Fixed issues with schema name collisions #5486
Conversation
|
OK. This looks interesting. I need to review properly. Can you describe what’s going on with LinkNode and the helper methods? How does this affect the output schema? How does it affect the rendered docs? 👍🏽 |
|
@carltongibson I optimized my solution a bit. So the issue was that nodes in paths tree could be either OrderedDict or Link - which was causing issue in two cases:
So what I've done, I've made every node in a tree a LinkNode (ordered dict with links array). So each node now can be both a link and a subpath. Also, because I've got array of links, now we can have multiple links in same node (same path said in other way). How it affects the schema? It doesnt use retrieve , read etc as keys now, its just get_0, get_1, post_0 and so on. How it affect docs? I've checked swagger only because this is what Im working with and it works fine, for other UIs - I dont know. |
tests/test_schemas.py
Outdated
| @@ -446,17 +446,17 @@ def test_schema_for_regular_views(self): | |||
| coreapi.Field('ordering', required=False, location='query', schema=coreschema.String(title='Ordering', description='Which field to use when ordering the results.')) | |||
| ] | |||
| ), | |||
| 'custom_list_action': coreapi.Link( | |||
| 'get_1': coreapi.Link( | |||
There was a problem hiding this comment.
Can we avoid all these renames?
There was a problem hiding this comment.
Surely not every time because that is a part of the issue, they clash sometimes. Are they used anywhere? I thought they are just ids
There was a problem hiding this comment.
They're used in the interactive API documentation.
Can we keep them as are if there is no collision and only append a number when there is more than one?
i.e. here, if there were a collision, we'd have custom_list_action and custom_list_action_1 and so on.
mlubimow
force-pushed
the
master---schema-naming-collision
branch
from
30553f7 to
50b3c05
Compare
|
@carltongibson should be fine now |
tests/test_schemas.py
Outdated
|
|
||
| def test_is_list_view_recognises_retrieve_view_subclasses(): |
There was a problem hiding this comment.
Why have you removed this test case? I assume it's a mistake...
It's for a separate issue. #5165
There was a problem hiding this comment.
OK. This is good.
Pulling it into a project using DRF's interactive docs, it behaves well.
Can you just replace the test case from #5165 at the bottom?
I think then we're good to go.
Great one. Thanks.
tests/test_schemas.py
Outdated
|
|
||
| def test_is_list_view_recognises_retrieve_view_subclasses(): |
There was a problem hiding this comment.
We need this test case.
There was a problem hiding this comment.
strange its got removed, I will add it ofc
|
@carltongibson Done! |
If urls include elements in
['retrieve', 'list', 'create', 'update', 'partial_update', 'destroy']schema generation can fail (if a nested item clashes with a previously definedLink) or replace other link.PR fixes #5484 and #4704.
Also took unit tests from another PR targeting this issue: #5464