parthchandra commented on issue #1371: URL: https://github.com/apache/datafusion-comet/issues/1371#issuecomment-4399518101
@coderfender go ahead. Also, from my notes - The incompatibility comes from rounding differences between Comet and Spark. Comet uses the binary representation of the numbers and sometimes these numbers are not represented exactly in IEEE 754. (e.g. 1.005 is actually 1.00499999999999989). Spark represents these as string. With rounding Comet ends up with 1.00, Spark ends up with 1.01. This happens for a very small set of numbers. For deterministic data however, it is a correctness issue. String representation comes at the cost of performance. So now the tradeoff is - faster, but sometimes not exact OR slower but always exact. I would like to have both, and default to the faster version, but others might have different opinions. -- 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]
