Allow FDWs and custom scan providers to replace joins with scans.

Foreign data wrappers can use this capability for so-called "join
pushdown"; that is, instead of executing two separate foreign scans
and then joining the results locally, they can generate a path which
performs the join on the remote server and then is scanned locally.
This commit does not extend postgres_fdw to take advantage of this
capability; it just provides the infrastructure.

Custom scan providers can use this in a similar way.  Previously,
it was only possible for a custom scan provider to scan a single
relation.  Now, it can scan an entire join tree, provided of course
that it knows how to produce the same results that the join would
have produced if executed normally.

KaiGai Kohei, reviewed by Shigeru Hanada, Ashutosh Bapat, and me.

doc/src/sgml/custom-scan.sgml

@@ -81,6 +81,28 @@ typedef struct CustomPath
     detailed below.
   </para>
 
+  <para>
+   A custom scan provider can also add join paths; in this case, the scan
+   must produce the same output as would normally be produced by the join
+   it replaces.  To do this, the join provider should set the following hook.
+   This hook may be invoked repeatedly for the same pair of relations, with
+   different combinations of inner and outer relations; it is the
+   responsibility of the hook to minimize duplicated work.
+<programlisting>
+typedef void (*set_join_pathlist_hook_type) (PlannerInfo *root,
+                                             RelOptInfo *joinrel,
+                                             RelOptInfo *outerrel,
+                                             RelOptInfo *innerrel,
+                                             List *restrictlist,
+                                             JoinType jointype,
+                                             SpecialJoinInfo *sjinfo,
+                                             SemiAntiJoinFactors *semifactors,
+                                             Relids param_source_rels,
+                                             Relids extra_lateral_rels);
+extern PGDLLIMPORT set_join_pathlist_hook_type set_join_pathlist_hook;
+</programlisting>
+  </para>
+
   <sect2 id="custom-scan-path-callbacks">
   <title>Custom Path Callbacks</title>
 
@@ -124,7 +146,9 @@ typedef struct CustomScan
     Scan      scan;
     uint32    flags;
     List     *custom_exprs;
+    List     *custom_ps_tlist;
     List     *custom_private;
+    List     *custom_relids;
     const CustomScanMethods *methods;
 } CustomScan;
 </programlisting>
@@ -141,11 +165,27 @@ typedef struct CustomScan
     is only used by the custom scan provider itself.  Plan trees must be able
     to be duplicated using <function>copyObject</>, so all the data stored
     within these two fields must consist of nodes that function can handle.
+    <literal>custom_relids</> is set by the core code to the set of relations
+    which this scan node must handle; except when this scan is replacing a
+    join, it will have only one member.
     <structfield>methods</> must point to a (usually statically allocated)
     object implementing the required custom scan methods, which are further
     detailed below.
   </para>
 
+  <para>
+   When a <structname>CustomScan</> scans a single relation,
+   <structfield>scan.scanrelid</> should be the range table index of the table
+   to be scanned, and <structfield>custom_ps_tlist</> should be
+   <literal>NULL</>.  When it replaces a join, <structfield>scan.scanrelid</>
+   should be zero, and <structfield>custom_ps_tlist</> should be a list of
+   <structname>TargetEntry</> nodes.  This is necessary because, when a join
+   is replaced, the target list cannot be constructed from the table
+   definition.  At execution time, this list will be used to initialize the
+   tuple descriptor of the <structname>TupleTableSlot</>.  It will also be
+   used by <command>EXPLAIN</>, when deparsing.
+  </para>
+
   <sect2 id="custom-scan-plan-callbacks">
    <title>Custom Scan Callbacks</title>
    <para>

doc/src/sgml/fdwhandler.sgml

@@ -598,6 +598,42 @@ IsForeignRelUpdatable (Relation rel);
 
    </sect2>
 
+   <sect2>
+    <title>FDW Routines For Remote Joins</title>
+    <para>
+<programlisting>
+void
+GetForeignJoinPaths(PlannerInfo *root,
+                    RelOptInfo *joinrel,
+                    RelOptInfo *outerrel,
+                    RelOptInfo *innerrel,
+                    List *restrictlist,
+                    JoinType jointype,
+                    SpecialJoinInfo *sjinfo,
+                    SemiAntiJoinFactors *semifactors,
+                    Relids param_source_rels,
+                    Relids extra_lateral_rels);
+</programlisting>
+     Create possible access paths for a join of two foreign tables managed
+     by the same foreign data wrapper.
+     This optional function is called during query planning.
+    </para>
+    <para>
+     This function the FDW to add <structname>ForeignScan</> paths for the
+     supplied <literal>joinrel</>.  Typically, the FDW will send the whole
+     join to the remote server as a single query, as performing the join
+     remotely rather than locally is typically much more efficient.
+    </para>
+    <para>
+     Since we cannot construct the slot descriptor for a remote join from
+     the catalogs, the FDW should set the <structfield>scanrelid</> of the
+     <structname>ForeignScan</> to zero and <structfield>fdw_ps_tlist</>
+     to an appropriate list of <structfield>TargetEntry</> nodes.
+     Junk entries will be ignored, but can be present for the benefit of
+     deparsing performed by <command>EXPLAIN</>.
+    </para>
+   </sect2>
+
    <sect2 id="fdw-callbacks-explain">
     <title>FDW Routines for <command>EXPLAIN</></title>
 

