CircularJSON ============ [![donate](https://img.shields.io/badge/$-donate-ff69b4.svg?maxAge=2592000&style=flat)](https://github.com/WebReflection/donate) ![Downloads](https://img.shields.io/npm/dm/circular-json.svg) [![Build Status](https://travis-ci.org/WebReflection/circular-json.svg?branch=master)](https://travis-ci.org/WebReflection/circular-json) [![Coverage Status](https://coveralls.io/repos/github/WebReflection/circular-json/badge.svg?branch=master)](https://coveralls.io/github/WebReflection/circular-json?branch=master) Serializes and deserializes otherwise valid JSON objects containing circular references into and from a specialized JSON format. - - - ## Check out [flatted](https://github.com/WebReflection/flatted#flatted) too !!! The future engine of CircularJSON is represented by [flatted](https://github.com/WebReflection/flatted#flatted) * smaller output, compared to every other alternative * fast, and usually much faster, than CircularJSON, on IE/Edge browser too * smaller in size, easier to maintain, 30 LOC proof of concept The next release of CircularJSON will be based on `flatted` representation which is, as the term indicates, a flat array of references. That means that produced and imported strings will **not be compatible** with CircularJSON one. If you care about improving serialization of your data, and you don't need extra JSON API arguments, try to see if `flatted` works already out of the box. - - - ### A Working Solution To A Common Problem A usage example: ```JavaScript var object = {}; object.arr = [ object, object ]; object.arr.push(object.arr); object.obj = object; var serialized = CircularJSON.stringify(object); // '{"arr":["~","~","~arr"],"obj":"~"}' // NOTE: CircularJSON DOES NOT parse JS // it handles receiver and reviver callbacks var unserialized = CircularJSON.parse(serialized); // { arr: [ [Circular], [Circular] ], // obj: [Circular] } unserialized.obj === unserialized; unserialized.arr[0] === unserialized; unserialized.arr.pop() === unserialized.arr; ``` A quick summary: * **new** in version `0.5`, you can specify a JSON parser different from JSON itself. `CircularJSON.parser = ABetterJSON;` is all you need. * uses `~` as a special prefix symbol to denote which parent the reference belongs to (i.e. `~root~child1~child2`) * reasonably fast in both serialization and deserialization * compact serialization for easier and slimmer transportation across environments * [tested and covered](test/circular-json.js) over nasty structures too * compatible with all JavaScript engines Node Installation & Usage ============ ```bash npm install --save circular-json ``` ```javascript 'use strict'; var CircularJSON = require('circular-json'), obj = { foo: 'bar' }, str ; obj.self = obj; str = CircularJSON.stringify(obj); ``` There are no dependencies. Browser Installation & Usage ================ * Global: * AMD: * CommonJS: (generated via [gitstrap](https://github.com/WebReflection/gitstrap)) ```html ``` ```javascript 'use strict'; var CircularJSON = window.CircularJSON , obj = { foo: 'bar' } , str ; obj.self = obj; str = CircularJSON.stringify(obj); ``` NOTE: Platforms without native JSON (i.e. MSIE <= 8) requires `json3.js` or similar. It is also *a bad idea* to `CircularJSON.parse(JSON.stringify(object))` because of those manipulation used in `CircularJSON.stringify()` able to make parsing safe and secure. As summary: `CircularJSON.parse(CircularJSON.stringify(object))` is the way to go, same is for `JSON.parse(JSON.stringify(object))`. API === It's the same as native JSON, except the fourth parameter `placeholder`, which circular references to be replaced with `"[Circular]"` (i.e. for logging). * CircularJSON.stringify(object, replacer, spacer, placeholder) * CircularJSON.parse(string, reviver) Bear in mind `JSON.parse(CircularJSON.stringify(object))` will work but not produce the expected output. Similar Libraries ======= ### Why Not the [@izs](https://twitter.com/izs) One The module [json-stringify-safe](https://github.com/isaacs/json-stringify-safe) seems to be for `console.log()` but it's completely pointless for `JSON.parse()`, being latter one unable to retrieve back the initial structure. Here an example: ```JavaScript // a logged object with circular references { "circularRef": "[Circular]", "list": [ "[Circular]", "[Circular]" ] } // what do we do with above output ? ``` Just type this in your `node` console: `var o = {}; o.a = o; console.log(o);`. The output will be `{ a: [Circular] }` ... good, but that ain't really solving the problem. However, if that's all you need, the function used to create that kind of output is probably faster than `CircularJSON` and surely fits in less lines of code. ### Why Not {{put random name}} Solution So here the thing: circular references can be wrong but, if there is a need for them, any attempt to ignore them or remove them can be considered just a failure. Not because the method is bad or it's not working, simply because the circular info, the one we needed and used in the first place, is lost! In this case, `CircularJSON` does even more than just solve circular and recursions: it maps all same objects so that less memory is used as well on deserialization as less bandwidth too! It's able to redefine those references back later on so the way we store is the way we retrieve and in a reasonably performant way, also trusting the snappy and native `JSON` methods to iterate.