danielcweeks commented on issue #1485: URL: https://github.com/apache/iceberg/issues/1485#issuecomment-698436564
It feels like the biggest issue here is that behavior of cache is not consistent or well-defined (as far as I can tell). I believe the operation originated from RDDs which were intended to be immutable. So that would imply that the cache would not change regardless of what happens externally. The intent was that a cache is fault tolerant and will be reconstituted from the original transforms. That would imply that Sparks behavior for built in tables is incorrect. (You're correct, I was only talking about the first case). Presto and Hive would treat a view as a view and always resolve back to the original table (the only case where this would not be true is in context of a transaction, but that's not implemented). The second case feels more clear to me. A view should always reflect the current state of the underlying sources, so I would expect what you laid out in the example to be the expected behavior (though it currently doesn't behave that way). ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
