Adding declaration of unions outside type expressions

[antoniogarrote]

Fixes #530

Extends the spec to make it possible to declare a union without using a type expression.

It introduces a new facet anyOf to declare the super types of the union.
It adds two restrictions to anyOf:

• It is mandatory for unions
• It cannot e empty.

[sichvoge]

And at least two elements right?

[petrochenko-pavel-a]

Yep,originally there was an items option for union types, but anyOf is definitely a better name

[sichvoge]

@antoniogarrote could you update your PR pls.

[sichvoge]

QQ: whats the value of each item inside the anyOf array? a simple string representing a reference to another type or also an object. For example, how do I express Person | Animal[]?

[antoniogarrote]

@sichvoge

In my opinion, any valid type declaration should be allowed:

anyOf:
  - Person | Animal[]

Expresses:
Union(Union(Persion,Animal[]))

anyOf:
  - Person 
  - Animal[]

Expresses:
Union(Person,Animal[])

anyOf:
  - type:
      Person
  - type:
      array
    items:
      Animal

Expresses:
Union(Person,Animal[])

Etc.

At the of the day, if you run the canonical form, all of them become:

{type: "union"
 anyOf:[
  {type: "object", properties:{... properties of person ...}},
  {type: "array", items: {type: "Object", properties: {... properties of animal here ...}}]
}

[sichvoge]

OK. Maybe we should adjust the description text in the spec with that. Its not really clear.

[usarid]

That will need to go into a new version of RAML, presumably 1.1, right ?