Allow SQL-language functions to return the output of an INSERT/UPDATE/DELETE

RETURNING clause, not just a SELECT as formerly.

A side effect of this patch is that when a set-returning SQL function is used
in a FROM clause, performance is improved because the output is collected into
a tuplestore within the function, rather than using the less efficient
value-per-call mechanism.

doc/src/sgml/xfunc.sgml

@@ -1,4 +1,4 @@
-<!-- $PostgreSQL: pgsql/doc/src/sgml/xfunc.sgml,v 1.132 2008/07/18 03:32:52 tgl Exp $ -->
+<!-- $PostgreSQL: pgsql/doc/src/sgml/xfunc.sgml,v 1.133 2008/10/31 19:37:56 tgl Exp $ -->
 
  <sect1 id="xfunc">
   <title>User-Defined Functions</title>
@@ -106,7 +106,9 @@
     The body of an SQL function must be a list of SQL
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
-    <type>void</>, the last statement must be a <command>SELECT</>.
+    <type>void</>, the last statement must be a <command>SELECT</>,
+    or an <command>INSERT</>, <command>UPDATE</>, or <command>DELETE</>
+    that has a <literal>RETURNING</> clause.
    </para>
 
     <para>
@@ -119,11 +121,11 @@
      <command>BEGIN</>, <command>COMMIT</>, <command>ROLLBACK</>, or
      <command>SAVEPOINT</> commands into a <acronym>SQL</acronym> function.)
      However, the final command
-     must be a <command>SELECT</command> that returns whatever is
+     must be a <command>SELECT</command> or have a <literal>RETURNING</>
+     clause that returns whatever is
      specified as the function's return type.  Alternatively, if you
      want to define a SQL function that performs actions but has no
      useful value to return, you can define it as returning <type>void</>.
-     In that case, the function body must not end with a <command>SELECT</command>.
      For example, this function removes rows with negative salaries from
      the <literal>emp</> table:
 
@@ -257,6 +259,16 @@ $$ LANGUAGE SQL;
 </programlisting>
 
      which adjusts the balance and returns the new balance.
+     The same thing could be done in one command using <literal>RETURNING</>:
+
+<programlisting>
+CREATE FUNCTION tf1 (integer, numeric) RETURNS numeric AS $$
+    UPDATE bank
+        SET balance = balance - $2
+        WHERE accountno = $1
+    RETURNING balance;
+$$ LANGUAGE SQL;
+</programlisting>
     </para>
    </sect2>
 
@@ -422,7 +434,7 @@ SELECT (new_emp()).name;
 
 <screen>
 SELECT new_emp().name;
-ERROR:  syntax error at or near "." at character 17
+ERROR:  syntax error at or near "."
 LINE 1: SELECT new_emp().name;
                         ^
 </screen>
@@ -705,7 +717,7 @@ SELECT *, upper(fooname) FROM getfoo(1) AS t1;
     <para>
      When an SQL function is declared as returning <literal>SETOF
      <replaceable>sometype</></literal>, the function's final
-     <command>SELECT</> query is executed to completion, and each row it
+     query is executed to completion, and each row it
      outputs is returned as an element of the result set.
     </para>
 
@@ -798,6 +810,18 @@ SELECT name, listchildren(name) FROM nodes;
      This happens because <function>listchildren</function> returns an empty set
      for those arguments, so no result rows are generated.
     </para>
+
+    <note>
+     <para>
+      If a function's last command is <command>INSERT</>, <command>UPDATE</>,
+      or <command>DELETE</> with <literal>RETURNING</>, that command will
+      always be executed to completion, even if the function is not declared
+      with <literal>SETOF</> or the calling query does not fetch all the
+      result rows.  Any extra rows produced by the <literal>RETURNING</>
+      clause are silently dropped, but the commanded table modifications
+      still happen (and are all completed before returning from the function).
+     </para>
+    </note>
    </sect2>
 
