Huge News!Announcing our $40M Series B led by Abstract Ventures.Learn More →

gubu

Package Overview

Dependencies

Advanced tools

Install Socket

Detect and block malicious and high-risk dependencies

Install

gubu

An object shape validation utility.

8.2.0
Source
npm

Version published: 5 months ago

Weekly downloads: 10K; increased by10.17%

Maintainers: 1

Weekly downloads

Created: 3 years ago

Source

Gubu: An object shape validation utility.

	This open source module is sponsored and supported by Voxgig.

This is a schema validator in the tradition of Joi or any JSON-Schema validator, but with a much nicer developer experience. It works in JavaScript and TypeScript, on the browser and the backend.

The big idea is that your schema looks (almost) exactly like your data.

That makes your schema much easier to read and reason about. You could call it:

"Schema By Example"

Let's say you want to have the some options for a module you are writing:

{
  port: 8080,
  host: 'localhost'
}

This is a valid Gubu specification. It means:

{
  port: 8080,        // Must be a number, is optional, and the default is 8080
  host: 'localhost'  // Must be a string, is optional, and the default is 'localhost'
}

If your user does not specify these options, then you get the defaults:

{} --> { port: 8080, host: 'localhost' }

In Gubu, the most common case for options is the easiest: everything is optional, and the default value defines the type you will accept. This also works for objects, which get "filled out" if not present:

const { Gubu } = require('gubu')

const shape = Gubu({
  server: {
    port: 8080,
    host: 'localhost'
  }
}

let options = {}
shape(options)

// And now options is:
{
  server: {
    port: 8080,
    host: 'localhost'
  }
}

Another big feature of Gubu is that you can fill out objects to any depth (unlike Object.assign or the ... spread operator).

You may have noticed that Gubu mutates the input to be validated (by injecting defaults as needed). This is deliberate. Cloning arbitrary values in JavaScript is complicated, so Gubu leaves that decision to calling code.

To make properties required, you use the standard JavaScript wrapper objects (Number, String, Boolean, ...):

const { Gubu } = import 'gubu' // `import` also works! And in browsers too.

const shape = Gubu({
  timeout: Number,
  message: String,
  debug: Boolean,
})

// All good here - these are valid options!
shape({
  timeout: 10000,
  message: 'Hello!',
  debug: true,
})

Required properties don't have defaults (how could they?), so you only need to specify the type.

To validate arrays, you provide an example element. All elements of the array must match the example element:

const shape = Gubu([Number])

// All good - we want numbers!
shape([100, 200, 300])

Motivation

Why write yet another validator? I've used Joi for a long time, but always found its schema definition a little verbose at the syntax level. I've never liked JSON-Schema - it's just too noisy to eyeball. What I do like is Vue.js property validation, but that only works at the top level. I did write a prototype deep Vue property validator using Joi, but it's pretty clunky.

This validator is motivated by two use cases: adding message validation to the Seneca microservices framework, and providing deep defaults for complex React and Vue.js components. I think it does both jobs rather nicely with an easy-to-read syntax.

On the philosophical side, this validator is motivated by the idea of query-by-example. This comes from the belief that software developers should not be subjected to unnecessary additional levels of abstraction. Abstraction is hard, takes you further from the problem you're getting paid to solve, and is hard to "eye-ball". I don't want to parse and run schemas "in my head", I just want to see the literal values and structure that I care about, right here, right now.

Quick Example

_^top


const { Gubu } = require('gubu')

// Property a is optional, must be a Number, and defaults to 1.
// Property b is required, and must be a String.
const shape = Gubu({ a: 1, b: String })

// Object shape is good! Prints `{ a: 99, b: 'foo' }`
console.log( shape({ a: 99, b: 'foo' }) )

// Object shape is also good. Prints `{ a: 1, b: 'foo' }`
console.log( shape({ b: 'foo' }) )

// Object shape is bad. Throws an exception with message:
//   Validation failed for property "a" with value "BAD" because the value is not of type number.
//   Validation failed for property "b" with value "" because the value is required.
console.log( shape({ a: 'BAD' }) )

// Object shape is bad. Throws an exception with message:
//    Validation failed for object "{b:foo,c:true}" because the property "c" is not allowed.
console.log( shape({ b: 'foo', c: true }) )

As shown above, you use the exported Gubu function to create a validation shape checker using an example object. Pass the value you want to validate to the shape checker. If the value is valid (matches the example object), the shape checker returns the value (with missing defaults injected). Otherwise the shape checker throws an exception listing all (not just the first!) of the validation errors.

Common Use Cases

_^top

Option defaults and validation

Let's say you have a server that needs to run on a given host and port, but by default should run on localhost on port 8080. The host should be a non-empty string, and the port should be a number.

const optionShape = Gubu({
  host: 'localhost',
  port: 8080
})

// These print: { host: 'localhost', port: 8080 }
console.log(optionShape())
console.log(optionShape({}))

// Prints: { host: 'localhost', port: 9090 }
console.log(optionShape({ port: 9090 }))

// All of these throw an error.
console.log(optionShape({ host: 9090 }))   // Not a string.
console.log(optionShape({ port: '9090' })) // Not a number.
console.log(optionShape({ host: '' }))     // The empty string is a host name!
console.log(optionShape({ hpst: 'foo' }))  // 'hpst' is not a valid property name.

