Github user srowen commented on the issue:
https://github.com/apache/spark/pull/15512
Hm, if the benchmark you give generalizes much that is certainly
compelling. I guess I'm surprised that instantiating the object can be so
expensive relative to deserialization since it just happens once per task.
But it is a fairly simple change. `ThreadLocal` avoids thread safety issues
though I do wonder if the serializers can hold on to state that would make this
a source of memory leak?
Maybe ... @squito or @zsxwing has a thought?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
