On 2017/01/13 0:43, Robert Haas wrote:
On Wed, Jan 11, 2017 at 2:45 AM, Etsuro Fujita
<fujita.ets...@lab.ntt.co.jp> wrote:
As I said before, that might be fine for 9.6, but I don't think it's a good
idea to search the pathlist because once we support parameterized foreign
join paths, which is on my TODOs, we would have to traverse through the
possibly-lengthy pathlist to find a local-join path, as mentioned in [3].


I'm not sure that's really going to be a problem.  The number of
possible parameterizations that need to be considered isn't usually
very big.  I bet the path list will have ten or a few tens of entries
in it, not a hundred or a thousand.  Traversing it isn't that
expensive.

That having been said, I haven't read the patches, so I'm not really
up to speed on the bigger issues here.  But surely it's more important
to get the overall design right than to worry about the cost of
walking the pathlist or worrying about the cost of an extra function
call (?!).

My biggest concern about GetExistingLocalJoinPath is that might not be extendable to the case of foreign-join paths with parameterization; in which case, fdw_outerpath for a given foreign-join path would need to have the same parameterization as the foreign-join path, but there might not be any existing paths with the same parameterization in the path list. You might think we could get the fdw_outerpath by getting an existing path with no parameterization as in GetExistingLocalJoinPath and then modifying the path's param_info to match the parameterization of the foreign-join path. I don't know that really works, but that might be inefficient.

What I have in mind to support foreign-join paths with parameterization for postgres_fdw like this: (1) generate parameterized paths from any joinable combination of the outer/inner cheapest-parameterized paths that have pushed down the outer/inner relation to the remote server, in a similar way as postgresGetForeignJoinPaths creates unparameterized paths, and (2) create fdw_outerpath for each parameterized path from the outer/inner paths used to generate the parameterized path, by create_nestloop_path (or, create_hashjoin_path or create_mergejoin_path if full join), so that the resulting fdw_outerpath has the same parameterization as the paramterized path. This would probably work and might be more efficient. And the patch I proposed would be easily extended to this, by replacing the outer/inner cheapest-total paths with the outer/inner cheapest-parameterized paths. Attached is the latest version of the patch.

Best regards,
Etsuro Fujita
*** a/contrib/postgres_fdw/expected/postgres_fdw.out
--- b/contrib/postgres_fdw/expected/postgres_fdw.out
***************
*** 1519,1540 **** SELECT t1.c1, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1) ORDER BY t1.c3, t
                 Output: t1.c1, t2.c1, t1.c3, t1.*, t2.*
                 Relations: (public.ft1 t1) INNER JOIN (public.ft2 t2)
                 Remote SQL: SELECT r1."C 1", r1.c3, CASE WHEN (r1.*)::text IS NOT NULL THEN ROW(r1."C 1", r1.c2, r1.c3, r1.c4, r1.c5, r1.c6, r1.c7, r1.c8) END, r2."C 1", CASE WHEN (r2.*)::text IS NOT NULL THEN ROW(r2."C 1", r2.c2, r2.c3, r2.c4, r2.c5, r2.c6, r2.c7, r2.c8) END FROM ("S 1"."T 1" r1 INNER JOIN "S 1"."T 1" r2 ON (((r1."C 1" = r2."C 1")))) ORDER BY r1.c3 ASC NULLS LAST, r1."C 1" ASC NULLS LAST FOR UPDATE OF r1
!                ->  Merge Join
                       Output: t1.c1, t1.c3, t1.*, t2.c1, t2.*
!                      Merge Cond: (t1.c1 = t2.c1)
!                      ->  Sort
                             Output: t1.c1, t1.c3, t1.*
!                            Sort Key: t1.c1
!                            ->  Foreign Scan on public.ft1 t1
!                                  Output: t1.c1, t1.c3, t1.*
!                                  Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1" FOR UPDATE
!                      ->  Sort
                             Output: t2.c1, t2.*
!                            Sort Key: t2.c1
!                            ->  Foreign Scan on public.ft2 t2
!                                  Output: t2.c1, t2.*
!                                  Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1"
! (23 rows)
  
