Validate Fields

A simple yet powerful JSON schema validator

Install

npm install validate-fields

Usage

Basic

let validate = require('validate-fields')(),
	schema = {
		name: String,
		age: 'uint'
	}, value = {
		name: 'Ann',
		age: 27
	}

// Validate the value against the schema definition
if (validate(schema, value)) {
	console.log('Valid!')
} else {
	console.log('Invalid', validate.lastError)
}

validate throws if it could not understand the schema and returns false if the schema is ok but the value doesn't match the schema.

validate.lastError, validate.lastErrorMessage and validate.lastErrorPath will let you know about the validation error cause

See the list of valid values for the schema bellow

Pre-parsing

// Parse the schema once and store it
let schema = validate.parse([String]) // an array of non-empty strings

// Set express route
// app here is an express instance, used only to ilustrate this example (not part of this module!)
app.post('/api/posts/by-tags', function (req, res, next) {
	// Validate using the parsed schema
	if (schema.validate(req.body)) {
		next()
	} else {
		// Answer with error (note that we're not using validate.lastError here)
		res.status(400).send(schema.lastError)
	}
}, handlePostsByTags)

This way is recommended because if the schema definition is invalid, validate.parse() will throw early and schema.validate() won't throw. You also gain in speed!

Escaping HTML

HTML strings can be escaped by sending one more argument to validate:

let v = ['html>']
validate([String], v)
v[0] === 'html>'

validate([String], v, {
	escape: true
})
v[0] === 'html>'

Or, using pre-parsing

let v = ['html>']
validate.parse([String]).validate(v, {
	escape: true
})
v[0] === 'html>'

Strict matching

By default, extra keys are not considered errors:

validate({}, {a: 2}) // true

This can be changed with the strict option:

validate({}, {a: 2}, {strict: true}) // false

Standard types

Hash map (object)

Example: {a: Number, 'b?': String, c: {d: 'int'}}

Keys that end in '?' are optional (b, in the example above). All others must be present and must not empty. Empty is null or undefined; for strings, '' is also considered empty.

If a value is optional and is empty it will be removed. Example:

let obj = {a: '', b: null}
validate({'a?': String, 'b?': 'int'}, obj) // true
obj // {}

Optional fields may declare a default value to use in its place when empty. Example:

let obj = {}
validate({'a=12': Number, 'b=[]': [String], 'c="hi"': String}, obj) // true
obj // {a: 12, b: [], c: 'hi'}

The value after the = character must be a valid JSON.

Array

Example: {books: [{title: String, author: String}]}

Other types

• Number: a double
• String: a string
• Object: any non-null object
• Array: any array
• Boolean
• Date: any date string accepted by Date constructor (ISO strings are better though)
• '*': anything
• 'number(-3.5,10)': a number x with -3.5 <= x <= 10
• 'number(-3.5,)': a number x with -3.5 <= x
• 'number(,10)': a number x with x <= 10
• 'numeric': a double value encoded in a string
• 'numeric(<...>)': like 'number(<...>)', but for numeric strings
• 'int': an integer between -2^51 and 2^51 (safe integer)
• 'int(-3,10)': an integer between -3 and 10. Lower and upper bounds are optional
• 'numericInt': like 'int', but for numeric strings
• 'uint': a natural number less than 2^51
• 'uint(3,10)': a natural number between 3 and 10. Lower and upper bounds are optional
• 'numericUint': like 'int', but for numeric strings
• 'string(17)': a string with exactly 17 chars
• 'string(,100)': at most 100 chars
• 'string(8,)': at least 8 chars
• 'string(8,100)': at most 100, at least 8
• 'hex': a hex string (even number of hex-chars)
• 'hex(12)': a hex-string with exactly 12 hex-chars (that is, 6 bytes)
• 'base64': a base64 string, with valid padding
• 'id': a mongo objectId as a 24-hex-char string
• 'email': a string that looks like an email address
• 'in(cat, dog, cow)': a string in the given set of strings
• 'numberIn(3, 1.4, -15)': a number in the given set of values
• 'numericIn(<...>)': like 'numberIn(<...>)', but for numeric strings
• /my-own-regex/: a string that matches the custom regexp

Custom types

The simplest way to define your own type is like this:

validate.typedef('name', {
	first: String,
	'last?': String
})
validate.typedef('zip-code', /^\d{5}([- ]\d{4})?$/)

validate('name', {last:'Souza'}) // false
validate.lastError // 'I was expecting a value in first'
validate(['zip-code'], ['12345', '12345-1234']) // true

Typedef works as a simple alias (like in C)

Every time you call require('validate-fields')() a new context is created. Each context is isolated from each other. Types defined in one context do not conflict with other contexts.

If you need more power, you can create the type from scratch and define every detail about the validation.