Deep structures

You're building a front end component that displays complex data from the back end, and you want to handle missing data gracefully, at any depth in the structure.

const productListShape = Gubu({
  products: [
    {
      name: String, // Product name is a required String
      img: 'generic.png' // Use a default image if not defined
    }
  ]
})

let result = productListShape({})

// No products, but our data structure still has an array where
// one is expected, so no `undefined` errors.
result === {
  products: []
}

// Fix data with missing fields
let result = productListShape({
  products: [
    { name: 'Apple', img: 'apple.png' },
    { name: 'Pear', img: 'pear.png' },
    { name: 'Banana' } // Missing image!
  ]
})

// Banana will not have a broken image.
result === {
  products: [
    { name: 'Apple', img: 'apple.png' },
    { name: 'Pear', img: 'pear.png' },
    { name: 'Banana', img: 'generic.png' }
  ]
}

Shape Building

For more specific shapes, such as required objects and arrays, you can use shape builder functions that still respect your data structure.

The Required shape builder makes a value explicitly required:

const { Gubu, Required } = require('gubu')

const userShape = Gubu({
  person: Required({  // person must be a defined object
    name: String,
    age: Number,
  })
})

// This will fail, with message:
//   Validation failed for property "person" with value "" because the value is required.
userShape({}) 

// This will pass, returning the object:
userShape({
  person: {
    name: 'Alice',
    age: 99
  }
})

Shape builders are exported by the Gubu module directly, and are also available as properties of the Gubu function:

const { Gubu, Required, Closed } = import 'gubu'

Required === Gubu.Required
Closed === Gubu.Closed

For the full list of shape builders, see the API reference.

Install

_^top

$ npm install gubu

Usage

The Gubu module has no dependencies. A function named Gubu is exported as the main entry point. Shape builders (such as Required) and utility functions are provided as properties of Gubu or can be imported separately.

TypeScript

Gubu is written in TypeScript, and can be imported naturally:

import { Gubu } from 'gubu'

Types are provided in gubu.d.ts.

Gubu tries to play nice with compile-time type validation of your shapes, and mostly succeeds.

Browser

A minified version is provided as gubu.min.js, which can be directly loaded into a web page and exports a Gubu global object.

However you're probably better off importing this module in the usual manner for your build process and bundling it together with everything else.

Shape Rules

_^top

The general principle of Gubu's design is that the schema shape should match a valid object or value as closely as possible.

For scalar values you can provide a native type object to make the value required:

Gubu(String) matches strings: 'foo'
Gubu(Number) matches numbers: 123
Gubu(Boolean) matches booleans: true

Or defaults to make the value optional:

Gubu('bar') matches strings: 'foo', and undefined
Gubu(0) matches numbers: 123, and undefined
Gubu(false) matches booleans: true, and undefined

If a value is optional and undefined, the default value is returned: Gubu('bar')() returns 'bar'.

The values null and NaN must match exactly. They are not the same as undefined. The value undefined is special - it literally means no value. To allow a property to be absent entirely, use the Skip shape builder.

Empty strings are not considered to be valid (this is usually what you want). To allow an empty string, use Gubu(Empty('foo')) or Gubu(Empty(String)) (where Empty is exported by the Gubu module).

For objects, write them as you want them:

let shape = Gubu({
  foo: {
    bar: {
      zed: String,
      qaz: Number,
    }
  }
})

The above shape will match:

{
  foo: {
    bar: {
      zed: 'x',
      qaz: 1
    }
  }
}

If you want to allow arbitrary properties in an object, you can use the Open shape builder:

let openObject = Gubu(Open({ a: 1 }))

// This now passes (normally property `b` would not be allowed).
openObject({ a: 11, b: 22 }))

For arrays, the first element is treated as the shape that all elements in the array must match:

Gubu([String]) matches ['a', 'b', 'c']
Gubu([{x: 1}]) matches [{x: 11}, {x: 22}, {x: 33}]

If you need specific elements to match specific shapes, use the Closed shape builder:

Gubu(Closed([String, Number])) matches ['a', 1].

You can specify custom validation functions using the Check shape builder:

Gubu({a: Check((v) => 10 < v) }): matches {a: 11} as 10 < 11

And you can manipulate the value if you need to:

Gubu({a: Check((v, u) => 10 < v ? (u.val = 2 * v, true) : false )}): matches {a: 11} as 10 < 11 and returns {a: 22}.

You can also compose validations together:

const shape = Gubu({ a: Gubu({ x: Number }) })

// Matches { a: { x: 1 } } as expected
shape({ a: { x: 1 } })

The shape builder functions (Required, Closed, etc.) are also available as properties of the main Gubu function, so you don't have to introduce them into your top level variable namespace.

As a convenience, you can chain most builders. Thus a Required and Closed object can be specified with:

const { Gubu } = require('gubu')

const shape = Gubu({
  a: Gubu.Closed({ x: 1 }).Required(),
  b: Gubu.Required({ x: 1 }).Closed(), // Same as line above
})

You can also use an optional syntax to avoid importing the shape builders entirely. This is useful if you want your schemas to be fully serializable to JSON.

// The key string format `name: builder-expression` is converted into 
// an application of the shape builders on the value.

const shape = Gubu({ 'a: Open': { x: 1 }}, { keyexpr: { active: true } })

// Matches as a.y is allowed as a is Open
shape({ a: { x: 1, y: 2 } })

// shape is equivalent to:
shape = Gubu({ a: Open({ x: 1 }) })

The builder expression syntax is the same as JavaScript function call syntax, but all values must be in JSON format.

const shape = Gubu({ 'a: Exact("red",1,true)': 'red'}, { keyexpr: { active: true } })

You can write your own builders—see the next section.