  SELECT t1.c1, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1) ORDER BY t1.c3, t1.c1 OFFSET 100 LIMIT 10 FOR UPDATE OF t1;
   c1  | c1  
--- 1519,1534 ----
                 Output: t1.c1, t2.c1, t1.c3, t1.*, t2.*
                 Relations: (public.ft1 t1) INNER JOIN (public.ft2 t2)
                 Remote SQL: SELECT r1."C 1", r1.c3, CASE WHEN (r1.*)::text IS NOT NULL THEN ROW(r1."C 1", r1.c2, r1.c3, r1.c4, r1.c5, r1.c6, r1.c7, r1.c8) END, r2."C 1", CASE WHEN (r2.*)::text IS NOT NULL THEN ROW(r2."C 1", r2.c2, r2.c3, r2.c4, r2.c5, r2.c6, r2.c7, r2.c8) END FROM ("S 1"."T 1" r1 INNER JOIN "S 1"."T 1" r2 ON (((r1."C 1" = r2."C 1")))) ORDER BY r1.c3 ASC NULLS LAST, r1."C 1" ASC NULLS LAST FOR UPDATE OF r1
!                ->  Nested Loop
                       Output: t1.c1, t1.c3, t1.*, t2.c1, t2.*
!                      Join Filter: (t1.c1 = t2.c1)
!                      ->  Foreign Scan on public.ft1 t1
                             Output: t1.c1, t1.c3, t1.*
!                            Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1" FOR UPDATE
!                      ->  Foreign Scan on public.ft2 t2
                             Output: t2.c1, t2.*
!                            Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1"
! (17 rows)
  
  SELECT t1.c1, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1) ORDER BY t1.c3, t1.c1 OFFSET 100 LIMIT 10 FOR UPDATE OF t1;
   c1  | c1  
***************
*** 1563,1584 **** SELECT t1.c1, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1) ORDER BY t1.c3, t
                 Output: t1.c1, t2.c1, t1.c3, t1.*, t2.*
                 Relations: (public.ft1 t1) INNER JOIN (public.ft2 t2)
                 Remote SQL: SELECT r1."C 1", r1.c3, CASE WHEN (r1.*)::text IS NOT NULL THEN ROW(r1."C 1", r1.c2, r1.c3, r1.c4, r1.c5, r1.c6, r1.c7, r1.c8) END, r2."C 1", CASE WHEN (r2.*)::text IS NOT NULL THEN ROW(r2."C 1", r2.c2, r2.c3, r2.c4, r2.c5, r2.c6, r2.c7, r2.c8) END FROM ("S 1"."T 1" r1 INNER JOIN "S 1"."T 1" r2 ON (((r1."C 1" = r2."C 1")))) ORDER BY r1.c3 ASC NULLS LAST, r1."C 1" ASC NULLS LAST FOR UPDATE OF r1 FOR UPDATE OF r2
!                ->  Merge Join
                       Output: t1.c1, t1.c3, t1.*, t2.c1, t2.*
!                      Merge Cond: (t1.c1 = t2.c1)
!                      ->  Sort
                             Output: t1.c1, t1.c3, t1.*
!                            Sort Key: t1.c1
!                            ->  Foreign Scan on public.ft1 t1
!                                  Output: t1.c1, t1.c3, t1.*
!                                  Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1" FOR UPDATE
!                      ->  Sort
                             Output: t2.c1, t2.*
!                            Sort Key: t2.c1
!                            ->  Foreign Scan on public.ft2 t2
!                                  Output: t2.c1, t2.*
!                                  Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1" FOR UPDATE
! (23 rows)
  
  SELECT t1.c1, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1) ORDER BY t1.c3, t1.c1 OFFSET 100 LIMIT 10 FOR UPDATE;
   c1  | c1  
--- 1557,1572 ----
                 Output: t1.c1, t2.c1, t1.c3, t1.*, t2.*
                 Relations: (public.ft1 t1) INNER JOIN (public.ft2 t2)
                 Remote SQL: SELECT r1."C 1", r1.c3, CASE WHEN (r1.*)::text IS NOT NULL THEN ROW(r1."C 1", r1.c2, r1.c3, r1.c4, r1.c5, r1.c6, r1.c7, r1.c8) END, r2."C 1", CASE WHEN (r2.*)::text IS NOT NULL THEN ROW(r2."C 1", r2.c2, r2.c3, r2.c4, r2.c5, r2.c6, r2.c7, r2.c8) END FROM ("S 1"."T 1" r1 INNER JOIN "S 1"."T 1" r2 ON (((r1."C 1" = r2."C 1")))) ORDER BY r1.c3 ASC NULLS LAST, r1."C 1" ASC NULLS LAST FOR UPDATE OF r1 FOR UPDATE OF r2
!                ->  Nested Loop
                       Output: t1.c1, t1.c3, t1.*, t2.c1, t2.*
!                      Join Filter: (t1.c1 = t2.c1)
!                      ->  Foreign Scan on public.ft1 t1
                             Output: t1.c1, t1.c3, t1.*
!                            Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1" FOR UPDATE
!                      ->  Foreign Scan on public.ft2 t2
                             Output: t2.c1, t2.*
!                            Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1" FOR UPDATE
! (17 rows)
  