src/backend/commands/explain.c

@@ -730,11 +730,17 @@ ExplainPreScanNode(PlanState *planstate, Bitmapset **rels_used)
 		case T_ValuesScan:
 		case T_CteScan:
 		case T_WorkTableScan:
-		case T_ForeignScan:
-		case T_CustomScan:
 			*rels_used = bms_add_member(*rels_used,
 										((Scan *) plan)->scanrelid);
 			break;
+		case T_ForeignScan:
+			*rels_used = bms_add_members(*rels_used,
+										 ((ForeignScan *) plan)->fdw_relids);
+			break;
+		case T_CustomScan:
+			*rels_used = bms_add_members(*rels_used,
+										 ((CustomScan *) plan)->custom_relids);
+			break;
 		case T_ModifyTable:
 			*rels_used = bms_add_member(*rels_used,
 									((ModifyTable *) plan)->nominalRelation);
@@ -1072,8 +1078,11 @@ ExplainNode(PlanState *planstate, List *ancestors,
 		case T_ValuesScan:
 		case T_CteScan:
 		case T_WorkTableScan:
+			ExplainScanTarget((Scan *) plan, es);
+			break;
 		case T_ForeignScan:
 		case T_CustomScan:
+			if (((Scan *) plan)->scanrelid > 0)
 				ExplainScanTarget((Scan *) plan, es);
 			break;
 		case T_IndexScan:

src/backend/executor/execScan.c

@@ -251,6 +251,12 @@ ExecAssignScanProjectionInfo(ScanState *node)
 	/* Vars in an index-only scan's tlist should be INDEX_VAR */
 	if (IsA(scan, IndexOnlyScan))
 		varno = INDEX_VAR;
+	/* Also foreign or custom scan on pseudo relation should be INDEX_VAR */
+	else if (scan->scanrelid == 0)
+	{
+		Assert(IsA(scan, ForeignScan) || IsA(scan, CustomScan));
+		varno = INDEX_VAR;
+	}
 	else
 		varno = scan->scanrelid;
 

src/backend/executor/nodeCustom.c

@@ -23,7 +23,7 @@ CustomScanState *
 ExecInitCustomScan(CustomScan *cscan, EState *estate, int eflags)
 {
 	CustomScanState    *css;
-	Relation			scan_rel;
+	Index				scan_relid = cscan->scan.scanrelid;
 
 	/* populate a CustomScanState according to the CustomScan */
 	css = (CustomScanState *) cscan->methods->CreateCustomScanState(cscan);
@@ -48,12 +48,26 @@ ExecInitCustomScan(CustomScan *cscan, EState *estate, int eflags)
 	ExecInitScanTupleSlot(estate, &css->ss);
 	ExecInitResultTupleSlot(estate, &css->ss.ps);
 
-	/* initialize scan relation */
-	scan_rel = ExecOpenScanRelation(estate, cscan->scan.scanrelid, eflags);
+	/*
+	 * open the base relation and acquire an appropriate lock on it;
+	 * also, get and assign the scan type
+	 */
+	if (scan_relid > 0)
+	{
+		Relation		scan_rel;
+
+		scan_rel = ExecOpenScanRelation(estate, scan_relid, eflags);
 		css->ss.ss_currentRelation = scan_rel;
 		css->ss.ss_currentScanDesc = NULL;	/* set by provider */
 		ExecAssignScanType(&css->ss, RelationGetDescr(scan_rel));
-
+	}
+	else
+	{
+		TupleDesc	ps_tupdesc;
+
+		ps_tupdesc = ExecCleanTypeFromTL(cscan->custom_ps_tlist, false);
+		ExecAssignScanType(&css->ss, ps_tupdesc);
+	}
 	css->ss.ps.ps_TupFromTlist = false;
 
 	/*
@@ -89,10 +103,10 @@ ExecEndCustomScan(CustomScanState *node)
 
 	/* Clean out the tuple table */
 	ExecClearTuple(node->ss.ps.ps_ResultTupleSlot);
-	if (node->ss.ss_ScanTupleSlot)
 	ExecClearTuple(node->ss.ss_ScanTupleSlot);
 
 	/* Close the heap relation */
+	if (node->ss.ss_currentRelation)
 		ExecCloseScanRelation(node->ss.ss_currentRelation);
 }
 

src/backend/executor/nodeForeignscan.c

@@ -102,7 +102,7 @@ ForeignScanState *
 ExecInitForeignScan(ForeignScan *node, EState *estate, int eflags)
 {
 	ForeignScanState *scanstate;
-	Relation	currentRelation;
+	Index		scanrelid = node->scan.scanrelid;
 	FdwRoutine *fdwroutine;
 
