Fix deepcopy of primitive types

[ndrluis]

The IcebergRootModel inherits from Pydantic RootModel, which has its own implementation of deepcopy. When deepcopy runs, it calls this __deepcopy__ method and ignores that it's a Singleton. So, my solution was to change the order of inheritance and implement a __deepcopy__ method for singletons that returns itself.

https://github.com/pydantic/pydantic/blob/f024d03b832d1bbcbadf76184ed14d92571a71ca/pydantic/root_model.py#L108-L116

[sungwy]

This is really cool 🔥 - thank you for putting in this bug fix so quickly!

This is a nit: I think we could add a test to test_metadata.py to deepcopy a TableMetadata object and ensure that the output is equal to the original one with a Schema representation that contains a mix of the issue affected types.

Theoretically this will work, but its never a bad thing to add an explicit test to demonstrate that the issue is resolved. But other than that, this looks great! Such a speedy resolution of a rather opaque issue 🚀

[ndrluis]

Thank you, @syun64. I have added the test and used an existing metadata definition to avoid duplication.

Regarding the solution, to be honest, it was not quick. It took me several weeks of learning and then stepping away from the problem for a few weeks to return with fresh ideas on how to solve it.

[sungwy]

Hi @HonahX @Fokko could we ask for your help in triggering this workflow?

[kevinjqliu]

Thanks for addressing this 🐛!

[kevinjqliu]

LGTM!

@syun64 wdyt

[sungwy]

Yep! Fantastic fix 💯

[ndrluis]

@HonahX @Fokko, can you please trigger the workflow? I have already run the tests in my fork and everything is green.

[Fokko]

@ndrluis Thanks for pinging me here. This approval should be gone once the first PR has been merged. Which should be soon 🚀

[HonahX]

Thanks for the great fix!

This approval should be gone once the first PR has been merged.

@Fokko I might have lost track. Which PR are you referring to as the first one?

[Fokko]

@HonahX Sorry for the ambigous comment there. I was referring to approving the CI to run, which is very annoying. Now @ndrluis has his first PR in, this approval should be gone and the CI should be kicked off right away.

[Fokko]

Thanks @ndrluis for working on this, and @syun64, @kevinjqliu and @HonahX for the quick review 🚀