oclif-dynamic-commands

oclif-dynamic-commands

oclif plugin that loads commands dynamically on run.

The current folder's package.json is read to find the folders the dynamic commands can be found in.

Installation

1. Install the library:

npm i oclif-dynamic-commands

1. Add it to your oclif plugins in pacakge.json:

  "oclif": {
    "plugins": [
      "oclif-dynamic-commands"
      ...
    ],
    ...
  }

Writing Commands

Commands should be standard oclif commands, with one exception - command IDs are not defined by folder location, but manually and statically defined in the class itself:

import { Command } from '@oclif/core';

export default class Example extends Command {
  static id = 'example';
  static summary = 'An example command.';
  static description = 'A longer description of your example command.';
  static examples = [
    '$ your-cli-command example',
  ];

  async run(): Promise<void> {
    this.log('Example command has been run!');
  }
}

If your command is a sub-command (it contains the topic separator in it, e.g. drive:inacar), you must ensure the "root command" (drive) is defined somewhere in your built-in or dynamic commands, or you will not be able to see it in the help menu or list of available commands. Users will still be able to run the command, but its discoverability will suffer.

Configuration

In the folder that is to container your dynamic commands, ensure the package.json file has the following setting:

  "oclif": {
    "dynamic-commands": {
      "folders": [
        "./lib/commands/**/*.ts",
        "./src/commands/**/*.ts"
      ]
    }
  }

Folders will be explored in the order defined, and any new commands loaded will overwrite any existing commands with the same ID. In this way, you can allow for dynamic commands to override built-in commands, or allow for commands to be loaded from a common library, but still have project-specific commands overwrite them.

Entries in the folders array can use glob patterns from glob.

Caveats & Cautions

Typescript may be necessary!

If you are allowing users to write dynamic commands in Typescript, you will need to ensure your CLI is being run through Typescript, not running a pre-compiled Javascript binary! Otherwise, you will get syntax errors when Node attempts to parse a Typescript file.

One way to (relatively) easily do that is to add dev.sh and run.sh shell files to your binary folder that look like this:

#!/usr/bin/env bash

DIR="$(dirname "$(readlink -f "$0")")"

"$DIR/../node_modules/tsx/dist/cli.js" --tsconfig ./tsconfig.json -- $DIR/dev.js "$@"

This requires installing tsx in the dependencies of your CLI, which is a very quick Typescript interpreter, though it has a few limitations.

This script will run the dev.js Javascript file in your CLI's bin file, ensuring the tsconfig.json file from the current working directory is loaded. Which leads us to...

TsConfig differences between your CLI and your dynamic commands

With the above shell file, the tsconfig.json of the current working directory is used to run your CLI. This is not strictly necessary, but it can reduce issues with customised tsconfig values needed by your dynamic commands. Specifically, if the project that your dynamic commands live within requires a special paths setting, you must use the tsconfig.json from the project's folder to ensure those paths are loaded and set correctly. Otherwise, any includes from your dynamic commands that rely on those paths being set will fail.

This raises an interesting question: How do you have combine a paths setting from your project's tsconfig.json with one from your CLI? The short answer is that you cannot. So if you expect your dynamic commands to rely on custom paths, you cannot rely on them in your CLI.

You likely also want to both a) try to rely on as few custom TsConfig settings as you can and b) inform any developers of dynamic commands not to significantly modify their TsConfig settings or they could encounter unexpected failures when running the CLI.