As I understand it, "additional", "constrain existing" and "add" are the key words there. Redefining is not specified or supported. Beyond that, it further says:

Conceptually, the behavior of extends can be seen as validating an
instance against all constraints in the extending schema as well as
the extended schema(s).

So in your example, you would require it to be both a string and an integer.

I see how it would be interpreted that way, but it does seem significantly less useful without being able to do basic overriding... no?

As @markolson said, the draft-03 specification (and the many discussions that occurred around it) state that extending a schema should be viewed as simply placing additional constraints on an existing schema.

While I agree that overriding an attribute has a lot of use, I am aiming for meeting the JSON Schema specification; the language for "extends" is currently the same in draft-04 as is it in draft-03, which is currently being discussed. There has been some movement towards an "inherits" attribute that allows for multiple schema inheritance with a defined schema merge strategy, but that is still in the design stages.

Understood, thanks for the response. I'll work around it for now.

On Jan 17, 2012, at 7:33 AM, Kenny Hoxworth wrote:

As @markolson said, the draft-03 specification (and the many discussions that occurred around it) state that extending a schema should be viewed as simply placing additional constraints on an existing schema.

While I agree that overriding an attribute has a lot of use, I am aiming for meeting the JSON Schema specification; the language for "extends" is currently the same in draft-04 as is it in draft-03, which is currently being discussed. There has been some movement towards an "inherits" attribute that allows for multiple schema inheritance with a defined schema merge strategy, but that is still in the design stages.

---

Reply to this email directly or view it on GitHub:
#22 (comment)

Look into http://en.wikipedia.org/wiki/Liskov_substitution_principle

"Extends" is only for substitutability, not for code reuse. If you want code reuse, build a code generator instead :) As for whether it is "useful" or not - building type coercion into a schema definition is begging for subtle bugs and giant headaches for any software that uses that schema definition.