Java support inheritance

[ahgittin]

as per #1015 (and #946) this is now optional and disabled by default. this results in extra unnecessary code, masking fields in some cases, and breaks code generated for jax-rs in certain cases (enums with inheritance).

as noted by @xhh at #1050 this is not a perfect fix, but it improves the situation significantly (restoring master branch behaviour, as the current situation is a regression; my spec builds fine with master but not with develop_2.0 ?).

[wing328]

@ahgittin do you mind sharing the spec with us assuming it can be shared publicly?

[ahgittin]

@wing328 here's a cut-down test case whose -l jaxrs makes for invalid java without this, valid java with this:

swagger: '2.0'
info:
  title: buggy inheritance for java
  version: "0.0.1"
definitions:
  Super:
    properties:
      my_enum:
        type: string
        enum: [ A, B ]

  Sub:
    allOf:
    - $ref: '#/definitions/Super'
    - type: object
    properties:
      another:
        type: string

paths:
  /home:
    get:
      summary: with develop_2.0 prior to #1094 the java generated fails

[ahgittin]

also noticed this had some strange commits; i've rebased so it's just the one-line change

[webron]

FWIW, that's not inheritance, that's composition. In this case, Sub and Super have no parent-child relationship. The end result in this case would be that Sub would include the same properties as Super but will not inherit from it.

[ahgittin]

sorry @webron i mis-typed the example, it's now correct, and it definitely is implemented as inheritance (maybe it shouldn't be):

public class Sub extends Super {

  public enum MyEnumEnum {
     A,  B, 
  }

redeclaring the enum breaks compilation as Super.getMyEnum() returns the parent's MyEnumEnum so Sub.getMyEnum() has an incompatible return type

HTH

[webron]

Not sure what's changed, but as long as there's no discriminator, there's no inheritance (from the spec's point of view).

[wing328]

If we look at the Java client in the "master" branch, we'll find that it implements allOf using extends in Java, which is not the correct implementation as Ron has pointed out that allOf is model composition (not inheritance)

@xhh previously attempt to fix allOf with #946 (copy parent properties) but it does not implement discriminator (should be implemented as extends in Java) due to time constraint.

@tandrup introduced supportsInheritance to let people choose the behaviour with #1015.

Before someone from the community implements discriminator, we can merge this PR for the time being to keep the old (master) behaviour if no one has any concern.

[webron]

I'd suggest putting a hold on the discriminator implementation as there's a technical detail with the spec we need to resolve first. It'll be done with in the upcoming weeks, and will come with an implementation (or suggestion for implementation). Sorry for the trouble on this one.

[ahgittin]

closing and will resubmit against master as per change to swagger PR process

[ahgittin]

closed in favour of #1120

[steverice]

there's a technical detail with the spec we need to resolve first

@webron are you referring to OAI/OpenAPI-Specification#403?

[webron]

That is correct. It should be resolved soon.