parmenides
v0.1.2
Published
[![styled with prettier](https://img.shields.io/badge/styled_with-prettier-ff69b4.svg)](https://github.com/prettier/prettier) [![Greenkeeper badge](https://badges.greenkeeper.io/alexjoverm/typescript-library-starter.svg)](https://greenkeeper.io/) [![Travi
Downloads
6
Readme
Parmenides
TypeScript library that performs dynamic type checking
TypeScript is great. It warns you of static type errors and hence earns you lots of time and headaches. But your program probably have entrypoints (network requests, file readings, etc.) that can not be trusted completely.
For instance, supose you read some configuration from a JSON
file:
import { readFile } from 'fs';
interface AwsConfig {
// definitions...
};
readFile('./my-aws-config.json', { encoding: 'utf8' }, (err, awsConfigStr) => {
if (err) {
console.error(err);
return;
}
const awsConfig: AwsConfig = JSON.parse(awsConfigStr);
});
In this example, TypeScript can not prevent errors if the read JSON doesn't have an expected property. These are some cases this library was created for.
Features
- Infers typings.
- Very lightweight (under 4kb without minifying).
- Expressive errors.
- Works both client and server-side.
Installation
npm install --save parmenides
Usage
The main concept behind the library is "contracts". A contract is an identity function that throws an error if the parameter doesn't have the expected type.
For example if you pass a string
to the str
contract, it will return the same value, for other types it will throw an error:
import { str } from 'parmenides';
str('Hello world'); // <- Returns 'Hello world'
str(8 as any); // <- Throws an error
Check for specific values
import { oneOf } from 'parmenides';
import { createInterface } from 'readline';
const isFruit = oneOf('apple', 'banana', 'strawberry', 'orange');
type Fruit = 'apple' | 'banana' | 'strawberry' | 'orange';
const readLine = createInterface({
input: process.stdin,
output: process.stdout
});
const aFruit: Fruit = 'cheese'; // <- static error. It will be warned by TypeScript itself.
readLine.question('Which is your favourite fruit?', (someFruit) => {
const favouriteFruit: Fruit = isFruit(someFruit); // <- Will throw an error if `someFruit` has any other value than 'apple', 'banana', 'strawberry' or 'orange'. It's a potential dynamic error and TypeScript could not detect it.
});
It's important to notice that while str
is a contract itself, oneOf
is not. oneOf
is a function that returns a contract. You can think of it like a contract builder.
There are some other functions that help you build contracts
. For instance, there is arrOf
:
import { arrOf, num, oneOf } from 'parmenides';
const onlyNumbers = arrOf(num);
onlyNumbers([1, 2, 3]); // <- Returns [1, 2, 3]
onlyNumbers(['Hello', 'world', 99] as any); // <- Throws an error
const onlyHobbits = arrOf(oneOf('Frodo', 'Bilbo', 'Merry', 'Pippin', 'Sam', 'Gollum'));
onlyHobbits(['Bilbo', 'Sam']); // <- Returns the array
onlyHobbits(['Frodo', 'Pippin', 'Gandalf']); // <- Throws an error
As you can see, arrOf
takes a contract as parameter and returns another contract.
Last, but not least, the objOf
function is perhaps the most usefull one:
import { objOf, bool, str, num, arrOf } from 'parmenides';
const personValidator = objOf({
name: str,
age: num,
profession: oneOf('student', 'employed', 'unemployed', 'retired'),
address: objOf({
street: str,
city: str,
state: str,
country: str
}),
driving_license: bool
});
const peopleValidator = arrOf(personValidator);
// xhr function from any library you like
xhr('/URI/to/people')
.then(peopleValidator)
.then(people => /* The `people` variable is guaranteed to have the shape you have defined... */);
Notice that the objOf
function takes an object that describes the shape of the expected objects as a parameter. That object's properties are contracts.
Type inference
It's important to mention that all the contracts are typed and TypeScript will prevent errors if the parameters are incorrect and will infer the output:
import { str, num, objOf } from 'parmenides';
str(9); // TypeScript will error ("string expected").
const aNumber = num(9); // TypeScript will infere it's a number.
const fooBarContract = objOf({
foo: str,
bar: num
});
fooBarContract({
baz: 'Hello'
}); // <- Typescript will error
const fooBar = fooBarContract({
foo: 'Hello',
bar: 100
}); // <- TypeScript will infer type {foo: string; bar: number;}
furthermore, you can use the type function ContractOf
to put the types of a contract inside a type. For example:
import { ContractOf, num, objOf, str } from 'parmenides';
const personContract = objOf({
name: str,
age: num
});
type Person = ContractOf<typeof personContract>;
const john: Person = {
name: "John Doe",
age: 27
};
This is very useful as you can define the types for your external sources from contracts, and then use it with its name.
