Simple callback-based JSON validator for complex use-cases
npm install nonvalidnonvalid is a validation library that relieves you of necessity to traverse data structures manually but keeps your validation logic completely custom. The workflow is as follows: you describe the expected shape of your data and provide a callback for each piece of it that you’d like to validate. If any of the callbacks return an error (that is, any truthy value), this becomes the result of the validation. Otherwise, false is returned, which means that there is no error and the validation is successful.
nonvalid is suited best for deep and self-referential validation. Want to check that the object supplied to your script contains only values that are either string arrays or names of other keys in this object? Need to verify that a structure of nested objects correctly describes a family tree? nonvalid is your guy!
nonvalid supports validating values of all JavaScript data types, including symbols and bigints. It does not, however, support things like Map or Set.
- Installation
- Basics
- Validator instances
- Matchers
- Automatic traversal
- Shape validation
- Paths
- Regular navigation
- Safe navigation
- Full API reference
If you will be using nonvalid as part of a Node.js project, install the library with npm i nonvalid. You can import nonvalid in one of the following ways:
CommonJS:
``js`
const nonvalid = require('nonvalid');
ES6 Modules:
`js`
import nonvalid from 'nonvalid';
Browser:
`html`
Generally, you call nonvalid with two arguments: the data you are validating and the schema that you’re validating it against. The simplest way to use nonvalid is to compare two values literally:
`js`
nonvalid({ value: 42 }, { value: 42 }); // returns false, as there is no error
`js`
nonvalid({ value: '42' }, { value: 42 }); // returns true, as there is a type mismatch
As the old saying goes, all valid values are alike, but each invalid one is invalid in its own way. This is why nonvalid represents validity as just false; validation errors, on the other hand, are allowed to assume any truthy value that best reflects the exact nature of the error.
Consider the following piece of code:
`js`
nonvalid({ value: 42.42 }, { value: v => {
if (typeof v !== 'number') {
return 'Value must be a number';
}
if (v < 42) {
return 'Value must not be less than 42';
}
if (v >= 43) {
return 'Value must not be much greater than 42';
}
} });
This call will return false, but were we to pass { value: 41 } as the first argument, it would have returned the string 'Value must not be less than 42'.
If you prefer, you can rewrite the callback using short-circuit evaluation and a matcher:
`js`
nonvalid({ value: 42.42 }, { value: v =>
!nonvalid.number(v) && 'Value must be a number' ||
v < 42 && 'Value must not be less than 42' ||
v >= 43 && 'Value must not be much greater than 42'
});
Note that earlier, when we tried to validate the wrongly typed { value: '42' } against { value: 42 }, there was no callback for nonvalid to call, so it simply returned the generic true for an error.
If you are going to validate more than one value with nonvalid, you must create a new validator instance every time before performing validation. Use nonvalid.instance() for this:
`js
function validate(value) {
const nv = nonvalid.instance();
return nv(value, v => !nv.number(v) || v < 42 || v >= 43);
}
console.log(validate(42.42));
console.log(validate('foo'));
console.log(validate(41));
console.log(validate(43));
`
After creating a dedicated instance, you must call all matchers and other methods on this instance rather than on nonvalid itself.
Some of the most basic checks needed for careful data validation often look ugly. For instance, to make sure that a value is a proper object, you would write typeof value === 'object' && value !== null && !Array.isArray(value). To simplify things, nonvalid comes with a (very limited) set of validators of this kind, called matchers: number, string, boolean, null, undefined, defined, bigint, symbol, function, array, and, finally, object (which is implemented exactly as shown above). For full reference on matchers, see below.
To use a matcher, simply pass a value to it:
`js`
const nv = nonvalid.instance();
console.log(nv.object({ foo: 'bar' })); // true
console.log(nv.object('foo')); // false
console.log(nv.object(null)); // false
When calling a matcher during validation, you can choose to omit the argument you pass to it. In this case, the matcher will use the innermost value that you are currently inspecting:
`js`
const nv = nonvalid.instance();
console.log(nv({ value: { foo: 'bar' } }, { value: () => !nv.object() }));
// This is the same as v => !nv.object(v); validation returns false
`js`
const nv = nonvalid.instance();
console.log(nv({ value: 42 }, { value: () => !nv.object() }));
// Still the same as v => !nv.object(v); validation returns true
You can also add your own matchers with addMatcher:
`js`
const nv = nonvalid.instance();
nv.addMatcher('positive', n => nv.number(n) && n > 0);
console.log(nv({ value: 42 }, { value: () => !nv.positive() })); // false
Note that in order to be made a matcher, a function has to accept exactly one parameter. However, after it has become a matcher, we can call it with no arguments during validation—just as we did when dealing with built-in matchers. Similarly, custom matchers will automagically be capable of safe navigation.
nonvalid automatically traverses objects and arrays and compares values residing in their corresponding keys/indices. When in the schema (the right-hand argument to nonvalid) a function is encountered, it is called with the corresponding left-hand value as an argument. The result the function returns is then treated as an error of validation: falsy values mean there was no error; truthy values mean there _is_ an error, and it is immediately propagated to the very top and returned as the outcome of the whole validation routine.
