aggregate(DISTINCT ...) works, per SQL spec.

Note this forces initdb because of change of Aggref node in stored rules.

1 files changed, 364 insertions, 198 deletions

diff --git a/src/backend/executor/nodeAgg.c b/src/backend/executor/nodeAgg.c
index 0956af455f1..0a95c92347f 100644
--- a/src/backend/executor/nodeAgg.c
+++ b/src/backend/executor/nodeAgg.c
@@ -3,15 +3,35 @@
  * nodeAgg.c
  *	  Routines to handle aggregate nodes.
  *
- * Copyright (c) 1994, Regents of the University of California
+ *	  ExecAgg evaluates each aggregate in the following steps: (initcond1,
+ *	  initcond2 are the initial values and sfunc1, sfunc2, and finalfunc are
+ *	  the transition functions.)
+ *
+ *		 value1 = initcond1
+ *		 value2 = initcond2
+ *		 foreach input_value do
+ *			value1 = sfunc1(value1, input_value)
+ *			value2 = sfunc2(value2)
+ *		 value1 = finalfunc(value1, value2)
+ *
+ *	  If initcond1 is NULL then the first non-NULL input_value is
+ *	  assigned directly to value1.  sfunc1 isn't applied until value1
+ *	  is non-NULL.
+ *
+ *	  sfunc1 is never applied when the current tuple's input_value is NULL.
+ *	  sfunc2 is applied for each tuple if the aggref is marked 'usenulls',
+ *	  otherwise it is only applied when input_value is not NULL.
+ *	  (usenulls was formerly used for COUNT(*), but is no longer needed for
+ *	  that purpose; as of 10/1999 the support for usenulls is dead code.
+ *	  I have not removed it because it seems like a potentially useful
+ *	  feature for user-defined aggregates.  We'd just need to add a
+ *	  flag column to pg_aggregate and a parameter to CREATE AGGREGATE...)
  *
  *
- * NOTE
- *	  The implementation of Agg node has been reworked to handle legal
- *	  SQL aggregates. (Do not expect POSTQUEL semantics.)	 -- ay 2/95
+ * Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *	  $Header: /cvsroot/pgsql/src/backend/executor/nodeAgg.c,v 1.59 1999/10/30 02:35:14 tgl Exp $
+ *	  $Header: /cvsroot/pgsql/src/backend/executor/nodeAgg.c,v 1.60 1999/12/13 01:26:52 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -20,11 +40,15 @@
 
 #include "access/heapam.h"
 #include "catalog/pg_aggregate.h"
+#include "catalog/pg_operator.h"
 #include "executor/executor.h"
 #include "executor/nodeAgg.h"
 #include "optimizer/clauses.h"
+#include "parser/parse_expr.h"
+#include "parser/parse_oper.h"
 #include "parser/parse_type.h"
 #include "utils/syscache.h"
+#include "utils/tuplesort.h"
 
 /*
  * AggStatePerAggData - per-aggregate working state for the Agg scan
@@ -36,6 +60,9 @@ typedef struct AggStatePerAggData
 	 * thereafter:
 	 */
 
+	/* Link to Aggref node this working state is for */
+	Aggref	   *aggref;
+
 	/* Oids of transfer functions */
 	Oid			xfn1_oid;
 	Oid			xfn2_oid;
@@ -48,6 +75,18 @@ typedef struct AggStatePerAggData
 	FmgrInfo	xfn2;
 	FmgrInfo	finalfn;
 	/*
+	 * Type of input data and Oid of sort operator to use for it;
+	 * only set/used when aggregate has DISTINCT flag.  (These are not
+	 * used directly by nodeAgg, but must be passed to the Tuplesort object.)
+	 */
+	Oid			inputType;
+	Oid			sortOperator;
+	/*
+	 * fmgr lookup data for input type's equality operator --- only set/used
+	 * when aggregate has DISTINCT flag.
+	 */
+	FmgrInfo	equalfn;
+	/*
 	 * initial values from pg_aggregate entry
 	 */
 	Datum		initValue1;		/* for transtype1 */
@@ -55,19 +94,29 @@ typedef struct AggStatePerAggData
 	bool		initValue1IsNull,
 				initValue2IsNull;
 	/*
-	 * We need the len and byval info for the agg's transition status types
-	 * in order to know how to copy/delete values.
+	 * We need the len and byval info for the agg's input and transition
+	 * data types in order to know how to copy/delete values.
 	 */
-	int			transtype1Len,
+	int			inputtypeLen,
+				transtype1Len,
 				transtype2Len;
-	bool		transtype1ByVal,
+	bool		inputtypeByVal,
+				transtype1ByVal,
 				transtype2ByVal;
 