    <sect2 id="xfunc-sql-functions-returning-table">
@@ -1459,16 +1483,13 @@ PG_MODULE_MAGIC;
     <para>
      By-value  types  can  only be 1, 2, or 4 bytes in length
      (also 8 bytes, if <literal>sizeof(Datum)</literal> is 8 on your machine).
-     You should be careful 
+     You should be careful to define your types such that they will be the
-     to define your types such that  they  will  be  the  same  
+     same size (in bytes) on all architectures.  For example, the
-     size (in bytes) on all architectures.  For example, the 
+     <literal>long</literal> type is dangerous because it is 4 bytes on some
-     <literal>long</literal> type is dangerous because  it  
+     machines and 8 bytes on others, whereas <type>int</type> type is 4 bytes
-     is 4 bytes on some machines and 8 bytes on others, whereas 
+     on most Unix machines.  A reasonable implementation of the
-     <type>int</type>  type  is  4  bytes  on  most  
+     <type>int4</type> type on Unix machines might be:
-     Unix machines.  A reasonable implementation of  
+
-     the  <type>int4</type>  type  on  Unix
-     machines might be:
-     
 <programlisting>
 /* 4-byte integer, passed by value */
 typedef int int4;
@@ -1479,7 +1500,7 @@ typedef int int4;
      On  the  other hand, fixed-length types of any size can
      be passed by-reference.  For example, here is a  sample
      implementation of a <productname>PostgreSQL</productname> type:
-     
+
 <programlisting>
 /* 16-byte structure, passed by reference */
 typedef struct
@@ -1502,7 +1523,7 @@ typedef struct
      Finally, all variable-length types must also be  passed
      by  reference.   All  variable-length  types must begin
      with a length field of exactly 4 bytes, and all data to
-     be  stored within that type must be located in the memory 
+     be  stored within that type must be located in the memory
      immediately  following  that  length  field.   The
      length field contains the total length of the structure,
      that is,  it  includes  the  size  of  the  length  field
@@ -1540,8 +1561,8 @@ typedef struct {
     </para>
 
     <para>
-     When manipulating 
+     When manipulating
-     variable-length types, we must  be  careful  to  allocate  
+     variable-length types, we must  be  careful  to  allocate
      the  correct amount  of memory and set the length field correctly.
      For example, if we wanted to  store  40  bytes  in  a <structname>text</>
      structure, we might use a code fragment like this:
@@ -1772,7 +1793,7 @@ memcpy(destination-&gt;data, buffer, 40);
 #include &lt;string.h&gt;
 
 /* by value */
-         
+
 int
 add_one(int arg)
 {
@@ -1787,7 +1808,7 @@ add_one_float8(float8 *arg)
     float8    *result = (float8 *) palloc(sizeof(float8));
 
     *result = *arg + 1.0;
-       
+
     return result;
 }
 
@@ -1798,7 +1819,7 @@ makepoint(Point *pointx, Point *pointy)
 
     new_point-&gt;x = pointx-&gt;x;
     new_point-&gt;y = pointy-&gt;y;
-       
+
     return new_point;
 }
 
@@ -1841,7 +1862,7 @@ concat_text(text *arg1, text *arg2)
      <filename>funcs.c</filename> and compiled into a shared object,
      we could define the functions to <productname>PostgreSQL</productname>
      with commands like this:
-     
+
 <programlisting>
 CREATE FUNCTION add_one(integer) RETURNS integer
      AS '<replaceable>DIRECTORY</replaceable>/funcs', 'add_one'
@@ -1855,7 +1876,7 @@ CREATE FUNCTION add_one(double precision) RETURNS double precision
 CREATE FUNCTION makepoint(point, point) RETURNS point
      AS '<replaceable>DIRECTORY</replaceable>/funcs', 'makepoint'
      LANGUAGE C STRICT;
-                         
+
 CREATE FUNCTION copytext(text) RETURNS text
      AS '<replaceable>DIRECTORY</replaceable>/funcs', 'copytext'
      LANGUAGE C STRICT;
@@ -1947,7 +1968,7 @@ PG_FUNCTION_INFO_V1(funcname);
 /* by value */
 