`js`
nv({ type: 'dragon', position: [12.3, 25.1] }, {
type: v => !['dragon', 'zombie'].includes(v) && 'Unknown character type',
position: [
v => !nv.number(v) && 'The X coordinate must be a number',
v => !nv.number(v) && 'The Y coordinate must be a number'
]
});
The cool thing about traversing data structures this way is that you can use recursion whenever there is a need of doing so. For example, if we want to allow null values for position, we might write:
`js`
nv({ type: 'dragon', position: [12.3, 25.1] }, {
type: v => !['dragon', 'zombie'].includes(v) && 'Unknown character type',
position: v => v !== null && nv(v, [
v => !nv.number(v) && 'The X coordinate must be a number',
v => !nv.number(v) && 'The Y coordinate must be a number'
])
});
In fact, matchers and recursive calls don’t even require passing them the current value, so the following code will do the job as well:
`js`
nv({ type: 'dragon', position: [12.3, 25.1] }, {
type: v => !['dragon', 'zombie'].includes(v) && 'Unknown character type',
position: () => !nv.null() && nv([
() => !nv.number() && 'The X coordinate must be a number',
() => !nv.number() && 'The Y coordinate must be a number'
])
});
Note that not having position at all (as opposed to having position: null) would make the object invalid. Basically, missing position is the same as having position: undefined. If we want to allow this as well, we need to write position: () => nv.defined() && !nv.null() && ... in the schema.
Keep in mind that the order in which you place validation schema properties in the code is the order in which the checks will be executed (with some qualifications). This means that if you pass in a non-existing character type with non-numerical coordinates, the specific error you will get in return depends on whether you put the type checker or the position checker first when describing the schema object.
If nonvalid encounters a number where the validation schema expects an object or, say, there is an object where the schema wants an array, validation will return true indicating a generic error:
`js`
nv({ position: [12.3, 25.1] }, { position: {
x: () => !nv.number() && 'X must be a number',
y: () => !nv.number() && 'Y must be a number'
} }); // returns true because we have an array on the left but an object on the right
If you want to customize this error, provide a truthy value in the [[nv.error]](#nv-error) field of the object / after the nv.end auxiliary value in the array:
`js`
nv({ position: [12.3, 25.1] }, { position: {
x: () => !nv.number() && 'X must be a number',
y: () => !nv.number() && 'Y must be a number',
[nv.error]: 'Position must be an object'
} });
`js`
nv({ position: { x: 12.3, y: 25.1 } }, { position: [
() => !nv.number() && 'X must be a number',
() => !nv.number() && 'Y must be a number',
nv.end,
'Position must be an array'
] });
Often you would want your arrays or objects to contain any number of values if only they satisfy certain requirements. By default, nonvalid will return true for any extra values that the validation schema does not account for.
To allow keys other than listed in the schema for an object, use [[nv.other]](#nv-other) in conjunction with nv.key():
`jsCode for the ${nv.key()} color is invalid
nv.addMatcher('color', v => nv.string() && v.match(/^#[0-9a-f]{6}$/));
nv({ default: '#ffffff', red: '#ff0000', white: '#ffffff', black: '#000000' }, {
default: () => !nv.color() && 'You must specify a valid default color',
[nv.other]: () => !nv.color() &&
});
For arrays, put the callback after nv.end and use nv.index() if necessary:
`jsColor code at position ${nv.index()} is invalid
nv.addMatcher('color', v => nv.string() && v.match(/^#[0-9a-f]{6}$/));
nv(['#ffffff', '#ff0000', '#000000'], [
() => !nv.color() && 'You must specify a valid default color as the first item',
nv.end,
() => !nv.color() &&
]);
Of course, you can provide both this callback and a shape mismatch custom error (we omit here the definition of the color matcher):
`jsCode for the ${nv.key()} color is invalid
nv({ default: '#ffffff', red: '#ff0000', white: '#ffffff', black: '#000000' }, {
default: () => !nv.color() && 'You must specify a valid default color',
[nv.other]: () => !nv.color() && ,`
[nv.error]: 'The value must be an object'
});
`jsColor code at position ${nv.index()} is invalid
nv(['#ffffff', '#ff0000', '#000000'], [
() => !nv.color() && 'You must specify a valid default color as the first item',
nv.end,
() => !nv.color() && ,`
'The value must be an array'
]);
Having no mandatory elements (like the default value above) can be handled too:
`jsCode for the ${nv.key()} color is invalid
nv({ red: '#ff0000', white: '#ffffff', black: '#000000' }, {
[nv.other]: () => !nv.color() && ,`
[nv.error]: 'The value must be an object'
});
`jsColor code at position ${nv.index()} is invalid
nv(['#ff0000', '#000000'], [
nv.end,
() => !nv.color() && ,`
'The value must be an array'
]);
Note the nv.end in front of the validation array. It is important to put it there, as otherwise the callback will only apply to the element at index 0, and the element at index 1 will be compared to the string 'The value must be an array'. That’s not what you want.
