Switch branches/tags
Nothing to show
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
242 lines (166 sloc) 7.52 KB

TYSON Tutorial

This tutorial gives a user-friendly introduction to the TYSON syntax.

Executive summary

TYSON extends JSON with type support, and that's all it does.

It is backward compatible with JSON.

It is designed to be compatible with any JSON Schema technology, meaning that it can be used as input and annotated output of such technologies with little work.

It is compatible with any host language and is fully agnostic in this respect.

JSON geek's summary

With TYSON, you can add type annotations to JSON documents by prefixing values with a type name in parentheses.

("person") {
  "name" : "Cooper",
  "first" : "Sheldon",
  "middle" : "Lee",
  "multiplication" : ("cell-division-kind") "MITOSIS",
  "birthdate" : ("date") "1980-02-26",
  "picture" : ("hexBinary") "0123456789abcdef",
  "friends" : ("ids") [ 1, 2, 4, 5 ]


The first important thing to know is that any JSON value is also a TYSON value.

This includes strings:


Numbers (We will see later on that there are several types thereof):







[ "foo", 1, null, true, [ 1, 2, 3, 4 ] ]

And objects:

  "foo" : "bar",
  "bar" : {
    "foo" : [ 1.2, 2, 4, 5, 2e5 ],
    "bar" : true,
    "foobar" : null

More atomic types

What is new with TYSON is type support beyond the basic types of JSON. Anybody can create types, give them a name, and go ahead and use them (how exactly these types are created, documented and agreed on is out of scope of TYSON, though).

For example, in a straightforward way, atomic types from XML Schema can be taken over with no effort:

For example, you can have dates:

("date") "2018-09-01" 

Binary values:

("hexBinary") "0123456789abcdef"

Types such as the above are not part of the TYSON specification itself, but they are introduced in JSound, a schema language that works with TYSON.

Object and array types can also be user-defined:

("array-of-booleans") [ true, true, false, true, false ]
("person") {
  "name" : "Cooper",
  "first" : "Sheldon",
  "birthdate" : ("date") "1980-02-26",
  "picture" : ("hexBinary") "0123456789abcdef",
  "friends" : ("ids") [ 1, 2, 4, 5 ]

A type annotation can be put in front of absolutely any value, as a string in parenthesis. And this is all there is to know about the TYSON syntax in itself, because there is nothing else.

Builtin types

TYSON explicitly introduces builtin types corresponding to JSON values:

  • object (all JSON objects)
  • array (all JSON arrays)
  • string (all JSON strings, sequences of Unicode codepoints)
  • boolean (true and false)
  • integer (all integers, with no limit or bound)
  • decimal (all decimal numbers, with no limit or bound)
  • double (all IEEE754 doubles)
  • null (a type that only matches the value null)

For backward compatibility and ease of use, these annotations are implicit. The document seen above

  "foo" : "bar",
  "bar" : {
    "foo" : [ 1.2, 2, 4, 5, 2e5 ],
    "bar" : true,
    "foobar" : null

is semantically (seen from the consuming application) exactly the same as the more explicit:

("object") {
  "foo" : ("string") "bar",
  "bar" : ("object") {
    "foo" : ("array") [ ("decimal") 1.2, ("integer") 2, ("integer") 4, ("integer") 5, ("double")2e5 ],
    "bar" : ("boolean") true,
    "foobar" : ("null") null

For numbers, anything with no dots (like 1) and no scientific notation is implicitly an integer, anything with a dot and no scientific notation (like 1.1) a decimal, anything in scientific notation (like 1e1) a double.

TYSON does not introduce any other type (not even date or hexBinary, which is left to schema technologies such as JSound).

Builtin type validation

TYSON performs validation against the above builtin types, and only them.

For example, the following values are not valid TYSON:

("string") {}
("array") "foo"
("boolean") "bar"

TYSON, however, does not know of any other types, and validation must be done separately. For example, because my-integer is not builtin, the following value is in conformant TYSON as far as the TYSON processor is concerned.

("my-integer") "1.1"

The TYSON processor will simply tell the consuming application that it encountered a lexical value "1.1" annotated with the type "my-integer". It is then up to the consuming application to take it from there.

In practice, TYSON users will exchange instances that use types that they know and agree on. If it is known and agreed that the value space of the type my-integer is the set of all integers, then the above value must be rejected -- by the consuming application, not the TYSON processor -- as invalid against that type. If somebody else in the world decides that his type my-integer is the set of all strings, then from their perspective the value is valid. How types are documented, named, known and standardized is out of scope of TYSON.

Number types

Thanks to the more fine-grained number types, it is possible to have more control about how some data appears to a host language, beyond just calling it a number:

("decimal") 2
("double") 3.14
("integer") 1e10

Plus, you can also have the special IEEE754 double values that are not supported by JSON itself:

("double") "NaN"
("double") "+INF"
("double") "-INF"

Define your own types

TYSON poses very little restrictions on types. All it requires for compatibility and soundness is that only user-defined types fulfilling the requirements below be instantiated, parsed with a TYSON processor and validated by the consuming application:

  • a user-defined object type must be associated with a value space (a subset of the set of all objects),

  • an array type must be associated with a value space (a subset of the set of all arrays),

  • an atomic type must be associated with a value space, a lexical space (the string representations of the values) and a lexical mapping that associates each lexical value with a typed value (the mapping may not be injective).

The details of creating a type, documenting it, sharing it with the world, agreeing on names, type discovery and so on is out of scope of TYSON. All TYSON does is provide syntax to losslessly store and share typed data once types are agreed on.

If the consuming application of a TYSON processor (for example, with JSound support) is provided with further types, and with their documentation in terms of value spaces, then it should validate lexical values annotated with these types, while the builtin types are validated by the TYSON processor itself.


For the purpose of validation, TYSON does not differentiate between quoted and unquoted values: it takes whatever there is as the lexical value, excluding the quotes if there are any. The consuming application will not even know whether the value was quoted or not.

The following two values are the same: the integer 1.

("integer") 1
("integer") "1"

In both cases, the consuming application receives the lexical value 1 annotated with the builtin type integer, along a green light that it is valid.

Note that this has nothing to do with casting. TYSON does not cast. It only looks whether the lexical value is in the lexical space of an atomic type.

This means that value that JSON allows to be unquoted do not need quotes, and can be used as lexical values of any atomic type.


For more information, or to know more on corner cases, the TYSON specification can be found at