 	/*
 	 * These values are working state that is initialized at the start
-	 * of an input tuple group and updated for each input tuple:
+	 * of an input tuple group and updated for each input tuple.
+	 *
+	 * For a simple (non DISTINCT) aggregate, we just feed the input values
+	 * straight to the transition functions.  If it's DISTINCT, we pass the
+	 * input values into a Tuplesort object; then at completion of the input
+	 * tuple group, we scan the sorted values, eliminate duplicates, and run
+	 * the transition functions on the rest.
 	 */
 
+	Tuplesortstate *sortstate;	/* sort object, if a DISTINCT agg */
+
 	Datum		value1,			/* current transfer values 1 and 2 */
 				value2;
 	bool		value1IsNull,
@@ -82,28 +131,248 @@ typedef struct AggStatePerAggData
 } AggStatePerAggData;
 
 
+static void initialize_aggregate (AggStatePerAgg peraggstate);
+static void advance_transition_functions (AggStatePerAgg peraggstate,
+										  Datum newVal, bool isNull);
+static void finalize_aggregate (AggStatePerAgg peraggstate,
+								Datum *resultVal, bool *resultIsNull);
+static Datum copyDatum(Datum val, int typLen, bool typByVal);
+
+
 /*
- * Helper routine to make a copy of a Datum.
- *
- * NB: input had better not be a NULL; might cause null-pointer dereference.
+ * Initialize one aggregate for a new set of input values.
  */
-static Datum
-copyDatum(Datum val, int typLen, bool typByVal)
+static void
+initialize_aggregate (AggStatePerAgg peraggstate)
 {
-	if (typByVal)
-		return val;
+	Aggref		   *aggref = peraggstate->aggref;
+
+	/*
+	 * Start a fresh sort operation for each DISTINCT aggregate.
+	 */
+	if (aggref->aggdistinct)
+	{
+		/* In case of rescan, maybe there could be an uncompleted
+		 * sort operation?  Clean it up if so.
+		 */
+		if (peraggstate->sortstate)
+			tuplesort_end(peraggstate->sortstate);
+
+		peraggstate->sortstate =
+			tuplesort_begin_datum(peraggstate->inputType,
+								  peraggstate->sortOperator,
+								  false);
+	}
+
+	/*
+	 * (Re)set value1 and value2 to their initial values.
+	 */
+	if (OidIsValid(peraggstate->xfn1_oid) &&
+		! peraggstate->initValue1IsNull)
+		peraggstate->value1 = copyDatum(peraggstate->initValue1, 
+										peraggstate->transtype1Len,
+										peraggstate->transtype1ByVal);
+	else
+		peraggstate->value1 = (Datum) NULL;
+	peraggstate->value1IsNull = peraggstate->initValue1IsNull;
+
+	if (OidIsValid(peraggstate->xfn2_oid) &&
+		! peraggstate->initValue2IsNull)
+		peraggstate->value2 = copyDatum(peraggstate->initValue2, 
+										peraggstate->transtype2Len,
+										peraggstate->transtype2ByVal);
 	else
+		peraggstate->value2 = (Datum) NULL;
+	peraggstate->value2IsNull = peraggstate->initValue2IsNull;
+
+	/* ------------------------------------------
+	 * If the initial value for the first transition function
+	 * doesn't exist in the pg_aggregate table then we will let
+	 * the first value returned from the outer procNode become
+	 * the initial value. (This is useful for aggregates like
+	 * max{} and min{}.)  The noInitValue flag signals that we
+	 * still need to do this.
+	 * ------------------------------------------
+	 */
+	peraggstate->noInitValue = peraggstate->initValue1IsNull;
+}
+
+/*
+ * Given a new input value, advance the transition functions of an aggregate.
+ *
+ * Note: if the agg does not have usenulls set, null inputs will be filtered
+ * out before reaching here.
+ */
+static void
+advance_transition_functions (AggStatePerAgg peraggstate,
+							  Datum newVal, bool isNull)
+{
+	Datum		args[2];
+
+	if (OidIsValid(peraggstate->xfn1_oid) && !isNull)
 	{
-		char   *newVal;
+		if (peraggstate->noInitValue)
+		{
+			/*
+			 * value1 has not been initialized. This is the first non-NULL
+			 * input value. We use it as the initial value for value1.
+			 *
+			 * XXX We assume, without having checked, that the agg's input
+			 * type is binary-compatible with its transtype1!
+			 *
+			 * We have to copy the datum since the tuple from which it came
+			 * will be freed on the next iteration of the scan.
+			 */
+			peraggstate->value1 = copyDatum(newVal,
+											peraggstate->transtype1Len,
+											peraggstate->transtype1ByVal);
+			peraggstate->value1IsNull = false;
+			peraggstate->noInitValue = false;
+		}
+		else
+		{
+			/* apply transition function 1 */
+			args[0] = peraggstate->value1;
+			args[1] = newVal;
+			newVal = (Datum) fmgr_c(&peraggstate->xfn1,
+									(FmgrValues *) args,
+									&isNull);
+			if (! peraggstate->transtype1ByVal)
+				pfree(peraggstate->value1);
+			peraggstate->value1 = newVal;
+		}
+	}
 
