JSON & Data

JSON is a language-independent text format for structured data. It allows objects with double-quoted string keys, arrays, strings, finite numbers, true, false, and null. It cannot represent functions, undefined, symbols, BigInt, dates, or comments, and every key and string must use double quotes.

JSON.parse turns a JSON string into the corresponding JavaScript value (object, array, string, number, boolean, or null). If the text is not valid JSON — wrong quotes, a trailing comma, undefined, a stray token — it throws a SyntaxError rather than returning null, so calls on untrusted input should be wrapped in try/catch.

JSON.stringify serializes a value to a JSON string. In objects it silently drops keys whose value is a function, undefined, or a symbol; in arrays those become null. NaN and Infinity serialize to null. A top-level undefined or function yields undefined (no string), not the text 'undefined'.

Serializing to JSON and parsing back produces a deep copy with no shared references for plain data. But it only round-trips JSON-representable values: functions, undefined, and symbols are dropped, Date becomes a string, Map/Set become {}, NaN/Infinity become null, BigInt throws, and a circular reference throws a TypeError.

eval runs an arbitrary string as code with the caller's privileges, so untrusted input becomes a code-injection hole. It also defeats engine optimization, can read and write surrounding scope, and makes code hard to reason about. To turn text into data use JSON.parse; to look up dynamic keys use bracket notation instead.

JSON cannot encode several JavaScript values. A circular reference makes JSON.stringify throw a TypeError, and so does a BigInt. undefined, functions, and symbols are dropped as object keys and become null in arrays. NaN and Infinity serialize to null. Date has no native type and is stringified via its toJSON.

JSON.stringify(value, replacer, space) lets you customize output. A replacer function transforms or drops each key/value pair; a replacer array whitelists which keys survive. If a value has a toJSON() method, its return value is serialized instead. The third space argument (a number or string) pretty-prints with that indentation.

structuredClone deep-copies circular graphs, Date, RegExp, Map, Set, and typed arrays, but throws on functions and DOM nodes. The JSON round-trip is simpler yet lossy: it drops functions, breaks on cycles and BigInt, and degrades Date/Map. A manual clone handles anything — including functions — but you must write and maintain that logic yourself.

JSON.parse(text, reviver) passes a function that is called for every key/value pair, bottom-up, after parsing. Its return value replaces the parsed value; returning undefined deletes that key. It is typically used to transform raw values back into rich types — for example turning ISO date strings into real Date objects during parsing.

Define a tagged schema: on serialization, a toJSON or a replacer converts each rich value into a plain JSON object carrying a type marker and a serializable payload, e.g. { __type: 'Map', value: [...] }. On parse, a reviver matches that marker and reconstructs the original type. The marker convention is the contract both sides share to survive the lossy JSON layer.