 PG_FUNCTION_INFO_V1(add_one);
-         
+
 Datum
 add_one(PG_FUNCTION_ARGS)
 {
@@ -1981,7 +2002,7 @@ makepoint(PG_FUNCTION_ARGS)
 
     new_point-&gt;x = pointx-&gt;x;
     new_point-&gt;y = pointy-&gt;y;
-       
+
     PG_RETURN_POINT_P(new_point);
 }
 
@@ -2447,7 +2468,7 @@ include $(PGXS)
      in the <literal>results/</literal> directory), and copying them to
      <literal>expected/</literal> if they match what you want from the test.
     </para>
-   
+
    </tip>
   </sect2>
 
@@ -2476,7 +2497,7 @@ SELECT name, c_overpaid(emp, 1500) AS overpaid
 
      Using call conventions version 0, we can define
      <function>c_overpaid</> as:
-     
+
 <programlisting>
 #include "postgres.h"
 #include "executor/executor.h"  /* for GetAttributeByName() */
@@ -2522,11 +2543,11 @@ c_overpaid(PG_FUNCTION_ARGS)
     </para>
 
     <para>
-     <function>GetAttributeByName</function> is the 
+     <function>GetAttributeByName</function> is the
      <productname>PostgreSQL</productname> system function that
      returns attributes out of the specified row.  It has
      three arguments: the argument of type <type>HeapTupleHeader</type> passed
-     into 
+     into
      the  function, the name of the desired attribute, and a
      return parameter that tells whether  the  attribute
      is  null.   <function>GetAttributeByName</function> returns a <type>Datum</type>
@@ -2733,7 +2754,7 @@ typedef struct
 {
     /*
      * Number of times we've been called before
-     * 
+     *
      * call_cntr is initialized to 0 for you by SRF_FIRSTCALL_INIT(), and
      * incremented for you every time SRF_RETURN_NEXT() is called.
      */
@@ -2750,7 +2771,7 @@ typedef struct
 
     /*
      * OPTIONAL pointer to result slot
-     * 
+     *
      * This is obsolete and only present for backwards compatibility, viz,
      * user-defined SRFs that use the deprecated TupleDescGetSlot().
      */
@@ -2758,7 +2779,7 @@ typedef struct
 
     /*
      * OPTIONAL pointer to miscellaneous user-provided context information
-     * 
+     *
      * user_fctx is for use as a pointer to your own data to retain
      * arbitrary context information between calls of your function.
      */
@@ -2766,7 +2787,7 @@ typedef struct
 
     /*
      * OPTIONAL pointer to struct containing attribute type input metadata
-     * 
+     *
      * attinmeta is for use when returning tuples (i.e., composite data types)
      * and is not used when returning base data types. It is only needed
      * if you intend to use BuildTupleFromCStrings() to create the return
@@ -2948,7 +2969,7 @@ retcomposite(PG_FUNCTION_ARGS)
     call_cntr = funcctx-&gt;call_cntr;
     max_calls = funcctx-&gt;max_calls;
     attinmeta = funcctx-&gt;attinmeta;
- 
+
     if (call_cntr &lt; max_calls)    /* do when there is more left to send */
     {
         char       **values;
@@ -3126,7 +3147,7 @@ CREATE FUNCTION make_array(anyelement) RETURNS anyarray
     <para>
      Add-ins can reserve LWLocks and an allocation of shared memory on server
      startup.  The add-in's shared library must be preloaded by specifying
-     it in 
+     it in
      <xref linkend="guc-shared-preload-libraries"><indexterm><primary>shared-preload-libraries</></>.
      Shared memory is reserved by calling:
 <programlisting>
@@ -3139,11 +3160,11 @@ void RequestAddinShmemSpace(int size)
 <programlisting>
 void RequestAddinLWLocks(int n)
 </programlisting>
-     from <function>_PG_init</>. 
+     from <function>_PG_init</>.
     </para>
     <para>
      To avoid possible race-conditions, each backend should use the LWLock
-     <function>AddinShmemInitLock</> when connecting to and initializing 
+     <function>AddinShmemInitLock</> when connecting to and initializing
      its allocation of shared memory, as shown here:
 <programlisting>
         static mystruct *ptr = NULL;