-		if (typLen == -1)		/* variable length type? */
-			typLen = VARSIZE((struct varlena *) DatumGetPointer(val));
-		newVal = (char *) palloc(typLen);
-		memcpy(newVal, DatumGetPointer(val), typLen);
-		return PointerGetDatum(newVal);
+	if (OidIsValid(peraggstate->xfn2_oid))
+	{
+		/* apply transition function 2 */
+		args[0] = peraggstate->value2;
+		isNull = false;			/* value2 cannot be null, currently */
+		newVal = (Datum) fmgr_c(&peraggstate->xfn2,
+								(FmgrValues *) args,
+								&isNull);
+		if (! peraggstate->transtype2ByVal)
+			pfree(peraggstate->value2);
+		peraggstate->value2 = newVal;
 	}
 }
 
+/*
+ * Compute the final value of one aggregate.
+ */
+static void
+finalize_aggregate (AggStatePerAgg peraggstate,
+					Datum *resultVal, bool *resultIsNull)
+{
+	Aggref	   *aggref = peraggstate->aggref;
+	char	   *args[2];
+
+	/*
+	 * If it's a DISTINCT aggregate, all we've done so far is to stuff the
+	 * input values into the sort object.  Complete the sort, then run
+	 * the transition functions on the non-duplicate values.  Note that
+	 * DISTINCT always suppresses nulls, per SQL spec, regardless of usenulls.
+	 */
+	if (aggref->aggdistinct)
+	{
+		Datum		oldVal = (Datum) 0;
+		bool		haveOldVal = false;
+		Datum		newVal;
+		bool		isNull;
+
+		tuplesort_performsort(peraggstate->sortstate);
+		while (tuplesort_getdatum(peraggstate->sortstate, true,
+								  &newVal, &isNull))
+		{
+			if (isNull)
+				continue;
+			if (haveOldVal)
+			{
+				Datum	equal;
+
+				equal = (Datum) (*fmgr_faddr(&peraggstate->equalfn)) (oldVal,
+																	  newVal);
+				if (DatumGetInt32(equal) != 0)
+				{
+					if (! peraggstate->inputtypeByVal)
+						pfree(DatumGetPointer(newVal));
+					continue;
+				}
+			}
+			advance_transition_functions(peraggstate, newVal, false);
+			if (haveOldVal && ! peraggstate->inputtypeByVal)
+				pfree(DatumGetPointer(oldVal));
+			oldVal = newVal;
+			haveOldVal = true;
+		}
+		if (haveOldVal && ! peraggstate->inputtypeByVal)
+			pfree(DatumGetPointer(oldVal));
+		tuplesort_end(peraggstate->sortstate);
+		peraggstate->sortstate = NULL;
+	}
+
+	/*
+	 * Now apply the agg's finalfn, or substitute the appropriate transition
+	 * value if there is no finalfn.
+	 *
+	 * XXX For now, only apply finalfn if we got at least one
+	 * non-null input value.  This prevents zero divide in AVG().
+	 * If we had cleaner handling of null inputs/results in functions,
+	 * we could probably take out this hack and define the result
+	 * for no inputs as whatever finalfn returns for null input.
+	 */
+	if (OidIsValid(peraggstate->finalfn_oid) &&
+		! peraggstate->noInitValue)
+	{
+		if (peraggstate->finalfn.fn_nargs > 1)
+		{
+			args[0] = (char *) peraggstate->value1;
+			args[1] = (char *) peraggstate->value2;
+		}
+		else if (OidIsValid(peraggstate->xfn1_oid))
+			args[0] = (char *) peraggstate->value1;