 	/* check for unsupported flags */
@@ -141,16 +141,24 @@ ExecInitForeignScan(ForeignScan *node, EState *estate, int eflags)
 	ExecInitScanTupleSlot(estate, &scanstate->ss);
 
 	/*
-	 * open the base relation and acquire appropriate lock on it.
+	 * open the base relation and acquire an appropriate lock on it;
+	 * also, get and assign the scan type
 	 */
-	currentRelation = ExecOpenScanRelation(estate, node->scan.scanrelid, eflags);
-	scanstate->ss.ss_currentRelation = currentRelation;
+	if (scanrelid > 0)
+	{
+		Relation	currentRelation;
 
-	/*
-	 * get the scan type from the relation descriptor.  (XXX at some point we
-	 * might want to let the FDW editorialize on the scan tupdesc.)
-	 */
+		currentRelation = ExecOpenScanRelation(estate, scanrelid, eflags);
+		scanstate->ss.ss_currentRelation = currentRelation;
 		ExecAssignScanType(&scanstate->ss, RelationGetDescr(currentRelation));
+	}
+	else
+	{
+		TupleDesc	ps_tupdesc;
+
+		ps_tupdesc = ExecCleanTypeFromTL(node->fdw_ps_tlist, false);
+		ExecAssignScanType(&scanstate->ss, ps_tupdesc);
+	}
 
 	/*
 	 * Initialize result tuple type and projection info.
@@ -161,7 +169,7 @@ ExecInitForeignScan(ForeignScan *node, EState *estate, int eflags)
 	/*
 	 * Acquire function pointers from the FDW's handler, and init fdw_state.
 	 */
-	fdwroutine = GetFdwRoutineForRelation(currentRelation, true);
+	fdwroutine = GetFdwRoutine(node->fdw_handler);
 	scanstate->fdwroutine = fdwroutine;
 	scanstate->fdw_state = NULL;
 
@@ -193,6 +201,7 @@ ExecEndForeignScan(ForeignScanState *node)
 	ExecClearTuple(node->ss.ss_ScanTupleSlot);
 
 	/* close the relation. */
+	if (node->ss.ss_currentRelation)
 		ExecCloseScanRelation(node->ss.ss_currentRelation);
 }
 

src/backend/foreign/foreign.c

@@ -304,11 +304,11 @@ GetFdwRoutine(Oid fdwhandler)
 
 
 /*
- * GetFdwRoutineByRelId - look up the handler of the foreign-data wrapper
- * for the given foreign table, and retrieve its FdwRoutine struct.
+ * GetFdwHandlerByRelId - look up the handler of the foreign-data wrapper
+ * for the given foreign table
  */
-FdwRoutine *
-GetFdwRoutineByRelId(Oid relid)
+Oid
+GetFdwHandlerByRelId(Oid relid)
 {
 	HeapTuple	tp;
 	Form_pg_foreign_data_wrapper fdwform;
@@ -350,7 +350,18 @@ GetFdwRoutineByRelId(Oid relid)
 
 	ReleaseSysCache(tp);
 
-	/* And finally, call the handler function. */
+	return fdwhandler;
+}
+
+/*
+ * GetFdwRoutineByRelId - look up the handler of the foreign-data wrapper
+ * for the given foreign table, and retrieve its FdwRoutine struct.
+ */
+FdwRoutine *
+GetFdwRoutineByRelId(Oid relid)
+{
+	Oid			fdwhandler = GetFdwHandlerByRelId(relid);
+
 	return GetFdwRoutine(fdwhandler);
 }
 

src/backend/nodes/copyfuncs.c

@@ -592,8 +592,11 @@ _copyForeignScan(const ForeignScan *from)
 	/*
 	 * copy remainder of node
 	 */
+	COPY_SCALAR_FIELD(fdw_handler);
 	COPY_NODE_FIELD(fdw_exprs);
+	COPY_NODE_FIELD(fdw_ps_tlist);
 	COPY_NODE_FIELD(fdw_private);
+	COPY_BITMAPSET_FIELD(fdw_relids);
 	COPY_SCALAR_FIELD(fsSystemCol);
 
 	return newnode;
@@ -617,7 +620,9 @@ _copyCustomScan(const CustomScan *from)
 	 */
 	COPY_SCALAR_FIELD(flags);
 	COPY_NODE_FIELD(custom_exprs);
+	COPY_NODE_FIELD(custom_ps_tlist);
 	COPY_NODE_FIELD(custom_private);
+	COPY_BITMAPSET_FIELD(custom_relids);
 
 	/*
 	 * NOTE: The method field of CustomScan is required to be a pointer to a

src/backend/nodes/outfuncs.c

@@ -558,8 +558,11 @@ _outForeignScan(StringInfo str, const ForeignScan *node)
 
 	_outScanInfo(str, (const Scan *) node);
 