  SELECT t1.c1, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1) ORDER BY t1.c3, t1.c1 OFFSET 100 LIMIT 10 FOR UPDATE;
   c1  | c1  
***************
*** 1608,1629 **** SELECT t1.c1, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1) ORDER BY t1.c3, t
                 Output: t1.c1, t2.c1, t1.c3, t1.*, t2.*
                 Relations: (public.ft1 t1) INNER JOIN (public.ft2 t2)
                 Remote SQL: SELECT r1."C 1", r1.c3, CASE WHEN (r1.*)::text IS NOT NULL THEN ROW(r1."C 1", r1.c2, r1.c3, r1.c4, r1.c5, r1.c6, r1.c7, r1.c8) END, r2."C 1", CASE WHEN (r2.*)::text IS NOT NULL THEN ROW(r2."C 1", r2.c2, r2.c3, r2.c4, r2.c5, r2.c6, r2.c7, r2.c8) END FROM ("S 1"."T 1" r1 INNER JOIN "S 1"."T 1" r2 ON (((r1."C 1" = r2."C 1")))) ORDER BY r1.c3 ASC NULLS LAST, r1."C 1" ASC NULLS LAST FOR SHARE OF r1
!                ->  Merge Join
                       Output: t1.c1, t1.c3, t1.*, t2.c1, t2.*
!                      Merge Cond: (t1.c1 = t2.c1)
!                      ->  Sort
                             Output: t1.c1, t1.c3, t1.*
!                            Sort Key: t1.c1
!                            ->  Foreign Scan on public.ft1 t1
!                                  Output: t1.c1, t1.c3, t1.*
!                                  Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1" FOR SHARE
!                      ->  Sort
                             Output: t2.c1, t2.*
!                            Sort Key: t2.c1
!                            ->  Foreign Scan on public.ft2 t2
!                                  Output: t2.c1, t2.*
!                                  Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1"
! (23 rows)
  
  SELECT t1.c1, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1) ORDER BY t1.c3, t1.c1 OFFSET 100 LIMIT 10 FOR SHARE OF t1;
   c1  | c1  
--- 1596,1611 ----
                 Output: t1.c1, t2.c1, t1.c3, t1.*, t2.*
                 Relations: (public.ft1 t1) INNER JOIN (public.ft2 t2)
                 Remote SQL: SELECT r1."C 1", r1.c3, CASE WHEN (r1.*)::text IS NOT NULL THEN ROW(r1."C 1", r1.c2, r1.c3, r1.c4, r1.c5, r1.c6, r1.c7, r1.c8) END, r2."C 1", CASE WHEN (r2.*)::text IS NOT NULL THEN ROW(r2."C 1", r2.c2, r2.c3, r2.c4, r2.c5, r2.c6, r2.c7, r2.c8) END FROM ("S 1"."T 1" r1 INNER JOIN "S 1"."T 1" r2 ON (((r1."C 1" = r2."C 1")))) ORDER BY r1.c3 ASC NULLS LAST, r1."C 1" ASC NULLS LAST FOR SHARE OF r1
!                ->  Nested Loop
                       Output: t1.c1, t1.c3, t1.*, t2.c1, t2.*
!                      Join Filter: (t1.c1 = t2.c1)
!                      ->  Foreign Scan on public.ft1 t1
                             Output: t1.c1, t1.c3, t1.*
!                            Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1" FOR SHARE
!                      ->  Foreign Scan on public.ft2 t2
                             Output: t2.c1, t2.*
!                            Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1"
! (17 rows)
  