+		else if (OidIsValid(peraggstate->xfn2_oid))
+			args[0] = (char *) peraggstate->value2;
+		else
+			elog(ERROR, "ExecAgg: no valid transition functions??");
+		*resultIsNull = false;
+		*resultVal = (Datum) fmgr_c(&peraggstate->finalfn,
+									(FmgrValues *) args,
+									resultIsNull);
+	}
+	else if (OidIsValid(peraggstate->xfn1_oid))
+	{
+		/* Return value1 */
+		*resultVal = peraggstate->value1;
+		*resultIsNull = peraggstate->value1IsNull;
+		/* prevent pfree below */
+		peraggstate->value1IsNull = true;
+	}
+	else if (OidIsValid(peraggstate->xfn2_oid))
+	{
+		/* Return value2 */
+		*resultVal = peraggstate->value2;
+		*resultIsNull = peraggstate->value2IsNull;
+		/* prevent pfree below */
+		peraggstate->value2IsNull = true;
+	}
+	else
+		elog(ERROR, "ExecAgg: no valid transition functions??");
+
+	/*
+	 * Release any per-group working storage, unless we're passing
+	 * it back as the result of the aggregate.
+	 */
+	if (OidIsValid(peraggstate->xfn1_oid) &&
+		! peraggstate->value1IsNull &&
+		! peraggstate->transtype1ByVal)
+		pfree(peraggstate->value1);
+	
+	if (OidIsValid(peraggstate->xfn2_oid) &&
+		! peraggstate->value2IsNull &&
+		! peraggstate->transtype2ByVal)
+		pfree(peraggstate->value2);
+}
 
 /* ---------------------------------------
  *
@@ -118,30 +387,6 @@ copyDatum(Datum val, int typLen, bool typByVal)
  *	  the expression context to be used when ExecProject evaluates the
  *	  result tuple.
  *
- *	  ExecAgg evaluates each aggregate in the following steps: (initcond1,
- *	  initcond2 are the initial values and sfunc1, sfunc2, and finalfunc are
- *	  the transition functions.)
- *
- *		 value1 = initcond1
- *		 value2 = initcond2
- *		 foreach tuple do
- *			value1 = sfunc1(value1, aggregated_value)
- *			value2 = sfunc2(value2)
- *		 value1 = finalfunc(value1, value2)
- *
- *	  If initcond1 is NULL then the first non-NULL aggregated_value is
- *	  assigned directly to value1.  sfunc1 isn't applied until value1
- *	  is non-NULL.
- *
- *	  sfunc1 is never applied when the current tuple's aggregated_value
- *	  is NULL.  sfunc2 is applied for each tuple if the aggref is marked
- *	  'usenulls', otherwise it is only applied when aggregated_value is
- *	  not NULL.  (usenulls was formerly used for COUNT(*), but is no longer
- *	  needed for that purpose; as of 10/1999 the support for usenulls is
- *	  dead code.  I have not removed it because it seems like a potentially
- *	  useful feature for user-defined aggregates.  We'd just need to add a
- *	  flag column to pg_aggregate and a parameter to CREATE AGGREGATE...)
- *
  *	  If the outer subplan is a Group node, ExecAgg returns as many tuples
  *	  as there are groups.
  *
@@ -161,7 +406,6 @@ ExecAgg(Agg *node)
 	TupleTableSlot *resultSlot;
 	HeapTuple	inputTuple;
 	int			aggno;
-	List	   *alist;
 	bool		isDone;
 	bool		isNull;
 