src/backend/executor/execQual.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/executor/execQual.c,v 1.235 2008/10/29 00:00:38 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/executor/execQual.c,v 1.236 2008/10/31 19:37:56 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -1334,7 +1334,8 @@ ExecMakeFunctionResult(FuncExprState *fcache,
 {
 	List	   *arguments;
 	Datum		result;
-	FunctionCallInfoData fcinfo;
+	FunctionCallInfoData fcinfo_data;
+	FunctionCallInfo fcinfo;
 	PgStat_FunctionCallUsage fcusage;
 	ReturnSetInfo rsinfo;		/* for functions returning sets */
 	ExprDoneCond argDone;
@@ -1384,6 +1385,20 @@ restart:
 		Assert(!fcache->setArgsValid);
 	}
 
+	/*
+	 * For non-set-returning functions, we just use a local-variable
+	 * FunctionCallInfoData.  For set-returning functions we keep the callinfo
+	 * record in fcache->setArgs so that it can survive across multiple
+	 * value-per-call invocations.  (The reason we don't just do the latter
+	 * all the time is that plpgsql expects to be able to use simple expression
+	 * trees re-entrantly.  Which might not be a good idea, but the penalty
+	 * for not doing so is high.)
+	 */
+	if (fcache->func.fn_retset)
+		fcinfo = &fcache->setArgs;
+	else
+		fcinfo = &fcinfo_data;
+
 	/*
 	 * arguments is a list of expressions to evaluate before passing to the
 	 * function manager.  We skip the evaluation if it was already done in the
@@ -1394,8 +1409,8 @@ restart:
 	if (!fcache->setArgsValid)
 	{
 		/* Need to prep callinfo structure */
-		InitFunctionCallInfoData(fcinfo, &(fcache->func), 0, NULL, NULL);
+		InitFunctionCallInfoData(*fcinfo, &(fcache->func), 0, NULL, NULL);
-		argDone = ExecEvalFuncArgs(&fcinfo, arguments, econtext);
+		argDone = ExecEvalFuncArgs(fcinfo, arguments, econtext);
 		if (argDone == ExprEndResult)
 		{
 			/* input is an empty set, so return an empty set. */
@@ -1412,8 +1427,7 @@ restart:
 	}
 	else
 	{
-		/* Copy callinfo from previous evaluation */
+		/* Re-use callinfo from previous evaluation */
-		memcpy(&fcinfo, &fcache->setArgs, sizeof(fcinfo));
 		hasSetArg = fcache->setHasSetArg;
 		/* Reset flag (we may set it again below) */
 		fcache->setArgsValid = false;
@@ -1424,12 +1438,12 @@ restart:
 	 */
 	if (fcache->func.fn_retset)
 	{
-		fcinfo.resultinfo = (Node *) &rsinfo;
+		fcinfo->resultinfo = (Node *) &rsinfo;
 		rsinfo.type = T_ReturnSetInfo;
 		rsinfo.econtext = econtext;
 		rsinfo.expectedDesc = fcache->funcResultDesc;
 		rsinfo.allowedModes = (int) (SFRM_ValuePerCall | SFRM_Materialize);
-		/* note we do not set SFRM_Materialize_Random */
+		/* note we do not set SFRM_Materialize_Random or _Preferred */
 		rsinfo.returnMode = SFRM_ValuePerCall;
 		/* isDone is filled below */
 		rsinfo.setResult = NULL;
@@ -1468,9 +1482,9 @@ restart:
 
 			if (fcache->func.fn_strict)
 			{
-				for (i = 0; i < fcinfo.nargs; i++)
+				for (i = 0; i < fcinfo->nargs; i++)
 				{
-					if (fcinfo.argnull[i])
+					if (fcinfo->argnull[i])
 					{
 						callit = false;
 						break;
@@ -1480,12 +1494,12 @@ restart:
 
 			if (callit)
 			{
-				pgstat_init_function_usage(&fcinfo, &fcusage);
+				pgstat_init_function_usage(fcinfo, &fcusage);
 