+	WRITE_OID_FIELD(fdw_handler);
 	WRITE_NODE_FIELD(fdw_exprs);
+	WRITE_NODE_FIELD(fdw_ps_tlist);
 	WRITE_NODE_FIELD(fdw_private);
+	WRITE_BITMAPSET_FIELD(fdw_relids);
 	WRITE_BOOL_FIELD(fsSystemCol);
 }
 
@@ -572,7 +575,9 @@ _outCustomScan(StringInfo str, const CustomScan *node)
 
 	WRITE_UINT_FIELD(flags);
 	WRITE_NODE_FIELD(custom_exprs);
+	WRITE_NODE_FIELD(custom_ps_tlist);
 	WRITE_NODE_FIELD(custom_private);
+	WRITE_BITMAPSET_FIELD(custom_relids);
 	appendStringInfoString(str, " :methods ");
 	_outToken(str, node->methods->CustomName);
 	if (node->methods->TextOutCustomScan)

src/backend/optimizer/path/joinpath.c

@@ -17,10 +17,13 @@
 #include <math.h>
 
 #include "executor/executor.h"
+#include "foreign/fdwapi.h"
 #include "optimizer/cost.h"
 #include "optimizer/pathnode.h"
 #include "optimizer/paths.h"
 
+/* Hook for plugins to get control in add_paths_to_joinrel() */
+set_join_pathlist_hook_type set_join_pathlist_hook = NULL;
 
 #define PATH_PARAM_BY_REL(path, rel)  \
 	((path)->param_info && bms_overlap(PATH_REQ_OUTER(path), (rel)->relids))
@@ -260,6 +263,27 @@ add_paths_to_joinrel(PlannerInfo *root,
 							 restrictlist, jointype,
 							 sjinfo, &semifactors,
 							 param_source_rels, extra_lateral_rels);
+
+	/*
+	 * 5. If both inner and outer relations are managed by the same FDW,
+	 * give it a chance to push down joins.
+	 */
+	if (joinrel->fdwroutine &&
+		joinrel->fdwroutine->GetForeignJoinPaths)
+		joinrel->fdwroutine->GetForeignJoinPaths(root, joinrel,
+												 outerrel, innerrel,
+												 restrictlist, jointype, sjinfo,
+												 &semifactors,
+												 param_source_rels,
+												 extra_lateral_rels);
+	/*
+	 * 6. Finally, give extensions a chance to manipulate the path list.
+	 */
+	if (set_join_pathlist_hook)
+		set_join_pathlist_hook(root, joinrel, outerrel, innerrel,
+							   restrictlist, jointype,
+							   sjinfo, &semifactors,
+							   param_source_rels, extra_lateral_rels);
 }
 
 /*

src/backend/optimizer/plan/createplan.c

@@ -44,7 +44,6 @@
 #include "utils/lsyscache.h"
 
 
-static Plan *create_plan_recurse(PlannerInfo *root, Path *best_path);
 static Plan *create_scan_plan(PlannerInfo *root, Path *best_path);
 static List *build_path_tlist(PlannerInfo *root, Path *path);
 static bool use_physical_tlist(PlannerInfo *root, RelOptInfo *rel);
@@ -220,7 +219,7 @@ create_plan(PlannerInfo *root, Path *best_path)
  * create_plan_recurse
  *	  Recursive guts of create_plan().
  */
-static Plan *
+Plan *
 create_plan_recurse(PlannerInfo *root, Path *best_path)
 {
 	Plan	   *plan;
@@ -1961,16 +1960,25 @@ create_foreignscan_plan(PlannerInfo *root, ForeignPath *best_path,
 	ForeignScan *scan_plan;
 	RelOptInfo *rel = best_path->path.parent;
 	Index		scan_relid = rel->relid;
-	RangeTblEntry *rte;
+	Oid			rel_oid = InvalidOid;
 	Bitmapset  *attrs_used = NULL;
 	ListCell   *lc;
 	int			i;
 
-	/* it should be a base rel... */
-	Assert(scan_relid > 0);
+	/*
+	 * If we're scanning a base relation, look up the OID.
+	 * (We can skip this if scanning a join relation.)
+	 */
+	if (scan_relid > 0)
+	{
+		RangeTblEntry *rte;
+
 		Assert(rel->rtekind == RTE_RELATION);
 		rte = planner_rt_fetch(scan_relid, root);
 		Assert(rte->rtekind == RTE_RELATION);
+		rel_oid = rte->relid;
+	}
+	Assert(rel->fdwroutine != NULL);
 