By the way, in case you were wondering, nv.error, nv.end, and nv.other are all symbols.
At any point during traversal, you can request the path in the validation tree to the currently inspected element. Call to nv.path() will return an array of keys/indices that, when chained to the root element of the tree, will lead to the current value. By calling nv.path(name), you can get a textual (JS-like) representation of the path, assuming that the name of the root element is name.
`js`
nv({ users: [{ name: 'John Doe' }, { name: 'Richard Roe' }] }, { users: [nv.end, () => nv({
name: v => {
if (v.startsWith('John')) {
console.log(nv.path()); // ['users', 0, 'name']
}
if (v.startsWith('Richard')) {
console.log(nv.path('data')); // 'data["users"][1]["name"]'
}
}
})] });
Similarly, if the inspected value turned out to be invalid, the path to the element which triggered the validation error can be retrieved via nv.errorPath() or nv.errorPath(name):
`js`
nv({ names: ['John Doe', 'Richard Roe', null] }, { names: [nv.end, () => !nv.string()] });
console.log(nv.errorPath()); // ['names', 2]
console.log(nv.errorPath('props')); // 'props["names"][2]'
nonvalid supports three methods for navigating the inspected tree of objects and arrays. During validation, you can call nv.root() to get the topmost value, nv.value() to retrieve the current (innermost) value, and nv.up(n) to get n levels up the tree from the current value, starting with its immediate parent:
`js`
nv({ a: { b: { c: 'd' } } }, { a: { b: { c: () => {
console.log(nv.root()); // { a: { b: { c: 'd' } } }
console.log(nv.value()); // 'd'
console.log(nv.up()); // { c: 'd' }
console.log(nv.up(0)); // { c: 'd' }
console.log(nv.up(1)); // { b: { c: 'd' } }
console.log(nv.up(2)); // { a: { b: { c: 'd' } } }
} } } });
`js`
nv(['a', ['b', ['c', 'd']]], ['a', ['b', () => {
console.log(nv.root()); // ['a', ['b', ['c', 'd']]]
console.log(nv.value()); // ['c', 'd']
console.log(nv.up()); // ['b', ['c', 'd']]
console.log(nv.up(0)); // ['b', ['c', 'd']]
console.log(nv.up(1)); // ['a', ['b', ['c', 'd']]]
console.log(nv.up(2)); // exception
}]]);
The methods work correctly regardless of whether you’re using recursion for validation or not. However, navigation will start to act funny if you skip levels for some reason:
`js`
nv({ a: { b: { c: 'd' } } }, { a: v => {
return nv(v.b.c, () => {
console.log(nv.root()); // { a: { b: { c: 'd' } } }, as should be
console.log(nv.value()); // 'd', as should be
console.log(nv.up()); // { a: { b: { c: 'd' } } }, because two levels were skipped
console.log(nv.up(1)); // exception
});
} });
nonvalid provides a built-in way to perform optional chaining on a tree that is being validated. To use it, create a function that calls either nv.root(), nv.up(), or nv.value(), performs regular chaining on the received value, and returns its result. Pass this function as the argument to any matcher. The matcher will call the function and treat its returned value as if you have passed it to the matcher directly. If the chaining was unsuccessful, the value will be undefined.