  SELECT t1.c1, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1) ORDER BY t1.c3, t1.c1 OFFSET 100 LIMIT 10 FOR SHARE OF t1;
   c1  | c1  
***************
*** 1652,1673 **** SELECT t1.c1, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1) ORDER BY t1.c3, t
                 Output: t1.c1, t2.c1, t1.c3, t1.*, t2.*
                 Relations: (public.ft1 t1) INNER JOIN (public.ft2 t2)
                 Remote SQL: SELECT r1."C 1", r1.c3, CASE WHEN (r1.*)::text IS NOT NULL THEN ROW(r1."C 1", r1.c2, r1.c3, r1.c4, r1.c5, r1.c6, r1.c7, r1.c8) END, r2."C 1", CASE WHEN (r2.*)::text IS NOT NULL THEN ROW(r2."C 1", r2.c2, r2.c3, r2.c4, r2.c5, r2.c6, r2.c7, r2.c8) END FROM ("S 1"."T 1" r1 INNER JOIN "S 1"."T 1" r2 ON (((r1."C 1" = r2."C 1")))) ORDER BY r1.c3 ASC NULLS LAST, r1."C 1" ASC NULLS LAST FOR SHARE OF r1 FOR SHARE OF r2
!                ->  Merge Join
                       Output: t1.c1, t1.c3, t1.*, t2.c1, t2.*
!                      Merge Cond: (t1.c1 = t2.c1)
!                      ->  Sort
                             Output: t1.c1, t1.c3, t1.*
!                            Sort Key: t1.c1
!                            ->  Foreign Scan on public.ft1 t1
!                                  Output: t1.c1, t1.c3, t1.*
!                                  Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1" FOR SHARE
!                      ->  Sort
                             Output: t2.c1, t2.*
!                            Sort Key: t2.c1
!                            ->  Foreign Scan on public.ft2 t2
!                                  Output: t2.c1, t2.*
!                                  Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1" FOR SHARE
! (23 rows)
  
  SELECT t1.c1, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1) ORDER BY t1.c3, t1.c1 OFFSET 100 LIMIT 10 FOR SHARE;
   c1  | c1  
--- 1634,1649 ----
                 Output: t1.c1, t2.c1, t1.c3, t1.*, t2.*
                 Relations: (public.ft1 t1) INNER JOIN (public.ft2 t2)
                 Remote SQL: SELECT r1."C 1", r1.c3, CASE WHEN (r1.*)::text IS NOT NULL THEN ROW(r1."C 1", r1.c2, r1.c3, r1.c4, r1.c5, r1.c6, r1.c7, r1.c8) END, r2."C 1", CASE WHEN (r2.*)::text IS NOT NULL THEN ROW(r2."C 1", r2.c2, r2.c3, r2.c4, r2.c5, r2.c6, r2.c7, r2.c8) END FROM ("S 1"."T 1" r1 INNER JOIN "S 1"."T 1" r2 ON (((r1."C 1" = r2."C 1")))) ORDER BY r1.c3 ASC NULLS LAST, r1."C 1" ASC NULLS LAST FOR SHARE OF r1 FOR SHARE OF r2
!                ->  Nested Loop
                       Output: t1.c1, t1.c3, t1.*, t2.c1, t2.*
!                      Join Filter: (t1.c1 = t2.c1)
!                      ->  Foreign Scan on public.ft1 t1
                             Output: t1.c1, t1.c3, t1.*
!                            Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1" FOR SHARE
!                      ->  Foreign Scan on public.ft2 t2
                             Output: t2.c1, t2.*
!                            Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1" FOR SHARE
! (17 rows)
  
  SELECT t1.c1, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1) ORDER BY t1.c3, t1.c1 OFFSET 100 LIMIT 10 FOR SHARE;
   c1  | c1  
***************
*** 1684,1689 **** SELECT t1.c1, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1) ORDER BY t1.c3, t
--- 1660,1726 ----
   110 | 110
  (10 rows)
  