-				fcinfo.isnull = false;
+				fcinfo->isnull = false;
 				rsinfo.isDone = ExprSingleResult;
-				result = FunctionCallInvoke(&fcinfo);
+				result = FunctionCallInvoke(fcinfo);
-				*isNull = fcinfo.isnull;
+				*isNull = fcinfo->isnull;
 				*isDone = rsinfo.isDone;
 
 				pgstat_end_function_usage(&fcusage,
@@ -1511,7 +1525,7 @@ restart:
 					if (fcache->func.fn_retset &&
 						*isDone == ExprMultipleResult)
 					{
-						memcpy(&fcache->setArgs, &fcinfo, sizeof(fcinfo));
+						Assert(fcinfo == &fcache->setArgs);
 						fcache->setHasSetArg = hasSetArg;
 						fcache->setArgsValid = true;
 						/* Register cleanup callback if we didn't already */
@@ -1567,7 +1581,7 @@ restart:
 				break;			/* input not a set, so done */
 
 			/* Re-eval args to get the next element of the input set */
-			argDone = ExecEvalFuncArgs(&fcinfo, arguments, econtext);
+			argDone = ExecEvalFuncArgs(fcinfo, arguments, econtext);
 
 			if (argDone != ExprMultipleResult)
 			{
@@ -1605,9 +1619,9 @@ restart:
 		 */
 		if (fcache->func.fn_strict)
 		{
-			for (i = 0; i < fcinfo.nargs; i++)
+			for (i = 0; i < fcinfo->nargs; i++)
 			{
-				if (fcinfo.argnull[i])
+				if (fcinfo->argnull[i])
 				{
 					*isNull = true;
 					return (Datum) 0;
@@ -1615,11 +1629,11 @@ restart:
 			}
 		}
 
-		pgstat_init_function_usage(&fcinfo, &fcusage);
+		pgstat_init_function_usage(fcinfo, &fcusage);
 
-		fcinfo.isnull = false;
+		fcinfo->isnull = false;
-		result = FunctionCallInvoke(&fcinfo);
+		result = FunctionCallInvoke(fcinfo);
-		*isNull = fcinfo.isnull;
+		*isNull = fcinfo->isnull;
 
 		pgstat_end_function_usage(&fcusage, true);
 	}
@@ -1737,7 +1751,7 @@ ExecMakeTableFunctionResult(ExprState *funcexpr,
 	rsinfo.type = T_ReturnSetInfo;
 	rsinfo.econtext = econtext;
 	rsinfo.expectedDesc = expectedDesc;
-	rsinfo.allowedModes = (int) (SFRM_ValuePerCall | SFRM_Materialize);
+	rsinfo.allowedModes = (int) (SFRM_ValuePerCall | SFRM_Materialize | SFRM_Materialize_Preferred);
 	if (randomAccess)
 		rsinfo.allowedModes |= (int) SFRM_Materialize_Random;
 	rsinfo.returnMode = SFRM_ValuePerCall;

src/backend/tcop/dest.c

@@ -8,7 +8,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/tcop/dest.c,v 1.72 2008/01/01 19:45:52 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/tcop/dest.c,v 1.73 2008/10/31 19:37:56 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -32,6 +32,7 @@
 #include "access/xact.h"
 #include "commands/copy.h"
 #include "executor/executor.h"
+#include "executor/functions.h"
 #include "executor/tstoreReceiver.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
@@ -132,6 +133,9 @@ CreateDestReceiver(CommandDest dest, Portal portal)
 
 		case DestCopyOut:
 			return CreateCopyDestReceiver();
+
+		case DestSQLFunction:
+			return CreateSQLFunctionDestReceiver();
 	}
 
