fix #1161 #1144 followup

[soyuka]

Q	A
Bug fix?	yes
New feature?	no
BC breaks?	no
Deprecations?	no
Tests pass?	yes
Fixed tickets	#1161
License	MIT
Doc PR	na

@teohhanhui I don't see the point of setting up a complex visited array for this task. I mean, I could populate an array like in your example, but if I then want to use it to compute the RouteCollection, won't I do the job twice?

[teohhanhui]

I hope we can add such tests to verify that it works properly. At a glance, it seems like we'd not get the intended results.

[soyuka]

Because the tracking array is per $rootResourceClass this should be just fine. Lmk what more tests you have in mind. I like the tests on the Routing\ApiLoader but I might be able to do better!

[teohhanhui]

I see! That means it should work the same.

The only problem is the order of traversal then. I imagined it would only work properly if we do a BFS instead of a DFS, because otherwise deeper paths would prevent shallower paths from being created. Take for example with the following relations to be exposed as subresources:

Product isRelatedTo Product
Product isSimilarTo Product

With BFS, we'd have:

• api_products_is_related_to_get_subresource
  /products/{id}/is_related_to
• api_products_is_similar_to_get_subresource
  /products/{id}/is_similar_to

But with DFS, we get instead:

• api_products_is_related_to_get_subresource
  /products/{id}/is_related_to
• api_products_is_related_to_is_similar_to_get_subresource
  /products/{id}/is_related_to/{isRelatedToId}/is_similar_to

Which is not what we want.

[teohhanhui]

Let's add spaces to aid debugging? IMO it's not worth the memory savings lol...

[soyuka]

sure lol

[dunglas]

Thanks @soyuka!