zabetak commented on a change in pull request #1875:
URL: https://github.com/apache/hive/pull/1875#discussion_r565377888



##########
File path: ql/src/java/org/apache/hadoop/hive/ql/parse/CalcitePlanner.java
##########
@@ -679,27 +680,15 @@ Operator genOPTree(ASTNode ast, PlannerContext 
plannerCtx) throws SemanticExcept
 
           // Determine if we should re-throw the exception OR if we try to 
mark plan as reAnayzeAST to retry
           // planning as non-CBO.
-          if (e instanceof CalciteSubquerySemanticException || e instanceof 
CalciteViewSemanticException
-              || e instanceof CalciteSubqueryRuntimeException) {
-            // Non-CBO path for CalciteSubquerySemanticException fails with 
completely different error
-            // and masks the original failure.
-            // Non-CBO path for CalciteViewSemanticException would fail in a 
similar way as CBO path.
-            throw new SemanticException(e);
-          }
-
-          boolean isHiveTest = conf.getBoolVar(ConfVars.HIVE_IN_TEST);
-          // At this point we retry with CBO off:
-          // 1) If this is not test mode (common case)
-          // 2) If we are in test mode and we are missing stats
-          // 3) if we are in test mode and a CalciteSemanticException is 
generated
-          reAnalyzeAST = (!isHiveTest || isMissingStats ||  e instanceof 
CalciteSemanticException);

Review comment:
       Personally, I would prefer to make the `CONSERVATIVE` mode fail for 
`RuntimeException`. This would given an incentive to the client to report the 
unknown problem (which could be serious) and at the same time it is not 
blocking since they could switch temporarily to the more permissive `ALWAYS` 
mode. If others feel the same then we can go for the change otherwise we leave 
it as is.




----------------------------------------------------------------
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]

Reply via email to