bandersnatch
Super lightweight and friendly CLI scaffolding for Node.js programs.
Features
- 🌊 Fluid syntax
- ➰ Built-in REPL
- 💬 Prompts for missing arguments
- 🔜 Autocompletes arguments
- 🤯 Fully typed
- ⚡ Uses the power of
yargs
and inquirer
It's built in TypeScript and command arguments are fully typed.
Bandersnatch is not designed to be used as a full CLI framework like oclif, and
tries to minimize the assumptions made about your program to make it easy and
intuitive to work with.
Table of contents
Getting started
Installation
yarn|npm add bandersnatch
Simple example
Let's create a simple program foo.ts
:
import { program, command } from 'bandersnatch'
const foo = command('foo')
.description('Outputs "bar".')
.action(() => console.log('bar'))
program().default(foo).run()
This creates a new program, adds a default command which logs "bar" to the
stdout, and runs the program.
Now try your program by running it:
$ ts-node foo.ts
bar
ℹ Assuming you have ts-node
installed.
Try running ts-node foo.ts help
to see the auto-generated help output:
$ ts-node foo.ts help
bin.js
Outputs "bar".
Commands:
bin.js Outputs "bar". [default]
Options:
--help Show help [boolean]
--version Show version number [boolean]
ℹ You see bin.js
here instead of foo.ts
because we're running the program
with ts-node
.
REPL example
A program can also show an interactive
REPL to make interacting
with more complex programs easier and to enable autocompleting of commands and
arguments.
Let's create a new program dice.ts
with a command to roll a dice:
import { program, command } from 'bandersnatch'
async function rng(bounds: [number, number]) {
const [min, max] = bounds
return Math.floor(Math.random() * (max - min + 1)) + min
}
const dice = program().add(
command('roll')
.option('min', { default: 1 })
.option('max', { default: 6 })
.action(async (args) => {
console.log(await rng([args.min, args.max]))
})
)
dice.repl()
This code defines a program dice
and a command roll
with two options, both
of which will inherit a default value. When the command is executed, it calls
an async random number generator (async only for illustrative purposes) and
writes its results to stdout.
The last line in our code runs the program as a interactive REPL, which means
it won't accept any arguments on the command line, but render a prompt instead.
This prompt will read any user input, parse it, and execute matching commands.
Try rolling the dice:
$ ts-node dice.ts
> roll
5
The REPL can autocomplete commands, arguments and options. Try typing only the
letter r
and then hit TAB. This works for options as well:
$ ts-node dice.ts
> r
[TAB]
> roll -
[TAB]
> roll --m
[TAB] [TAB]
--min --max
Prompt
Bandersnatch can also ask a user for input if arguments were not provided on the
command line:
Let's say we want to write a program pizza.ts
which takes pizza orders:
import { program, command } from 'bandersnatch'
const cmd = command()
.argument('address', {
prompt: 'Your address',
})
.option('name', {
description: 'Your name',
default: 'anonymous',
prompt: true,
})
.option('size', {
description: 'Choose pizza size',
choices: ['small', 'medium', 'large'],
default: 'medium',
prompt: true,
})
.option('toppings', {
description: 'Pick some toppings',
choices: ['mozzarella', 'pepperoni', 'veggies'],
default: ['mozzarella'],
prompt: true,
})
.option('confirmed', {
description: 'Order pizza?',
default: true,
prompt: true,
})
.action((args) => {
console.log(args)
})
program().description('Order a pizza').default(cmd).run()
And run it:
$ ts-node pizza.ts
? Your address The Netherlands
? Your name Joram
? Choose pizza size small
? Pick some toppings veggies
? Order pizza? Yes
{
name: 'Joram',
size: 'small',
toppings: [ 'veggies' ],
confirmed: true,
address: 'The Netherlands'
}
You can choose to specify parameters on the command line, in which case you
won't get a prompt for these options:
$ ts-node pizza.ts "The Netherlands" --name Joram --confirmed
? Choose pizza size small
? Pick some toppings veggies
? Order pizza? Yes
{
name: 'Joram',
size: 'small',
toppings: [ 'veggies' ],
confirmed: true,
address: 'The Netherlands'
}
⚠ Please note that even though --confirmed
was specified on the command line,
it was still being prompted. This is a known issue. In this case, the default
value was the same as the input, in which case bandersnatch doesn't know whether
a value was explicitly passed in or inherited from the default value.
