Github user srowen commented on the pull request:
https://github.com/apache/spark/pull/2217#issuecomment-54281202
@viirya I also don't quite get why mutability helps the use case you
describe. You seem to want some persistent mutable state per executor. But that
isn't the purpose of a Broadcast variable. I think it sort of looks related
since a Broadcast variable may be shared across many tasks within one JVM, but
that works exactly because it is immutable and it doesn't matter which copy of
the immutable data you look at.
Persistent state would have its own costs. You can no longer freely
schedule tasks on any available executor. You have to deal with loss of the
executor and that state. You could always implement that yourself with an
external storage system if it made sense. It seems at odds with Spark's design
though.
---
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]
