v7-0001-Allow-Sort-nodes-to-use-the-fast-single-datum-tuples.patch

diff --git a/src/backend/executor/nodeSort.c b/src/backend/executor/nodeSort.c
index b99027e0d7..e1adecafaa 100644
--- a/src/backend/executor/nodeSort.c
+++ b/src/backend/executor/nodeSort.c
@@ -29,6 +29,10 @@
  *		which saves the results in a temporary file or memory. After the
  *		initial call, returns a tuple from the file with each call.
  *
+ *		The tuplesort can either occur on the whole tuple (this is the nominal
+ *		case) or, when the input / output tuple consists of only one attribute,
+ *		we switch to the tuplesort_*_datum API, optimized for that specific case.
+ *
  *		Conditions:
  *		  -- none.
  *
@@ -86,7 +90,13 @@ ExecSort(PlanState *pstate)
 		outerNode = outerPlanState(node);
 		tupDesc = ExecGetResultType(outerNode);
 
-		tuplesortstate = tuplesort_begin_heap(tupDesc,
+		/*
+		 * Switch to the tuplesort_*_datum interface when we have only one
+		 * key, as it is much more efficient especially when the type is
+		 * pass-by-value.
+		 */
+		if (tupDesc->natts != 1)
+			tuplesortstate = tuplesort_begin_heap(tupDesc,
 											  plannode->numCols,
 											  plannode->sortColIdx,
 											  plannode->sortOperators,
@@ -95,23 +105,49 @@ ExecSort(PlanState *pstate)
 											  work_mem,
 											  NULL,
 											  node->randomAccess);
+		else
+		{
+			node->tupleSort = false;
+			tuplesortstate = tuplesort_begin_datum(TupleDescAttr(tupDesc, 0)->atttypid,
+												   plannode->sortOperators[0],
+												   plannode->collations[0],
+												   plannode->nullsFirst[0],
+												   work_mem,
+												   NULL,
+												   node->randomAccess);
+		}
+
 		if (node->bounded)
 			tuplesort_set_bound(tuplesortstate, node->bound);
 		node->tuplesortstate = (void *) tuplesortstate;
 
 		/*
-		 * Scan the subplan and feed all the tuples to tuplesort.
++		 * Scan the subplan and feed all the tuples to tuplesort, using either
++		 * the _putdatum or _puttupleslot API as appropriate.
 		 */
 
-		for (;;)
-		{
-			slot = ExecProcNode(outerNode);
-
-			if (TupIsNull(slot))
-				break;
-
-			tuplesort_puttupleslot(tuplesortstate, slot);
-		}
+		if (node->tupleSort)
+			for (;;)
+			{
+				slot = ExecProcNode(outerNode);
+
+				if (TupIsNull(slot))
+					break;
+
+				tuplesort_puttupleslot(tuplesortstate, slot);
+			}
+		else
+			for (;;)
+			{
+				slot = ExecProcNode(outerNode);
+
+				if (TupIsNull(slot))
+					break;
+				slot_getsomeattrs(slot, 1);
+				tuplesort_putdatum(tuplesortstate,
+								   slot->tts_values[0],
+								   slot->tts_isnull[0]);
+			}
 
 		/*
 		 * Complete the sort.
@@ -150,9 +186,18 @@ ExecSort(PlanState *pstate)
 	 * next fetch from the tuplesort.
 	 */
 	slot = node->ss.ps.ps_ResultTupleSlot;
-	(void) tuplesort_gettupleslot(tuplesortstate,
+	if (node->tupleSort)
+		(void) tuplesort_gettupleslot(tuplesortstate,
 								  ScanDirectionIsForward(dir),
 								  false, slot, NULL);
+	else
+	{
+		ExecClearTuple(slot);
+		if (tuplesort_getdatum(tuplesortstate, ScanDirectionIsForward(dir),
+							   &(slot->tts_values[0]), &(slot->tts_isnull[0]), NULL))
+			ExecStoreVirtualTuple(slot);
+	}
+
 	return slot;
 }
 
@@ -191,6 +236,7 @@ ExecInitSort(Sort *node, EState *estate, int eflags)
 	sortstate->bounded = false;
 	sortstate->sort_Done = false;
 	sortstate->tuplesortstate = NULL;
+	sortstate->tupleSort = true;
 
 	/*
 	 * Miscellaneous initialization
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
index 0ec5509e7e..4c41ab80b1 100644
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -2151,6 +2151,7 @@ typedef struct SortState
 	int64		bound_Done;		/* value of bound we did the sort with */
 	void	   *tuplesortstate; /* private state of tuplesort.c */
 	bool		am_worker;		/* are we a worker? */
+	bool		tupleSort;  /* are we *not* using the single value optimization ? */	
 	SharedSortInfo *shared_info;	/* one entry per worker */
 } SortState;