Matt Schoen [email protected] of Defective Studios

I came across the need to send structured data to and from a server on one of my projects, and figured it would be worth my while to use JSON. When I looked into the issue, I tried a few of the C# implementations listed on http://json.org, but found them to be too complicated to work with and expand upon. So, I've written a very simple JSONObject class, which can be generically used to encode/decode data into a simple container. This page assumes that you know what JSON is, and how it works. It's rather simple, just go to json.org for a visual description of the encoding format.

As an aside, this class is pretty central to the AssetCloud content management system, from Defective Studios.

Update: The code has been updated to version 1.4 to incorporate user-submitted patches and bug reports. This fixes issues dealing with whitespace in the format, as well as empty arrays and objects, and escaped quotes within strings.

Users should not have to modify the JSONObject class themselves, and must follow the very simple proceedures outlined below:

Sample data (in JSON format):

{
    "TestObject": {
        "SomeText": "Blah",
        "SomeObject": {
            "SomeNumber": 42,
            "SomeBool": true,
            "SomeNull": null
        },
        
        "SomeEmptyObject": { },
        "SomeEmptyArray": [ ],
        "EmbeddedObject": "{\"field\":\"Value with \\\"escaped quotes\\\"\"}"
    }
}

• Decode JSON-formatted strings into a usable data structure
• Encode structured data into a JSON-formatted string
• Interoperable with Dictionary and WWWForm
• Optimized parse/stringify functions -- minimal (unavoidable) garbage creation
• Asynchronous stringify function for serializing lots of data without frame drops
• MaxDepth parsing will skip over nested data that you don't need
• Special (non-compliant) Baked object type can store stringified data within parsed objects
• Copy to new JSONObject
• Merge with another JSONObject (experimental)
• Random access (with int or string)
• ToString() returns JSON data with optional "pretty" flag to include newlines and tabs
• Switch between double and float for numeric storage depending on level of precision needed (and to ensure that numbers are parsed/stringified correctly)
• Supports Infinity and NaN values
• JSONTemplates static class provides serialization functions for common classes like Vector3, Matrix4x4
• Object pool implementation (experimental)
• Handy JSONChecker window to test parsing on sample data

It should be pretty obvious what this parser can and cannot do. If anyone reading this is a JSON buff (is there such a thing?) please feel free to expand and modify the parser to be more compliant. Currently I am using the .NET System.Convert namespace functions for parsing the data itself. It parses strings and numbers, which was all that I needed of it, but unless the formatting is supported by System.Convert, it may not incorporate all proper JSON strings. Also, having never written a JSON parser before, I don't doubt that I could improve the efficiency or correctness of the parser. It serves my purpose, and hopefully will help you with your project! Let me know if you make any improvements :)

Also, you JSON buffs (really, who would admit to being a JSON buff...) might also notice from my feature list that this thing isn't exactly to specifications. Here is where it differs:

• "a string" is considered valid JSON. There is an optional "strict" parameter to the parser which will bomb out on such input, in case that matters to you.
• The Baked mode is totally made up.
• The MaxDepth parsing is totally made up.
• NaN and Infinity aren't officially supported by JSON ([http://stackoverflow.com/questions/1423081/json-left-out-infinity-and-nan-json-status-in-ecmascript read more] about this issue... I lol'd @ the first comment on the first answer)
• I have no idea about edge cases in my parsing strategy. I have been using this code for about 3 years now and have only had to modify the parser because other people's use cases (still valid JSON) didn't parse correctly. In my experience, anything that this code generates is parsed just fine.

Encoding is something of a hard-coded process. This is because I have no idea what your data is! It would be great if this were some sort of interface for taking an entire class and encoding it's number/string fields, but it's not. I've come up with a few clever ways of using loops and/or recursive methods to cut down of the amount of code I have to write when I use this tool, but they're pretty project-specific.

Note: This section used to be WRONG! And now it's OLD! Will update later... this will all still work, but there are now a couple of ways to skin this cat.