Let’s consider the following validation scenario:
`js`
nv.addMatcher('color', v => nv.string() && v.match(/^#[0-9a-f]{6}$/));
nv({ shades: { red: '#ff0000', white: '#ffffff' }, color: 'red' }, {
shades: () => nv.defined() && nv({ [nv.other]: () => !nv.color() }),
color: v => !nv.color() && nv.undefined(() => nv.root().shades[v]) // safe navigation
});
So, we want the object to have a field named color which holds either a valid color code or a name of a color whose code is indexed in the shades nested object. But we don’t care too much about shades: it may as well be undefined as long as color is already a valid code. To achieve this, we use safe navigation by writing nv.undefined(() => nv.root().shades[v]). If we were to navigate simply by nv.undefined(nv.root().shades[v]), the code would throw an exception in the unfortunate case when shades are not present while color is not really a color code.
To facilitate safe navigation, there is an additional matcher for you, get. It is useful in that it allows you to directly get the value of a safe navigation chain:
`js`
nv({
redirects: {
server: { home: 'news', news: 'news/latest' },
},
page: 'home'
}, {
page: v => {
console.log(nv.get(() => nv.up().redirects.server[v])); // 'news'
console.log(nv.get(() => nv.up().redirects.client[v])); // undefined
console.log(nv.get(() => {
return nv.up().redirects.server[nv.up().redirects.server[v]];
})); // 'news/latest'
console.log(nv.get(() => {
return nv.up().redirects.client[nv.up().redirects.client[v]];
})); // undefined
console.log(nv.get(() => nv.value())); // 'home'
console.log(nv.get(() => nv.value().some.field)); // undefined
console.log(nv.get(() => v.some.field)); // exception
}
});
Notice that we’ve used nested safe navigation here: that is, the value of a safe navigation chain serves as an index/key for another chain. Also, the last line causes an exception, because the variable v is not a subject to safe navigation (only nv.value(), nv.up(), and nv.root() can be).
One small benefit of using nonvalid’s nested safe navigation is that you don’t need to worry about keys named 'undefined':
`js`
nv({
types: {
'string': 'the most straightforward type',
'boolean': 'just yay or nay',
'number': 'infinitely complicated',
'bigint': 'finitely complicated',
'object': 'somewhat curly',
'null': 'almost like an object',
'undefined': 'like the future of JS',
'symbol': 'not like anything else'
}
}, {
type: v => {
// v is undefined, because the object does not hold a key named 'type'
console.log(v); // undefined
console.log(nv.value()); // undefined
console.log(nv.get(() => nv.root().types[v])); // 'like the future of JS'
console.log(nv.get(() => nv.root().types[nv.value()])); // undefined
}
});
Magic!
Creates and returns a new validator instance. Use this method every time you need to validate a new value. See Validator instances.
#### nv(value, schema) or nv(schema)
If called as nv(value, schema), validates the value against the schema. If called as nv(schema), validates the currently inspected value against the schema (that is, the call is identical to nv(nv.value(), schema)). Validation occurs through automatic tree traversal, strict value comparison and running callbacks. Each callback receives two parameters: the current value (what would nv.value() return) and the current key (what would nv.key() or nv.index() return; undefined if not inside any object or array).
Validation traverses an object’s properties in the following order:
1. Validation schema keys that are non-negative integers (or rather strings that represent them), in ascending order.
2. Other validation schema keys that are strings, in the order in which they were defined.
3. Validation schema keys that are symbols, in the order in which they were defined.
4. The inspected object’s remaining keys that are non-negative integers (or rather strings that represent them), in ascending order.
5. The inspected object’s other remaining keys that are strings, in the order in which they were defined.
6. The inspected object’s remaining keys that are symbols, in the order in which they were defined.
If the inspected object doesn’t have a key encountered in steps 1, 2, or 3, the corresponding value is considered undefined. Keys/values encountered in steps 4, 5, 6, if any, are sent to the validation schema’s callback provided under the key [[nv.other]](#nv-other). If such a callback is not present, true is returned.
Validation traverses an array’s elements in the following order:
1. “Regular” validation schema elements (that is, elements before nv.end), in ascending order of their indices.
2. Remaining elements of the inspected array, in ascending order of their indices.
If the inspected array doesn’t contain an element at a position examined during stage 1, its value is considered undefined. Elements encountered in stage 2, if any, are sent to the validation schema’s callback provided after nv.end. If no such callback is specified, true is returned.
The method returns false if the value is valid (even if a callback has returned some other falsy value, like null); otherwise returns a truthy error. Can, and often should, be called recursively. See Automatic traversal.
#### nv.addMatcher(name, func) or nv.addMatcher(namedFunc)
Adds a function to the list of available matchers. The matcher can later be called with nv.matcherName() or nv.matcherName(v), where nv is the instance to which the matcher was added and matcherName represents the name of the matcher. If the name is already taken by another matcher or a method of the instance, the call to addMatcher will throw.
