model_rebuild calls for top level fragments

[mat-sop]

This pr restores model_rebuild calls for top-level fragment models. When a fragment is parsed to many models only 1 (with the same name as the fragment) will get model_rebuild call.

Implementation is complete, but I'm marking this as a draft because I think it's a good opportunity to add a test that checks if generated models can be imported and used with .model_validate(...). Maybe as a part of generated_client or new test that imports models from expected_clients?

resolves #253

[bombsimon]

Thanks for this fix! I updated to 0.11.0 today and hit this issue. Do you want any assistance with writing tests for this? I'm also curious if you have any suggested workaround for this until this is merged, should I just rebuild them manually?

[rafalp]

@bombsimon I'll be taking over this (and other) PRs from Matt who left the company. I'll need to catch up with the code and internals first, but this PR is a priority for me.

[bombsimon]

@rafalp Oh, sad to hear you're losing a great colleague but congrats to Matt I hope! 🎈 And luckily for us using this the codebase is still in good hands with you!

Understandable that this will take some time, let me know if I can do anything to help with the process! And FYI I based my branch with issues on this branch and that works just fine so this solves the problem for me.

[rafalp]

Hello @bombsimon. If you want to pick up this PR to add tests to it, please feel free to do so.

Apologies for taking this long to return to this.

[bombsimon]

@rafalp I'll try to pick this up next week and I'll let you know if I won't make it so you're not waiting for me!

Don't mention it, all your effort is greatly apprecaited!

[bombsimon]

I don't have access to push to this branch so I crated #267 and rebased on main.

I opted to import a model from expected_clients that would not have been fully built without calling model_rebuild() and I'm asserting both on the expected JSON schema and that we can validate all our types as expected. I added a somewhat unrelated but negative test to ensure we don't blindly validate invalid models. Maybe a bit out of scope since this should be tested in the Pydantic module but it might be helpful enough to show what we're trying to assert.

Since there's no way for me to turn of the calls to model_rebuild this test will solely serve as a check to not remove these by accident in the future. I think this one test case should be enough to confirm we generate an expected client.

[rafalp]

Maybe a bit out of scope since this should be tested

Tests like this are useful, they codify assumptions/requirements on the dependency's behavior in tests suite.

[rafalp]

Superseded by #267