[Bug] Update Pydantic converter example to reflect `with_child_unrestricted` changes

[gregbrowndev]

What are you really trying to do?

I want to use Pydantic and datetime objects as input/outputs/signals/queries to workflows and activities.

Describe the bug

The recommended reference example for the Pydantic converter creates a new sandbox to work around a bug in Pydantic.

There is a TODO in the code below:

https://github.com/temporalio/samples-python/blob/main/pydantic_converter/worker.py

that says:

TODO(cretz): Use with_child_unrestricted when temporalio/sdk-python#254 is fixed and released

The issue in question has now been merged but the example hasn't been updated.

Minimal Reproduction

https://github.com/temporalio/samples-python/blob/main/pydantic_converter/worker.py

Environment/Versions

n/a

Additional context

n/a

Thanks!

[cretz]

Thanks! Agreed we can now perform that TODO.

[baegmon]

Any update on this one? Currently I'm using Pydantic v2 models with:

from pydantic_core import to_jsonable_python

and

workflow_runner=SandboxedWorkflowRunner(
    restrictions=SandboxRestrictions.default.with_passthrough_modules(
        "pydantic",
    )
)

Would love to see an example without seeing warnings from https://github.com/temporalio/sdk-python/blob/main/temporalio/converter.py#L561

[cretz]

That warning is unrelated to this issue. This issue is simply changing how the sandbox is created. That warning is because presumably you are not using a custom converter like https://github.com/temporalio/samples-python/blob/main/pydantic_converter/converter.py and configuring it on the client.

[baegmon]

That warning is unrelated to this issue. This issue is simply changing how the sandbox is created. That warning is because presumably you are not using a custom converter like https://github.com/temporalio/samples-python/blob/main/pydantic_converter/converter.py and configuring it on the client.

I've tried that example actually, with just one difference which is using from pydantic_core import to_jsonable_python over pydantic_encoder for v2 on the converter: https://github.com/temporalio/samples-python/blob/main/pydantic_converter/converter.py#L32

I am still getting the issue on temporalio==1.6.0, do you not get this error?
This is just on a default python 3.12 devcontainer.

[cretz]

We have not tested or created a sample for v2. There is a separate issue at #97 for that. But I think the issue to resolve this unrelated TODO for v1 is unrelated to your issue.

(also transferring this issue to the samples repo since it doesn't belong in primary repo)

[dandavison]

Update: the SDK now contains official Pydantic v2 support, which supports datetime and many other field types, and the sample has been updated. This is the sample to follow for Pydantic v2: https://github.com/temporalio/samples-python/tree/main/pydantic_converter. The old pydantic v1 sample is still in the repo, but if at all possible please use pydantic v2 with the official SDK support.