KSON is a simple data interchange format based on JSON. Its serialized representation doesn't contain the redundant keys of typical json documents and more compact representations of values are possible through an extensible encoding/decoding mechanism.

KSON uses a simple schema format to describe arbitrarrily nested objects and encoded types. The minified and gzipped javascript library is < 1K.

// Load schema definitions

var movie_schemas = [
  {
    "id": "role",
    "fields": ["name", "character"],
    "meta": [0, 0]
  },{
    "id": "movie",
    "fields": ["title", "year", "rating", "cover", "actors"],
    "meta": [
      0,
      0,
      0,
      "prefix:http\://movies.db/covers/|suffix:.jpg",
      "[]role"
    ]
  }
];
KSON.addSchema(movie_schemas);

var movies = [
  {
   "title": "Forrest Gump",
   "year": 1994,
   "rating": 8.7,
   "cover": "http://movies.db/covers/8.jpg",
   "actors": [
    {"name": "Tom Hanks", "character": "Forest Gump"},
    {"name": "Robin Wright", "character": "Jenny Curran"},
    {"name": "Gary Sinise", "character": "Lieutenant Dan Taylor"}
   ]
  },
  {
    "title": "Toy Story",
    ...
  }
  ...
];

// pretty printing is not part of the library
console.log(KSON.stringify(movies, "[]movie"));
[
  "[]movie",
  "Forest Gump",
  1994,
  8.7,
  "8",
  [
    ["Tom Hanks", "Forest Gump"],
    ["Robin Wright", "Jenny Curran"],
    ["Gary Sinise", "Lieutenant Dan Taylor"]
  ],
  "Toy Story",
  ...
]

// Of course the "movie" and "role" schema definitions may also be
// serialized to KSON using the "schema" schema:

var schema_data = KSON.stringify(movie_schemas, "[]schema")
console.log(schema_data);

["[]schema","role",["name","character"],[0, 0],"movie",
  ["title","year","rating","cover","actors"],
  [0,0,0,"prefix:http\://movies.db/covers/|suffix:.jpg","[]role"]]

// Likewise schemas can be initialized with a kson document
KSON.addSchema(schema_data);

• Q: Why should I use this instead of JSON/MessagePack/Thrift/...?

• A: Unless you are targeting the browser you probably shouldn't. If you are, then the size and speed of the javascript libraries for these formats may make KSON a favorable option.

• Q: Is KSON faster/smaller/better than JSON?

• A: Tests using node.js indicate that parsing is marginally slower. This should be outweighed by faster transmission speed of signifigantly less data. See the node.js benchmarks for more info.

• Q: Is this a drop in replacement for JSON?

• A: No, you will need to write schema definitions for your data. Automatic schema detection based on example data may help in getting started. Also see "restrictions" in the Specification.

• Q: Why yet another a schema format for JSON?

• A: Existing JSON schema formats include extrenious information to the purpose of KSON and would require larger javascript library for parsing and serialization.

• Q: Is there language support for php/ruby/java/c...?

• A: Currently javascript and python are supported. Porting to other languages with existing JSON support should be fairly easy and patches are very welcome.

A KSON schema consists of three fields:

id        *string* identifier
fields    *array* of field names
meta      *array* signifying the types associated with
            corresponding elements in the *fields* array.

A schema only defines one object. Nesting is accomplished by referencing subschemas via their id in the meta array.

An element of meta specifies the content of the element of fields at the same index. The lengths of the fields and meta arrays must be equal. An element of meta must be one of the following:

0               field contains a plain javascript value of type
                string, boolean, number, null.
"[]"            array of plain values
"schema-id"     string specifying the schema of a nested object
"[]schema-id"   string specifying the schema of an array of nested
                objects
"codec-id"
"[]codec-id"

The top level elements of a KSON document must be either an array of objects or an object. A plain value or an array of plain values cannot be serialized.

Subschemas and codecs use the same syntax so you shouldn't load a schema and install a codec which both use the same name. If you do, the behaviour is undefined.

Decoders are executed in the order of their definition in the meta field. The decoder is passed the value to decode as the first argument and the intermediate data object as the second. The intermediate object will have the values of all previous fields decoded and applied.

Included codecs:

enum:a:b:c:d   "a" -> 0, "b" -> 1, ...
prefix:egg     "eggbacon" -> "bacon"
suffix:bacon   "eggsausagebacon" -> "eggsausage"
bool           true -> 1, false -> 0
date           new Date() -> 1364938727390
int36          1364938727390 -> "hf1lcnka"

Codec chaining

prefix:egg|suffix:bacon "eggsausagebacon" -> "sausage" date|int36 new Date() -> "hf1lcnka"

Custom codecs are registered using factory functions in the following manner.

KSON.coders["my_coder_id"] = function(args) {
    // args is a string array from a schemas meta field
    // enum:a:b:c:d -> args == ["a", "b", "c", "d"]

    return function (val, encode) {
        // The actual encoding/decoding function
        return encode ? encode_op(val) : decode_op(val);
    }
}

The python library validates schemas and codec definitions, the javascript library does not. Invalid schema definition loaded with the javascript library may parse or stringify bad data without any throwing any errors.

Common errors include: - Loading a schema before the custom codecs it uses have been registered. - Loading a schema before the subschemas on which it depends have been loaded. Note that schemas are loaded in the order they appear in a file, so dependent schmas schould come after those they depend uppon. - Circularly dependent schmas (including schemas which depend on themselves) are possible by using a stub schema, which only specifies an id and no fields. This can be replaced by the propper schema with the same id, after its dependencies have been loaded. - Loading a json encoded schema. The argument of the KSON.addSchema function may be either a KSON encoded schema or schemas, an already decoded schema or an array of schemas. It is however fairly simple to load json encoded schemas: KSON.addSchema(JSON.parse(raw_json_schemas));

The python library will warn about these and other issues, so use it to

$ pip install kson

Automatic schema detection:

$ kson --auto-detect data.json --id schema_id

$ git clone [email protected]:mbarkhau/kson.git
$ npm install uglify-js -g
$ npm install benchmark microtime

$ pip install py.test

$ make test
$ make bench
$ make minify