TypeScript
Bandersnatch works perfectly well with non-TypeScript codebases. However, when
you do use TypeScript the command arguments are fully typed.
Let's slightly improve the example program above to illustrate this:
.option('size', {
description: 'Choose pizza size',
- choices: ['small', 'medium', 'large'],
+ choices: ['small', 'medium', 'large'] as const,
default: 'medium',
prompt: true,
})
.option('toppings', {
description: 'Pick some toppings',
- choices: ['mozzarella', 'pepperoni', 'veggies'],
+ choices: ['mozzarella', 'pepperoni', 'veggies'] as const,
default: ['mozzarella'],
prompt: true,
})
The first argument passed to the action handler function is now typed like this:
type Args = {
address: string
name: string
size: 'small' | 'medium' | 'large'
toppings: ('mozzarella' | 'pepperoni' | 'veggies')[]
confirmed: boolean
}
ℹ More examples in the
examples
directory.
API
All methods are chainable unless the docs mention otherwise.
program(options)
Creates a new program. Options (object, optional) can contain these keys:
description
(string, optional) is used in --help output.prompt
(string, default: >
) use this prompt prefix when in REPL mode.help
(boolean, default: true) adds help
and --help
to the program which
displays program usage information.version
(boolean, default: true) adds version
and --version
to the
program which displays program version from package.json.
program.description(description)
Sets the program description (string, required) used in --help output.
program.prompt(prompt)
Use this prompt prefix (string, required) when in REPL mode.
program.add(command)
Adds a command to the program.
program().add(command(...))
program.default(command)
Adds a default command to the program. Shorthand for:
program().add(command(...).default())
program.run(command)
Uses process.argv or passed in command (string, optional) to match and execute
command. Returns promise.
program()
.add(command(...))
.run()
program.repl()
Start a read-eval-print loop. Returns promise-like repl instance.
program()
.add(command(...))
.repl()
program.runOrRepl()
Invokes run()
if process.argv is set, repl()
otherwise. Returns promise or
promise-like repl instance.
program()
.add(command(...))
.runOrRepl()
program.isRepl()
Returns true
if program is running a repl loop, false
otherwise.
program.on(event, listener)
Attaches a listener function for the event. Currently, these events are
supported:
program().on('run', (cmd) => logger.debug(`Running ${cmd}`))
command(name, options)
Creates a new command.
- Name (string, optional) is used to invoke a command. When not used as the
default command, a name is required.
- Options (object, optional) can contain these keys:
description
(string, optional) is used in --help output.
command.argument(name, options)
Adds a positional argument to the command.
- Name (string, required) is used to identify the argument.
- Options can be provided to change the behavior of the argument. Object with
any of these keys:
description
(string) is used in --help output.optional
(boolean) makes this argument optional.variadic
(boolean) eagerly take all remaining arguments and parse as an
array. Only valid for the last argument.type
(string) one of "boolean"|"number"|"string"
which determines the
runtime type of the argument.default
(any) default value for the argument.choices
(array) any input value should be included in the array, or it
will be rejected.prompt
(boolean|string) prompts for missing arguments. If it is true, it
will use the arguments description or name as the question text. If it is a
string, it will be used as the question text.alias
(string|array) alias or aliases for the argument.coerce
(function) transform function for this argument value (untyped).
command.option(name, options)
Adds an option to the command.
- Name (string, required) is used to identify the option.
- Options (object, optional) can be provided to change the behavior of the
option. Object with any of these keys:
description
(string, optional) is used in --help output.type
(string) one of "array"|"boolean"|"count"|"number"|"string"
which
determines the runtime type of the argument. Use count for the number of
times an option was provided (e.g. verbosity levels).default
(any) default value for the argument.choices
(array) any input value should be included in the array, or it
will be rejected.prompt
(boolean|string) prompts for missing arguments. If it is true, it
will use the arguments description or name as the question text. If it is a
string, it will be used as the question text.alias
(string|array) alias or aliases for the option.coerce
(function) transform function for this option value (untyped).
command.command(command)
Adds a sub-command to the command.
command.default()
Mark command as default. Default commands are executed immediately and don't
require a name.
command.action(function)
Function which executes when the command is invoked. Is called with these
arguments:
- Args (object) is an object containing key/value pairs of parsed arguments and
options.
