[Swift3] decoders call parents decoders to fix issue with class inheritance

[julienfouilhe]

PR checklist

• Read the contribution guidelines.
• Ran the shell/batch script under ./bin/ to update Petstore sample so that CIs can verify the change. (For instance, only need to run ./bin/{LANG}-petstore.sh and ./bin/security/{LANG}-petstore.sh if updating the {LANG} (e.g. php, ruby, python, etc) code generator or {LANG} client's mustache templates)
• Filed the PR against the correct branch: master for non-breaking changes and 2.3.0 branch for breaking (non-backward compatible) changes.

Description of the PR

I was having problems with classes inheriting from other classes, fields defined in parent classes were nil. I fixed it by calling the parent's decoder in the child's decoder and passing the child's instance.

[wing328]

@julienfouilhe thanks for the PR.

cc @jaz-ah @Edubits @jgavris @hexelon

[jaz-ah]

@julienfouilhe could you regenerate the sample files and put in here as well? I wanted to see what the output looks like now...

[julienfouilhe]

@jaz-ah, it should be in the commit. Most of it is in Models.swift, and the diff is not displayed on github unless you click because it's too large. :)

Edit: Oh, you mean those other files in the samples. Give me a few secs

[jaz-ah]

@julienfouilhe ah I meant the actual sample output -i.e. the sample/test project checked into the tree...

[julienfouilhe]

@jaz-ah I ran swift3-petstore-all.sh. Tell me if that's not what you needed. I didn't push those at first because I thought it was just reversing the order of some args in the sample files.

[jaz-ah]

@julienfouilhe I must have missed it - I see it now - looks good - thx! +1 @wing328

[julienfouilhe]

@jaz-ah Yes I just ran them and amended my commit. ;)

I don't think there's a test for that though. If you want to see where the bug is, you would need to add an endpoint to the Petstore API /cats returning Cat objects and not Pet. Before this commit, it didn't go through the Pet decoder, and thus didn't fill fields described in Pet

[wing328]

@julienfouilhe thanks for the PR.

@jaz-ah thanks for reviewing the change.

[wuf810]

I am double checking but the changes in this PR have broken my swagger generated code .

This line in the template

_ = Decoders.decode(clazz: {{parent}}.self, source: source, instance: result)

generates the following code for classes in the Models.swift file :

_ = Decoders.decode(clazz: String.self, source: source, instance: result)

As a result the decode function is looking for a class decoder named "String".

NB The classes this fails on do use AdditionalProperties (which @danwoodel and I are currently fixing)…and I think this might be the issue as the current implementation of AdditionalProperties handling in the Swift3 codegen is completely broken in that it tries to make the class a subclass of a string collection.

This I would guess is what is causing the {{parent}} to mistakenly return it the type as String.

Comments/thoughts anyone?

[dwoodel]

I think it might be a good idea to add some models in the PetStore example with some additional properties.

[julienfouilhe]

We should check if decoders["\({{parent}}.self)"] exists too. I should have done that.

Done in #5416

[wing328]

I think it might be a good idea to add some models in the PetStore example with some additional properties.

We do have that in petstore-with-fake-endpoints-models-for-testing.yaml, which covers a lot more edge cases.

Currently, the Swift Petstore is using petstore.yaml (the basic one) and I'll create a task to track updating Swift Petstore to use petstore-with-fake-endpoints-models-for-testing.yaml

[wing328]

Created #5419 for tracking

[wing328]

@dwoodel actually swift3 petstore is already using the fake petstore. Please feel free to submit a PR to add additional test cases to fake petstore so as to cover more edge cases.

[wing328]

@julienfouilhe when you've time I wonder if you can make the same change against 2.3.0 branch instead as I got merge conflict when syncing the changes from master to 2.3.0. Thanks!

[julienfouilhe]

@wing328 Done!