 	/*
 	 * Sort clauses into best execution order.  We do this first since the FDW
@@ -1985,13 +1993,39 @@ create_foreignscan_plan(PlannerInfo *root, ForeignPath *best_path,
 	 * has selected some join clauses for remote use but also wants them
 	 * rechecked locally).
 	 */
-	scan_plan = rel->fdwroutine->GetForeignPlan(root, rel, rte->relid,
+	scan_plan = rel->fdwroutine->GetForeignPlan(root, rel, rel_oid,
 												best_path,
 												tlist, scan_clauses);
+	/*
+	 * Sanity check.  There may be resjunk entries in fdw_ps_tlist that
+	 * are included only to help EXPLAIN deparse plans properly. We require
+	 * that these are at the end, so that when the executor builds the scan
+	 * descriptor based on the non-junk entries, it gets the attribute
+	 * numbers correct.
+	 */
+	if (scan_plan->scan.scanrelid == 0)
+	{
+		bool	found_resjunk = false;
+
+		foreach (lc, scan_plan->fdw_ps_tlist)
+		{
+			TargetEntry	   *tle = lfirst(lc);
+
+			if (tle->resjunk)
+				found_resjunk = true;
+			else if (found_resjunk)
+				elog(ERROR, "junk TLE should not apper prior to valid one");
+		}
+	}
+	/* Set the relids that are represented by this foreign scan for Explain */
+	scan_plan->fdw_relids = best_path->path.parent->relids;
 
 	/* Copy cost data from Path to Plan; no need to make FDW do this */
 	copy_path_costsize(&scan_plan->scan.plan, &best_path->path);
 
+	/* Track FDW server-id; no need to make FDW do this */
+	scan_plan->fdw_handler = rel->fdw_handler;
+
 	/*
 	 * Replace any outer-relation variables with nestloop params in the qual
 	 * and fdw_exprs expressions.  We do this last so that the FDW doesn't
@@ -2053,12 +2087,7 @@ create_customscan_plan(PlannerInfo *root, CustomPath *best_path,
 {
 	CustomScan *cplan;
 	RelOptInfo *rel = best_path->path.parent;
-
-	/*
-	 * Right now, all we can support is CustomScan node which is associated
-	 * with a particular base relation to be scanned.
-	 */
-	Assert(rel && rel->reloptkind == RELOPT_BASEREL);
+	ListCell   *lc;
 
 	/*
 	 * Sort clauses into the best execution order, although custom-scan
@@ -2077,6 +2106,30 @@ create_customscan_plan(PlannerInfo *root, CustomPath *best_path,
 															  scan_clauses);
 	Assert(IsA(cplan, CustomScan));
 
+	/*
+	 * Sanity check.  There may be resjunk entries in custom_ps_tlist that
+	 * are included only to help EXPLAIN deparse plans properly. We require
+	 * that these are at the end, so that when the executor builds the scan
+	 * descriptor based on the non-junk entries, it gets the attribute
+	 * numbers correct.
+	 */
+	if (cplan->scan.scanrelid == 0)
+	{
+		bool	found_resjunk = false;
+
+		foreach (lc, cplan->custom_ps_tlist)
+		{
+			TargetEntry	   *tle = lfirst(lc);
+
+			if (tle->resjunk)
+				found_resjunk = true;
+			else if (found_resjunk)
+				elog(ERROR, "junk TLE should not apper prior to valid one");
+		}
+	}
+	/* Set the relids that are represented by this custom scan for Explain */
+	cplan->custom_relids = best_path->path.parent->relids;
+
 	/*
 	 * Copy cost data from Path to Plan; no need to make custom-plan providers
 	 * do this

src/backend/optimizer/plan/setrefs.c

@@ -86,6 +86,12 @@ static void flatten_unplanned_rtes(PlannerGlobal *glob, RangeTblEntry *rte);
 static bool flatten_rtes_walker(Node *node, PlannerGlobal *glob);
 static void add_rte_to_flat_rtable(PlannerGlobal *glob, RangeTblEntry *rte);
 static Plan *set_plan_refs(PlannerInfo *root, Plan *plan, int rtoffset);
+static void set_foreignscan_references(PlannerInfo *root,
+									   ForeignScan *fscan,
+									   int rtoffset);
+static void set_customscan_references(PlannerInfo *root,
+									  CustomScan *cscan,
+									  int rtoffset);
 static Plan *set_indexonlyscan_references(PlannerInfo *root,
 							 IndexOnlyScan *plan,
 							 int rtoffset);
@@ -565,31 +571,11 @@ set_plan_refs(PlannerInfo *root, Plan *plan, int rtoffset)
 			}
 			break;
 		case T_ForeignScan:
-			{
-				ForeignScan *splan = (ForeignScan *) plan;
-
-				splan->scan.scanrelid += rtoffset;
-				splan->scan.plan.targetlist =
-					fix_scan_list(root, splan->scan.plan.targetlist, rtoffset);
-				splan->scan.plan.qual =
-					fix_scan_list(root, splan->scan.plan.qual, rtoffset);
-				splan->fdw_exprs =
-					fix_scan_list(root, splan->fdw_exprs, rtoffset);
-			}
+			set_foreignscan_references(root, (ForeignScan *) plan, rtoffset);
 			break;
 
 		case T_CustomScan:
-			{
-				CustomScan *splan = (CustomScan *) plan;
-
-				splan->scan.scanrelid += rtoffset;
-				splan->scan.plan.targetlist =
-					fix_scan_list(root, splan->scan.plan.targetlist, rtoffset);
-				splan->scan.plan.qual =
-					fix_scan_list(root, splan->scan.plan.qual, rtoffset);
-				splan->custom_exprs =
-					fix_scan_list(root, splan->custom_exprs, rtoffset);
-			}
+			set_customscan_references(root, (CustomScan *) plan, rtoffset);
 			break;
 
 		case T_NestLoop:
@@ -876,6 +862,121 @@ set_plan_refs(PlannerInfo *root, Plan *plan, int rtoffset)
 	return plan;
 }
 
