As brought up on the mailing list: https://lists.apache.org/thread/hscjqyw7lpt95vlkzoslm6pyhy9x6wso
When we resume producing in the source node, we should run the continuation (which starts scanning again) as a new task. This will avoid potential stack overflow (if we pause and start many times) and help make for more understandable stack traces. The continuation is unlikely to have anything to do with the task that is calling resume and so there is not as much downside to a context switch.
Reporter: Weston Pace / @westonpace
Note: This issue was originally created as ARROW-17180. Please see the migration documentation for further details.
As brought up on the mailing list: https://lists.apache.org/thread/hscjqyw7lpt95vlkzoslm6pyhy9x6wso
When we resume producing in the source node, we should run the continuation (which starts scanning again) as a new task. This will avoid potential stack overflow (if we pause and start many times) and help make for more understandable stack traces. The continuation is unlikely to have anything to do with the task that is calling resume and so there is not as much downside to a context switch.
Reporter: Weston Pace / @westonpace
Note: This issue was originally created as ARROW-17180. Please see the migration documentation for further details.