In addition to this README, the unit tests are comprehensive and provide many usage examples.

Shape Builders

The built-in shape builders help you match the following shapes:

Existence:
- Required: Make a value required.
- Optional: Make a value optional.
- Skip: Make a value skippable (no default value is injected if missing).
- Default Specify a default builder.
Value constraints:
- Empty: Allow string values to be empty.
- Exact: The value must match one of an exact list of literal values.
- All: All shapes must match value.
- Some: Some shapes (at least one) must match value.
- One: Exactly one shape (and no more) must match value.
- Any: This shape will match any value.
- Never: This shape will never match anything.
Length constraints (operate on values with a length or numeric value):
- Below: Match a value (or length of value) less than the given amount.
- Max: Match a value (or length of value) less than or equal to the given amount.
- Min: Match a value (or length of value) greater than or equal to the given amount.
- Len: Match a value (or length of value) exactly equal to the given amount.
- Above: Match a value (or length of value) greater than the given amount.
General constraints:
- Closed: Allow only explicitly defined elements in an array.
- Open: Allow arbitrary properties in an object (no constraint on their value).
- Child: All non-explicit child values of an object must match this shape.
- Func: The value is explicitly a function.
Mutations:
- Rename: Rename the key of a property.
- Define: Define a name for a value.
- Refer: Refer to a defined value by name.
Customizations:
- Check: Define a general custom validation function for the value (recommended).
- Before: Define a custom validation function called before a value is processed (advanced use).
- After: Define a custom validation function called after a value is processed (advanced use).
- Key: The key (or path) of the current object is injected as the value.

Regular Expressions

The Check shape builder will also accept a regular expression (instead of a function). The value will be converted to a string (using String(...)), and will be valid if it matches the regular expression. The values null, undefined and NaN are not converted to strings and will always fail this check.

let shape = Gubu({ countryCode: Check(/^[A-Z][A-Z]$/) })

shape({ countryCode: 'IE' })) // PASS.
shape({ countryCode: 'BAD' })) // FAIL: 'Validation failed for property "countryCode" with value "BAD" because check "/^[A-Z][A-Z]$/" failed.'

Recursive Shapes

You can use the Define and Refer shape builders to validate recursive shapes. Use Define first to name a given shape. Then use Refer to apply the definition of the shape.

let tree = Gubu({
  root: Define('BRANCH', {
    value: String,
    left: Refer('BRANCH'),
    right: Refer('BRANCH'),
  })
})

// This passes, returning the object.
tree({
  root: {
    value: 'A',
    left: {
      value: 'AB',
      left: {
        value: 'ABC'
      },
      right: {
        value: 'ABD'
      },
    },
    right: {
      value: 'AE',
      left: {
        value: 'AEF'
      },
    },
  }
})

// This fails with error:
// "Validation failed for property "root.left.left.left.value" with value "123" because the value is not of type string."

tree({
  root: {
    value: 'A',
    left: {
      value: 'AB',
      left: {
        value: 'ABC',
        left: {
          value: 123
        },
      },
    },
  }
})

API

_^top

Shape

API | Shapes | Errors | Custom Validation | Builders

A shape specification can either be at the root of a JSON structure, an array element, or the value of an object property. For a value to pass validation, it is compared to the shape, and must match the constraints of the shape. If the shape has elements or properties, these must also match, and are validated recursively in a depth-first manner ¹.

Required Values

The value must be of the indicated type, and must exist.

String: match any string, but not the empty string ².
Number: match any number, but not BigInt values.
Boolean: match any boolean.
Function: match any function.
Object: match any object.
Array: match any array.
Symbol: match any symbol.
BigInt: match any BigInt (including the 1n syntax form).
Error: match an object created with new Error(...)
Date: match an object created with new Date(...)
RegExp: match an object created with /.../ or new RegExp(...)

You can require an instance of any class (that is, an object created with new) by using the class name as the shape.

Optional Values with Defaults

The value must be of the indicated type, and is derived from the given default. If the value does not exist, the default value is inserted.

foo: match any string, but replace an empty string ³.
123: match any number, but not BigInt values.
true: match any boolean.
new Object(): match any object.
new Array(): match any array.
new Symbol('bar'): match any symbol.
new BigInt(456): match any BigInt (including the 1n syntax form).
new Error(): match an object created with new Error(...)
new Date(): match an object created with new Date(...)
/x/: match an object created with /.../ or new RegExp(...)

You can provide an instance of any class as a default. The value, if present, must be an instance of the same class.

Note that new Function() does not match anonymous functions, and should not be used ⁴.

Objects

Plain objects can be specified directly as they appear in values to be validated. If you want an object {foo: 123}, then the shape is also {foo: 123}, meaning any object with a foo property (and no other properties) that is a number. The foo property is optional, and will default to the value 123.

You can define plain objects to any depth. The shape { bar: { foo: 123 } } defines an object that optionally contains another object as the value of the property bar.

As objects and sub-objects are often referenced directly in data structures (using dot notation), Gubu will construct missing objects by default, and fill in the missing child values (which may themselves be objects). These protect your code from undefined value errors in default cases.