// Note: your data can only be numbers and strings.
// This is not a solution for object serialization
// or anything like that.
JSONObject j = new JSONObject(JSONObject.Type.OBJECT);
// number
j.AddField("field1", 0.5);
// string
j.AddField("field2", "sampletext");
// array
JSONObject arr = new JSONObject(JSONObject.Type.ARRAY);
j.AddField("field3", arr);

arr.Add(1);
arr.Add(2);
arr.Add(3);

string encodedString = j.print();

NEW! The constructor, Add, and AddField functions now support a nested delegate structure. This is useful if you need to create a nested JSONObject in a single line. For example:

DoRequest(URL, new JSONObject(delegate(JSONObject request) {
    request.AddField("sort", delegate(JSONObject sort) {
        sort.AddField("_timestamp", "desc");
    });
    request.AddField("query", new JSONObject(delegate(JSONObject query) {
        query.AddField("match_all", JSONObject.obj);
    }));
    request.AddField("fields", delegate(JSONObject fields) {
        fields.Add("_timestamp");
    });
}).ToString());

Decoding is much simpler on the input end, and again, what you do with the JSONObject will vary on a per-project basis. One of the more complicated way to extract the data is with a recursive function, as drafted below. Calling the constructor with a properly formatted JSON string will return the root object (or array) containing all of its children, in one neat reference! The data is in a public ArrayList called list, with a matching key list (called keys!) if the root is an Object. If that's confusing, take a glance over the following code and the print() method in the JSONObject class. If there is an error in the JSON formatting (or if there's an error with my code!) the debug console will read "improper JSON formatting".

string encodedString = "{\"field1\": 0.5,\"field2\": \"sampletext\",\"field3\": [1,2,3]}";
JSONObject j = new JSONObject(encodedString);
accessData(j);
//access data (and print it)
void accessData(JSONObject obj){
    switch(obj.type){
        case JSONObject.Type.OBJECT:
            for(int i = 0; i < obj.list.Count; i++){
                string key = (string)obj.keys[i];
                JSONObject j = (JSONObject)obj.list[i];
                Debug.Log(key);
                accessData(j);
            }
            break;
        case JSONObject.Type.ARRAY:
            foreach(JSONObject j in obj.list){
                accessData(j);
            }
            break;
        case JSONObject.Type.STRING:
            Debug.Log(obj.str);
            break;
        case JSONObject.Type.NUMBER:
            Debug.Log(obj.n);
            break;
        case JSONObject.Type.BOOL:
            Debug.Log(obj.b);
            break;
        case JSONObject.Type.NULL:
            Debug.Log("NULL");
            break;
        
    }
}

NEW! Decoding now also supports a delegate format which will automatically check if a field exists before processing the data, providing an optional parameter for an OnFieldNotFound response. For example:

new JSONObject(data);
list.GetField("hits", delegate(JSONObject hits) {
	hits.GetField("hits", delegate(JSONObject hits2) {
		foreach (JSONObject gameSession in hits2.list) {
			Debug.Log(gameSession);
		}
	});
}, delegate(string name) {	//"name" will be equal to the name of the missing field.  In this case, "hits"
	Debug.LogWarning("no game sessions");
});

I've added a string and int [] index to the class, so you can now retrieve data as such (from above):

JSONObject arr = obj["field3"];
Debug.log(arr[2].n); //Should ouptut "3"

Big update!

• Better GC performance. Enough of that garbage!
  • Remaining culprits are internal garbage from StringBuilder.Append/AppendFormat, String.Substring, List.Add/GrowIfNeeded, Single.ToString
• Added asynchronous Stringily function for serializing large amounts of data at runtime without frame drops
• Added Baked type
• Added MaxDepth to parsing function
• Various cleanup refactors recommended by ReSharper

• Added support for NaN
• Added strict mode to fail on purpose for improper formatting. Right now this just means that if the parse string doesn't start with [ or {, it will print a warning and return a null JSONObject.
• Changed infinity and NaN implementation to use float and double instead of Mathf
• Handles empty objects/arrays better
• Added a flag to print and ToString to turn on/off pretty print. The define on top is now an override to system-wide disable