timsaucer commented on PR #1311: URL: https://github.com/apache/datafusion-python/pull/1311#issuecomment-3583151474
I have a suspicion that what is happening here is that we're uncovering this feature working in pyo3 0.25 but not as intended. I suspect that what is happening is that we're getting a keyboard exception raised in the record batch reader rather than in the `wait_for_future`. My reasoning is this: If it were our code that is catching the keyboard raised then we *should* be getting the datafusion wrapped error and not the keyboard exception. By putting in these functions that are holding the gil longer we actually able to catch those signals. I bet that there is some interaction happening in which our tokio runtime attaching to the gil is actually letting the python thread get that keyboard interrupt into the record batch reader somehow. This is more of a feeling than anything I've discovered in the code changes from 0.25 to 0.26. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