In the example bellow, we create a type defined by 'divBy3' that matches JSON numbers. The third param is the check function, that should throw when the value is invalid. It may return a new value, that will replace the original value in the object

validate.registerType('divBy3', 'number', function (value) {
	if (value%3 !== 0) {
		// Note: you should throw a string
		// If you throw an Error instance it won't be caught!
		throw 'I was expecting a number divisible by 3'
	}
	return value/3
})

let obj = {n: 12}
validate({n: 'divBy3'}, obj) // true
obj.n // 4

In the example bellow, we implement the general case of the above. Instead of a fixed string (divBy3) we define a syntax like 'divBy(n)' where n is a number. The "argument" list is sent as the third parameter to the check function.

validate.registerTaggedType({
	tag: 'divBy',
    jsonType: 'number',
    minArgs: 1, // default = 0. If 0, 'tag' and 'tag()' are equal
    maxArgs: 1, // default = 0 = no limit
    sparse: false, // default = false. If true, let arguments be skipped: 'tag(1,,2)'
    numeric: true // default = false. If true, parse all arguments as numbers
}, function (value, args) {
	let n = args[0]
	if (value%n !== 0) {
		throw 'I was expecting a number divisible by '+n
	}
	return value/n
})
validate('divBy(17)', 35) // false
validate('divBy(35)', 35) // true

See more examples in the folder types. All core types are defined there

Finally, you can extend a type by providing pre and post hooks:

validate.typedef('time-in', {
	hour: /^\d\d$/,
	minute: /^\d\d$/
}, null, value => {
	// Convert that object to a string like 'HH:MM'
	return value.hour + ':' + value.minute
})

validate.typedef('time-out', {
	hour: /^\d\d$/,
	minute: /^\d\d$/
}, value => ({
	// Convert 'HH:MM' to an object
	hour: value.slice(0, 2),
	minute: value.slice(3, 5)
}))

Pre-hooks will be executed before the base type checks. Post-hooks will be run afterwards.

Interchangeable format

A parsed Field can be serialized to JSON with JSON.stringify():

let fields = validate.parse({name: String, age: 'uint', 'birth?': Date})
JSON.stringify(fields) // '{"name":"$String","age":"uint","birth?":"$Date"}'

Note that custom types can't be serialized, they are replaced by their JSON-type:

let fields = validate.parse({myType: 'divBy(7)'}) // defined above
JSON.stringify(fields) // '{"myType":"$Number"}'

To get back a Field instance, JSON.parse() should be used with a reviver and the result then parsed:

let serializedFields = '{"name":"$String","age":"uint","birth?":"$Date"}'
let definition = JSON.parse(serializedFields, validate.reviver) // {name: String, age: 'uint', 'birth?': Date}
let fields = validate.parse(definition)

fields.validate({name: 'John', age: 12}) // true

JSON Schema

A parsed Field can be exported as JSON Schema:

let fields = validate.parse({name: String, age: 'uint', 'birth?': Date})
fields.toJSONSchema()
/* {
	type: 'object',
	properties: {
		name: {
			type: 'string'
		},
		age: {
			type: 'integer',
			minimum: 0
		},
		birth: {
			type: 'string',
			format: 'date-time'
		}
	},
	required: ['name', 'age']
} */

Partial validation

A schema can be used for validating only a subset of fields, for example:

let schema = validate.parse({label: String, creation: Date})
schema.validate({label: 'earth'}) // false
schema.validate({label: 'earth'}, {partial: ['label']}) // true

This is useful when fields in the value can be dynamically selected, like in an endpoint when the caller only cares about two or three fields from a normally bigger schema.

Multiple paths should be passed as elements of the options.partial array. Paths can also recurse into inner properties:

validate({
	post: {title: String, creation: Date},
	user: {name: String, age: 'uint'}
}, {
	post: {title: 'Top 10 ways to create top listings'},
	user: {age: 30}
}, {
	partial: ['post.title', 'user.age']
}) // true

In arrays, partial paths enumerate over its elements:

validate({
	table: {
		rows: [{
			cells: [{value: Number, text: String}]
		}]
	}
}, {
	table: {
		rows: [{
			cells: [{value: 3}, {value: 13}]
		}, {
			cells: [{value: 14}, {value: '15'}]
		}]
	}
}, {
	partial: ['table.rows.cells.value']
}) // false
validate.lastError // I was expecting number and you gave me string in table.rows.1.cells.1.value

Custom values should implement partial validation accordingly, using the value received over options._partialTree. Example:

validate.registerType('custom-with-partial', 'object', (value, extra, options) => {
	options._partialTree // {c: {d: false}}
})
validate(
	{a: {b: 'custom-with-partial'}},
	{a: {b: 'some data'},
	{partial: ['a.b.c.d']}
)

In strict mode, any field in the value not listed in the partial will result in an error. Likewise, any field in the partial not in the schema will result in an error.