 	/* should never get here */
@@ -158,6 +162,7 @@ EndCommand(const char *commandTag, CommandDest dest)
 		case DestTuplestore:
 		case DestIntoRel:
 		case DestCopyOut:
+		case DestSQLFunction:
 			break;
 	}
 }
@@ -198,6 +203,7 @@ NullCommand(CommandDest dest)
 		case DestTuplestore:
 		case DestIntoRel:
 		case DestCopyOut:
+		case DestSQLFunction:
 			break;
 	}
 }
@@ -240,6 +246,7 @@ ReadyForQuery(CommandDest dest)
 		case DestTuplestore:
 		case DestIntoRel:
 		case DestCopyOut:
+		case DestSQLFunction:
 			break;
 	}
 }

src/backend/utils/fmgr/README

@@ -1,4 +1,4 @@
-$PostgreSQL: pgsql/src/backend/utils/fmgr/README,v 1.15 2008/10/29 00:00:38 tgl Exp $
+$PostgreSQL: pgsql/src/backend/utils/fmgr/README,v 1.16 2008/10/31 19:37:56 tgl Exp $
 
 Function Manager
 ================
@@ -434,7 +434,9 @@ and returns null.  isDone is not used and should be left at ExprSingleResult.
 
 The Tuplestore must be created with randomAccess = true if
 SFRM_Materialize_Random is set in allowedModes, but it can (and preferably
-should) be created with randomAccess = false if not.
+should) be created with randomAccess = false if not.  Callers that can support
+both ValuePerCall and Materialize mode will set SFRM_Materialize_Preferred,
+or not, depending on which mode they prefer.
 
 If available, the expected tuple descriptor is passed in ReturnSetInfo;
 in other contexts the expectedDesc field will be NULL.  The function need

src/include/executor/functions.h

@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/executor/functions.h,v 1.31 2008/03/18 22:04:14 tgl Exp $
+ * $PostgreSQL: pgsql/src/include/executor/functions.h,v 1.32 2008/10/31 19:37:56 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -15,6 +15,7 @@
 #define FUNCTIONS_H
 
 #include "nodes/execnodes.h"
+#include "tcop/dest.h"
 
 
 extern Datum fmgr_sql(PG_FUNCTION_ARGS);
@@ -24,4 +25,6 @@ extern bool check_sql_fn_retval(Oid func_id, Oid rettype,
 					bool insertRelabels,
 					JunkFilter **junkFilter);
 
+extern DestReceiver *CreateSQLFunctionDestReceiver(void);
+
 #endif   /* FUNCTIONS_H */

src/include/nodes/execnodes.h

@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/nodes/execnodes.h,v 1.193 2008/10/29 00:00:39 tgl Exp $
+ * $PostgreSQL: pgsql/src/include/nodes/execnodes.h,v 1.194 2008/10/31 19:37:56 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -151,13 +151,15 @@ typedef enum
 /*
  * Return modes for functions returning sets.  Note values must be chosen
  * as separate bits so that a bitmask can be formed to indicate supported
- * modes.
+ * modes.  SFRM_Materialize_Random and SFRM_Materialize_Preferred are
+ * auxiliary flags about SFRM_Materialize mode, rather than separate modes.
  */
 typedef enum
 {
 	SFRM_ValuePerCall = 0x01,	/* one value returned per call */
 	SFRM_Materialize = 0x02,	/* result set instantiated in Tuplestore */
-	SFRM_Materialize_Random = 0x04		/* Tuplestore needs randomAccess */
+	SFRM_Materialize_Random = 0x04,		/* Tuplestore needs randomAccess */
+	SFRM_Materialize_Preferred = 0x08	/* caller prefers Tuplestore */
 } SetFunctionReturnMode;
 