The general form of an object shape is:

{ 
  'propName0': <SHAPE>,
  'propName1': <SHAPE>,
  ...
  'propNameN': <SHAPE>,
}

where propNameX is any string (the quotes may be omitted if the property name is a valid JavaScript identifier—this is just normal JavaScript syntax after all).

The <SHAPE> can be any valid Gubu shape definition.

Required Properties (Object)

To mark an object property as required, use the required value shapes (such as String), or use the shape builder Required:

const { Required } = Gubu

let shape = Gubu({
  foo: Number,
  bar: Required({
    zed: Boolean
  })
})

// This passes, returning the value unchanged.
shape({ foo: 1, bar: { zed: false } })

// These fail, throwing an Error.
shape({ bar: { zed: false } }) // foo is required
shape({ foo: 'abc', bar: { zed: false } }) // foo is not a number
shape({ foo: 1 }) // bar is required
shape({ foo: 1, bar: {} }) // bar.zed is required
shape({ foo: 1, bar: { zed: false, baz: 2 }, qaz: 3 }) // new properties are not allowed

Object properties that are required must always be provided, even if they are children of optional objects—they wouldn't be required otherwise! To allow such deep required properties to be missing, use an explicit Skip shape builder:

let strictShape = Gubu({ a: { b: String } })

// Passes
strictShape({ a: { b: 'ABC' } })

// Fails, even though a is not required, because a.b is required.
strictShape({})


let easyShape = Gubu({ a: Skip({ b: String }) })

// Now both pass
easyShape({ a: { b: 'ABC' } })
easyShape({})

// This still fails, as `a` is now defined, and needs `b`
easyShape({ a: {} })

Open Objects

Normally, objects can only contain explicitly defined properties. To allow an object to have an unrestricted set of properties, use the Open shape builder:

const { Open } = Gubu

let shape = Gubu(Open({
  a: 1
}))

shape({ a: 11, b: 22 }) // PASS: returns { a: 11, b: 22 }
shape({ b: 22, c: 'foo' }) // PASS: returns { a: 1, b: 22, c: 'foo' }

shape({ a: 'foo' }) // FAIL: property `a` must still be a number

The Open shape builder applies only to the object it wraps, and does not apply to child objects. You need to use Open explicitly for each object that can have arbitrary properties.

let shape = Gubu(Open({
  a: Open({
    b: 1
  })
}))


shape({ a: { b: 11, c: 22 }, d: 33 }) // PASS, returns object

An empty object shape ({}) is automatically open and will allow any properties.

The Open shape builder can also be used without adding the Open builder function to your imports. Instead, use the optional keyexpr feature and specify that an object is open with:

let shape = Gubu({
  'a: Open': {
    b: 1
  }
}, { keyexpr: { active: true } })

Optional Objects

Objects are optional by default, and will be created if not present. To prevent this, use the shape builder Skip to indicate that you do not wish an object to be inserted when it is missing—it can be skipped.

const { Skip } = Gubu
let shape = Gubu({
  a: { x: 1 },
  b: Skip({ y: 2 }),
  c: Skip({ z: Skip({ k: 3 }) }),
})

// Skippable properties are not inserted as defaults if missing.
shape({}) // returns { a: { x: 1 } }, b and c are missing entirely
shape({ b: {} }) // returns { a: { x: 1 }, b: { y: 2 } }, defaults for b are inserted
shape({ c: {} }) // returns { a: { x: 1 }, c: {} }, c has no non-optional defaults
shape({ c: { z: {} } }) // returns { a: { x: 1 }, c: { z: { k: 3 } } } // z has defaults

If the object value is present but empty, any default values will be inserted.

Object Values

You can define a general shape for all non-explicit object values using the Child shape builder:

const { Value } = Gubu

let shape = Gubu(Child(String, {
  a: 123,
}))

// All non-explicit properties must be a String
shape({ a: 11, b: 'abc' }) // b is a string
shape({ c: 'foo', d: 'bar' }) // c and d are strings

// These fail
shape({ a: 'abc' }) // a must be a number
shape({ b: { x: 1 } }) // b must be a string

Using the Child shape builder in this way automatically makes the object open, but constrains the values that can be used for non-explicit properties.

The general shape can be any valid shape:

const { Required, Value } = Gubu

let shape = Gubu({
  people: Required({}).Value({ name: String, age: Number })
})

// This passes:
shape({ people: { 
  alice: { name: 'Alice', age:99 }, 
  bob:   { name: 'Bob',   age:98 },
} })

// These fail:
shape({ people: { 
  alice: { name: 'Alice', age:99 }, 
  bob:   { name: 'Bob' }, // age is a required number
} })
shape({}) // people is a required object

Arrays

Arrays can be specified directly using the first element as the shape that each element of the value array must match. If you want an array of numbers ([ 1, 2, 3 ], say), then the shape is [Number].

let shape = Gubu([Number])
shape() // PASS: returns [] (the array itself is optional)
shape([]) // PASS: returns [] (empty arrays pass)
shape([1]) // PASS: element matches Number shape
shape([1, 2]) // PASS: all elements match Number shape

shape([1, 2, 'bad']) // FAILS; element 2 is not a number


// Array elements can be any complex shape.
shape = Gubu([{x: 1}])
shape([{ x: 123 }, { x: 456 }]) // PASS: elements match {x: 1}
shape([{}]) // PASS:  returns [{x: 1}]

shape([{x: 123}, {x: 'a'}]) // FAILS; element 1 does not match {x: 1}

You can also define special shapes for individual elements at specific indexes, using the Closed shape builder, as shown below.