+ -- FOR UPDATE/SHARE in situations where a full join is pushed down
+ EXPLAIN (VERBOSE, COSTS OFF)
+ SELECT t1.c1, t2.c1, t3.c1 FROM (SELECT c1 FROM "S 1"."T 3" WHERE c1 = 50) t1 INNER JOIN (ft4 t2 FULL JOIN ft5 t3 ON (t2.c1 = t3.c1)) ON (TRUE) ORDER BY t1.c1, t2.c1, t3.c1 FOR UPDATE OF t1;
+                                                                                                                                                      QUERY PLAN                                                                                                                                                      
+ ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+  LockRows
+    Output: "T 3".c1, t2.c1, t3.c1, "T 3".ctid, t2.*, t3.*
+    ->  Nested Loop
+          Output: "T 3".c1, t2.c1, t3.c1, "T 3".ctid, t2.*, t3.*
+          ->  Foreign Scan
+                Output: t2.c1, t2.*, t3.c1, t3.*
+                Relations: (public.ft4 t2) FULL JOIN (public.ft5 t3)
+                Remote SQL: SELECT r2.c1, CASE WHEN (r2.*)::text IS NOT NULL THEN ROW(r2.c1, r2.c2, r2.c3) END, r3.c1, CASE WHEN (r3.*)::text IS NOT NULL THEN ROW(r3.c1, r3.c2, r3.c3) END FROM ("S 1"."T 3" r2 FULL JOIN "S 1"."T 4" r3 ON (((r2.c1 = r3.c1)))) ORDER BY r2.c1 ASC NULLS LAST, r3.c1 ASC NULLS LAST
+                ->  Hash Full Join
+                      Output: t2.c1, t2.*, t3.c1, t3.*
+                      Hash Cond: (t2.c1 = t3.c1)
+                      ->  Foreign Scan on public.ft4 t2
+                            Output: t2.c1, t2.*
+                            Remote SQL: SELECT c1, c2, c3 FROM "S 1"."T 3"
+                      ->  Hash
+                            Output: t3.c1, t3.*
+                            ->  Foreign Scan on public.ft5 t3
+                                  Output: t3.c1, t3.*
+                                  Remote SQL: SELECT c1, c2, c3 FROM "S 1"."T 4"
+          ->  Materialize
+                Output: "T 3".c1, "T 3".ctid
+                ->  Seq Scan on "S 1"."T 3"
+                      Output: "T 3".c1, "T 3".ctid
+                      Filter: ("T 3".c1 = 50)
+ (24 rows)
+ 
+ EXPLAIN (VERBOSE, COSTS OFF)
+ SELECT t1.c1, t2.c1, t3.c1 FROM (SELECT c1 FROM "S 1"."T 3" WHERE c1 = 50) t1 INNER JOIN (ft4 t2 FULL JOIN ft5 t3 ON (FALSE)) ON (TRUE) ORDER BY t1.c1, t2.c1, t3.c1 FOR UPDATE OF t1;
+                                                                                                                                                 QUERY PLAN                                                                                                                                                 
+ -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+  LockRows
+    Output: "T 3".c1, t2.c1, t3.c1, "T 3".ctid, t2.*, t3.*
+    ->  Nested Loop
+          Output: "T 3".c1, t2.c1, t3.c1, "T 3".ctid, t2.*, t3.*
+          ->  Foreign Scan
+                Output: t2.c1, t2.*, t3.c1, t3.*
+                Relations: (public.ft4 t2) FULL JOIN (public.ft5 t3)
+                Remote SQL: SELECT r2.c1, CASE WHEN (r2.*)::text IS NOT NULL THEN ROW(r2.c1, r2.c2, r2.c3) END, r3.c1, CASE WHEN (r3.*)::text IS NOT NULL THEN ROW(r3.c1, r3.c2, r3.c3) END FROM ("S 1"."T 3" r2 FULL JOIN "S 1"."T 4" r3 ON ((false))) ORDER BY r2.c1 ASC NULLS LAST, r3.c1 ASC NULLS LAST
+                ->  Merge Full Join
+                      Output: t2.c1, t2.*, t3.c1, t3.*
+                      Join Filter: false
+                      ->  Foreign Scan on public.ft4 t2
+                            Output: t2.c1, t2.*
+                            Remote SQL: SELECT c1, c2, c3 FROM "S 1"."T 3"
+                      ->  Materialize
+                            Output: t3.c1, t3.*
+                            ->  Foreign Scan on public.ft5 t3
+                                  Output: t3.c1, t3.*
+                                  Remote SQL: SELECT c1, c2, c3 FROM "S 1"."T 4"
+          ->  Materialize
+                Output: "T 3".c1, "T 3".ctid
+                ->  Seq Scan on "S 1"."T 3"
+                      Output: "T 3".c1, "T 3".ctid
+                      Filter: ("T 3".c1 = 50)
+ (24 rows)
+ 
  -- join in CTE
  EXPLAIN (VERBOSE, COSTS OFF)
  WITH t (c1_1, c1_3, c2_1) AS (SELECT t1.c1, t1.c3, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1)) SELECT c1_1, c2_1 FROM t ORDER BY c1_3, c1_1 OFFSET 100 LIMIT 10;
***************
*** 4016,4033 **** UPDATE ft2 SET c2 = ft2.c2 + 500, c3 = ft2.c3 || '_update9', c7 = DEFAULT
           Output: ft2.c1, (ft2.c2 + 500), NULL::integer, (ft2.c3 || '_update9'::text), ft2.c4, ft2.c5, ft2.c6, 'ft2       '::character(10), ft2.c8, ft2.ctid, ft1.*
           Relations: (public.ft2) INNER JOIN (public.ft1)
           Remote SQL: SELECT r1."C 1", r1.c2, r1.c3, r1.c4, r1.c5, r1.c6, r1.c8, r1.ctid, CASE WHEN (r2.*)::text IS NOT NULL THEN ROW(r2."C 1", r2.c2, r2.c3, r2.c4, r2.c5, r2.c6, r2.c7, r2.c8) END FROM ("S 1"."T 1" r1 INNER JOIN "S 1"."T 1" r2 ON (((r1.c2 = r2."C 1")) AND (((r2."C 1" % 10) = 9)))) FOR UPDATE OF r1
!          ->  Hash Join
                 Output: ft2.c1, ft2.c2, ft2.c3, ft2.c4, ft2.c5, ft2.c6, ft2.c8, ft2.ctid, ft1.*
!                Hash Cond: (ft2.c2 = ft1.c1)
                 ->  Foreign Scan on public.ft2
                       Output: ft2.c1, ft2.c2, ft2.c3, ft2.c4, ft2.c5, ft2.c6, ft2.c8, ft2.ctid
                       Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c8, ctid FROM "S 1"."T 1" FOR UPDATE
!                ->  Hash
                       Output: ft1.*, ft1.c1
!                      ->  Foreign Scan on public.ft1
!                            Output: ft1.*, ft1.c1
!                            Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1" WHERE ((("C 1" % 10) = 9))
! (17 rows)
  
