GeezerVison

This module is ESM 🔆. Please read this.

📖  Full documentation & guides:  onury.io/notation

A utility for reading, modifying, and filtering the contents of JavaScript objects and arrays — using object/bracket notation strings or glob patterns.

Notation.create({ x: 1 }).set('some.prop', true).filter(['*.prop']).value
// { some: { prop: true } }

• Usage
• Notation
• Glob Notation
• Filtering Data with Glob patterns
• Object and Bracket Notation Syntax
• Globs and Data Integrity
• Source Object Mutation
• Documentation
• Quality

Install via NPM:

npm i notation

import { Notation } from 'notation';

Notation is a class for modifying or inspecting the contents (property keys and values) of a data object or array.

When reading or inspecting an enumerable property value such as obj.very.deep.prop; with pure JS, you would have to do several checks:

if (obj 
        && obj.hasOwnProperty('very') 
        && obj.very.hasOwnProperty('deep')  
        && obj.very.deep.hasOwnProperty('prop')
    ) {
    return obj.very.deep.prop === undefined ? defaultValue : obj.very.deep.prop;
}

With Notation, you can do this:

const notate = Notation.create;
return notate(obj).get('very.deep.prop', defaultValue);

You can also inspect & get the value:

console.log(notate(obj).inspectGet('very.deep.prop'));
// {
//     notation: 'very.deep.prop',
//     has: true,
//     value: 'some value',
//     type: 'string',
//     level: 3,
//     lastNote: 'prop'
// }

To modify or build a data object:

const notate = Notation.create;
const obj = { car: { brand: "Dodge", model: "Charger" }, dog: { breed: "Akita" } };
notate(obj)                          // initialize. equivalent to `new Notation(obj)`
    .set('car.color', 'red')         // { car: { brand: "Dodge", model: "Charger", color: "red" }, dog: { breed: "Akita" } }
    .remove('car.model')             // { car: { brand: "Dodge", color: "red" }, dog: { breed: "Akita" } }
    .filter(['*', '!car'])           // { dog: { breed: "Akita" } } // equivalent to .filter(['dog'])
    .flatten()                       // { "dog.breed": "Akita" }
    .expand()                        // { dog: { breed: "Akita" } }
    .merge({ 'dog.color': 'white' }) // { dog: { breed: "Akita", color: "white" } }
    .copyFrom(other, 'boat.name')    // { dog: { breed: "Akita", color: "white" }, boat: { name: "Mojo" } }
    .rename('boat.name', 'dog.name') // { dog: { breed: "Akita", color: "white", name: "Mojo" } }
    .value;                          // result object ^

See the documentation for more...

With a glob-notation, you can use wildcard stars * and bang ! prefix. A wildcard star will include all the properties at that level and a bang prefix negates that notation for exclusion.

• Only Notation#filter() method accepts glob notations. Regular notations (without any wildcard * or ! prefix) should be used with all other members of the Notation class.
• For raw Glob operations, you can use the NotationGlob class.

Removes duplicates, redundant items and logically sorts the array:

import { NotationGlob } from 'notation';

const globs = ['*', '!id', 'name', 'car.model', '!car.*', 'id', 'name', 'age'];
console.log(NotationGlob.normalize(globs));
// ——» ['*', '!car.*', '!id', 'car.model']

In the normalized result ['*', '!car.*', '!id', 'car.model']:

• id is removed and !id (negated version) is kept. (In normalization, negated always wins over the positive, if both are same).
• Duplicate glob, name is removed. The remaining name is also removed bec. * renders it redundant; which covers all possible notations.
• (In non-restrictive mode) car.model is kept (although * matches it) bec. it's explicitly defined while we have a negated glob that also matches it: !car.*.

console.log(NotationGlob.normalize(globs, { restrictive: true }));
// ——» ['*', '!car.*', '!id']

• In restrictive mode, negated removes every match.

Unites two glob arrays optimistically and sorts the result array logically:

const globsA = ['*', '!car.model', 'car.brand', '!*.age'];
const globsB = ['car.model', 'user.age', 'user.name'];
const union = NotationGlob.union(globsA, globsB); 
console.log(union);
// ——» ['*', '!*.age', 'user.age']

In the united result ['*', '!*.age', 'user.age']:

• (negated) !car.model of globsA is removed because globsB has the exact positive version of it. (In union, positive wins over the negated, if both are same.)
• But then, car.model is redundant and removed bec. we have * wildcard, which covers all possible non-negated notations.
• Same applies to other redundant globs except user.age bec. we have a !*.age in globsA, which matches user.age. So both are kept in the final array.

When filtering a data object with a globs array; properties that are explicitly defined with globs or implied with wildcards, will be included. Any matching negated-pattern will be excluded. The resulting object is created from scratch without mutating the original.

const data = {
    car: {
        brand: 'Ford',
        model: 'Mustang',
        age: 52
    },
    user: {
        name: 'John',
        age: 40
    }
};
const globs = ['*', '!*.age', 'user.age'];
const filtered = Notation.create(data).filter(globs).value;
console.log(filtered);
// ——»
// {
//     car: {
//         brand: 'Ford',
//         model: 'Mustang'
//     },
//     user: {
//         name: 'John',
//         age: 40
//     }
// }

In non-restrictive mode; even though we have the !*.age negated glob; user.age is still included in the result because it's explicitly defined.