The general form of an array shape is:

[
  <SHAPE>,
]

where <SHAPE> is any valid Gubu shape definition. All elements must match <SHAPE>, and the empty array is allowed. This is the standard case and usually what you want.

For special cases, where elements at specific indexes must match specific shapes, you can define these arrays using the Closed shape builder.

let shape = Gubu(Closed([Number, String, Boolean]))

// This passes, returning the array as is.
shape([ 123, 'abc', true ])

// These fail.
shape([ 'bad' ]) // Index 0 must be a number
shape([ 123 ]) // Index 1 and 2 are missing
shape([ 123, 'abc', true, 'extra' ]) // Too many elements

As a shortcut, a shape array with two or more elements is considered closed, and all elements are considered special. Thus Closed([String, Number]) is the same as [String, Number]. For a single element closed array, you must use the Closed shape builder to close the array. Thus Closed([Number]) means the array can only ever have one element, a number.

As arrays are often referenced directly in data structures, Gubu will construct missing arrays by default, and for closed arrays, fill in the missing element values if there are empty or undefined array entries.

Required Properties (Array)

To mark an array element as required, use the required value shapes (such as String), or use the shape builder Required.

let shape = Gubu([{ x: 1 }, Required({ y: true })])

// These pass
shape([{ x: 2 }, { y: false }])
shape([undefined, { y: false }]) // returns [{ x: 1 }, { y: false }]
shape([{ x: 2 }, {}]) // returns [{ x: 2 }, { y: true }]

// These fail
shape([{ x: 2 }, undefined]) // Element 1 is required
shape([{ x: 2 }]) // Element 1 is required

Array Properties

JavaScript allows arrays to have properties: let a = []; a.foo = 1. Matching against array properties in the current version must be done by writing a custom validator using the Check shape builder.

Length Constraints

You can control the allowed length of an array using the shape builders Min, Max, Above, Below, and Len to restrict the length of the array.

let { Min } = Gubu
let shape = Gubu(Min(2, [Number]))

// These pass
shape([11,22]) // length is 2, >= minimum 2
shape([11,22,33]) // length is 3, >= minimum 2

// These fail
shape([11]) // length is 1, not >= minimum 2
shape([]) // length is 0, not >= minimum 2

The length constraining shape builders (Min, Max, Above, Below, and Len) work in a sensible way depending on the data type. For strings they control character length, for numbers they control magnitude, for objects they control property count, and for arrays, they control length:

let { Max } = Gubu

// Maximum string length
shape = Gubu(Max(2, String))
shape('a') // PASS
shape('ab') // PASS
shape('abc') // FAIL: string longer than 2 characters

// Maximum object size
shape = Gubu(Max(2, {})) // An empty object is open, so can accept any properties
shape({ a: 1 }) // PASS
shape({ a: 1, b: 2 }) // PASS
shape({ a: 1, b: 2, c: 3 }) // FAIL: more than 2 properties in object

Functions

Literal function value operate in the same way as any other literal values, defining an optional value shape with a default. This allows you to provide default functions for your module options, if you are using Gubu as an option validator.

let shape = Gubu({
  fn: () => true
})

// This passes
shape({ 
  fn: () => false 
})

// This injects the default function
shape({)) === {
  fn: () => true 
})

To require a function, use the shape Function, (Gubu(Function)(()=>true) will pass).

Key Expressions

If you wish to avoid importing the shape builder functions, or you would like your shape schema to be serializable to JSON, then you can use key expressions.

Key expressions allow you to specify the shape builders in the property key string. They must be enabled with an optional flag when creating the shape.

let shape = Gubu({
  'a: Skip': 1
}, { keyexpr: { active: true } })

This syntax is equivalent to:

let shape = Gubu({
  a: Skip(1)
})

When key epxressions are active, any key string of the form:

key:builder-expression

is evaluated as a key expression.

The syntax of key expressions is the same as JavaScript function call syntax. However, literal values must be written as JSON, and commas between arguments are optional. Sequential shape builder functions are evaluated left-to-right.

The following are all equivalent:

let shape = Gubu({
  a: Min(1, Max(4, Number)), // A number x, where 1 <= x <= 4.
  'b: Min(1, Max(4))': Number, // The property value (in this case, Number) is implicit,
  'c: Min(1) Max(4)': Number, // Same effect, since both operate on Number.
  'd: Min(1).Max(4)': Number, // Chaining is also supported.

}, { keyexpr: { active: true } })

Not all shape builders are supported, only those that can accept literal values as arguments. Thus custom Check functions cannot be specified as key expressions.

The literal value arguments are parsed with JSON.parse, so must be valid JSON. In addition, regular expressions, and the values undefined and NaN are also recognized.

Key expressions can only be used with object keys, and are not supported at the top level.

Custom Validation

API | Shapes | Errors | Custom Validation | Builders

You can define custom validators by providing a function to the Check shape builder. The first argument to this function will provide the value to validate. Return true if the value is valid, and false if not.

import { Gubu, Check } from 'gubu'

let shape = Gubu({ a: Check((v) => 10 < v) })
shape({ a: 11 }) // passes, as 10 < 11 is true
shape({ a: 9 })  // fails, as 10 < 9 is false

You can modify the value using the second argument to the custom validation function, by assigning a new value to the val property:

let shape = Gubu({ 
  a: Check((value, update) => {
    update.val = value * 2 
    return true // Remember to return true to indicate value is valid!
  })
})

shape({ a: 3 }) // returns { a: 6 }

As a special case, if you want to explicitly set the value to undefined or NaN, use the property uval.