  UPDATE ft2 SET c2 = ft2.c2 + 500, c3 = ft2.c3 || '_update9', c7 = DEFAULT
    FROM ft1 WHERE ft1.c1 = ft2.c2 AND ft1.c1 % 10 = 9;
--- 4053,4068 ----
           Output: ft2.c1, (ft2.c2 + 500), NULL::integer, (ft2.c3 || '_update9'::text), ft2.c4, ft2.c5, ft2.c6, 'ft2       '::character(10), ft2.c8, ft2.ctid, ft1.*
           Relations: (public.ft2) INNER JOIN (public.ft1)
           Remote SQL: SELECT r1."C 1", r1.c2, r1.c3, r1.c4, r1.c5, r1.c6, r1.c8, r1.ctid, CASE WHEN (r2.*)::text IS NOT NULL THEN ROW(r2."C 1", r2.c2, r2.c3, r2.c4, r2.c5, r2.c6, r2.c7, r2.c8) END FROM ("S 1"."T 1" r1 INNER JOIN "S 1"."T 1" r2 ON (((r1.c2 = r2."C 1")) AND (((r2."C 1" % 10) = 9)))) FOR UPDATE OF r1
!          ->  Nested Loop
                 Output: ft2.c1, ft2.c2, ft2.c3, ft2.c4, ft2.c5, ft2.c6, ft2.c8, ft2.ctid, ft1.*
!                Join Filter: (ft2.c2 = ft1.c1)
                 ->  Foreign Scan on public.ft2
                       Output: ft2.c1, ft2.c2, ft2.c3, ft2.c4, ft2.c5, ft2.c6, ft2.c8, ft2.ctid
                       Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c8, ctid FROM "S 1"."T 1" FOR UPDATE
!                ->  Foreign Scan on public.ft1
                       Output: ft1.*, ft1.c1
!                      Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1" WHERE ((("C 1" % 10) = 9))
! (15 rows)
  
  UPDATE ft2 SET c2 = ft2.c2 + 500, c3 = ft2.c3 || '_update9', c7 = DEFAULT
    FROM ft1 WHERE ft1.c1 = ft2.c2 AND ft1.c1 % 10 = 9;
***************
*** 4159,4176 **** DELETE FROM ft2 USING ft1 WHERE ft1.c1 = ft2.c2 AND ft1.c1 % 10 = 2;
           Output: ft2.ctid, ft1.*
           Relations: (public.ft2) INNER JOIN (public.ft1)
           Remote SQL: SELECT r1.ctid, CASE WHEN (r2.*)::text IS NOT NULL THEN ROW(r2."C 1", r2.c2, r2.c3, r2.c4, r2.c5, r2.c6, r2.c7, r2.c8) END FROM ("S 1"."T 1" r1 INNER JOIN "S 1"."T 1" r2 ON (((r1.c2 = r2."C 1")) AND (((r2."C 1" % 10) = 2)))) FOR UPDATE OF r1
!          ->  Hash Join
                 Output: ft2.ctid, ft1.*
!                Hash Cond: (ft2.c2 = ft1.c1)
                 ->  Foreign Scan on public.ft2
                       Output: ft2.ctid, ft2.c2
                       Remote SQL: SELECT c2, ctid FROM "S 1"."T 1" FOR UPDATE
!                ->  Hash
                       Output: ft1.*, ft1.c1
!                      ->  Foreign Scan on public.ft1
!                            Output: ft1.*, ft1.c1
!                            Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1" WHERE ((("C 1" % 10) = 2))
! (17 rows)
  