+/*
+ * set_foreignscan_references
+ *     Do set_plan_references processing on an ForeignScan
+ */
+static void
+set_foreignscan_references(PlannerInfo *root,
+						   ForeignScan *fscan,
+						   int rtoffset)
+{
+	if (rtoffset > 0)
+	{
+		Bitmapset  *tempset = NULL;
+		int			x = -1;
+
+		while ((x = bms_next_member(fscan->fdw_relids, x)) >= 0)
+			tempset = bms_add_member(tempset, x + rtoffset);
+		fscan->fdw_relids = tempset;
+	}
+
+	if (fscan->scan.scanrelid == 0)
+	{
+		indexed_tlist *pscan_itlist = build_tlist_index(fscan->fdw_ps_tlist);
+
+		fscan->scan.plan.targetlist = (List *)
+			fix_upper_expr(root,
+						   (Node *) fscan->scan.plan.targetlist,
+						   pscan_itlist,
+						   INDEX_VAR,
+						   rtoffset);
+		fscan->scan.plan.qual = (List *)
+			fix_upper_expr(root,
+						   (Node *) fscan->scan.plan.qual,
+						   pscan_itlist,
+						   INDEX_VAR,
+						   rtoffset);
+		fscan->fdw_exprs = (List *)
+			fix_upper_expr(root,
+						   (Node *) fscan->fdw_exprs,
+						   pscan_itlist,
+						   INDEX_VAR,
+						   rtoffset);
+		fscan->fdw_ps_tlist =
+			fix_scan_list(root, fscan->fdw_ps_tlist, rtoffset);
+		pfree(pscan_itlist);
+	}
+	else
+	{
+		fscan->scan.scanrelid += rtoffset;
+		fscan->scan.plan.targetlist =
+			fix_scan_list(root, fscan->scan.plan.targetlist, rtoffset);
+		fscan->scan.plan.qual =
+			fix_scan_list(root, fscan->scan.plan.qual, rtoffset);
+		fscan->fdw_exprs =
+			fix_scan_list(root, fscan->fdw_exprs, rtoffset);
+	}
+}
+
+/*
+ * set_customscan_references
+ *     Do set_plan_references processing on an CustomScan
+ */
+static void
+set_customscan_references(PlannerInfo *root,
+						  CustomScan *cscan,
+						  int rtoffset)
+{
+	if (rtoffset > 0)
+	{
+		Bitmapset  *tempset = NULL;
+		int			x = -1;
+
+		while ((x = bms_next_member(cscan->custom_relids, x)) >= 0)
+			tempset = bms_add_member(tempset, x + rtoffset);
+		cscan->custom_relids = tempset;
+	}
+
+	if (cscan->scan.scanrelid == 0)
+	{
+		indexed_tlist *pscan_itlist =
+			build_tlist_index(cscan->custom_ps_tlist);
+
+		cscan->scan.plan.targetlist = (List *)
+			fix_upper_expr(root,
+						   (Node *) cscan->scan.plan.targetlist,
+						   pscan_itlist,
+						   INDEX_VAR,
+						   rtoffset);
+		cscan->scan.plan.qual = (List *)
+			fix_upper_expr(root,
+						   (Node *) cscan->scan.plan.qual,
+						   pscan_itlist,
+						   INDEX_VAR,
+						   rtoffset);
+		cscan->custom_exprs = (List *)
+			fix_upper_expr(root,
+						   (Node *) cscan->custom_exprs,
+						   pscan_itlist,
+						   INDEX_VAR,
+						   rtoffset);
+		cscan->custom_ps_tlist =
+			fix_scan_list(root, cscan->custom_ps_tlist, rtoffset);
+		pfree(pscan_itlist);
+	}
+	else
+	{
+		cscan->scan.scanrelid += rtoffset;
+		cscan->scan.plan.targetlist =
+			fix_scan_list(root, cscan->scan.plan.targetlist, rtoffset);
+		cscan->scan.plan.qual =
+			fix_scan_list(root, cscan->scan.plan.qual, rtoffset);
+		cscan->custom_exprs =
+			fix_scan_list(root, cscan->custom_exprs, rtoffset);
+	}
+}
+
 /*
  * set_indexonlyscan_references
  *		Do set_plan_references processing on an IndexOnlyScan

src/backend/optimizer/util/plancat.c

@@ -379,10 +379,15 @@ get_relation_info(PlannerInfo *root, Oid relationObjectId, bool inhparent,
 
