWithBoolean(passedParam, options.boolean);
    } else if (options?.number) {
      this.runWithNumber(passedParam, options.number);
    } else if (options?.string) {
      this.runWithString(passedParam, options.string);
    } else {
      this.runWithNone(passedParam);
    }
  }

  @Option({
    flags: '-n, --number [number]',
    description: 'A basic number parser'
  })
  parseNumber(val: string): number {
    return Number(val);
  }

  @Option({
    flags: '-s, --string [string]',
    description: 'A string return'
  })
  parseString(val: string): string {
    return val;
  }

  @Option({
    flags: '-b, --boolean [boolean]',
    description: 'A boolean parser'
  })
  parseBoolean(val: string): boolean {
    return JSON.parse(val);
  }

  runWithString(param: string[], option: string): void {
    this.logService.log({ param, string: option });
  }

  runWithNumber(param: string[], option: number): void {
    this.logService.log({ param, number: option });
  }

  runWithBoolean(param: string[], option: boolean): void {
    this.logService.log({ param, boolean: option });
  }

  runWithNone(param: string[]): void {
    this.logService.log({ param });
  }
}

Make sure the command class is added to a module

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

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

async function bootstrap() {
  await CommandFactory.run(AppModule);
}

bootstrap();

And just like that, you've got a command line application. All that's left is to run your build command (usually nest build) and run start like normal (node dist/main). If you're looking to package the command line app for other devs consumption (making somethng like the @nestjs/cli or jest), then you can add the bin property to the package.json and map the command appropriately.