JSON Schema Validator

This validator implements a recursive descent algorithm over the JSON Schema keyword set from Draft-07. Given a schema object and a data value, it evaluates every applicable keyword and collects all errors, rather than stopping at the first failure. Error paths are accumulated by passing the current path string (root, root.propertyName, root.arrayName[index]) down through recursive calls.

### Supported keywords by category

Type and value identity
- type — accepts a string or array of strings; "integer" is treated as a subtype of "number" and additionally requires Number.isInteger(value) to be true
- enum — value must deep-equal one of the listed items (compared via JSON.stringify)
- const — value must deep-equal the specified constant

String constraints
- minLength, maxLength — checked against string.length (character count)
- pattern — tested with new RegExp(pattern).test(value)

Number constraints
- minimum, maximum — inclusive bounds
- exclusiveMinimum, exclusiveMaximum — exclusive bounds (Draft-07 number form)
- multipleOf — checked via Number.isInteger(value / multipleOf)

Array constraints
- items — if an object schema, applied to every element; if an array of schemas, applied positionally
- minItems, maxItems — checked against array.length
- uniqueItems — all elements compared via JSON.stringify; first duplicate reported with its index
- contains — at least one array element must match the sub-schema

Object constraints
- required — each listed key must be present in the object
- properties — each defined property is validated recursively against its sub-schema if present
- additionalProperties — if false, any key not in properties is rejected; if an object schema, applied to all keys not in properties
- patternProperties — each key matching the pattern regexp is validated against the associated sub-schema
- minProperties, maxProperties — checked against Object.keys(object).length

Logical combinators
- allOf — data must pass every sub-schema in the array
- anyOf — data must pass at least one sub-schema
- oneOf — data must pass exactly one sub-schema
- not — data must fail the sub-schema

### Working example

Schema — requires an object with a username (string, 3–20 characters) and a role (one of "admin", "editor", "viewer"):

json {   "type": "object",   "required": ["username", "role"],   "properties": {     "username": { "type": "string", "minLength": 3, "maxLength": 20 },     "role": { "type": "string", "enum": ["admin", "editor", "viewer"] }   },   "additionalProperties": false } 

Data that fails — username is too short and role is not in the enum:

json { "username": "ab", "role": "superuser" } 

Errors produced:

root.username: string length 2 < minLength 3 root.role: value must be one of "admin", "editor", "viewer"

### Scope and limitations

This tool does not support:
- $ref — schema references are not resolved; the keyword is silently ignored
- $defs / definitions — referenced sub-schemas via $ref are not inlined
- format — the keyword is accepted but not validated
- if / then / else — conditional application keywords from Draft-07 are not implemented
- Schema vocabularies, $schema declaration enforcement, or annotation collection

For production-grade validation with full Draft-07 support including $ref resolution, use Ajv in your codebase.