node package manager


A streaming JSON parser that sometimes might be faster than v8 JSON.parse


This is a streaminig json parser, that only parsers what you tell it, based on dominictarr's JSONStream idea and creationix's jsonparse code(mostly his, didn't change much).

How to use?

var JSONParser = require('json--'),
    parser = new JSONParser(['rows', true, 'name']);

parser.onValue = console.log.bind(console);
parser.write(new Buffer('{"rows":[{'));
parser.write(new Buffer('"lolz": "wow", "name" : "sweet"'));
parser.write(new Buffer('}]}'));

Note that the parser.write only accepts a Buffer instance.


require('json--') will return a JSONParser class.

new JSONParser(path)

  • Array<Any> path: Should contain at least one element. If it's a function, it will be called with the key name and it's value will be converted to a Boolean to see if the key is needed or not.

It will return an instance of json--.


This function will be called on any value that matches the path completely.


  • Buffer buffer

How fast is it?

Here's some numbers for parsing a subset of npm registry. The test file is test.js and the json file is npm.json. Tested under Ubuntu 12.04 64bit.

In Node 0.6.19:

JSON--, path: rows: 3892ms
JSON--, path: rows,true: 4661ms
JSON--, path: rows,true,doc: 2433ms
JSON--, path: rows,true,doc,_id: 543ms
toString: 180ms
JSON: 627ms

In Node v0.8.6:

JSON--, path: rows: 4702ms
JSON--, path: rows,true: 4049ms
JSON--, path: rows,true,doc: 2417ms
JSON--, path: rows,true,doc,_id: 825ms
toString: 190ms
JSON: 653ms

The toString is the amount of the time that it takes for the buffer to be converted to utf8(JSON.parse needs it to be in utf8 string), but JSON-- only operates on a buffer(it converts it internally).

What about memory?

As of memory, since it all keeps what it needs to emit, only the objects in the specified path(unlike creationix's jsonparse, which kept everything to have a complete root object.), it uses much less memory.

Why so fast?

The trick that made this sped up is that, it only parses the parts that are in the path, and it skips the parts that is not in the specified path with a simple code without verifying it. This is very experimental, and this skipping might result in an undefined state if it tries to parse an invalid JSON string.

What's up with the name?

Since it doesn't parses fully, its not a complete JSON parser, so decrease one from it. If you have a better name, I'm open to suggestions.