Allow get_serializer_class to be overwritten for related urls without defining serializer_class fallback

[SafaAlfulaij]

Description of the Change

This fixes the fact that getting related view's serializer gets it from the super class, rather than the current class.
Bug can be reproduced by using:

    url(
        r"^entries/(?P<pk>[^/.]+)/(?P<related_field>\w+)/$",
        EntryViewSet.as_view({"get": "retrieve_related"}),
        name="entry-related",
    ),

instead of:

    url(
        r"entries/(?P<entry_pk>[^/.]+)/comments$",
        CommentViewSet.as_view({"get": "list"}),
        name="entry-comments",
    ),

Navigating to entries/1/comments/ now will result in an error stating that no get_serializer_class has been defined, even though it exists.

Checklist

• PR only contains one change (considered splitting up PR)
• unit-test added updated
• documentation updated
• CHANGELOG.md updated (only for user relevant changes)
• author name in AUTHORS

[codecov]

Codecov Report

Merging #904 (9275ae1) into master (51daed1) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #904   +/-   ##
=======================================
  Coverage   97.67%   97.67%           
=======================================
  Files          58       58           
  Lines        3101     3101           
=======================================
  Hits         3029     3029           
  Misses         72       72           

Impacted Files	Coverage Δ
tests/test_views.py	100.00% <0.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 51daed1...9275ae1. Read the comment docs.

[sliverc]

This fixes the fact that getting related view's serializer gets it from the super class, rather than the current class.

The PR description states that this PR is addressing issue that related view does not properly get parent serializer class (it seems to be another clean up after #860). I guess this change addresses this issue.

So at a quick glance for me it is not quite clear why introducing of a new links parameters is needed to fix this issue. Can you elaborate?

[SafaAlfulaij]

Explanation 2:
We need to add this parameter to specify the the related link for hyperlink fields (which can, most of the time, have a different related name).
For example in the example serializers, the field_name is comments_hyperlinked:

comments_hyperlinked = relations.HyperlinkedRelatedField(
        related_link_view_name="entry-comments",

And the related view is:

    url(
        r"entries/(?P<entry_pk>[^/.]+)/comments$",
        CommentViewSet.as_view({"get": "list"}),
        name="entry-comments",
    ),

So comments_hyperlinked relates to comments.
But here, we get the field_names, which will create such a url: /entries/1/comments_hyperlinked, which isn't really what we want.

[SafaAlfulaij]

Explanation 3:
This was found out by testing, and I think it's because of dynamic fields.
For example, here we remove the "featured" field when the serializer is initialized, but it exists in the serializer class itself, and therefore being checked.

[sliverc]

Thanks for your explanation. When reading through it seems that you are trying to address 3 different issues in one PR (each explanation 1).

1. the actual issue of the super class
2. support of hyperlinked related field
3. another bug

To ensure that each single issue is properly tested and better way of reviewing could you split up this PR? Out of urgency point number 1 would be high priority as it is a regression (I will see whether I will get to that as well at some point later on if you won't get to it before me).

[sliverc]

I added a test (resp. updated as before the test was invalid) and added changelog entry for the parent super class call issue.

Merging this. For the other two issue feel free to either create another PR or issue. Thanks for your testing effort - really appreciated that you found those hidden away bugs in the latest version.

[SafaAlfulaij]

I checked the issue with openapi schema generation, and it seems to be with related views as well.
I don't know now how to reproduce it, or if it's even valid.
And as for the hyperlinked related field, I'm not using those a lot, and as they are also related to related views, which I don't prefer, I'm not confident enough to work on it.

I'll open to issues about those problems, and let experts handle them :)