*** pgsql/src/backend/executor/nodeHashjoin.c	2009/01/01 17:23:41	1.97
--- pgsql/src/backend/executor/nodeHashjoin.c	2009/03/21 00:04:38	1.98
***************
*** 8,14 ****
   *
   *
   * IDENTIFICATION
!  *	  $PostgreSQL: pgsql/src/backend/executor/nodeHashjoin.c,v 1.96 2008/10/23 14:34:34 tgl Exp $
   *
   *-------------------------------------------------------------------------
   */
--- 8,14 ----
   *
   *
   * IDENTIFICATION
!  *	  $PostgreSQL: pgsql/src/backend/executor/nodeHashjoin.c,v 1.97 2009/01/01 17:23:41 momjian Exp $
   *
   *-------------------------------------------------------------------------
   */
*************** ExecHashJoin(HashJoinState *node)
*** 198,216 ****
  			node->hj_MatchedOuter = false;
  
  			/*
! 			 * now we have an outer tuple, find the corresponding bucket for
! 			 * this tuple from the hash table
  			 */
  			node->hj_CurHashValue = hashvalue;
  			ExecHashGetBucketAndBatch(hashtable, hashvalue,
  									  &node->hj_CurBucketNo, &batchno);
  			node->hj_CurTuple = NULL;
  
  			/*
  			 * Now we've got an outer tuple and the corresponding hash bucket,
! 			 * but this tuple may not belong to the current batch.
  			 */
! 			if (batchno != hashtable->curbatch)
  			{
  				/*
  				 * Need to postpone this outer tuple to a later batch. Save it
--- 198,220 ----
  			node->hj_MatchedOuter = false;
  
  			/*
! 			 * Now we have an outer tuple; find the corresponding bucket for
! 			 * this tuple in the main hash table or skew hash table.
  			 */
  			node->hj_CurHashValue = hashvalue;
  			ExecHashGetBucketAndBatch(hashtable, hashvalue,
  									  &node->hj_CurBucketNo, &batchno);
+ 			node->hj_CurSkewBucketNo = ExecHashGetSkewBucket(hashtable,
+ 															 hashvalue);
  			node->hj_CurTuple = NULL;
  
  			/*
  			 * Now we've got an outer tuple and the corresponding hash bucket,
! 			 * but it might not belong to the current batch, or it might
! 			 * match a skew bucket.
  			 */
! 			if (batchno != hashtable->curbatch &&
! 				node->hj_CurSkewBucketNo == INVALID_SKEW_BUCKET_NO)
  			{
  				/*
  				 * Need to postpone this outer tuple to a later batch. Save it
*************** ExecInitHashJoin(HashJoin *node, EState
*** 452,457 ****
--- 456,462 ----
  
  	hjstate->hj_CurHashValue = 0;
  	hjstate->hj_CurBucketNo = 0;
+ 	hjstate->hj_CurSkewBucketNo = INVALID_SKEW_BUCKET_NO;
  	hjstate->hj_CurTuple = NULL;
  
  	/*
*************** start_over:
*** 651,656 ****
--- 656,674 ----
  			BufFileClose(hashtable->outerBatchFile[curbatch]);
  		hashtable->outerBatchFile[curbatch] = NULL;
  	}
+ 	else						/* we just finished the first batch */
+ 	{
+ 		/*
+ 		 * Reset some of the skew optimization state variables, since we
+ 		 * no longer need to consider skew tuples after the first batch.
+ 		 * The memory context reset we are about to do will release the
+ 		 * skew hashtable itself.
+ 		 */
+ 		hashtable->skewEnabled = false;
+ 		hashtable->skewBucket = NULL;
+ 		hashtable->skewBucketNums = NULL;
+ 		hashtable->spaceUsedSkew = 0;
+ 	}
  
  	/*
  	 * We can always skip over any batches that are completely empty on both
*************** ExecReScanHashJoin(HashJoinState *node,
*** 880,885 ****
--- 898,904 ----
  	/* Always reset intra-tuple state */
  	node->hj_CurHashValue = 0;
  	node->hj_CurBucketNo = 0;
+ 	node->hj_CurSkewBucketNo = INVALID_SKEW_BUCKET_NO;
  	node->hj_CurTuple = NULL;
  
  	node->js.ps.ps_TupFromTlist = false;