 /*

src/include/tcop/dest.h

@@ -54,7 +54,7 @@
  * Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/tcop/dest.h,v 1.54 2008/01/01 19:45:59 momjian Exp $
+ * $PostgreSQL: pgsql/src/include/tcop/dest.h,v 1.55 2008/10/31 19:37:56 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -86,7 +86,8 @@ typedef enum
 	DestSPI,					/* results sent to SPI manager */
 	DestTuplestore,				/* results sent to Tuplestore */
 	DestIntoRel,				/* results sent to relation (SELECT INTO) */
-	DestCopyOut					/* results sent to COPY TO code */
+	DestCopyOut,				/* results sent to COPY TO code */
+	DestSQLFunction				/* results sent to SQL-language func mgr */
 } CommandDest;
 
 /* ----------------

src/test/regress/expected/rangefuncs.out

@@ -567,3 +567,179 @@ SELECT * FROM foo(3);
 (9 rows)
 
 DROP FUNCTION foo(int);
+--
+-- some tests on SQL functions with RETURNING
+--
+create temp table tt(f1 serial, data text);
+NOTICE:  CREATE TABLE will create implicit sequence "tt_f1_seq" for serial column "tt.f1"
+create function insert_tt(text) returns int as
+$$ insert into tt(data) values($1) returning f1 $$
+language sql;
+select insert_tt('foo');
+ insert_tt 
+-----------
+         1
+(1 row)
+
+select insert_tt('bar');
+ insert_tt 
+-----------
+         2
+(1 row)
+
+select * from tt;
+ f1 | data 
+----+------
+  1 | foo
+  2 | bar
+(2 rows)
+
+-- insert will execute to completion even if function needs just 1 row
+create or replace function insert_tt(text) returns int as
+$$ insert into tt(data) values($1),($1||$1) returning f1 $$
+language sql;
+select insert_tt('fool');
+ insert_tt 
+-----------
+         3
+(1 row)
+
+select * from tt;
+ f1 |   data   
+----+----------
+  1 | foo
+  2 | bar
+  3 | fool
+  4 | foolfool
+(4 rows)
+
+-- setof does what's expected
+create or replace function insert_tt2(text,text) returns setof int as
+$$ insert into tt(data) values($1),($2) returning f1 $$
+language sql;
+select insert_tt2('foolish','barrish');
+ insert_tt2 
+------------
+          5
+          6
+(2 rows)
+
+select * from insert_tt2('baz','quux');
+ insert_tt2 
+------------
+          7
+          8
+(2 rows)
+
+select * from tt;
+ f1 |   data   
+----+----------
+  1 | foo
+  2 | bar
+  3 | fool
+  4 | foolfool
+  5 | foolish
+  6 | barrish
+  7 | baz
+  8 | quux
+(8 rows)
+
+-- limit doesn't prevent execution to completion
+select insert_tt2('foolish','barrish') limit 1;
+ insert_tt2 
+------------
+          9
+(1 row)
+
+select * from tt;
+ f1 |   data   
+----+----------
+  1 | foo
+  2 | bar
+  3 | fool
+  4 | foolfool
+  5 | foolish
+  6 | barrish
+  7 | baz
+  8 | quux
+  9 | foolish
+ 10 | barrish
+(10 rows)
+
+-- triggers will fire, too
+create function noticetrigger() returns trigger as $$
+begin
+  raise notice 'noticetrigger % %', new.f1, new.data;
+  return null;
+end $$ language plpgsql;
+create trigger tnoticetrigger after insert on tt for each row
+execute procedure noticetrigger();
+select insert_tt2('foolme','barme') limit 1;
+NOTICE:  noticetrigger 11 foolme
+CONTEXT:  SQL function "insert_tt2" statement 1
+NOTICE:  noticetrigger 12 barme
+CONTEXT:  SQL function "insert_tt2" statement 1
+ insert_tt2 
+------------
+         11
+(1 row)
+
+select * from tt;
+ f1 |   data   
+----+----------
+  1 | foo
+  2 | bar
+  3 | fool
+  4 | foolfool
+  5 | foolish
+  6 | barrish
+  7 | baz
+  8 | quux
+  9 | foolish
+ 10 | barrish
+ 11 | foolme
+ 12 | barme
+(12 rows)
+
+-- and rules work
+create temp table tt_log(f1 int, data text);
+create rule insert_tt_rule as on insert to tt do also
+  insert into tt_log values(new.*);
+select insert_tt2('foollog','barlog') limit 1;
+NOTICE:  noticetrigger 13 foollog
+CONTEXT:  SQL function "insert_tt2" statement 1
+NOTICE:  noticetrigger 14 barlog
+CONTEXT:  SQL function "insert_tt2" statement 1
+ insert_tt2 
+------------
+         13
+(1 row)
+
+select * from tt;
+ f1 |   data   
+----+----------
+  1 | foo
+  2 | bar
+  3 | fool
+  4 | foolfool
+  5 | foolish
+  6 | barrish
+  7 | baz
+  8 | quux
+  9 | foolish
+ 10 | barrish
+ 11 | foolme
+ 12 | barme
+ 13 | foollog
+ 14 | barlog
+(14 rows)
+
+-- note that nextval() gets executed a second time in the rule expansion,
+-- which is expected.
+select * from tt_log;
+ f1 |  data   
+----+---------
+ 15 | foollog
+ 16 | barlog
+(2 rows)
+

