[Pulp-dev] serializers in pulp 3 can't use write_only fields

[Daniel Alley]

I just want to point out that we have a somewhat similar kind of feature
implemented, "minimal serializers", which uses the 2-serializer approach.
It is a very tiny amount of extra code/work.

e.g.
https://github.com/pulp/pulpcore/blob/master/pulpcore/app/serializers/task.py#L119
https://github.com/pulp/pulp_rpm/blob/master/pulp_rpm/app/serializers.py#L290

the implementation
https://github.com/pulp/pulpcore/blob/master/pulpcore/app/viewsets/base.py#L114-L139

I lean towards option 2 as well for consistency.

On Mon, Apr 27, 2020 at 5:20 AM Ina Panova <ipanova at redhat.com> wrote:

> I like option number 2 more, even if there is more work behind the scene
> it is more explicit and does not leave opened questions like ' why certain
> fields are null'
> --------
> Regards,
>
> Ina Panova
> Senior Software Engineer| Pulp| Red Hat Inc.
>
> "Do not go where the path may lead,
>  go instead where there is no path and leave a trail."
>
>
> On Fri, Apr 24, 2020 at 5:24 PM Dennis Kliban <dkliban at redhat.com> wrote:
>
>> Today during open floor we concluded that write_only fields cannot be
>> used in serializers because any automated handling of them during OpenAPI
>> schema generation will cause mutations to the schema which will result in
>> backwards incompatible changes to our bindings users.
>>
>> This change should only affect the Content serializers that have a 'file'
>> and 'repository' write_only fields that can be used for uploading content
>> into a repository in a single request. Two solutions were proposed:
>>
>> 1) Use a single serializer for GET and POST and leave 'file' and
>> 'repository' null for GET requests. This solution is the simplest for
>> plugin writers because they don't have to do anything extra. However, it
>> does cause users to wonder why those fields are present and null on GET
>> requests.
>>
>> 2) Require plugin writers to provide 2 serializers for Content - one for
>> GET and another for POST. This is more work for plugin writers, but will
>> make it clear to users what fields are expected in each case.
>>
>> Even though I initially proposed 1, I am in favor of adopting 2. What
>> does everyone else think?
>>
>> We will need to provide validation at startup that will check serializers
>> for presence of write_only fields. The application should fail to start if
>> any are present. This way plugin writers will know not to use the fields.
>> We will also need to provide plugin writer docs that clearly state the
>> guidelines around write_only fields and suggestions for getting by without
>> them.
>> _______________________________________________
>> Pulp-dev mailing list
>> Pulp-dev at redhat.com
>> https://www.redhat.com/mailman/listinfo/pulp-dev
>>
> _______________________________________________
> Pulp-dev mailing list
> Pulp-dev at redhat.com
> https://www.redhat.com/mailman/listinfo/pulp-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/pulp-dev/attachments/20200427/5bd26384/attachment.htm>