[ 
https://issues.apache.org/jira/browse/DERBY-3872?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12633378#action_12633378
 ] 

Mamta A. Satoor commented on DERBY-3872:
----------------------------------------

While trying to run through the 10.2 codeline and trunk, one of the things I 
noticed is that impl.sql.compile.SelectNode:preprocess method has code around 
line number 1000 that looks like following in 10.2 codeline
        /* A valid group by without any aggregates is equivalent to 
         * a distinct without the group by.  We do the transformation
         * in order to simplify the group by code.
         */
        if (groupByList != null &&
                selectAggregates.size() == 0 &&
                whereAggregates.size() == 0)
        {
                isDistinct = true;
                groupByList = null;
        }

The corresponding code in trunk has additional check in the if condition as 
shown below
        /* A valid group by without any aggregates or a having clause
         * is equivalent to a distinct without the group by.  We do the 
transformation
         * in order to simplify the group by code.
         */
        if (groupByList != null &&
                havingClause == null &&
                selectAggregates.size() == 0 &&
                whereAggregates.size() == 0)
        {
                isDistinct = true;
                groupByList = null;
                wasGroupBy = true;
        }

Just to see what would happen I removed the (havingClause == null &&) check in 
the trunk codeline and ran our queries that are running in NPE in this jira 
entry and they ran fine with no NPE. But of course, the solution couldn't be 
this easy because this change causes existing tests in junit suite to fail. I 
will try to undestand what was the logic behind the changes that went in for 
DERBY-681 but any other tips at all to make progress on this jira entry will be 
greatly appreciated. I am not sure if the queries in question in this jira 
entry should follow the new code path introduced by DERBY-681.

I am wondering if anyone who has looked at other regressions caused by 
DERBY-681 know why removing the havingClause == null && check makes the queries 
in this jira entry run with no NPE. 

