A library serializer `__init__` calls `kwargs['context']['request']` - how to skip/ignore?

Describe the bug We use a package called django_comments_xtd. In it, django_comments_xtd.api.serializers.ReadCommentSerializer.__init__ calls self.request = kwargs['context']['request']. How can I work round this or simply skip introspection of this class?

from drf_spectacular.plumbing import get_lib_doc_excludes as original_get_lib_doc_excludes

def get_lib_doc_excludes():
    from django_comments_xtd.api import serializers as comment_serializers
    from django_comments_xtd.api import views as comment_views

    return (
        original_get_lib_doc_excludes()
        + [getattr(comment_serializers, c) for c in dir(comment_serializers) if isclass(c)]
        + [getattr(comment_views, c) for c in dir(comment_views) if isclass(c)]
    )

SPECTACULAR_SETTINGS = {
    "GET_LIB_DOC_EXCLUDES": "path.to.above.build_docs.get_lib_doc_excludes",
}

To Reproduce pip install django_comments_xtd ./manage.py spectacular --file openapi.yml

Expected behavior I need a way of skipping this serializer. I have tried returning the class from a custom get_lib_doc_excludes, or all of the package’s view classes. I have tried to figure out a way with OpenApiSerializerExtension. I have even tried monkey-patching. Apologies if I missed something.

[chrispreee]

Amazing thank you. Have updated my pip requirement to last commit on master:

git+https://github.com/tfranzel/drf-spectacular.git@3292d9db957aaa60ccfeb8e14d91ac9ab42d4110

Will post back here once this is released (will be version of drf-spectacular after 0.24.2).

I gather the extension isn’t a wholesale replacement of the class then, given even with the init fix it still raises the context KeyError on 0.24.2.

I can break packages as we only ever use drf-spectacular as a management command (as our API is too large to introspect live), and I only apply these fixes then, but forgot this is not standard usage.

Enormous thanks for sticking with this, your help is very much appreciated, and I’ve learned several useful new things.

[tfranzel]

you also may want to use

SPECTACULAR_SETTINGS = {
    'PREPROCESSING_HOOKS': ['drf_spectacular.hooks.preprocess_exclude_path_format'],
}

to exclude the duplicated format endpoints.