CMDed

CMDed is a command line argument parser made for humans.

CMDed makes it easy to build CLI tools, or any command that needs arguments. Unlike other command argument parsers, CMDed aims at keeping the interface straight forward and simple, while also remaining very simple (and small) in nature.

Install

npm:

$ npm i --save cmded

yarn:

$ yarn add cmded

Documentation

The full documentation can be found here.

How it works

The core concept that makes CMDed works is "consuming" arguments combined with argument "scanning".

Any arguments that are validly parsed against a matching pattern are "consumed". Being consumed, they are then ignored by further processing. This makes parsing complex combinations straight forward and simple.

Take for example the following command:

./my-echo hello world --use-system-echo --duplicate

All arguments are scanned (in the order you specify) and marked as "consumed". So, when specifying the "boolean" arguments --use-system-echo and --duplicate in our code first, the arguments would be scanned, and the "boolean" arguments would be marked as "consumed" (visualized with .....):

./my-echo hello world ................. ...........

Now, inside your argument "context" there are the values { 'useSystemEcho': true, 'duplicate': true }, and then the remaining arguments are up for parsing next. This makes it straight forward for the parser to figure out what is going on, even if the argument order is changed. For example, if we place the "boolean" flags first, nothing changes:

./my-echo --duplicate --use-system-echo hello world

...when looking at the consumed arguments:

./my-echo ........... ................. hello world

In this case, we want all remaining arguments to be echoed to the user. CMDed has a special context property named _remaining which are all remaining "unconsumed" arguments after parsing has completed.

With all of this in mind, we can create the above example using CMDed:

const { CMDed, Types, showHelp } = require('../');
const { spawn } = require('child_process');

// Define our help for this command
const help = {
  '@usage': 'my-echo [options] ...arg1 ...arg2 ...argN',
  '@examples': [
    './my-echo hello world',
    './my-echo --duplicate hello world',
    './my-echo hello world --use-system-echo --duplicate',
  ],
  '--use-system-echo': 'Spawn system "echo", handing off arguments to the child process',
  '--duplicate': 'Duplicate all provided arguments',
};

// Parse our arguments.
// All remaining arguments will be placed into `args._remaining`.
let args = CMDed(({ $, store }) => {
  // Runners should always return `true` if they succeed,
  // so the `|| store(...)` syntax is one way for us to
  // set "default values" for each of our options. If they
  // fail for example (because no such argument was provided),
  // then the `|| store(...)` path will be followed, and the
  // value will be set to `false` (the default).
  $('--use-system-echo', Types.BOOLEAN()) || store({ 'useSystemEcho': false });
  $('--duplicate', Types.BOOLEAN()) || store({ 'duplicate': false });

  // This informs CMDed that everything went okay.
  // If a "Runner" ever returns `false`, then CMDed
  // will know that the Runner has failed... this isn't
  // always a bad thing, but it generally means that it
  // is time to show the help. In this case however,
  // both our arguments are optional, so if they weren't
  // parsed, then that is okay.
  return true;
}, { help });

// If no arguments were provided, then show the help
let remainingArgs = args._remaining;
if (remainingArgs.length === 0) {
  showHelp(help);
  process.exit(1);
}

// Duplicate args if requested
if (args.duplicate)
  remainingArgs = Array.prototype.concat.apply([], remainingArgs.map((arg) => [ arg, arg ]));

// Now let's echo
if (args.useSystemEcho) {
  spawn('echo', remainingArgs, { stdio: 'inherit' });
} else {
  console.log(remainingArgs.join(' '));
}

Examples

Check out the examples directory in the repository to see other examples for CMDed.

Runners

CMDed calls argument "handlers" Runners. There is a few built-in default Runners, such as BOOLEAN, INTEGER, DECIMAL, OCTAL, BYTES, and STRING. These can be accessed from the Types import.

Match and $

The $ dollar sign is an alias for match. match is the core of CMDed. It will match against the argument patterns specified, in combination with the parser, and will "consume" any matches found. It does so by scanning all the arguments. For example, if we specify a '--size' argument, it will scan all provided arguments until it finds this match, and once found, it will mark that argument as "consumed", and call the Runner, passing the parsed results to the Runner as the second argument.

There are a number of different ways match can be used. First, match can simply provided a string for a pattern to match against. If it is provided a string, then the match must be exact.