> NullPoinerException thrown when INTEGER function used as a predicate in a 
> WHERE clause of a SELECT .. GROUP BY ..  HAVING statement
> -----------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: DERBY-3872
>                 URL: https://issues.apache.org/jira/browse/DERBY-3872
>             Project: Derby
>          Issue Type: Bug
>    Affects Versions: 10.3.3.0
>            Reporter: Stan Bradbury
>            Assignee: Mamta A. Satoor
>         Attachments: NPE_Reproduction.sql, QueryWithoutTruePred.out, 
> QueryWithTruePred.out
>
>
> Use attached SQL script to create two tables , execute the following SQL and 
> throw the exception and stack trace below.  NOTE:  removing the 'always true' 
> clause '.. ( integer (1.1) = 1) .." from the SQL and the query does not fail. 
>  Releated??
> select  q1."DEPTNO" from DEPTTAB q1, EMPTAB q2 where  ( integer (1.1) = 1)  
> and  ( q2."DEPT_DEPTNO" =  q1."DEPTNO")  
> GROUP BY q1."DEPTNO" 
> HAVING  max( q2."SALARY") >=  ( select  q3."SALARY" from EMPTAB q3 where  ( 
> q3."EMPID" =  q1."DEPTNO") ) 
> ERROR 38000: The exception 'java.lang.NullPointerException' was thrown while 
> evaluating an expression.
> . . .derby.iapi.error.StandardException.newException
> . . .derby.iapi.error.StandardException.unexpectedUserException
> . . .derby.impl.services.reflect.DirectCall.invoke
> . . .derby.impl.sql.execute.ProjectRestrictResultSet.getNextRowCore
> . . .derby.impl.sql.execute.BasicNoPutResultSetImpl.getNextRow
> . . .derby.impl.jdbc.EmbedResultSet.movePosition
> . . .derby.impl.jdbc.EmbedResultSet.next
> . . .derby.tools.JDBCDisplayUtil.indent_DisplayResults
> . . .derby.tools.JDBCDisplayUtil.indent_DisplayResults
> . . .derby.tools.JDBCDisplayUtil.indent_DisplayResults
> . . .derby.tools.JDBCDisplayUtil.DisplayResults
> . . .derby.impl.tools.ij.utilMain.displayResult
> . . .derby.impl.tools.ij.utilMain.doCatch
> . . .derby.impl.tools.ij.utilMain.runScriptGuts
> . . .derby.impl.tools.ij.utilMain.go
> . . .derby.impl.tools.ij.Main.go
> . . .derby.impl.tools.ij.Main.mainCore
> . . .derby.impl.tools.ij.Main14.main
> . . .derby.tools.ij.main
> Caused by: java.lang.NullPointerException
> . . .derby.iapi.types.NumberDataType.compare
> . . .derby.impl.store.access.btree.ControlRow.compareIndexRowFromPageToKey
> . . .derby.impl.store.access.btree.ControlRow.searchForEntry
> . . .derby.impl.store.access.btree.LeafControlRow.search
> . . .derby.impl.store.access.btree.BTreeScan.positionAtStartForForwardScan
> . . .derby.impl.store.access.btree.BTreeForwardScan.positionAtStartPosition
> . . .derby.impl.store.access.btree.BTreeForwardScan.fetchRows
> . . .derby.impl.store.access.btree.BTreeScan.fetchNext
> . . .derby.impl.sql.execute.TableScanResultSet.getNextRowCore
> . . .derby.impl.sql.execute.IndexRowToBaseRowResultSet.getNextRowCore
> . . .derby.impl.sql.execute.ProjectRestrictResultSet.getNextRowCore
> . . .derby.impl.sql.execute.OnceResultSet.getNextRowCore
> . . .derby.exe.ac601a400fx011cx480cx5eacx00000010d8100.g0
> . . .derby.exe.ac601a400fx011cx480cx5eacx00000010d8100.e6
>       ... 17 more
> ============= begin nested exception, level (1) ===========
> java.lang.NullPointerException
> . . .derby.iapi.types.NumberDataType.compare
> . . .derby.impl.store.access.btree.ControlRow.compareIndexRowFromPageToKey
> . . .derby.impl.store.access.btree.ControlRow.searchForEntry
> . . .derby.impl.store.access.btree.LeafControlRow.search
> . . .derby.impl.store.access.btree.BTreeScan.positionAtStartForForwardScan
> . . .derby.impl.store.access.btree.BTreeForwardScan.positionAtStartPosition
> . . .derby.impl.store.access.btree.BTreeForwardScan.fetchRows
> . . .derby.impl.store.access.btree.BTreeScan.fetchNext
> . . .derby.impl.sql.execute.TableScanResultSet.getNextRowCore
> . . .derby.impl.sql.execute.IndexRowToBaseRowResultSet.getNextRowCore
> . . .derby.impl.sql.execute.ProjectRestrictResultSet.getNextRowCore
> . . .derby.impl.sql.execute.OnceResultSet.getNextRowCore
> . . .derby.exe.ac601a400fx011cx480cx5eacx00000010d8100.g0
> . . .derby.exe.ac601a400fx011cx480cx5eacx00000010d8100.e6
> . . .derby.impl.services.reflect.DirectCall.invoke
> . . .derby.impl.sql.execute.ProjectRestrictResultSet.getNextRowCore
> . . .derby.impl.sql.execute.BasicNoPutResultSetImpl.getNextRow
> . . .derby.impl.jdbc.EmbedResultSet.movePosition
> . . .derby.impl.jdbc.EmbedResultSet.next
> . . .derby.tools.JDBCDisplayUtil.indent_DisplayResults
> . . .derby.tools.JDBCDisplayUtil.indent_DisplayResults
> . . .derby.tools.JDBCDisplayUtil.indent_DisplayResults
> . . .derby.tools.JDBCDisplayUtil.DisplayResults
> . . .derby.impl.tools.ij.utilMain.displayResult
> . . .derby.impl.tools.ij.utilMain.doCatch
> . . .derby.impl.tools.ij.utilMain.runScriptGuts
> . . .derby.impl.tools.ij.utilMain.go
> . . .derby.impl.tools.ij.Main.go
> . . .derby.impl.tools.ij.Main.mainCore
> . . .derby.impl.tools.ij.Main14.main
> . . .derby.tools.ij.main
> ============= end nested exception, level (1) ===========
> Cleanup action completed

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

Reply via email to