  DELETE FROM ft2 USING ft1 WHERE ft1.c1 = ft2.c2 AND ft1.c1 % 10 = 2;
  SELECT c1,c2,c3,c4 FROM ft2 ORDER BY c1;
--- 4194,4209 ----
           Output: ft2.ctid, ft1.*
           Relations: (public.ft2) INNER JOIN (public.ft1)
           Remote SQL: SELECT r1.ctid, CASE WHEN (r2.*)::text IS NOT NULL THEN ROW(r2."C 1", r2.c2, r2.c3, r2.c4, r2.c5, r2.c6, r2.c7, r2.c8) END FROM ("S 1"."T 1" r1 INNER JOIN "S 1"."T 1" r2 ON (((r1.c2 = r2."C 1")) AND (((r2."C 1" % 10) = 2)))) FOR UPDATE OF r1
!          ->  Nested Loop
                 Output: ft2.ctid, ft1.*
!                Join Filter: (ft2.c2 = ft1.c1)
                 ->  Foreign Scan on public.ft2
                       Output: ft2.ctid, ft2.c2
                       Remote SQL: SELECT c2, ctid FROM "S 1"."T 1" FOR UPDATE
!                ->  Foreign Scan on public.ft1
                       Output: ft1.*, ft1.c1
!                      Remote SQL: SELECT "C 1", c2, c3, c4, c5, c6, c7, c8 FROM "S 1"."T 1" WHERE ((("C 1" % 10) = 2))
! (15 rows)
  
