fix: handle ServiceConflictError when reusing Actor across sequential context

[Mantisus]

Description

• handle ServiceConflictError when reusing Actor across sequential context

Issues

• Closes: Make Actor consistently reusable without raising ServiceConflictError #678

Testing

• Added new unit test

[Copilot]

Pull request overview

This PR fixes ServiceConflictError when reusing Actor across sequential async context manager blocks. The issue occurred when mixing Actor (singleton proxy) and Actor() (new instances) in sequential contexts, as new instances would try to register services in the global service_locator that were already set by previous instances.

Changes:

• Added error handling in event_manager property to catch ServiceConflictError and reuse existing event manager from service_locator
• Added parametrized test covering all four combinations of Actor/Actor() usage in sequential contexts

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
src/apify/_actor.py	Wraps service_locator.set_event_manager() in try/except to handle ServiceConflictError and reuse existing event manager
tests/unit/actor/test_actor_lifecycle.py	Adds parametrized test for all Actor/Actor() combinations in sequential contexts

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Pijukatel]

I am not so sure about the whole issue. I am afraid that allowing this implicitly leads to more edge cases. I would rather be explicit about such use case. Why not solve the actual issue (problem with example-code-runner-py running Actor() twice due to its unique use case)

apify/crawlee-python#1759
and
https://github.com/apify/example-code-runner-py/compare/master...reset-service-locator-if-custom-actor

[Mantisus]

Perhaps @vlada knows of some additional use cases so that we can better shape the expected behavior.

I followed the general pattern for resources in service_locator. But if the second context manager is expected to be opened with clean resources, then service_locator should be completely cleared when the Actor's context manager is closed.

[janbuchar]

Why not solve the actual issue (problem with example-code-runner-py running Actor() twice due to its unique use case)

apify/crawlee-python#1759 and apify/example-code-runner-py@master...reset-service-locator-if-custom-actor

Now imagine that somebody copies a snippet with repeated async with Actor: statements into their codebase and it stops working because the code depends on the example-code-runner-py. That aside, I think that the example code runner is just one example (haha), perhaps there are other valid use cases for this.

[Pijukatel]

Now imagine that somebody copies a snippet with repeated async with Actor: statements into their codebase and it stops working because the code depends on the example-code-runner-py.

Do we have such a snippet somewhere? example-code-runner-py is private, so I hope nobody ever gets the idea about repeated async with Actor: being some good pattern.

As far as I know, we do not advertise this usage anywhere, or?