 	/* Grab the fdwroutine info using the relcache, while we have it */
 	if (relation->rd_rel->relkind == RELKIND_FOREIGN_TABLE)
+	{
+		rel->fdw_handler = GetFdwHandlerByRelId(RelationGetRelid(relation));
 		rel->fdwroutine = GetFdwRoutineForRelation(relation, true);
+	}
 	else
+	{
+		rel->fdw_handler = InvalidOid;
 		rel->fdwroutine = NULL;
-
+	}
 	heap_close(relation, NoLock);
 
 	/*

src/backend/optimizer/util/relnode.c

@@ -14,6 +14,7 @@
  */
 #include "postgres.h"
 
+#include "foreign/fdwapi.h"
 #include "optimizer/cost.h"
 #include "optimizer/pathnode.h"
 #include "optimizer/paths.h"
@@ -122,6 +123,7 @@ build_simple_rel(PlannerInfo *root, int relid, RelOptKind reloptkind)
 	rel->subroot = NULL;
 	rel->subplan_params = NIL;
 	rel->fdwroutine = NULL;
+	rel->fdw_handler = InvalidOid;
 	rel->fdw_private = NULL;
 	rel->baserestrictinfo = NIL;
 	rel->baserestrictcost.startup = 0;
@@ -426,6 +428,18 @@ build_join_rel(PlannerInfo *root,
 	set_joinrel_size_estimates(root, joinrel, outer_rel, inner_rel,
 							   sjinfo, restrictlist);
 
+	/*
+	 * Set FDW handler and routine if both outer and inner relation
+	 * are managed by same FDW driver.
+	 */
+	if (OidIsValid(outer_rel->fdw_handler) &&
+		OidIsValid(inner_rel->fdw_handler) &&
+		outer_rel->fdw_handler == inner_rel->fdw_handler)
+	{
+		joinrel->fdw_handler = outer_rel->fdw_handler;
+		joinrel->fdwroutine = GetFdwRoutine(joinrel->fdw_handler);
+	}
+
 	/*
 	 * Add the joinrel to the query's joinrel list, and store it into the
 	 * auxiliary hashtable if there is one.  NB: GEQO requires us to append

src/backend/utils/adt/ruleutils.c

@@ -3862,6 +3862,10 @@ set_deparse_planstate(deparse_namespace *dpns, PlanState *ps)
 	/* index_tlist is set only if it's an IndexOnlyScan */
 	if (IsA(ps->plan, IndexOnlyScan))
 		dpns->index_tlist = ((IndexOnlyScan *) ps->plan)->indextlist;
+	else if (IsA(ps->plan, ForeignScan))
+		dpns->index_tlist = ((ForeignScan *) ps->plan)->fdw_ps_tlist;
+	else if (IsA(ps->plan, CustomScan))
+		dpns->index_tlist = ((CustomScan *) ps->plan)->custom_ps_tlist;
 	else
 		dpns->index_tlist = NIL;
 }

src/include/foreign/fdwapi.h

@@ -82,6 +82,17 @@ typedef void (*EndForeignModify_function) (EState *estate,
 
 typedef int (*IsForeignRelUpdatable_function) (Relation rel);
 
+typedef void (*GetForeignJoinPaths_function) (PlannerInfo *root,
+											  RelOptInfo *joinrel,
+											  RelOptInfo *outerrel,
+											  RelOptInfo *innerrel,
+											  List *restrictlist,
+											  JoinType jointype,
+											  SpecialJoinInfo *sjinfo,
+											  SemiAntiJoinFactors *semifactors,
+											  Relids param_source_rels,
+											  Relids extra_lateral_rels);
+
 typedef void (*ExplainForeignScan_function) (ForeignScanState *node,
 													struct ExplainState *es);
 
@@ -150,10 +161,14 @@ typedef struct FdwRoutine
 
 	/* Support functions for IMPORT FOREIGN SCHEMA */
 	ImportForeignSchema_function ImportForeignSchema;
+
+	/* Support functions for join push-down */
+	GetForeignJoinPaths_function GetForeignJoinPaths;
 } FdwRoutine;
 
 
 /* Functions in foreign/foreign.c */
+extern Oid GetFdwHandlerByRelId(Oid relid);
 extern FdwRoutine *GetFdwRoutine(Oid fdwhandler);
 extern FdwRoutine *GetFdwRoutineByRelId(Oid relid);
 extern FdwRoutine *GetFdwRoutineForRelation(Relation relation, bool makecopy);