@@ -190,42 +434,11 @@ ExecAgg(Agg *node)
 		/*
 		 * Initialize working state for a new input tuple group
 		 */
-		aggno = -1;
-		foreach(alist, aggstate->aggs)
+		for (aggno = 0; aggno < aggstate->numaggs; aggno++)
 		{
-			AggStatePerAgg	peraggstate = &peragg[++aggno];
+			AggStatePerAgg	peraggstate = &peragg[aggno];
 
-			/*
-			 * (Re)set value1 and value2 to their initial values.
-			 */
-			if (OidIsValid(peraggstate->xfn1_oid) &&
-				! peraggstate->initValue1IsNull)
-				peraggstate->value1 = copyDatum(peraggstate->initValue1, 
-												peraggstate->transtype1Len,
-												peraggstate->transtype1ByVal);
-			else
-				peraggstate->value1 = (Datum) NULL;
-			peraggstate->value1IsNull = peraggstate->initValue1IsNull;
-
-			if (OidIsValid(peraggstate->xfn2_oid) &&
-				! peraggstate->initValue2IsNull)
-				peraggstate->value2 = copyDatum(peraggstate->initValue2, 
-												peraggstate->transtype2Len,
-												peraggstate->transtype2ByVal);
-			else
-				peraggstate->value2 = (Datum) NULL;
-			peraggstate->value2IsNull = peraggstate->initValue2IsNull;
-
-			/* ------------------------------------------
-			 * If the initial value for the first transition function
-			 * doesn't exist in the pg_aggregate table then we will let
-			 * the first value returned from the outer procNode become
-			 * the initial value. (This is useful for aggregates like
-			 * max{} and min{}.)  The noInitValue flag signals that we
-			 * still need to do this.
-			 * ------------------------------------------
-			 */
-			peraggstate->noInitValue = peraggstate->initValue1IsNull;
+			initialize_aggregate(peraggstate);
 		}
 
 		inputTuple = NULL;		/* no saved input tuple yet */
@@ -243,13 +456,11 @@ ExecAgg(Agg *node)
 				break;
 			econtext->ecxt_scantuple = outerslot;
 
-			aggno = -1;
-			foreach(alist, aggstate->aggs)
+			for (aggno = 0; aggno < aggstate->numaggs; aggno++)
 			{
-				Aggref		   *aggref = (Aggref *) lfirst(alist);
-				AggStatePerAgg	peraggstate = &peragg[++aggno];
+				AggStatePerAgg	peraggstate = &peragg[aggno];
+				Aggref		   *aggref = peraggstate->aggref;
 				Datum			newVal;
-				Datum			args[2];
 
 				newVal = ExecEvalExpr(aggref->target, econtext,
 									  &isNull, &isDone);
@@ -257,53 +468,12 @@ ExecAgg(Agg *node)
 				if (isNull && !aggref->usenulls)
 					continue;	/* ignore this tuple for this agg */
 
