fix: revert Pydantic custom serializer

[korikuzma]

• The Pydantic custom serializer resulted in model_dump to only include minimal fields for ga4gh serialization. This caused issues in downstream apps that leverage FastAPI

[korikuzma]

Oh, I forgot to check __hash__

[ahwagner]

Yeah, I guess... 😞

It makes sense to not muck about with the "as json" return, though ideally we would be able to leverage the Rust code in pydantic core for serialization, and not need to introduce a new dependency.

That said, usability is secondary to performance, so I agree this is an improvement. 👍

[korikuzma]

@ahwagner I updated the hash. Lmk if this is what you think it should be?

[ahwagner]

@korikuzma what you wrote is fine. __hash__ is helpful for comparison operations and indexing, but doesn't play a role in any of the GA4GH identifier methods as far as I can remember. That this now reflects the ga4gh_digest method takes a little more compute that what might be strictly necessary, but I doubt it will be meaningful in terms of regular usage. I think we can move forward as is.

[korikuzma]

@ahwagner I can revert and use a different hash function, I was just basing off of the current implementation in main. @theferrit32 originally made the issue in #225 , so do you have any thoughts?

[theferrit32]

The __hash__ function is used by python's set type.

@korikuzma I think this is enough for #225 as long as the only types that will be in arrays with uniqueItems: true in the schema are identifiable ones. Actually at the moment I'm not seeing any uniqueItems: true in the schemas. I think previously it was used in the Haplotype and Genotype classes.

[korikuzma]

@theferrit32 ok, I will go ahead and merge and close #225