Fix inheritance

[klokane]

Reworked inheritance:

• fix correct order in extend element for inheritance from older ancestor to newest
• allow inheritance for primitive members
• fixed inheritance for RenderJSONVisitor
• introduce ElementMerger to handle correctly/standardized way inheritance resolve
• refactoring ExpandVisitor
• move members out of IElement (it now provide just Interface and some helpers)
• little refactoring in TypeQueryVisitor (expose enum ast typedef)
• correct order "data structures" according to membership/inheritance (most dependent objects are registered latest)

To be resolved:

• cripled some schema test after rebase (done)
• add test for primitive types inheritance (done)
• add test for complex membership/inheritance
  (test are created, just needs to added to repository and check if schema is generated correctly)

Resolved issues:

• fix MSON - named type based on primitive is wrongly rendered #98
• fix Regression - enum inline definition pick into JSON "array" instead of member #176

[kylef]

Curious, any particular reason you've moved this down? In most languages it preferred to put system includes before local includes.

[klokane]

In short? In C/C++ is preferred place system includes below specific :)

In long:

// file a.h
std::string fn();

// file a.cc
#include <string>
#include "a.h"

std::string fn() { return "hello"; }

a.cc will be compiled ok

but

// file: b.cc
#include "a.h"

// something...

b.cc compilation will fail with something like unknown type 'std::string'

[kylef]

Ah that's a good point. It's no longer been a problem in Objective-C with the introduction of clang modules. I completely forgot it's not standard/default across all compilers.

👍

[klokane]

Last commit (103c0bf) is for discussion, it fix problems with schema fixtures, but fixtures definition in schema are not correct. There should be selected one value for enum instead of array of possible values.

Correct implementation is commented out by #if 0 block

There still missing one schema to fix.

[klokane]

Hmm interesting, sent patch (commented out part) fixes #176 :)

[klokane]

Curent status

• @w-vi is checking changes in JSON Schema rendering
• I'm working on fix: https://github.com/apiaryio/drafter/blob/klokane/fix-inheritance/test/fixtures/mson/primitive-variables.apib

(member c8 - key should be expanded as number - because it is variable)

[klokane]

@w-vi review addressed. There are no tests :( in current tests implementation it is not possible to prepare test with comparable output.

We need implement querying for refract to build tests for similar cases

[w-vi]

My part is done, I think it's ready for merge ... so leaving it to @pksunkara if he has anything to comment on

[klokane]

BTW why "continuous-integration/appveyor" CI is (again and again) yellow? I mean, we should remove it form CI proces it it does not work.

[pksunkara]

We should be relying on the other 2 green things, not the continuous-integration/appveyor which is just an aggregator.

 continuous-integration/appveyor/branch
 continuous-integration/appveyor/pr

The yellow one is an issue but hopefully appveyor will fix it soon. We can try to forget about it until then since we need the other CI statuses.

[klokane]

@pksunkara yes, I know about it. But if I remember it is not working for few months. It is confusing

[pksunkara]

yes, I know about it. But if I remember it is not working for few months. It is confusing

I know and I feel the same. 😭 But we really do need the windows CI. 😞

[pksunkara]

Did you mean to have 2012-12-02 here?

[klokane]

Probably, yes. But it is not important, so it can stay as is.
In final inheritance is used sample from Date3 so test is able to check if it is inherited in required way

[pksunkara]

@klokane Finished review. Just minor stuff to fix.

[klokane]

Review addressed, fixedup on to previous commit Fix typo && push --force