- Command runner (function) can be invoked with one (string) parameter to
execute another command.
Design principles
In general, bandersnatch is designed to create
twelve-factor apps.
Errors
The bandersnatch API allows to catch errors in a promise-like way. The run
and
repl
program methods return either a promise or promise-like object which can
be used to handle program errors:
program()
.default(
command()
.description('This command will always fail')
.action(function () {
throw new Error('Whoops')
})
)
.runOrRepl()
.catch((error: any) => {
console.error('[failed]', String(error))
if (!app.isRepl()) {
process.exit(1)
}
})
Output
Programs are encouraged to use the following conventions with regards to output,
based on the
POSIX standard.
- When a program is designed to be used in a scripting environment and its
output should be available as stdin for other programs, use stdout for
printing output and stderr for diagnostic output (e.g. progress and/or error
messages).
- When a program is designed to be used as a service (twelve-factor app), use
stdout/stderr as a logging mechanism for informative messages/error and
diagnostic messages.
Bandersnatch has no built-in method for writing to stdout/stderr. Node.js
provides everything you need.
Bundle
There are many options to bundle your application for distribution. We'll
discuss a common pattern.
ℹ An example can be found in the
examples/bundle
directory.
Init a package.json
if needed:
mkdir echo && cd echo
yarn init
Install dependencies:
yarn add bandersnatch
yarn add typescript pkg --dev
And create an example app in src/cli.ts
:
import { program, command } from 'bandersnatch'
export default program()
.withHelp()
.default(
command('echo', 'Echo something in the terminal')
.argument('words', 'Say some kind words', { variadic: true })
.action(console.log)
)
Building your app with TypeScript is very powerful, but runtime compilation is
slow so we compile the code ahead of time.
Add a tsconfig.json
, similar to:
{
"include": ["./src"],
"compilerOptions": {
"target": "es2017",
"module": "commonjs",
"lib": ["es2017"],
"declaration": true,
"outDir": "lib",
"rootDir": "src",
"strict": true,
"allowSyntheticDefaultImports": true,
"esModuleInterop": true,
"moduleResolution": "node"
}
}
Add these scripts to your package.json
:
{
"name": "echo",
"version": "1.0.0",
"main": "index.js",
"license": "MIT",
+ "scripts": {
+ "prepublishOnly": "yarn build",
+ "build": "tsc",
+ },
"dependencies": {
"bandersnatch": "^1.0.0"
},
"devDependencies": {
"pkg": "^4.4.2",
"typescript": "^3.7.3"
}
}
And compile now by running yarn build
.
Next, we need to create a simple entry point echo.js
, which can be run with
node:
#!/usr/bin/env node
require('./lib/cli').default.run()
To run your app, users may want to run yarn global add echo
. For this to
work, we need to make a small adjustment to package.json
:
{
"name": "echo",
"version": "1.0.0",
- "main": "index.js",
+ "bin": "echo.js",
+ "files": [
+ "lib"
+ ],
"license": "MIT",
"scripts": {
"prepublishOnly": "yarn build",
"build": "tsc",
},
"dependencies": {
"bandersnatch": "^1.0.0"
},
"devDependencies": {
"pkg": "^4.4.2",
"typescript": "^3.7.3"
}
}
You can now npm publish
.
To create a binary (your app with Node.js bundled), add this script to
package.json
:
{
"name": "echo",
"version": "1.0.0",
"bin": "echo.js",
"files": [
"lib"
],
"license": "MIT",
"scripts": {
"prepublishOnly": "yarn build",
"build": "tsc",
+ "bundle": "yarn build && pkg -t host ."
},
"dependencies": {
"bandersnatch": "^1.0.0"
},
"devDependencies": {
"pkg": "^4.4.2",
"typescript": "^3.7.3"
}
}
ℹ Omit -t host
to create binaries for all platforms.
Run yarn bundle
and then ./echo --help
. 💪
Optionally deploy to GitHub, S3, etc. using your preferred CD method if needed.
Todo
Contributing
Contributions are very welcome.
git clone git@github.com:hongaar/bandersnatch.git
cd bandersnatch
yarn
yarn start examples/foo.ts
Please use
conventional commits.
License
Copyright (c) 2020 Joram van den Boezem. Licensed under the MIT license.
Inspired by vorpal