If called as nv.addMatcher(name, func), name becomes the name of the matcher (regardless of the name of func). name can be of type string or symbol.
If called as nv.addMatcher(namedFunc), the name of namedFunc becomes the name of the matcher. Don’t add matchers this way if you’re planning to minify your code, as minification may change or remove function names.
func (or namedFunc) must accept exactly one parameter. The function will be wrapped by nonvalid to allow usage described in the following paragraphs.
When during validation called as nv.matcherName() with no arguments, nonvalid will actually call the underlying function with the current value (that is, the call will be identical to nv.matcherName(nv.value())).
When during validation called as nv.matcherName(f), where f is a function distinct from the current value (f !== nv.value()), it will:
1. Switch the validator instance to a safe mode of execution. In this mode, calls to nv.value(), nv.root(), and nv.up() return safely wrapped values instead of the actual elements.
2. Call f in this mode.f
3. Switch back and unwrap the value that returned.func
4. Call the initial function, or namedFunc, passing it the unwrapped value.
See Safe navigation and Matchers.
#### nv.path() or nv.path(name)
Can be called during validation only. If called as nv.path(), returns a chain (array) of keys/indices that lead to the current value. Keys can be strings or symbols; indices are numbers. If current value is at the topmost level, returns an empty array. If called as nv.path(name) where name is a string, returns a formatted string that looks like a JS expression to retrieve the value from the topmost object/array, assuming name is its name. See Paths.
#### nv.errorPath() or nv.errorPath(name)
Can be called after validation only. If the inspected value was deemed valid or if the validation threw, returns null. Otherwise, the method returns exactly what nv.path() or nv.path(name) would have returned at the point in time when the validation error occurred. See Paths.
Can be called during validation only. If called while validating a value inside an object, returns this value’s key (a string or a symbol). Otherwise throws. See Shape validation.
Can be called during validation only. If called while validating an element of an array, returns this element’s index (a number). Otherwise throws. See Shape validation.
Can be called during validation only. Returns the (innermost) currently inspected value. See Regular navigation and Safe navigation.
Can be called during validation only. Returns the outermost object or array that the current value belongs to. If the current value is not inside any object or array, it throws.
See Regular navigation and Safe navigation.
Can be called during validation only. Goes n levels up the tree of objects/arrays that the current value is nested in. If called as nv.up(), n is considered to be zero. Zero levels mean the method should get the immediate parent of the current value; one level gets the parent of the parent, etc. If the current value is nested less than, or exactly, n levels deep in the tree, the method throws.
See Regular navigation and Safe navigation.
Can be a key of an object anywhere inside the validation schema. The value of this key must be a function. The function will be used as a callback for validating any keys (and respective values) of the inspected (sub)object that are not explicitly defined in the schema. See Shape validation.
Can be a key of an object anywhere inside the validation schema. The value of this key must be truthy and cannot be a function. It defines what will be returned as an error if the inspected (sub)value is not an object. See Shape validation.
Can be an element of an array anywhere inside the validation schema. At most two elements can follow this value in the array. If one of them is a function, it will be used as a callback for validating any values of the inspected (sub)array whose position in the array is after the last “regular” schema value (that is, starting where nv.end is in the schema). If one of the values that follow nv.end in the schema is truthy and not a function, it will be returned as an error if the corresponding inspected (sub)value is not an array. See Shape validation.
nv.end is interchangeable with an empty slot in an array literal: if you prefer, you can write [3, 2, 1,, v => v !== 0] instead of [3, 2, 1, nv.end, v => v !== 0]; or [, 'Value must be an array'] instead of [nv.end, 'Value must be an array'].
Built-in matchers are simple functions that work as though they were added via addMatcher. See the entry on this method for details about the inner workings of matchers.
Also see Matchers.
Returns true if v is of type number and is not Infinity, -Infinity, or NaN. Otherwise returns false.
Returns true if v is of type string. Otherwise returns false.
Returns true if v is of type boolean (i.e., it’s either true or false). Otherwise returns false.
Returns true if v is null. Otherwise returns false.
Returns true if v is undefined. Otherwise returns false.
Returns true if v is not undefined. Otherwise returns false.
Returns true if v is of type bigint. Otherwise returns false.
Returns true if v is of type symbol. Otherwise returns false.
Returns true if v is a function. Otherwise returns false.
Returns true if v is an array. Otherwise returns false.
Returns true if v is a proper object (i.e., of type object, but not an array or function). Otherwise returns false.
Returns v itself. Can be used for safe navigation. You can also call nv.get() as a non-safe alias for nv.value()`.