src/test/regress/output/create_function_1.source

@@ -61,7 +61,7 @@ LINE 2:     AS 'not even SQL';
 CREATE FUNCTION test1 (int) RETURNS int LANGUAGE SQL
     AS 'SELECT 1, 2, 3;';
 ERROR:  return type mismatch in function declared to return integer
-DETAIL:  Final SELECT must return exactly one column.
+DETAIL:  Final statement must return exactly one column.
 CONTEXT:  SQL function "test1"
 CREATE FUNCTION test1 (int) RETURNS int LANGUAGE SQL
     AS 'SELECT $2;';

src/test/regress/sql/rangefuncs.sql

@@ -279,3 +279,62 @@ AS $$ SELECT a, b
               generate_series(1,$1) b(b) $$ LANGUAGE sql;
 SELECT * FROM foo(3);
 DROP FUNCTION foo(int);
+
+--
+-- some tests on SQL functions with RETURNING
+--
+
+create temp table tt(f1 serial, data text);
+
+create function insert_tt(text) returns int as
+$$ insert into tt(data) values($1) returning f1 $$
+language sql;
+
+select insert_tt('foo');
+select insert_tt('bar');
+select * from tt;
+
+-- insert will execute to completion even if function needs just 1 row
+create or replace function insert_tt(text) returns int as
+$$ insert into tt(data) values($1),($1||$1) returning f1 $$
+language sql;
+
+select insert_tt('fool');
+select * from tt;
+
+-- setof does what's expected
+create or replace function insert_tt2(text,text) returns setof int as
+$$ insert into tt(data) values($1),($2) returning f1 $$
+language sql;
+
+select insert_tt2('foolish','barrish');
+select * from insert_tt2('baz','quux');
+select * from tt;
+
+-- limit doesn't prevent execution to completion
+select insert_tt2('foolish','barrish') limit 1;
+select * from tt;
+
+-- triggers will fire, too
+create function noticetrigger() returns trigger as $$
+begin
+  raise notice 'noticetrigger % %', new.f1, new.data;
+  return null;
+end $$ language plpgsql;
+create trigger tnoticetrigger after insert on tt for each row
+execute procedure noticetrigger();
+
+select insert_tt2('foolme','barme') limit 1;
+select * from tt;
+
+-- and rules work
+create temp table tt_log(f1 int, data text);
+
+create rule insert_tt_rule as on insert to tt do also
+  insert into tt_log values(new.*);
+
+select insert_tt2('foollog','barlog') limit 1;
+select * from tt;
+-- note that nextval() gets executed a second time in the rule expansion,
+-- which is expected.
+select * from tt_log;