Bug: Router context not being available when using AppSync with async resolvers

[bjorger]

Expected Behaviour

When accessing the router.context in a resolver, with async resolvers, it should not be an empty dict, but much rather expose the context injected with app.append_context

Current Behaviour

When using router.context in async resolvers, the router.context is an empty dict

Code snippet

def lambda_handler(event, context):
    token = extract_auth_token_from_event(event)
    logger.info("token", extra={"token": token})
    app.append_context(
        isAdmin=True,
    )
    result = app.resolve(event, context)

    return asyncio.run(result)

@router.resolver(type_name="Mutation", field_name="addPet")
async def addPet(pet: dict):
    logger.info("Handling add pet mutation")
    logger.info("Context", extra={"resolver_context": router.context}) // this shows that router.context is an empty dict.
    
    return { "name": "Peter", "type": "Cat" }

Possible Solution

No response

Steps to Reproduce

Just plug in the lambda_handler and mutation I attached above, deploy it and run the mutation via the AppSync console

Powertools for AWS Lambda (Python) version

latest

AWS Lambda function runtime

3.8

Packaging format used

Lambda Layers

Debugging logs

No response

[boring-cyborg]

Thanks for opening your first issue here! We'll come back to you as soon as we can.
In the meantime, check out the #python channel on our Powertools for AWS Lambda Discord: Invite link

[leandrodamascena]

Hey @bjorger! Thanks for opening this bug. I was able to reproduce this error and I can confirm this is a bug.

The problem is because we need to clean the context after the execution to prevent it from remaining in memory on the next run, but it looks like we have a problem with coroutines because it is being cleaned up before the coroutine runs.

I'm adding this to the backlog to investigate more this or next week. If you have time and found a solution before me, I'd love you submit a PR to fix this.

[leandrodamascena]

Hi @bjorger! I spent some time today investigating this issue and I don't know if we can fix it without a refactoring that will be a breaking change for our customers. The problem here is because we have no control over the evenloop or tasks that will be scheduled by asyncio and we clean up the context before the coroutine starts. I would try to start a new eventloop to control it, but I have no idea about the consequences and I don't know if that works.

To be honest I think this is an error in our implementation, we should have control over the async execution and execute it inside the AppSyncResolver and not the customer doing that. If you see our async_batch_resolver function we do what I said and we don't have this bug.

But I have an alternative for this. Not the best developer experience, but it solves the problem and we can refactor this in v4.

1 - I will check if the function resolver is a coroutine here and if so, I will not call self.clear_context(). Then we can preserve the context.
2 - This solution creates a side effect that the context can persist across multiple invocations, but I don't see this breaking customers today because this is not working anyway in async invocation.
3 - To solve the problem in the item 2, if the customer does not want to preserve the context across multiple invocations (there are cases where customers can preserve it), the customer should invoke self.contex.clear() after the resolver is executed.
4 - I will update the documentation to make it clear for customers.

I'm still open to hearing other opinions/ideas to fix this, but if we don't have any, what do you think of this solution?

Thanks