You can also supply a regular expression (RegExp) to the Runner as the first argument (the "pattern"). If a RegExp is supplied, then the results of the matching regular expression will be passed as the second argument (the parser results) to the Runner. All of the built-in Runners won't know what to do with this "RegExp result", so if you are passing it into a built-in or non-custom Runner, then you will need to format the parser results first. You can use the formatParsedResult option callback to do this. Whatever this method returns will be what is passed directly to the Runner. In order to be compatible with the default and built-in Runner interface, this must be an object, and it must have at least the keys name and value for the argument parsed. For RegExp patterns, only one argument will ever be "consumed" upon match. If you need to match against/parse more than a single argument, then you will need to create a Runner, or use a Function pattern instead.

You can also supply a Function to the Runner as a pattern matcher. It is then up to the this pattern matcher to parse and consume as many arguments as it wants. This pattern matcher must consume all arguments positively matched. The underlying system can not know what was parsed, and therefor can not mark parsed arguments. Don't forget to consume each argument positively matched, or you will have some strange bugs later on down the line. A Function pattern matcher will be considered a successful match if a truthy value is returned. If this is the case, then the Runner will be called with the truthy result (after being passed through formatParsedResult if any was specified). As with all matchers, the minimum requirement to work with the default and built-in Runners is to provide an object with at least name and value properties.

Runner parameters

All Runners by default need no parameters. You can however optionally supply parameters to Runners. Here are the parameters that can be passed to all built-in Runners.

Note: Custom Runners might have different parameters that can be supplied.

{
  formatParsedResult: (value: any, context: RunnerContext, options: object) => {
    // Format the result returned by the parser, the RegExp pattern matcher, or the function pattern matcher. "options" are the options supplied to the Runner.
  },
  validate: (value: any, context: RunnerContext): boolean => {
    // validate the parsed "value"
  },
  format: (value: any): any => {
    // Format the value after parsing and validation
  },
  solo: true, // Solo forces the parser to only parse a single argument
}

For example, any Runner can have a validate method supplied to it, like so:

CMDed(({ $ }) => {
  $('--size', Types.INTEGER({
    validate: (value, { showHelp, exit }) => {
      if (value < 0) {
        console.error('"--size" must be positive');
        showHelp();
        exit(1);
      }

      return true;
    },
  }));
});

As you can see, the validate method will allow you to validate a parsed argument, and if a bad value was provided, do something about it. The exit method will simply call process.exit with the provided status code. If showHelp is called from the RunnerContext provided, then it need not be provided the help object.

Or, maybe you want to use your own custom RegExp pattern matcher, and format the matching results with formatParsedResult:

CMDed(({ $ }) => {
  $(
    (/(\d+),?/g), // Parse an array of numbers
    ({ store }, { name, value }) => {
      // Did we get a valid array?
      if (!Array.isArray(value) || value.length === 0)
        return false;

      // Are all numbers parsed finite values?
      let allFinite = value.some((number) => !isFinite(number));
      if (!allFinite)
        return false;

      // Our Runner will store the final result
      store({ [name]: value });

      // Success
      return true;
    },
    {
      // Format our parsed results
      formatParsedResult: (value, _context, options) => {
        return {
          name: options.name || 'numbers',
          value: value.map((numberStr) => parseInt(numberStr, 10)),
        };
      },
    }
  );
});

All Runners by default (except for BOOLEAN) will parse two arguments if they need to. The default parser will parse arguments with the following patterns name=value or name value. Arguments can include prefixes (but they don't need to). A prefix is any non-word character that comes first. For example, --name would have the prefix --.

To turn off this default behavior of parsing two different formats, and up to two arguments, you can pass the { solo: true } parameter to a Runner. Doing so will force the parser to only parse and consume a single argument. For example, if you wanted to parse a single argument that you knew was an integer, then you could Types.INTEGER({ solo: true }), and this would parse an integer argument like 10, or 10e4, or -10. It would also parse something like name=10, or name=10e4, or name=-10. It wouldn't however parse something like name 10. This would need to parse and consume two arguments, which is disallowed when in solo mode.

There are many instances when solo mode is useful. For example, if you wanted to parse an array of arguments, you could use a built-in parser such as Types.INTEGER, and simply match against it as many times as needed while in solo mode.

RunnerContext

Each Runner is passed a RunnerContext, and many other methods are also provide this context. This context gives a handful of useful methods you can use to interact with the current argument context, for example store, and fetch to update the user context, exit to exit the program immediately, hasMatch to see if anything matched in the current Runner, the rootOptions provided to the command parser, and args if you need to access the arguments directly.

store and fetch

store and fetch allow you to interact with the user context. The "user context" is where the parsed argument values are being stored. Both methods support deep paths. For example you can fetch a value like fetch('some.deep.property'), or fetch({ 'some.deep.property': defaultValue }). You can set deep values like store({ some: { deep: { value: value } } }), or store('some.deep.property', value).

