boroknagyz commented on pull request #1751: URL: https://github.com/apache/iceberg/pull/1751#issuecomment-725676838
Thanks @rdblue for your response. Impala already has read support for Iceberg tables (that are stored in HadoopTables or HadoopCatalog). (Btw Impala can already write non-partitioned Iceberg tables). Impala uses HMS to store table metadata and uses the InputFormat class of the storage descriptor to determine the table format. So it doesn't use these classes, but only their names to identify the table type. Currently Impala uses class names from Hiveberg to identify Iceberg tables: https://github.com/apache/impala/blob/ea13e744977c71206c11a54f65d77aeecb55703d/fe/src/main/java/org/apache/impala/catalog/HdfsFileFormat.java#L80 But there's already an upstream change request to use the new classes from Iceberg: https://gerrit.cloudera.org/#/c/16707/ I don't think this change would introduce missing classes in Hive because HiveTableOperations already sets org.apache.iceberg.mr.hive.HiveIcebergSerDe which is in the same package. I think this change would help interoperability between Impala and Hive as they would create tables the same way, What do you think? ---------------------------------------------------------------- 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]
