[jira] [Work logged] (HIVE-26771) Use DDLTask to created Iceberg table when running ctas statement

[ASF GitHub Bot (Jira)]

     [ 
https://issues.apache.org/jira/browse/HIVE-26771?focusedWorklogId=830047&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-830047
 ]

ASF GitHub Bot logged work on HIVE-26771:
-----------------------------------------

                Author: ASF GitHub Bot
            Created on: 30/Nov/22 12:58
            Start Date: 30/Nov/22 12:58
    Worklog Time Spent: 10m 
      Work Description: kasakrisz commented on code in PR #3802:
URL: https://github.com/apache/hive/pull/3802#discussion_r1035938326


##########
iceberg/iceberg-handler/src/main/java/org/apache/iceberg/mr/hive/HiveIcebergMetaHook.java:
##########
@@ -236,11 +236,25 @@ public void 
commitCreateTable(org.apache.hadoop.hive.metastore.api.Table hmsTabl
       
setFileFormat(catalogProperties.getProperty(TableProperties.DEFAULT_FILE_FORMAT));
 
       String metadataLocation = 
hmsTable.getParameters().get(BaseMetastoreTableOperations.METADATA_LOCATION_PROP);
+      Table table;
       if (metadataLocation != null) {
-        Catalogs.registerTable(conf, catalogProperties, metadataLocation);
+        table = Catalogs.registerTable(conf, catalogProperties, 
metadataLocation);
       } else {
-        Catalogs.createTable(conf, catalogProperties);
+        table = Catalogs.createTable(conf, catalogProperties);
       }
+
+      if 
(!Boolean.parseBoolean(catalogProperties.getProperty(hive_metastoreConstants.TABLE_IS_CTAS)))
 {
+        return;
+      }
+
+      // set this in the query state so that we can rollback the table in the 
lifecycle hook in case of failures
+      SessionStateUtil.addResource(conf, InputFormatConfig.CTAS_TABLE_NAME,
+              catalogProperties.getProperty(Catalogs.NAME));
+
+      String tableIdentifier = catalogProperties.getProperty(Catalogs.NAME);

Review Comment:
   done



##########
iceberg/iceberg-handler/src/main/java/org/apache/iceberg/mr/hive/HiveIcebergSerDe.java:
##########
@@ -120,16 +114,11 @@ public void initialize(@Nullable Configuration 
configuration, Properties serDePr
         this.tableSchema = hiveSchemaOrThrow(e, autoConversion);
         // This is only for table creation, it is ok to have an empty 
partition column list
         this.partitionColumns = ImmutableList.of();
-        // create table for CTAS
-        if (e instanceof NoSuchTableException &&
-                
Boolean.parseBoolean(serDeProperties.getProperty(hive_metastoreConstants.TABLE_IS_CTAS)))
 {
-          if (!Catalogs.hiveCatalog(configuration, serDeProperties)) {
-            throw new SerDeException(CTAS_EXCEPTION_MSG);
-          }
 
-          if (!serDeProperties.containsKey(Constants.EXPLAIN_CTAS_LOCATION)) {
-            createTableForCTAS(configuration, serDeProperties);
-          }
+        if (e instanceof NoSuchTableException &&
+                
Boolean.parseBoolean(serDeProperties.getProperty(hive_metastoreConstants.TABLE_IS_CTAS))
 &&

Review Comment:
   done





Issue Time Tracking
-------------------

    Worklog Id:     (was: 830047)
    Time Spent: 2h 10m  (was: 2h)

> Use DDLTask to created Iceberg table when running ctas statement
> ----------------------------------------------------------------
>
>                 Key: HIVE-26771
>                 URL: https://issues.apache.org/jira/browse/HIVE-26771
>             Project: Hive
>          Issue Type: Improvement
>          Components: Iceberg integration
>            Reporter: Krisztian Kasa
>            Assignee: Krisztian Kasa
>            Priority: Major
>              Labels: pull-request-available
>          Time Spent: 2h 10m
>  Remaining Estimate: 0h
>
> When Iceberg table is created via ctas statement the table is created in 
> HiveIcebergSerDe and no DDL task is executed.
> Negative effects of this workflow:
> * Missing table properties: format-version, write.[delete|update|merge].mode
> * Default privileges of the new table are not granted.
> * The new Iceberg table can be seen by other transactions at compile time of 
> ctas.
> * Table creation and table properties are not shown in explain ctas output.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)