src/include/nodes/plannodes.h

@@ -471,7 +471,11 @@ typedef struct WorkTableScan
  * fdw_exprs and fdw_private are both under the control of the foreign-data
  * wrapper, but fdw_exprs is presumed to contain expression trees and will
  * be post-processed accordingly by the planner; fdw_private won't be.
- * Note that everything in both lists must be copiable by copyObject().
+ * An optional fdw_ps_tlist is used to map a reference to an attribute of
+ * underlying relation(s) onto a pair of INDEX_VAR and alternative varattno.
+ * When fdw_ps_tlist is used, this represents a remote join, and the FDW
+ * is responsible for setting this field to an appropriate value.
+ * Note that everything in above lists must be copiable by copyObject().
  * One way to store an arbitrary blob of bytes is to represent it as a bytea
  * Const.  Usually, though, you'll be better off choosing a representation
  * that can be dumped usefully by nodeToString().
@@ -480,18 +484,22 @@ typedef struct WorkTableScan
 typedef struct ForeignScan
 {
 	Scan		scan;
+	Oid			fdw_handler;	/* OID of FDW handler */
 	List	   *fdw_exprs;		/* expressions that FDW may evaluate */
+	List	   *fdw_ps_tlist;	/* tlist, if replacing a join */
 	List	   *fdw_private;	/* private data for FDW */
+	Bitmapset  *fdw_relids;		/* RTIs generated by this scan */
 	bool		fsSystemCol;	/* true if any "system column" is needed */
 } ForeignScan;
 
 /* ----------------
  *	   CustomScan node
  *
- * The comments for ForeignScan's fdw_exprs and fdw_private fields apply
- * equally to custom_exprs and custom_private.  Note that since Plan trees
- * can be copied, custom scan providers *must* fit all plan data they need
- * into those fields; embedding CustomScan in a larger struct will not work.
+ * The comments for ForeignScan's fdw_exprs, fdw_varmap and fdw_private fields
+ * apply equally to custom_exprs, custom_ps_tlist and custom_private.
+ * Note that since Plan trees can be copied, custom scan providers *must*
+ * fit all plan data they need into those fields; embedding CustomScan in
+ * a larger struct will not work.
  * ----------------
  */
 struct CustomScan;
@@ -512,7 +520,9 @@ typedef struct CustomScan
 	Scan		scan;
 	uint32		flags;			/* mask of CUSTOMPATH_* flags, see relation.h */
 	List	   *custom_exprs;	/* expressions that custom code may evaluate */
+	List	   *custom_ps_tlist;/* tlist, if replacing a join */
 	List	   *custom_private; /* private data for custom code */
+	Bitmapset  *custom_relids;	/* RTIs generated by this scan */
 	const CustomScanMethods *methods;
 } CustomScan;
 

src/include/nodes/relation.h

@@ -366,6 +366,7 @@ typedef struct PlannerInfo
  *		subroot - PlannerInfo for subquery (NULL if it's not a subquery)
  *		subplan_params - list of PlannerParamItems to be passed to subquery
  *		fdwroutine - function hooks for FDW, if foreign table (else NULL)
+ *		fdw_handler - OID of FDW handler, if foreign table (else InvalidOid)
  *		fdw_private - private state for FDW, if foreign table (else NULL)
  *
  *		Note: for a subquery, tuples, subplan, subroot are not set immediately
@@ -461,6 +462,7 @@ typedef struct RelOptInfo
 	List	   *subplan_params; /* if subquery */
 	/* use "struct FdwRoutine" to avoid including fdwapi.h here */
 	struct FdwRoutine *fdwroutine;		/* if foreign table */
+	Oid			fdw_handler;	/* if foreign table */
 	void	   *fdw_private;	/* if foreign table */
 
 	/* used by various scans and joins: */

src/include/optimizer/paths.h

@@ -30,6 +30,19 @@ typedef void (*set_rel_pathlist_hook_type) (PlannerInfo *root,
 														RangeTblEntry *rte);
 extern PGDLLIMPORT set_rel_pathlist_hook_type set_rel_pathlist_hook;
 
+/* Hook for plugins to get control in add_paths_to_joinrel() */
+typedef void (*set_join_pathlist_hook_type) (PlannerInfo *root,
+											 RelOptInfo *joinrel,
+											 RelOptInfo *outerrel,
+											 RelOptInfo *innerrel,
+											 List *restrictlist,
+											 JoinType jointype,
+											 SpecialJoinInfo *sjinfo,
+											 SemiAntiJoinFactors *semifactors,
+											 Relids param_source_rels,
+											 Relids extra_lateral_rels);
+extern PGDLLIMPORT set_join_pathlist_hook_type set_join_pathlist_hook;
+
 /* Hook for plugins to replace standard_join_search() */
 typedef RelOptInfo *(*join_search_hook_type) (PlannerInfo *root,
 														  int levels_needed,

src/include/optimizer/planmain.h

@@ -41,6 +41,7 @@ extern Plan *optimize_minmax_aggregates(PlannerInfo *root, List *tlist,
  * prototypes for plan/createplan.c
  */
 extern Plan *create_plan(PlannerInfo *root, Path *best_path);
+extern Plan *create_plan_recurse(PlannerInfo *root, Path *best_path);
 extern SubqueryScan *make_subqueryscan(List *qptlist, List *qpqual,
 				  Index scanrelid, Plan *subplan);
 extern ForeignScan *make_foreignscan(List *qptlist, List *qpqual,