But you can also do restrictive filtering. Let's take the same example:

const globs = ['*', '!*.age', 'user.age'];
const filtered = Notation.create(data).filter(globs, { restrictive: true }).value;
console.log(filtered);
// ——»
// {
//     car: {
//         brand: 'Ford',
//         model: 'Mustang'
//     },
//     user: {
//         name: 'John'
//     }
// }

Note that in restrictive mode, user.age is removed this time; due to !*.age pattern.

Each note (level) of a notation is validated against EcmaScript variable syntax, array index notation and object bracket notation.

• x[y], x.1, x.y-z, x.@ are incorrect and will never match.
• x["y"], x['1'], x["y-z"], x['@'] are correct object bracket notations.

• [0].x indicates x property of the first item of the root array.
• x[1] indicates second item of x property of the root object.

• * is valid wildcard for glob notation. Indicates all properties of an object.
• [*] is valid wildcard for glob notation. Indicates all items of an array.
• x[*] is valid wildcard for glob notation. Indicates all items of x property which should be an array.
• x['*'] just indicates a property/key (star), not a wildcard. Valid regular notation.
• x.* is valid wildcard for glob notation.
• x, x.* and x.*.* (and so on) are all equivalent globs. All normalize to x.
• Negated versions are NOT equivalent.
  • !x indicates removal of x.
  • !x.* only indicates removal of all first-level properties of x but not itself (empty object).
  • !x.*.* only indicates removal of all second-level properties of x; but not itself and its first-level properties (x.*).
  • Same rule applies for bracket notation or mixed notations.
    • [0] = [0][*] but ![0] ≠ ![0][*]
    • x = x[*] but !x ≠ !x[*]
    • [*] = [*].* but ![*] ≠ ![*].*

Below, we filter to;

• keep all properties of the source object,
• remove the second item of colors property (which is an array),
• and empty my-colors property (which is an object).

const source = {
    name: 'Jack',
    colors: ['blue', 'green', 'red'],
    'my-colors': { '1': 'yellow' }     // non-standard name "my-colors"
};
const globs = ['*', '!colors[1]', '!["my-colors"].*'];
console.log(Notation.create(source).filter(globs).value);
// —» 
// {
//     name: 'Jack',
//     colors: ['blue', 'red'],
//     'my-colors': {}
// }

In the example above, colors item at index 1 is emptied.

In a glob list, you cannot have both object and array notations for root level. The root level implies the source type which is either an object or array; never both.

For example, ['[*]', '!x.y'] will throw because when you filter a source array with this glob list; !x.y will never match since the root x indicates an object property (e.g. source.x).

Each glob you use should conform with the given source object.

For example:

const obj = { x: { y: 1 } };
const globs = ['*', '!x.*'];
console.log(Notation.create(obj).filter(globs).value);
// ——» { x: {} }

Here, we used !x.* negated glob to remove all the properties of x but not itself. So the result object has an x property with an empty object as its value. All good.

But in the source object; if the actual value of x is not an object, using the same glob list would throw:

const obj = { x: 1 }; // x is number
const globs = ['*', '!x.*'];
console.log(Notation.create(obj).filter(globs).value);
// ——» ERROR

This kind of type mismatch is critical so it will throw. The value 1 is a Number not an object, so it cannot be emptied with !x.*. (But we could have removed it instead, with glob !x.)

The source object or array will be mutated by default (except the #filter() method). To prevent mutation; you can call #clone() method before calling any method that modifies the object. The source object will be cloned deeply.

const notate = Notation.create;

const mutated = notate(source1).set('newProp', true).value;
console.log(source1.newProp); // ——» true

const cloned = notate(source2).clone().set('newProp', true).value;
console.log('newProp' in source2); // ——» false
console.log(cloned.newProp); // ——» true

Read the full documentation — guides, concepts and the API reference. For what changed in v3, see the change log.

Read the CHANGELOG.

• 100% test coverage (statements, branches, functions, lines) — enforced via Vitest thresholds. Run npm run cover.
• ~84% mutation score via StrykerJS. Run npm run mutation.

Mutation testing goes beyond coverage: it makes hundreds of small edits ("mutants") to the source — flipping > to >=, && to ||, returning undefined, etc. — and checks that a test fails for each. It catches tests that execute code without actually asserting its behavior (the trap where function getPositive(x){ return x } reaches 100% coverage but verifies nothing). Most of the surviving mutants here are equivalent mutants in the glob normalization/cover/union logic — redundant-but-harmless branches that produce identical results — plus environment-defensive guards. These can't be killed by definition, so the realistic, healthy target is a high score, not 100%.

• accesscontrol — Role and attribute based access control (RBAC + ABAC) with conditions, enforced ownership, custom actions and mandatory gates.
• configuard — Turn flat config rows from a database table into a nested, typed configuration object — with ${...} templating and accessor-based (ABAC) filtering.

© 2026, Onur Yıldırım. MIT License.