When fetch is in object form, then each property value is the "default value" if the context key being requested is undefined. For example fetch({ hello: 'world' }) would return { hello: internalValue || 'world' }, where 'world' would only be the value if internalValue is undefined. This is useful to get a few context variables at the same time via destructuring, while still providing default values if one or more is not present.

exit

Calling exit will call process.exit and terminate the program immediately. You can supply a process exit code (status code) as the first argument to exit.

hasMatch

Calling hasMatch will return true if any of the matchers inside the current Runner had a successful match, or false otherwise. This is useful to call as the return value of your Runner.

rootOptions

rootOptions are the root options provided to CMDed when it was called.

args

All provided command line arguments via the Arguments interface. The Arguments interface tracks which arguments have been consumed.

context

Get the raw "user context".

scope

You can at any time spawn a new sub scope in the "user context" by calling scope. It will create a sub scope within the user context, under the name you provide, and then all Runners under this scope will place their parsed values into this sub scope. For example:

const { CMDed, Types } = require('cmded');

let userContext = CMDed(({ $, scope }) => {
  return $('sub-command', ({ scope }) => {
    return scope('subCommand', ({ $ }) => {
      return $('--enabled', Types.BOOLEAN());
    });
  });
});

console.log(userContext);
// output: { subCommand: { enabled: true } }

parse

parse can be used at any time to manually invoke the configured parser.

$ and match

Match against a specified argument pattern, and place results into the "user context".

formatName

Call the configured property name formatter manually on a specified value.

markConsumed

Mark a specified array of argument indexes as consumed. An argument can be manually marked as "consumed" or "unconsumed" through the args interface.

showHelp

Output to stdout the help defined for the current command, or current Runner. By default, if this is called inside a Runner, it will try to find the "sub section" of the defined help for the Runner it was called from. If not matching "sub section" is found, then it will simply show the full help. The output of any showHelp call can be overloaded by providing your own showHelp method as a rootOption to your CMDed call.

Help

Help has a nice output formatter built-in for you, but must be manually defined. In the design process for CMDed, it was decided to sacrifice "auto help" or "easy help" generation for interface simplicity. However, this doesn't mean that it is difficult to specify help for your command. Simply use the help parameter on the CMDed call to provide help for your command. If you format this help object correctly, then CMDed will give you nice help when you ask for it. See the examples directory in the repository for more examples.

@usage

If an @usage key is specified in a scope or sub scope of the help, then it will be use to show the Usage: title of the help output.

@title

If an @title key is specified in a scope or sub scope of the help, then it will be shown directly under the Usage: at the top of the help. This should describe what the command or sub-command does. If not provided for a sub-command, this will just fallback to some sane default, showing the user how to invoke the sub-command.

@examples

An @examples key, with a value that is an array of strings, will show a list of examples at the end of your help output.

@notes

An @notes key, with a value that is an array of strings, will show a list of notes at the end of your help output.

@see

Used for sub commands. This specifies some short example on how to show more help for the sub command. If not provided, this will just fallback to some sane default, letting the user know to invoke --help on the sub command itself for more help.

all other keys

All other keys in the help scope will be listed as arguments to the command. The property value for each key will be used as the description for the command. For example: { '--enable': 'Enable a powerful feature.' }.

You can provide aliases or alternate formats by using the vertical pipe character |. For example: { '--enable=true|enable=false|--enable': 'Enable a powerful feature.' }.

If a help property has an object as a value, then it will be treated as a "sub scope", and will have a short description listed for it. For more information you will need to specifically request so by asking for help with the --help argument on the scope. For example, if your sub scope (or sub command) was named run, and you wanted more help on it, you would execute my-command run --help, or my-command --help run. This would then show the full run sub scope of the help.

The --help argument and helpArgPattern

By default, the internal "help" request for your command will be triggered with a --help argument. You can however change this to whatever you want by specifying the helpArgPattern parameter as a rootOption to your CMDed call. Right now this only supports a single pattern.

Note: if you want aliases for your --help argument, then you can always specify them yourself as Runners that will call showHelp, or you can submit a PR or issue request informing the CMDed team that you would like such a feature.

Async Runners

Asynchronous runners are supported. All you have to do is use the async/await syntax everywhere, and away you go! Nothing changes, you just need to await on all matchers/runners, scope calls, and the call to CMDed itself.

CMDed

CMDed is the main entry point for parsing your arguments. It takes two arguments, an entry point function (this is not a Runner, it is just the entry point to start invoking matchers/Runners), and a rootOptions object, specifying the root options for the process.

