[FLINK-16055] Avoid catalog functions when listing Hive built-in functions

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.11.0
• Component/s: Connectors / Hive, Tests
• Labels: pull-request-available
• Assignee: Rui Li
• Reporter: Rui Li
• Created: 14/Feb/20 09:50
• Updated: 18/Feb/20 18:34
• Resolved: 18/Feb/20 18:34

Description

When we create catalog functions with hiveShell during tests, the functions get registered with FunctionRegistry. If later on we try to list Hive built-in functions, FunctionRegistry::getFunctionInfo will try to register the catalog functions with a session registry. And therefore fails if we don't have a SessionState in the test. The following example exposes this issue.

	@Test
	public void test() throws Exception {
		hiveShell.execute("create function hiveudtf as 'org.apache.hadoop.hive.ql.udf.generic.GenericUDTFExplode'");
		HiveModule hiveModule = new HiveModule(hiveCatalog.getHiveVersion());
		hiveModule.listFunctions();
	}

Fails with:

Caused by: java.lang.RuntimeException: Function registery for session is not initialized
	at org.apache.hadoop.hive.ql.session.SessionState.getRegistryForWrite(SessionState.java:979)
	at org.apache.hadoop.hive.ql.exec.Registry.registerToSessionRegistry(Registry.java:657)
	at org.apache.hadoop.hive.ql.exec.Registry.getQualifiedFunctionInfoUnderLock(Registry.java:616)
	at org.apache.hadoop.hive.ql.exec.Registry.getFunctionInfo(Registry.java:321)
	at org.apache.hadoop.hive.ql.exec.FunctionRegistry.getFunctionInfo(FunctionRegistry.java:549)
	at org.apache.flink.table.catalog.hive.client.HiveShimV120.getFunctionInfo(HiveShimV120.java:162)
	at org.apache.flink.table.catalog.hive.client.HiveShimV120.lambda$listBuiltInFunctions$0(HiveShimV120.java:142)

Issue Links

• links to: GitHub Pull Request #11093