[
https://issues.apache.org/jira/browse/HIVE-26417?focusedWorklogId=794165&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-794165
]
ASF GitHub Bot logged work on HIVE-26417:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 22/Jul/22 10:31
Start Date: 22/Jul/22 10:31
Worklog Time Spent: 10m
Work Description: kasakrisz commented on code in PR #3461:
URL: https://github.com/apache/hive/pull/3461#discussion_r927521734
##########
ql/src/java/org/apache/hadoop/hive/ql/parse/MergeSemanticAnalyzer.java:
##########
@@ -62,11 +62,10 @@ protected ASTNode getTargetTableNode(ASTNode tree) {
@Override
public void analyze(ASTNode tree, Table targetTable, ASTNode tableNameNode)
throws SemanticException {
- if (tree.getToken().getType() != HiveParser.TOK_MERGE) {
- throw new RuntimeException("Asked to parse token " + tree.getName() + "
in " +
- "MergeSemanticAnalyzer");
+ boolean nonNativeAcid = AcidUtils.isNonNativeAcidTable(targetTable);
+ if (nonNativeAcid) {
Review Comment:
It is already implemented and used independently from this patch.
`HiveIcebergStorageHandler.fallbackToNonVectorizedModeBasedOnProperties` is
called when FSOperator "augments the plan"
https://github.com/apache/hive/blob/4146d220bfc0a29d117676844a4c7021fcc6a30d/ql/src/java/org/apache/hadoop/hive/ql/exec/FileSinkOperator.java#L1612
Issue Time Tracking
-------------------
Worklog Id: (was: 794165)
Time Spent: 0.5h (was: 20m)
> Iceberg integration: disable update and merge iceberg table when split update
> is off
> ------------------------------------------------------------------------------------
>
> Key: HIVE-26417
> URL: https://issues.apache.org/jira/browse/HIVE-26417
> Project: Hive
> Issue Type: Improvement
> Components: File Formats
> Reporter: Krisztian Kasa
> Assignee: Krisztian Kasa
> Priority: Major
> Labels: pull-request-available
> Fix For: 4.0.0
>
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> Iceberg table update and merge is implemented using split update early by
> HIVE-26319 and HIVE-26385.
> Without split update early deleted records has to be buffered in memory when
> updating iceberg tables. With split update early deleted records are
> processed by a separate reducer and no buffering is required. The ReduceSink
> operator also sorts the records.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