The rootOptions has the following shape:

{
  strict?: boolean;
  argv?: Array<string> | null;
  parser?: (context: RunnerContext, options?: object, index?: number) => object | undefined;
  formatter?: (name: string, context?: RunnerContext) => string;
  showHelp?: (subHelp: object | undefined, help: object | undefined, helpPath: string, context: RunnerContext) => void;
  help?: object | null;
  helpArgPattern?: string | null;
}

Parameters

strict = false

The strict parameter, if true, will "panic", and show the help for the command if any arguments are remaining that are unconsumed when the process has completed.

argv = process.argv.slice(2)

Specify the arguments to parse. By default this will be process.argv.slice(2), but can be any array of strings you wish to provide.

parser = defaultParser

Specify a parser for parsing all arguments. By default, this is the defaultParser provided by CMDed, that will parse the following argument patterns: {prefix}{name}={value} (one argument) or {prefix}{name} {value} (two arguments). This parser will not be invoked for RegExp and custom Function pattern matchers.

formatter = defaultFormatter

The default formatter to apply to argument names to convert them into "user context" key names. For example, the argument named --use-system-echo will be converted into camel case useSystemEcho. You can provide any formatter you want to format the name of your arguments and turn them in to user context key names.

showHelp = undefined

A custom --help output function that will write the command's help to stdout. This is a complete replacement of the built-in showHelp method. If you supply this, you must output all the help for your command. However, the help object will be provided to this call, so you will have something to work with to output the help.

context = {}

The default "user context" to supply. This can be any object... but it should be an object.

help = undefined

The help to provide for your command. This should be an object, where each key is an argument pattern, and each value a description of what that argument does.

Note: See the Help section and check out the examples for more information.

helpArgPattern = '--help'

The pattern that will trigger the internal help path for the command. This can be anything you want it to be. By default it is --help.

Types

CMDed comes with the following built-in types:

Types.BOOLEAN (solo = only consumes at most one argument)

A boolean type. This is a solo type by default, meaning it won't ever parse more than a single argument. However, it is valid to use the name=value syntax for your arguments, allowing you to set the boolean to any value. By default, a "no value" argument will be assumed to be true, such as --enabled. You can specifically set the value like so: --enabled=false, or --enabled=0, or --enabled=true, or --enabled=1.

Types.INTEGER (multi = consumes at most two arguments)

An integer type. This will parse a non-decimal, non-real number... an "integer" value. It will fail if there is a decimal place in the number provided. It does however support exponential notation, and can be either negative or positive. By default this is a multi command, so it can parse up to two arguments. It will however only parse a single argument if the "name=value" format is used for the argument. Examples: --size=10, or --size=10e4, or --size=-5, or --size 10, or --size 10e4, or --size -5.

Types.DECIMAL (multi = consumes at most two arguments)

A "floating point" type. This will parse a "real" number... and "decimal" floating point number. It supports exponential notation, and can be either negative or positive. By default this is a multi command, so it can parse up to two arguments. It will however only parse a single argument if the "name=value" format is used for the argument. Examples: --size=10.55, or --size=10.5e4, or --size=-5.123, or --size 10.55, or --size 10.5e4, or --size -5.123.

Types.HEX (multi = consumes at most two arguments)

A hex type. This will parse an integer value in hexadecimal notation. The hex value can be either negative or positive. By default this is a multi command, so it can parse up to two arguments. It will however only parse a single argument if the "name=value" format is used for the argument. Examples: --size=0xF, or --size=-0xAF, or --size 0xF, or --size -0xAF.

Types.OCTAL (multi = consumes at most two arguments)

An octal type. This will parse an integer value in octal notation. The octal value can be either negative or positive. By default this is a multi command, so it can parse up to two arguments. It will however only parse a single argument if the "name=value" format is used for the argument. Examples: --size=0o777, or --size=-66, or --size 777, or --size -0o66.

Types.BYTES (multi = consumes at most two arguments)

A size in bytes. This will parse a size in the number of bytes. This value must be positive, or the match will fail. You can use the b, k, kb, m, mb, g, gb, t, and tb postfixes to specify the absolute size in bytes (case insensitive). Floating point or decimal values can be used. For example, 1.5mb is valid. By default this is a multi command, so it can parse up to two arguments. It will however only parse a single argument if the "name=value" format is used for the argument. Examples: --bytes=10mb, or --bytes=32kb, or --bytes 1gb, or --bytes 1.5tb.

Types.STRING (multi = consumes at most two arguments)

A string type. This will parse any string value. By default this is a multi command, so it can parse up to two arguments. It will however only parse a single argument if the "name=value" format is used for the argument. Examples: --name=Bob, or --name Bob.