-				if (OidIsValid(peraggstate->xfn1_oid) && !isNull)
-				{
-					if (peraggstate->noInitValue)
-					{
-						/*
-						 * value1 has not been initialized. This is the
-						 * first non-NULL input value. We use it as the
-						 * initial value for value1.  XXX We assume,
-						 * without having checked, that the agg's input type
-						 * is binary-compatible with its transtype1!
-						 *
-						 * We have to copy the datum since the tuple from
-						 * which it came will be freed on the next iteration
-						 * of the scan.  
-						 */
-						peraggstate->value1 = copyDatum(newVal,
-												peraggstate->transtype1Len,
-												peraggstate->transtype1ByVal);
-						peraggstate->value1IsNull = false;
-						peraggstate->noInitValue = false;
-					}
-					else
-					{
-						/* apply transition function 1 */
-						args[0] = peraggstate->value1;
-						args[1] = newVal;
-						newVal = (Datum) fmgr_c(&peraggstate->xfn1,
-												(FmgrValues *) args,
-												&isNull);
-						if (! peraggstate->transtype1ByVal)
-							pfree(peraggstate->value1);
-						peraggstate->value1 = newVal;
-					}
-				}
-
-				if (OidIsValid(peraggstate->xfn2_oid))
-				{
-					/* apply transition function 2 */
-					args[0] = peraggstate->value2;
-					isNull = false;	/* value2 cannot be null, currently */
-					newVal = (Datum) fmgr_c(&peraggstate->xfn2,
-											(FmgrValues *) args,
-											&isNull);
-					if (! peraggstate->transtype2ByVal)
-						pfree(peraggstate->value2);
-					peraggstate->value2 = newVal;
-				}
+				if (aggref->aggdistinct)
+					tuplesort_putdatum(peraggstate->sortstate,
+									   newVal, isNull);
+				else
+					advance_transition_functions(peraggstate,
+												 newVal, isNull);
 			}
 
 			/*
@@ -320,70 +490,12 @@ ExecAgg(Agg *node)
 		 * Done scanning input tuple group.
 		 * Finalize each aggregate calculation.
 		 */
-		aggno = -1;
-		foreach(alist, aggstate->aggs)
+		for (aggno = 0; aggno < aggstate->numaggs; aggno++)
 		{
-			AggStatePerAgg	peraggstate = &peragg[++aggno];
-			char		   *args[2];
-
-			/*
-			 * XXX For now, only apply finalfn if we got at least one
-			 * non-null input value.  This prevents zero divide in AVG().
-			 * If we had cleaner handling of null inputs/results in functions,
-			 * we could probably take out this hack and define the result
-			 * for no inputs as whatever finalfn returns for null input.
-			 */
-			if (OidIsValid(peraggstate->finalfn_oid) &&
-				! peraggstate->noInitValue)
-			{
-				if (peraggstate->finalfn.fn_nargs > 1)
-				{
-					args[0] = (char *) peraggstate->value1;
-					args[1] = (char *) peraggstate->value2;
-				}
-				else if (OidIsValid(peraggstate->xfn1_oid))
-					args[0] = (char *) peraggstate->value1;
-				else if (OidIsValid(peraggstate->xfn2_oid))
-					args[0] = (char *) peraggstate->value2;
-				else
-					elog(ERROR, "ExecAgg: no valid transition functions??");
-				aggnulls[aggno] = false;
-				aggvalues[aggno] = (Datum) fmgr_c(&peraggstate->finalfn,
-												  (FmgrValues *) args,
-												  &(aggnulls[aggno]));
-			}
-			else if (OidIsValid(peraggstate->xfn1_oid))
-			{
-				/* Return value1 */
-				aggvalues[aggno] = peraggstate->value1;
-				aggnulls[aggno] = peraggstate->value1IsNull;
-				/* prevent pfree below */
-				peraggstate->value1IsNull = true;
-			}
-			else if (OidIsValid(peraggstate->xfn2_oid))
-			{
-				/* Return value2 */
-				aggvalues[aggno] = peraggstate->value2;
-				aggnulls[aggno] = peraggstate->value2IsNull;
-				/* prevent pfree below */
-				peraggstate->value2IsNull = true;
-			}
-			else
-				elog(ERROR, "ExecAgg: no valid transition functions??");
-
-			/*
-			 * Release any per-group working storage, unless we're passing
-			 * it back as the result of the aggregate.
-			 */
-			if (OidIsValid(peraggstate->xfn1_oid) &&
-				! peraggstate->value1IsNull &&
-				! peraggstate->transtype1ByVal)
-				pfree(peraggstate->value1);
+			AggStatePerAgg	peraggstate = &peragg[aggno];
 