You can also provide a custom error message using the update argument:

let shape = Gubu({ 
  a: Check((value, update) => {
    update.err = 'BAD VALUE $VALUE AT $PATH'
    return false // always fails
  })
})
shape({ a: 3 }) // throws "BAD VALUE 3 AT a"

The special replacement tags $VALUE and $PATH are replaced with the value, and the property path to the value, respectively.

The third argument to a custom validator is the internal state of the validation process. This is provided for special cases and workarounds, and the internal set of properties should not be considered stable. Review the source code to see what is available.

shape = Gubu({
  a: Check((value: any, update: any, state: any) => {
    update.val = value + ` KEY=${state.key}`
    return true
  })
})
shape({ a: 3 }) // returns { a: '3 KEY=a'}

The shape builders Before and After operate in a similar manner to the Check. They allow you to perform your custom validation before, or after, normal validation, respectively. They do not support regular expressions as an argument.

In general, you should use the Check shape builder for custom validation. The Before and After builders are provided for advanced usage.

Gubu function

To construct a shape use the Gubu function exported by this module:

// Using require
const { Gubu } = require('gubu')

// Using import
import { Gubu } from 'gubu'

let shape = Gubu({ x: 1 })

In the browser, Gubu adds itself directly to the window object for immediate use (if you directly load this module using a script tag). However you'll probably just want to import Gubu in the usual way into your own source code and let your package builder look after things.

The Gubu function has two arguments:

shape (optional): any valid shape definition ('abc', String, { x: 123 }, etc.).
options (optional): an options object.

The shape argument can be anything, and is used to define the validation shape expected. See the sections above for descriptions of the various shapes.

The options are:

name: a string defining a custom name for this shape (useful for debugging).

The Gubu function provides all built-in shape builders (Required, Closed, etc.) as properties. These are also exported directly from the module, so the following are equivalent:

const { Gubu, Required } = import 'gubu'

const { Gubu } = require('gubu')
const { Require } = Gubu

If you are concerned about namespacing the builders (if the names clash with your own names), the shape builders are also available with a 'G' prefix as an alias:

Gubu.GRequired === Gubu.Required

GubuShape function

When you create a shape using Gubu, a GubuShape shape validator function is returned:

// TypeScript
import { Gubu, GubuShape } from 'gubu'

// GubuShape is inferred:
const shape = Gubu(123) // `shape` is a validator function

The shape validator function has two arguments:

value: the value to validate (and modify with defaults).
context: (optional) a context object containing your own data.

The value can be anything. It is not duplicated and will be mutated if defaults are inserted.

If you do not wish the value to be mutated, you must clone it yourself first ⁵.

The context is a general purpose store for anything you might want to use in custom validation builders. It may also be used by builders to hold state information (the name of the builder is used for namespacing).

The context has one reserved name:

err: an array of validation errors.

If you provide a context with the property err as an empty array, any validation errors will be added to this array, and an Error will not be thrown:

let ctx = { err: [] }
Gubu(Number)('abc', ctx)  // does not throw
console.log(err[0]) // prints error description (number was expected)

The error descriptions are plain objects, not Error objects.

The GubuShape function has the following methods:

valid(value: any, context?: any): boolean
- returns true if the value matches the shape
- injects defaults into value
- does not throw Errors, use context = { err: [] } to get any errors
- can be used as type guard in TypeScript
match(value: any, context?: any): boolean
- returns true if the value matches the shape
- does not inject defaults,
- does not throw Errors, use context = { err: [] } to get any errors
- can be used as type guard in TypeScript
toString()
- returns a short string describing this GubuShape instance
[Util.inspect.custom]()
- same as toString
spec()
- returns a declarative description of the shape

The shape description provided by spec can be passed to Gubu to generate a new separate shape instance (see the Shape Nodes section).

Many shapes can be fully serialized to JSON, but those with custom validator functions are not serializable in the current version.

A GubuShape can be used be used as part of new shape definition. They are intended to be composable.

Errors

API | Shapes | Errors | Custom Validation | Builders

Gubu will attempt a full validation of the input value, collect all errors, and throw an Error if any validation failed. The error object will be an instance of GubuError, which extends TypeError with the following extra properties:

gubu: A marker value, always true.
code: Top level error code, in this version, always 'shape'
desc(): Call this function to get a more detailed description of the error.

The error description returned by desc() has the properties:

name: Always 'GubuError'
code: Top level error code, in this version, always 'shape'
err: An array of ErrDesc objects (these are all the errors that occurred).
ctx: The context object (if any) passed to this GubuShape validation call.

The message string of GubuError is a human readable generic description of the validation failure (with one issue per line) that is usable in your own application as-is. You can use the ErrDesc object instead to create entirely custom messages. Custom messages for specific custom errors can also be defined (see below).

