@o3/typed v0.1.3

Typed

Simplistic and minimal runtime type checking system for Node.js applications.

Typed is a minimal wrapper based type checking library for JavaScript to ensure user/developer facing functions are passed (and also return) their expected types. This is done in an attempt to reduce the amount human error in runtime environments.

Installing

via npm

npm i -S @o3/typed
# For fully featured runtime checking also install
npm i -S @o3/prop-types

How to use

Shared syntax pattern

You can define a subset of functions to accept ambiguous arguments like so:

// define your base library and/or utility functions
const add = (x, y) => x + y
const addOptional = (x, y = 3) => x + y;

And then wrap them using the typed functionality.

Allowing you to restrict your public functions to only accept specific input types and throw type errors when improper values are provided

const typed = require('@o3/typed/full');

module.exports = {
    // first two arguments must be string
    concat: typed(add, typed.string, typed.string)
    // first two arguments must be number
    addNumbers: typed(add, typed.number, typed.number),
    // first argument must be a number, while second can be ommited
    // but if passed must be a number
    addThree: addOptional(add, typed.number, typed.number.isOptional),
};

NOTE: typed/full uses a dependency called @o3/prop-types, which needs to be installed as a peer dependency. A more minimal version is exported at root level, which does not require this dependency, and instead opts to use a minimized lodash type build.

Using a alternative syntax, it is also possible to specify an expected resolving (return) type for a given function. For more complex functions this will ensure that input error won't return any unwanted inputs.

function complexAdd(x, y) {
    if (x === 0) {
    // zero sucks so let's do something awful
        // mostly for demonstrative purposes, I don't recommended writing
        // any production code like this ;)
        return null
    }

    // return forced integer add (base10)
    return parseInt(x, 10) + parseInt(y, 10);
}

// create (string or number) type to use in our definition
const stringOrNumber = typed.oneOfType([
    typed.string,
    typed.number,
]);

const addOnlyReturnNumber = typed(complexAdd, [
    // first and second param can be either a number or string
    stringOrNumber, stringOrNumber,
    // but must return a number
], typed.number);

addOnlyReturnNumber(2, 2) // returns expected type number (4)
addOnlyReturnNumber('2', 2) // returns expected type number (4)
addOnlyReturnNumber(0, 2) // throws due to invalid return value 'null'

Full format (recommended)

All the above examples are written in the full syntax, which requires the additional single dependency of @o3/prop-types, an adaptation of the Facebook React prop-types library. A more detailed write up of accepted syntax will come eventually, but I highly recommend you stick with their official documentation here.

The modifications made remove unneeded support for React components, add enforced isRequired on all known type definitions. To opt out of this, passing types.typeName.isOptional will allow voided values to be passed.

Basic format (minimal, no dependencies required)

While I'm making progress to reduce the amount of redundant abstractions within the @o3/prop-types library, by default this library exports a minimal version of the shown syntax, which relies on a minimized and tree-shaken port of all of lodash's known is{Type} exports.

The benefits of using this syntax are more strict type checking, since needing to rely on primitive checking that can sometimes be prone to strange quirks. The downside is that optional arguments are not really well supported due to it's highly strict nature.

Here is an example of the alternative basic format:

Example

const typed = require('@o3/typed');

const add = (x, y) => x + y;

// ensure first two paramaters meet expectation isNumber
// shorthand syntax
const addNumbers = typed(add, 'isNumber', 'isNumber');

// ensure that the parameter passed meets either expectation
const numberOrString = {
    anyOf: ['isNumber', 'isString'],
};

// longhand syntax
const addOnlyReturnNumber = typed(add, [numberOrString, numberOrString], 'isNumber');

Why make this

While searching the Internet for a solution to a problem I needed solving, I kept finding dynamic typing systems like TypeScript and Flow.

While these tools are great for ensuring that your internal code isn't calling any other code incorrectly, they don't seem to be orientated towards ensuring type compliance at runtime, which is an issue when allowing external developers to interface with your code.

I was also able to find several Babel runtime plug-ins that, to be frank, did not even meet the most minimal of my own expectations. I was underwhelmed by the quality of the runtime type-checking, since it seemed that the plugins would haphazardly enforce shallow type checking on all methods that included definitions, which in the case of Flow and TypeScript would be most methods within the codebase.

Performance wise this is not good as it adds at least several additional calls to every function invocation, which compounds quickly). Additionally, it also adds unneeded points of failure to each function call that was injected.

The solution employed by this library is intended to be applied to developer and user facing methods and calls, since those functions are most prone to input error, and should be the only functions (in a language like JavaScript) that realistically need runtime type checking.