Pregenerate PK

[tsx]

Make PK available before saving object.

[tsx]

Using this increases chances of accidentally creating inconsistency.

[tsx]

@jtharpla would like to use an ID for saving data to S3 before creating corresponding objects in mongo.

[thomasst]

What's the exact use case? Note that we want to move to shorter IDs (UUIDs) for any new collection. The new IDField already supports this behavior with the autogenerate=True option.

https://github.com/closeio/flask-common/blob/master/flask_common/fields/id.py

[wojcikstefan]

Just to make the behavior of the IDField more obvious, here's how it can be used in the wild:

In [4]: class TestDoc(Document):
   ...:     id = IDField(autogenerate=True, prefix='test', primary_key=True)
   ...:     name = StringField()
   ...:

In [5]: TestDoc.objects.create(name='aaa')
Out[5]: <TestDoc: TestDoc object>

In [6]: TestDoc.objects.create(name='bbb')
Out[6]: <TestDoc: TestDoc object>

In [7]: [t.pk for t in TestDoc.objects]
Out[7]: ['test_0Q9FASnMdGR4cVGtbJeRZG', 'test_6S2aaGjfjfefaCPePdroZB']

In [8]: new = TestDoc()

In [9]: new.pk
Out[9]: 'test_0hNrWgYDJ0HqyUdssV92SX'

That last line shows how you can access the PK before you save the document.

[jtharpla]

I ended up coding my stuff a different way where I didn't need to pre-generate the ID

[thomasst]

Let's just reopen if we need this again.

[thomasst]

@tsx Reason for reopen?

[tsx]

I'm thinking I might need it for fixing 10510 in the main repo.

While I see now we have IDField that works better, I can't easily migrate existing documents that use the old RandomPKDocument.

Still digging though...