David Hewitt
David Hewitt
Maybe merging #1480 will give us all we need to measure here.
Ungh, will investigate the test failures when I get a chance.
Having just investigated here, my summary is: 1. IMO this is working as intended; it makes all fields of all models serialize as any. 2. But this has the downside...
Yes, I think let's rebase and seek to merge. What we should do longer term is: - Move `pydantic-core` into the `pydantic` repo so we can lean into more aggressive...
Ok, this looks like it should be good to merge now. What I've done is introduced indirection in `CombinedSerializer` so that any call to a serializer checks if the runtime...
Now it does 😉
Even weirder, the schema we build for `Wrong` is a union of tagged unions: ``` {'type': 'union', 'choices': [{'type': 'tagged-union', 'choices': {'A': {'type': 'definition-ref', 'schema_ref': '__main__.A1:93926943463008', 'metadata': {}}}, 'discriminator': 'tag',...
Performance diffs do not seem concerning to me; clicking through to codspeed there is a mix of positive and negative changes to about the same scale. Serialization was probably impacted...
This is getting big enough that I'm going to halt here; I'm just doing mechanical changes so I'm convinced the upgrade is ok. Will continue it after PyO3 0.23 actually...
All is now merged into main, with last piece at #1584