marton-bod commented on a change in pull request #2191:
URL: https://github.com/apache/iceberg/pull/2191#discussion_r569313944
##########
File path:
mr/src/test/java/org/apache/iceberg/mr/hive/TestHiveIcebergStorageHandlerLocalScan.java
##########
@@ -253,6 +253,7 @@ public void testCreateTableWithColumnSpecification() throws
IOException {
testTables.locationForCreateTableSQL(identifier);
runCreateAndReadTest(identifier, createSql,
HiveIcebergStorageHandlerTestUtils.CUSTOMER_SCHEMA,
PartitionSpec.unpartitioned(), data);
+ shell.executeStatement("DROP TABLE " + identifier);
Review comment:
> We already have a test case for testCreateDropTable, so I think this
part is covered. Do I miss something? Maybe we should create one where we add
some data as well, if it is relevant
That test case does not detect the difference of behaviour between Hive2/3
and Hive4 with regards to purge=TRUE, because if we create a table and not to
do any additional snapshot-creating operations on the table, then
`metadata.snapshots()` will give back an empty collection, hence the
deleteFiles operation will not run during `dropTableData` (and you won't get
FileNotFoundExc for Hive4). That's why I added the DROP TABLE to some tests
where additional operations (like inserts) also take place.
On the other hand, not sure why `metadata.snapshots()` gives back an empty
collection in this case.
> Could we create specific test cases instead for dropping the Iceberg table
backing Hive table and then dropping the Hive table? (should only run for
non-HiveCatalog tests)
Yes, will do.
----------------------------------------------------------------
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]