Gubu(Number)('abc') // throws an Error with message:
`
Validation failed for value "abc" because the value is not of type number.'
`

let shape = Gubu({ 
  top: { 
    foo: String, 
    bar: Number 
  }
})

shape({ top: { foo: 123, bar: 'abc' }}) // throws an Error with message:
`
Validation failed for property "top.foo" with value "123" because the value is not of type string.
Validation failed for property "top.bar" with value "abc" because the value is not of type number.
`

ErrDesc Object

The ErrDesc object is the internal representation of an error, containing the full details of the error, which you can use for customization. The properties are:

k: string : Key of failing value (or empty string at top level).
n: Node : Failing shape node.
v: any : Failing value.
p: string : Key path to value.
w: string : Error code ("why").
m: number : Unique error mark for debugging (search in source code of gubu.ts).
t: string : Error message text.
u: any : User custom info.

The property n is the shape node whose validation failed.

try {
    Gubu({x: Number})({x: 'abc'})
}
catch(gubuError) {
  gubuError.desc()) === {
    name: 'GubuError',
    code: 'shape',
    err: [
      {
        k: undefined,
        n: { 
          '$': { 'gubu$': Symbol(gubu$), 'v$': '<VERSION>' },
          t: 'number',
          v: 0,
          r: true,
          o: false,
          d: 0,
          u: {},
          a: [],
          b: []
        },
        v: 'x',
        p: '',
        w: 'type',
        m: 1050,
        t: 'Validation failed for property "" with value "x" because the value is not of type number.',
        u: {},
      }
    ],
    ctx: {}
  }
}

From this description, you can determine that:

There was one err: err.length === 1
A type constraint failed: err[0].w === 'type'
The value was required: err[0].n.r === true
The value should be a number: err[0].n.t === 'number'
The value occurred at the top level: err[0].p = ''

The path of an error is the chain of properties that you follow to reach the failing value. For example, the path of the value of c in { a: { b: { c: 123 } }} is a.b.c.

In the case of arrays, the index is used as the property value. Thus, the path of the second element (=== 'y') of { a: ['x','y'] } is a.1. Paths do not used [] notation for arrays.

Values are converted to strings for the error message by using JSON.stringify. Circular values are handled safely. Long values are truncated to 30 characters.

The mark value (property m) is a numeric code that uniquely identifies the generation point of the error in the source code of gubu.ts, and should be quoted in bug reports (or indeed you can use it yourself to inspect the source code).

Error Collection

Instead of throwing validation errors, you can collect them using the reserved property err in the context argument:

let ctx = { err: [] }
Gubu(Number)('abc', ctx)  // does not throw
// ctx.err now contains an array of ErrDesc objects

The return value from Gubu in this case (and the value passed in!) should be considered corrupted (defaults may only be partially applied). If you want to retain the original value, you must clone it yourself before passing it to Gubu ⁵.

You can also set the context err property to false. In this case errors are not collected at all, and they are ignored, so that the full shape depth is always validated. The GubuShape.spec method uses this feature to generate a normalized validation Node hierarchy against the undefined value.

Custom Errors

When using a custom validator you can provide a custom error message using the Update.err property.

let shape = Gubu({ a: (value, update) => {
  update.err = 'BAD VALUE $VALUE AT $PATH'
  return false // always fails
})
shape({ a: 3 }) // throws "BAD VALUE 3 AT a"

Where $VALUE and $PATH are replaced by the value and path to the value, respectively.

TypeScript Types

Gubu makes a best-effort to support TypeScript types. The intersection of the type of the shape and the type of the value is used as the return type. This almost always does what you want, especially with optional default values (from which types will be inferred).

The GubuShape function also contains a property function valid with form:

valid(value: any, context?: any): boolean

This can be used as a type guard:

const shape = Gubu({ x: 1, y: 'Y' })
let data = { x: 2 }

if (shape.valid(data)) {
  console.log(data) // prints { x: 2, y: 'Y', z: true }
  console.log(data.x) // no type errors; prints 2
  console.log(data.y) // no type errors; prints 'Y'
  console.log(data.q) // type error! does not compile
}

The valid function does not throw, but you can optionally collect errors in the usual way with:

...
context = { err: [] }
if (shape.valid(data, context)) {
  ...
}

// failed
else {
  // context.err has the errors!
}

Where TypeScript cannot infer your types properly, you'll need to manually define them:

let shape = Gubu(Open({ x: 1}) as unknown as { x: number })
let data = { z: true }

if (shape.valid(data)) {
  console.log(data) // prints{ x: 1, z: true })
  console.log(data.x) // no type errors; prints 1
  console.log(data.z) // no type errors; prints true 
  console.log(data.q) // type error! does not compile
}

The holy grail would be for Gubu to use your type definitions directly:

interface User {
  name: string
  age: number
  job?: string
}

// DOES NOT WORK!
let shape = Gubu(User)

Sadly TypeScript does not provide runtime type information at present—it should!

If you're really keen on being ultra-DRY, and really want to avoid duplicating type definitions into almost, but not quite, the same shape definitions, here are your options:

Create an instance of your type, and use that as the shape definition:

Class Car {
  // These defaults become the shape definitions
  make: string = ''
  model: string = ''
}
const shape = Gubu({ ...new Car() })

Use code generation. Perhaps you are already building types from a SQL Schema? Use the same approach to build the shapes.
Parse the d.ts type definitions at runtime use those to dynamically define your shapes.

None of these options are that great. For moment, I recommend that you use the instance trick above if you can (option 1), and live with some manual fix up.

One more thing: at the moment I don't plan to support definitions in the other direction, going from shapes to TypeScript types. That would just be building a bad copy of the TypeScript type system using different syntax. That said, never say never, and if TypeScript inference can support it, I may look at it again.

