Overall improvements to parser.

[amw]

This PR is something that I've mentioned long time ago in #290. It adds support for sending list data in views other than RelationshipView allowing users to create bulk operation views (Bulk operations are one of JSON API extensions). But it also fixes a few "500 Server Error" instances where parser tried to use data like a dictionary even when it wasn't one.

Related issue by another user: #367

• Less exceptions caused by data type assumptions
• Better handling of list data
• Does not special-case RelationshipView
• Support meta in Resource Identifier Objects (per spec)

[codecov-io]

Codecov Report

Merging #393 into master will decrease coverage by 0.06%.
The diff coverage is 90%.

@@            Coverage Diff             @@
##           master     #393      +/-   ##
==========================================
- Coverage   91.75%   91.68%   -0.07%     
==========================================
  Files          55       55              
  Lines        2923     2936      +13     
==========================================
+ Hits         2682     2692      +10     
- Misses        241      244       +3

Impacted Files	Coverage Δ
rest_framework_json_api/parsers.py	93.97% <90%> (-3.17%)	⬇️

---

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 e501b04...277990f. Read the comment docs.

[mblayman]

@amw, thanks, this looks pretty cool. If I'm reading your description correctly, this is adding new behavior to the JSONParser. Do you think some new tests could be included to hit some of the extra lines in the parser? Looking at the codecov report, lines 127 and 139 do not appear to be covered.

[amw]

Sure, I'll add some tests. Just wanted to wait for initial feedback. I might have a bit of time during the holiday break.

[sliverc]

Thanks @amw for bringing this PR forward. I have also struggled with bulk operation support with json:api for a while and have been following the different discussions.

What you are suggesting is to implement the experimental bulk extension as described here

This seems to have been deprecated and json:api is properly moving to a completely different approach in version 1.1 as discussed in json-api/json-api#1197

I think though it is still legit to implement bulk extension in DJA as DJA currently targets json:api version "1.0". I think though as it is an experimental feature and media type is different ('application/vnd.api+json; ext=bulk') bulk support logic should be separated into its own parser. Reason being as the standard doesn't allow list primary data when creating or updating resources so the jsonapi parser shouldn't allow it either.

Instead a bulk parser could be designed like this:

import io
import json

from django.conf import settings
from rest_framework_json_api import parsers

class JSONAPIBulkParser(parsers.JSONParser):
    media_type = 'application/vnd.api+json; ext=bulk'

    def parse(self, stream, media_type=None, parser_context=None):
        data = []
        encoding = parser_context.get('encoding', settings.DEFAULT_CHARSET)
        data_collection = json.loads(stream.read().decode(encoding))

        for single_data in data_collection['data']:
            single_data = {'data': single_data}
            sub_data = super(JSONAPIBulkParser, self).parse(
                io.BytesIO(bytes(json.dumps(single_data), encoding)),
                media_type=media_type,
                parser_context=parser_context
            )
            data.append(sub_data)

        return data

For this parser to work a view mixin would be needed which could look something like this:

class BulkExtensionMixin(object):
     def get_serializer(self, *args, **kwargs):
        is_bulk = isinstance(self.request.data, list)
        kwargs['many'] = kwargs.pop('many', is_bulk)
        return super().get_serializer(*args, **kwargs)

This way if a user of DJA wants to use the experimental bulk feature it would need to add new parser to DEFAULT_PARSER_CLASSES and extend views with mixin which actually need bulk support. In the respective class it could be clearly documented that this is an experimental feature.

What do you think?

[sliverc]

Thanks for this.

There are quite a few things happening in this PR and I think do get a better picture it would be better to split it up.

• One PR for better exception handling in parser
• and another PR for a special bulk parser as outlined in above comment

What do you think? Would you have time to work on this?

[sliverc]

@amw
There hasn't been any progress on this since quite a while so I am closing this PR. If you still want to work on this please comment and I can reopen. Or if someone else wants to work on this change feel free to open a new PR.