Prevent panicking when __dict__ changes during iteration

[alexmojaki]

Change Summary

Gather the items of __dict__ into a SmallVec before iterating during serialization, so that modifications to the __dict__ in the process (e.g. from a @cached_property called by a field serializer) don't cause iteration to panic.

Replaces #1093

Related issue number

FIxes pydantic/pydantic#7832

Checklist

• Unit tests for the changes exist
• Documentation reflects the changes where applicable
• Pydantic tests pass with this pydantic-core (except for expected changes)
• My PR is ready to review, please add a comment including the phrase "please review" to assign reviewers

[codspeed-hq]

CodSpeed Performance Report

Merging #1196 will degrade performances by 16.76%

_{Comparing dict-size2 (e560eef) with main (cf42511)}

Summary

❌ 1 regressions
✅ 147 untouched benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	main	dict-size2	Change
❌	test_small_class_core_dict	32.2 µs	38.7 µs	-16.76%

[samuelcolvin]

I agree panicking isn't good, but I don't like the performance regression, and also worried about hampering future performance improvements.

@davidhewitt could we get this to raise an exception instead of panicking? Would that be a better way forward?

[samuelcolvin]

Sorry, I've reviewed the issue now.

Maybe this is the right idea, then we work on FastModel to improve performance significantly.

[samuelcolvin]

Oh, also the benchmark that's changed is unrelated.

[davidhewitt]

I think this is the right approach too.

In PyO3 upstream the initial reaction is that this sort of panic is acceptable because it is a bug in user code (in this case, the bug was here in Pydantic) so I think for now this is the path forward.