Shape Nodes

The data structure returned by GubuShape.spec() is the internal representation of the validation shape. This is a hierarchical data structure where the validation for each key-value pair is defined by a shape Node, which has the following structure:

$: typeof GUBU : Special marker to indicate a Gubu Node object.
t: ValType : Value type name (see below).
d: number : Depth of the object tree.
v: any : Default value.
r: boolean : Value is required.
p: boolean : Value is skippable (if key is absent, no default is injected).
n: number : Number of keys in default value.
c: any : Default child shape (for array elements and open objects).
u: Record<string, any> : Custom user meta data.
b: Validate[] : Custom before-validation functions.
a: Validate[] : Custom after-validation functions.
s?: string : Custom stringification of the value (mostly for error messages).

The ValType is string with exactly one of these values:

'any' : Any type.
'array' : An array.
'bigint' : A BigInt value.
'boolean' : The values true or false.
'custom' : Custom type defined by a validation function.
'function' : A function.
'instance' : An instance of a constructed object.
'list' : A list of types under a given logical rule.
'nan' : The NaN value.
'never' : No type.
'null' : The null value.
'number' : A number.
'object' : A plain object.
'string' : A string (but not the empty string).
'symbol' : A symbol reference.
'undefined' : The undefined value.

This structure is deliberately terse (hence the one character property names) to make eye-balling deep structure debugging print-outs easier.

As noted above, in the current version this structure is only fully serializable to JSON if there are no custom validations, and the custom user meta data is serializable.

This structure can be accessed in custom validators via the state.node parameter, and in shape builders via the before and after hook functions. It is also provided in error messages under the n property.

Shape Builder Usage

The validation rules for each value shape can be modified using shape builders. These are wrapping functions that add additional constraints to the value shape.

For example, the Required shape builder marks a value as required. This is most useful for objects and array, which are by default optional:

const { Gubu, Required } = require('gubu') // shaper builders are exported
let easier = Gubu({ x: 1 })
let stricter = Gubu(Required({ x: 1 }))

easier() // returns { x: 1 }

stricter() // fails
stricter({}) // returns { x: 1 } (x itself is an optional default)

Most shape builders can also be chained. For example, the Open shape builder allows additional properties to be added to an object. To also make the object required you can use either of these expressions:

const { Required, Open } = Gubu // shape builders are also properties of Gubu
Gubu(Open({ a: 1, b: 2 }).Required())
Gubu(Required({ a: 1, b: 2 }).Open())

Most shape builders can be composed (check their expected arguments!), so the following are also equivalent:

Gubu(Open(Required({ a: 1, b: 2 })))
Gubu(Required(Open({ a: 1, b: 2 })))

This flexibility allows you to adjust shapes without too much refactoring or "schema noise".

Shape Builder Reference

_^top

API | Shapes | Errors | Custom Validation | Shape Rules

The built-in shape builders are:

Above: Match a value (or length of value) greater than the given amount.
After: Define a custom validation function called after a value is processed.
All: All shapes must match value.
Any: This shape will match any value.
Before: Define a custom validation function called before a value is processed.
Below: Match a value (or length of value) less than the given amount.
Check: Check value with a custom validation function or regular expression.
Child: All non-explicit child values of an object must match this shape.
Closed: Allow only explicitly defined elements in an array.
Default: Specify a default value.
Define: Define a name for a value.
Empty: Allow string values to be empty.
Exact: The value must one of an exact list of values.
Func: The value is explicitly a function.
Key: The key (or path) of the current object is injected as the value.
Max: Match a value (or length of value) less than or equal to the given amount.
Min: Match a value (or length of value) greater than or equal to the given amount.
Never: This shape will never match anything.
Len: Match a value (or length of value) exactly equal to the given amount.
One: Exactly one shape (and no more) must match value.
Open: Allow arbitrary properties in an object.
Optional: Make a value optional.
Skip: Make a value explicitly optional (no default created).
Refer: Refer to a defined value by name.
Rename: Rename the key of a property.
Required: Make a value required.
Some: Some shapes (at least one) must match value.

gubu

Gubu: An object shape validation utility.

Motivation

Quick Example

Common Use Cases

Option defaults and validation

Deep structures

Shape Building

Install

Usage

TypeScript

Browser

Shape Rules

Shape Builders

Regular Expressions

Recursive Shapes

API

Shape

Required Values

Optional Values with Defaults

Objects

Required Properties (Object)

Open Objects

Optional Objects

Object Values

Arrays

Required Properties (Array)

Array Properties

Length Constraints

Functions

Key Expressions

Custom Validation

Gubu function

GubuShape function

Errors

ErrDesc Object

Error Collection

Custom Errors

TypeScript Types

Shape Nodes

Shape Builder Usage

Shape Builder Reference

Above Builder

After Builder

All Builder

Any Builder

Before Builder

Below Builder

Check Builder

Closed Builder

Define Builder

Empty Builder

Exact Builder

Max Builder

Min Builder

Min Builder Example

Min Builder Behaviour

Len Builder

Never Builder

One Builder

Skip Builder

Refer Builder

Rename Builder

Required Builder

Optional Builder

Default Builder

Some Builder

Value Builder

Child Builder

Func Builder

Key Builder

Custom Builders

Edge Cases

Empty Strings

Implementation

Contributing

Credits

GUBU

License

Footnotes