openinx commented on issue #1951: URL: https://github.com/apache/iceberg/issues/1951#issuecomment-748746287
I toke a look at the https://github.com/apache/iceberg/pull/1956/files, I don't think `spark`'s `Shim` policy will resolve all flink's incompatibility problem, because not all the incompatibility issues are happen in a single interface (or a separate services). Currently, we iceberg flink module depends on few `internal` interfaces (such as `RowDataTypeInfo`, it was removed in flink 1.12.0), which does not guarantee the interface compatibility from apache flink. The correct way is decoupling dependencies on flink's internal API interfaces, for the `TableColumn` & `RowDataTypeInfo` & `TableResult` issues we may need to change the code design so that we won't depend on those internal API interfaces. The idea way is: upgrading the flink version from 1.11.x to 1.12.x (passing all checks) without changing one line code. As the iceberg 0.11.0 release is coming, I'd rather to upgrade the flink version after release get finished, because all currently development and testing work are based on flink 1.11.x. I'm afraid that upgrading to 1.12.x will introduce extra instability which may delay the 0.11.0 release. ---------------------------------------------------------------- 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]