Because of the difference between type
and interface
, when you use the first, the type will be an alias for { name: string, age: number}
, and if you use an editor like VSCode, that's what you'll see when you hover over it. If you want to give it a name, you can use the following trick:
import { ContractOf, num, objOf, str } from 'parmenides';
const personContract = objOf({
name: str,
age: num
});
interface Person extends ContractOf<typeof personContract> {}
const john: Person = {
name: "John Doe",
age: 27
};
API
Built-in contracts
| Function | Type | Example |
| --------- | -------------------------------------- | -------------------------------- |
| bool
| IContract<boolean>
| bool(true);
|
| num
| IContract<number>
| num(89);
|
| str
| IContract<string>
| str('Hello world');
|
| undef
| IContract<undefined>
| undef(undefined);
|
| nil
| IContract<null>
| nil(null);
|
| arr
| <T> IContract<T[]>
| arr([1, 2, 3]);
|
| obj
| <T extends object> IContract<T>
| bool({foo: 'foo'});
|
| regExp
| IContract<RegExp>
| regExp(/^hello/i);
|
| date
| IContract<Date>
| date(new Date());
|
| anything
| <T> IContract<T>
| anything(4);
|
| never
| IContract<never>
| never(4 as never);
|
A note on anything
anything
is just an identity function that will never throw an error. Its static type will be inferred from the value if possible or will default to any
. It's useful with another functions like objOf
(view below). For instance you can define a contract like:
const objHasFooContract = objOf({
foo: anything
});
A note on never
You may think the never
contract is useless. But it can be used to do an exhaustive check:
const reactToSemaphore = (semaphoreLight: 'red' | 'yellow' | 'green') => {
switch (semaphoreLight) {
case 'red':
return 'stop';
case 'yellow':
return 'hurry';
case 'green':
return 'go';
default:
never(semaphoreLight);
}
};
The function reactToSemaphore
will fail in runtime if passed another value than 'red' | 'yellow' | 'green'
, but also with statically check that you aren't missing a case
in the switch
statement.
You can read more about the use of never
here.
contract
builders
optional
<T> (IContract<T>) -> IContract<T | undefined>
Takes a contract
and returns a new one that matches like the first one but also matches undefined
values.
const optionalNumber = optional(num);
// All the following are valid:
optionalNumber(9);
optionalNumber(undefined);
optionalNumber();
nullable
<T> (IContract<T>) -> IContract<T | null>
Takes a contract
and returns a new one that matches like the first one but also matches null
values.
const nullableNumber = nullable(num);
// The following are valid
nullableNumber(9);
nullableNumber(null);
oneOf
(...(string | number | boolean)[]) -> IContract<union of valid values>
It is used to validate unum
-like values. You specify the valid values and it returns a contract
that will check against them. Example:
const osContract = oneOf('Linux', 'Mac OS', 'Windows', 'Other');
const os = osContract('Linux'); // os's type is 'Linux' | 'Mac OS' | 'Windows' | 'Other'
TypeScript will infere the contract
's return value as the union of the literal types passed (up to 10 parameters, then behaves like <T extends string | number | boolean> IContract<T>
).
union
...(IContract) _> IContract<union of valid values>
It takes contracts as arguments and returns a new contract that matches if any of the them matches.
const numOrStr = union(num, str);
numOrStr(9);
numOrStr('nine');
TypeScript will infere the contract
's return value as the union of the return values of the contracts passed (up to 10 parameters, then behaves like IContract<any>
).
arrOf
<T> (IContract<T>) -> IContract<T[]>
It takes a contract
"C
" as a parameter and returns another contract
that expects an array
of elements that match C
.
const arrOfNumbersContract = arrOf(num);
const numbers = arrOfNumbersContract([1, 2, 3]);
objOf
<T> (IMapOfContracts<T>) -> IContract<T>
Takes an object that describes the shape of the objects
you want to validate and returns a contract
with that validation. That object's values must be contracts
.
const petContract = objOf(
name: str,
species: oneOf('dog', 'cat', 'golden fish', 'parrot', 'other'),
age: number,
gender: oneOf('male', 'female')
);
// <3
const oddy = petContract({
name: 'Oddy',
species: 'dog',
age: 8,
gender: 'female'
});
strictObjOf
<T> (IMapOfContracts<T>) -> IContract<T>
It is the same than objOf
function, but also checks that the target doesn't have extra properties.
// It only matches empty objects
const emptyObjectContract = strictObjOf({});
const emptyObject = emptyObjectContract({});
instanceOf
<C> (constructor: C) -> IContract<I>
(I
is instance of C
)
It takes a class or a or a constructor function and returns a contract
of instances of that class or constructor.
class Foo {}
const instanceOfFooContract = instanceOf(Foo);
const foo = instanceOfFooContract(new Foo());
dictionaryOf
<T> (IContract<T>) -> IContract<Record<string, T>>
It takes a contract
"C
" as a parameter and returns another contract
that expects a dictionary
of elements that match C
.
const contactsContract = dictionaryOf(
objOf({
name: str,
age: num
})
);
const contacts = contactsContract({
john: { name: 'john', age: 27},
jane: { name: 'jane', age: 27}
});
Publish
This library is published through TravisCI when merged to master, the version number is calculated automatically by semantic release.
It is important for the calculation of the release number that the previous npm release is found in the master branch and that the PR are merged without squashing.
The normal flow of development should be to create PR's against the development branch, merge them using the merge commit option. Then, once all the needed features/fixes are merged, we can create a PM from development to master that should be merged with fast-forward. We still need to configure github to disallow other type of merges and ease the release cycle.
Credits
Made from the tsall/typescript-library-starter
.
Thanks goes to these wonderful people (emoji key):
| Gonzalo Gluzman💻 🎨 📖 💡 🤔 📦 ⚠️ | Hernan Rajchert🎨 🤔 💻 | | :---: | :---: |
This project follows the all-contributors specification. Contributions of any kind welcome!