UUID representation mismatch with pymongo 4.0.1

Working with Data Drivers
1

I’m upgrading from pymongo 3.12 to pymongo 4.0.1, and I’m running into some weird behaviour with the UUID representation. The database connection is configured with CodecOptions(uuid_representation=4).

There is a difference between return values when using normal insertion and when using bulk insertion:

# bulk gives binary representation
doc = {'_id': uuid.uuid4()}
result = collection.bulk_write([ReplaceOne(doc, doc, upsert=True)])
result.inserted_ids
> {0: Binary(b'\xb6\r%\xf2\xadOMo\x919\xc4\x98P\xe5\n\x02', 4)}
# normal insertion gives UUID representation
result = collection.insert_one({'_id': uuid.uuid4()})
result.inserted_id
> UUID('1ec97f53-b1ae-471b-9e0c-157ca9b02953')

I was wondering if this is by design and if there is a way to make the bulk result also return the python UUID representation.

2

Thanks for reporting this issue, it is indeed a bug. I’ve opened a bug report here: https://jira.mongodb.org/browse/PYTHON-3075

Note that PyMongo is still sending the expected UUID format to the server, it’s just the client side upserted_ids field which is unexpected. A temporary workaround for this bug would be to convert the Binary into the expected type:

import uuid
from typing import Any, Dict, Mapping
from bson.binary import Binary, UuidRepresentation
from pymongo import MongoClient
from pymongo.collection import Collection
from pymongo.operations import ReplaceOne

def convert_ids(upserted_ids: Mapping[int, Any], coll: Collection) -> Dict[int, Any]:
    """Temporary workaround for https://jira.mongodb.org/browse/PYTHON-3075.

    Use like this::

        result = collection.bulk_write([...])
        upserted_ids = convert_ids(result.upserted_ids, collection)

    """
    res = {}
    rep = coll.codec_options.uuid_representation
    for idx, _id in upserted_ids.items():
        if rep == UuidRepresentation.UNSPECIFIED:
            if isinstance(_id, uuid.UUID):
                _id = Binary.from_uuid(_id)
        elif isinstance(_id, Binary):
            _id = _id.as_uuid(rep)
        res[idx] = _id
    return res

client = MongoClient(uuidRepresentation='standard')
collection = client.test.test
doc = {'_id': uuid.uuid4()}
result = collection.bulk_write([ReplaceOne(doc, doc, upsert=True)])
for _id in convert_ids(result.upserted_ids, collection).values():
    assert isinstance(_id, uuid.UUID)

Please follow the jira ticket for updates.

One question though, are you using pymongo directly or with a wrapper library? I ask because BulkWriteResult doesn’t have an inserted_ids property (it does have an upserted_ids property).

2 Likes
3

Thank you for the workaround and the bug report! You’re right it was upserted_ids, I apparently did something weird with the copy pasting and formatting to make it look nice…