JSON-dry
JSON-dry allows you to stringify objects containing circular references,
dates, regexes, ...
It can also be used to serialize and revive instances of your own classes.
Table of contents
Installation
$ npm install json-dry
Usage
Basic example
This is a basic example of stringifying an object (containing multiple references to the same object) and parsing it again.
var Dry = require('json-dry');
var obj = {};
var ref = {
date : new Date(),
regex : /test/i
};
obj.reference_one = ref;
obj.reference_two = ref;
obj.date = ref.date;
var dried = Dry.stringify(obj);
var undried = Dry.parse(dried);
undried.reference_one == undried.reference_two;
undried.reference_one.date == undried.date;
Implementing methods for serializing & reviving instances
Let's create an example class you might want to serialize and revive:
function Person(options) {
this.firstname = options.firstname;
this.lastname = options.lastname;
}
Person.prototype.fullname = function fullname() {
return this.firstname + ' ' + this.lastname;
};
var jelle = new Person({firstname: 'Jelle', lastname: 'De Loecker'});
jelle.fullname();
So now we've created a very basic class, let's register the class and add the 2 required methods for serializing & reviving.
Dry.registerClass(Person);
Person.prototype.toDry = function toDry() {
return {
value: {
firstname : this.firstname,
lastname : this.lastname
}
};
};
Person.unDry = function unDry(value) {
var result = new Person(value);
return result;
};
Now let's try stringifying it:
var dried = Dry.stringify(jelle);
var undried = Dry.parse(dried);
undried.fullname();
Serializing & reviving instances with circular references
Some classes contain references to each other, for example:
let alpha = new Alpha(),
beta = new Beta();
alpha.beta = beta;
beta.alpha = alpha;
The problem is that when you serialize & then try to revive this, one of the unDry
methods will receive an un-revived placeholder. This can obviously cause issues, especially when setting the property has side-effects. So a new argument whenDone
has been added to the unDry
method, like so:
Alpha.prototype.unDry = function unDry(obj, custom_method, whenDone) {
let alpha = new Alpha();
whenDone(function() {
alpha.beta = obj.beta;
});
return alpha;
}
whenDone
functions will be called just before the Dry.undry()
function exits, so all the references will have been revived by then.
toObject
While Dry.stringify
will return you with a json-valid string, Dry.toObject
will give you a valid simplified object.
In fact: Dry.stringify
is just a function that performs JON.stringify
on Dry.toObject
's output.
Why would you want to use this? Things like Workers
and IndexedDB
communicate data using the structured clone algorithm. So instead of performing expensive stringify operations you can just use these objects.
Cloning objects & instances
JSON-Dry offers a specialized clone
method. While in theory you could clone an object by drying end reviving it, like so:
var cloned = Dry.parse(Dry.toObject(jelle))
This is 14x slower than using clone
, because toObject
needs to generate paths, escape certain string values and create wrapper objects. These expensive things can be ignored when cloning:
var cloned = Dry.clone(jelle);
Clone methods
If you've added a toDry
and unDry
method to your class, by default the clone
method will use those to create the clone.
However, you can also create another method that gets precedence:
dryClone
Person.prototype.dryClone = function dryClone(seen_map, custom_method) {
return new Person({
firstname : this.firstname,
lastname : this.lastname
});
}
Custom clone methods
The clone
method takes an extra parameter called custom_method
. If you're cloning something that has a function property with the same name, that'll be used.
This can be used when you want to redact certain parts, for example:
Person.prototype.specialOccasionClone = function specialOccasionClone(seen_map, custom_method) {
return new Person({
firstname : this.firstname[0] + '.',
lastname : this.lastname
});
};
var special_clone = Dry.clone(jelle, 'specialOccasionClone');
special_clone.fullname();
Project history
Earlier versions of the project were heavily based on circular-json, a small library that adds (circular) reference support to JSON.
A lot of the JavaScript code on my websites was already shared between the server & client side, but I also wanted an easy way of sending data to the client while retaining references & class instances, so I started adding features to circular-json and called it json-dry
(dry as in don't repeat yourself).
After version 0.1.6 I integrated json-dry
into my protoblast library, and development continued in there. But now it has deserved its own repository once again.
This version is a rewrite of earlier versions. circular-json
used (and still uses) multiple arrays to keep track of already used objects, but json-dry
now uses WeakMap
s, something that makes the code easier to maintain and is also faster.
circular-json
was also implemented as a replacer
and reviver
function to JSON.stringify
and JSON.parse
respectively. json-dry
actually creates a new object before stringifying
it.
Because multiple references are represented as ~paths~to~the~first~reference
, the size of the JSON string can be a lot smaller. Can be, though, because sometimes reference paths are longer than the object they are refering to.
Because of this, as soon as json-dry
encounters a new path that is smaller than the previous one, it'll use that in the future. This helps a bit, though more improvements could be made in the future.
Project future
- Possibly use objects or arrays instead of string primitives for references. This would speed up serializing and parsing, but be a bit more verbose. Tell me what you think in issue #2
Versioning
We use SemVer for versioning. For the versions available, see the tags on this repository.
License
This project is licensed under the MIT License - see the LICENSE file for details
Acknowledgments
Many thanks to WebReflection, whose circular-json was the basis for earlier versions of this project.