-			if (OidIsValid(peraggstate->xfn2_oid) &&
-				! peraggstate->value2IsNull &&
-				! peraggstate->transtype2ByVal)
-				pfree(peraggstate->value2);
+			finalize_aggregate(peraggstate,
+							   & aggvalues[aggno], & aggnulls[aggno]);
 		}
 
 		/*
@@ -458,14 +570,14 @@ ExecAgg(Agg *node)
 
 		/*
 		 * Form a projection tuple using the aggregate results and the
-		 * representative input tuple.  Store it in the result tuple slot,
-		 * and return it if it meets my qual condition.
+		 * representative input tuple.  Store it in the result tuple slot.
 		 */
 		resultSlot = ExecProject(projInfo, &isDone);
 
 		/*
 		 * If the completed tuple does not match the qualifications,
 		 * it is ignored and we loop back to try to process another group.
+		 * Otherwise, return the tuple.
 		 */
 	}
 	while (! ExecQual(node->plan.qual, econtext));
@@ -505,6 +617,11 @@ ExecInitAgg(Agg *node, EState *estate, Plan *parent)
 
 	/*
 	 * find aggregates in targetlist and quals
+	 *
+	 * Note: pull_agg_clauses also checks that no aggs contain other agg
+	 * calls in their arguments.  This would make no sense under SQL semantics
+	 * anyway (and it's forbidden by the spec).  Because that is true, we
+	 * don't need to worry about evaluating the aggs in any particular order.
 	 */
 	aggstate->aggs = nconc(pull_agg_clause((Node *) node->plan.targetlist),
 						   pull_agg_clause((Node *) node->plan.qual));
@@ -588,6 +705,9 @@ ExecInitAgg(Agg *node, EState *estate, Plan *parent)
 		/* Mark Aggref node with its associated index in the result array */
 		aggref->aggno = aggno;
 
+		/* Fill in the peraggstate data */
+		peraggstate->aggref = aggref;
+
 		aggTuple = SearchSysCacheTuple(AGGNAME,
 									   PointerGetDatum(aggname),
 									   ObjectIdGetDatum(aggref->basetype),
@@ -644,6 +764,29 @@ ExecInitAgg(Agg *node, EState *estate, Plan *parent)
 		{
 			fmgr_info(finalfn_oid, &peraggstate->finalfn);
 		}
+
+		if (aggref->aggdistinct)
+		{
+			Oid			inputType = exprType(aggref->target);
+			Operator	eq_operator;
+			Form_pg_operator pgopform;
+
+			peraggstate->inputType = inputType;
+			typeInfo = typeidType(inputType);
+			peraggstate->inputtypeLen = typeLen(typeInfo);
+			peraggstate->inputtypeByVal = typeByVal(typeInfo);
+
+			eq_operator = oper("=", inputType, inputType, true);
+			if (!HeapTupleIsValid(eq_operator))
+			{
+				elog(ERROR, "Unable to identify an equality operator for type '%s'",
+					 typeidTypeName(inputType));
+			}
+			pgopform = (Form_pg_operator) GETSTRUCT(eq_operator);
+			fmgr_info(pgopform->oprcode, &(peraggstate->equalfn));
+			peraggstate->sortOperator = any_ordering_op(inputType);
+			peraggstate->sortstate = NULL;
+		}
 	}
 
 	return TRUE;
@@ -690,3 +833,26 @@ ExecReScanAgg(Agg *node, ExprContext *exprCtxt, Plan *parent)
 		ExecReScan(((Plan *) node)->lefttree, exprCtxt, (Plan *) node);
 
 }
+
+
+/*
+ * Helper routine to make a copy of a Datum.
+ *
+ * NB: input had better not be a NULL; might cause null-pointer dereference.
+ */
+static Datum
+copyDatum(Datum val, int typLen, bool typByVal)
+{
+	if (typByVal)
+		return val;
+	else
+	{
+		char   *newVal;
+
+		if (typLen == -1)		/* variable length type? */
+			typLen = VARSIZE((struct varlena *) DatumGetPointer(val));
+		newVal = (char *) palloc(typLen);
+		memcpy(newVal, DatumGetPointer(val), typLen);
+		return PointerGetDatum(newVal);
+	}
+}