  DELETE FROM ft2 USING ft1 WHERE ft1.c1 = ft2.c2 AND ft1.c1 % 10 = 2;
  SELECT c1,c2,c3,c4 FROM ft2 ORDER BY c1;
*** a/contrib/postgres_fdw/postgres_fdw.c
--- b/contrib/postgres_fdw/postgres_fdw.c
***************
*** 4365,4374 **** postgresGetForeignJoinPaths(PlannerInfo *root,
  		root->parse->commandType == CMD_UPDATE ||
  		root->rowMarks)
  	{
! 		epq_path = GetExistingLocalJoinPath(joinrel);
  		if (!epq_path)
  		{
! 			elog(DEBUG3, "could not push down foreign join because a local path suitable for EPQ checks was not found");
  			return;
  		}
  	}
--- 4365,4375 ----
  		root->parse->commandType == CMD_UPDATE ||
  		root->rowMarks)
  	{
! 		epq_path = CreateLocalJoinPath(root, joinrel, outerrel, innerrel,
! 									   jointype, extra);
  		if (!epq_path)
  		{
! 			elog(DEBUG3, "could not push down foreign join because a local path suitable for EPQ checks was not created");
  			return;
  		}
  	}
*** a/contrib/postgres_fdw/sql/postgres_fdw.sql
--- b/contrib/postgres_fdw/sql/postgres_fdw.sql
***************
*** 445,450 **** SELECT t1.c1, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1) ORDER BY t1.c3, t
--- 445,455 ----
  EXPLAIN (VERBOSE, COSTS OFF)
  SELECT t1.c1, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1) ORDER BY t1.c3, t1.c1 OFFSET 100 LIMIT 10 FOR SHARE;
  SELECT t1.c1, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1) ORDER BY t1.c3, t1.c1 OFFSET 100 LIMIT 10 FOR SHARE;
+ -- FOR UPDATE/SHARE in situations where a full join is pushed down
+ EXPLAIN (VERBOSE, COSTS OFF)
+ SELECT t1.c1, t2.c1, t3.c1 FROM (SELECT c1 FROM "S 1"."T 3" WHERE c1 = 50) t1 INNER JOIN (ft4 t2 FULL JOIN ft5 t3 ON (t2.c1 = t3.c1)) ON (TRUE) ORDER BY t1.c1, t2.c1, t3.c1 FOR UPDATE OF t1;
+ EXPLAIN (VERBOSE, COSTS OFF)
+ SELECT t1.c1, t2.c1, t3.c1 FROM (SELECT c1 FROM "S 1"."T 3" WHERE c1 = 50) t1 INNER JOIN (ft4 t2 FULL JOIN ft5 t3 ON (FALSE)) ON (TRUE) ORDER BY t1.c1, t2.c1, t3.c1 FOR UPDATE OF t1;
  -- join in CTE
  EXPLAIN (VERBOSE, COSTS OFF)
  WITH t (c1_1, c1_3, c2_1) AS (SELECT t1.c1, t1.c3, t2.c1 FROM ft1 t1 JOIN ft2 t2 ON (t1.c1 = t2.c1)) SELECT c1_1, c2_1 FROM t ORDER BY c1_3, c1_1 OFFSET 100 LIMIT 10;
*** a/doc/src/sgml/fdwhandler.sgml
--- b/doc/src/sgml/fdwhandler.sgml
***************
*** 995,1005 **** RecheckForeignScan (ForeignScanState *node, TupleTableSlot *slot);
       can be executed and the resulting tuple can be stored in the slot.
       This plan need not be efficient since no base table will return more
       than one row; for example, it may implement all joins as nested loops.
!      The function <literal>GetExistingLocalJoinPath</> may be used to search
!      existing paths for a suitable local join path, which can be used as the
!      alternative local join plan.  <literal>GetExistingLocalJoinPath</>
!      searches for an unparameterized path in the path list of the specified
!      join relation.  (If it does not find such a path, it returns NULL, in
       which case a foreign data wrapper may build the local path by itself or
       may choose not to create access paths for that join.)
      </para>
--- 995,1003 ----
       can be executed and the resulting tuple can be stored in the slot.
       This plan need not be efficient since no base table will return more
       than one row; for example, it may implement all joins as nested loops.
!      The function <literal>CreateLocalJoinPath</> may be used to build
!      a suitable local join path, which can be used as the alternative local
!      join plan.  (If it does not build such a path, it returns NULL, in
       which case a foreign data wrapper may build the local path by itself or
       may choose not to create access paths for that join.)
      </para>
*** a/src/backend/foreign/foreign.c
--- b/src/backend/foreign/foreign.c
***************
*** 22,27 ****
--- 22,29 ----
  #include "foreign/foreign.h"
  #include "lib/stringinfo.h"
  #include "miscadmin.h"
+ #include "optimizer/cost.h"
+ #include "optimizer/pathnode.h"
  #include "utils/builtins.h"
  #include "utils/memutils.h"
  #include "utils/rel.h"
***************
*** 693,805 **** get_foreign_server_oid(const char *servername, bool missing_ok)
  	return oid;
  }
  
  /*
!  * Get a copy of an existing local path for a given join relation.
!  *
!  * This function is usually helpful to obtain an alternate local path for EPQ
!  * checks.
!  *
!  * Right now, this function only supports unparameterized foreign joins, so we
!  * only search for unparameterized path in the given list of paths. Since we
!  * are searching for a path which can be used to construct an alternative local
!  * plan for a foreign join, we look for only MergeJoin, HashJoin or NestLoop
!  * paths.
   *
!  * If the inner or outer subpath of the chosen path is a ForeignScan, we
!  * replace it with its outer subpath.  For this reason, and also because the
!  * planner might free the original path later, the path returned by this
!  * function is a shallow copy of the original.  There's no need to copy
!  * the substructure, so we don't.
!  *
!  * Since the plan created using this path will presumably only be used to
!  * execute EPQ checks, efficiency of the path is not a concern. But since the
!  * path list in RelOptInfo is anyway sorted by total cost we are likely to
!  * choose the most efficient path, which is all for the best.
   */
! extern Path *
! GetExistingLocalJoinPath(RelOptInfo *joinrel)
  {
! 	ListCell   *lc;
! 
! 	Assert(joinrel->reloptkind == RELOPT_JOINREL);
  
! 	foreach(lc, joinrel->pathlist)
  	{
! 		Path	   *path = (Path *) lfirst(lc);
! 		JoinPath   *joinpath = NULL;
! 
! 		/* Skip parameterised paths. */
! 		if (path->param_info != NULL)
! 			continue;
! 
! 		switch (path->pathtype)
! 		{
! 			case T_HashJoin:
! 				{
! 					HashPath   *hash_path = makeNode(HashPath);
! 
! 					memcpy(hash_path, path, sizeof(HashPath));
! 					joinpath = (JoinPath *) hash_path;
! 				}
! 				break;
! 
! 			case T_NestLoop:
  				{
! 					NestPath   *nest_path = makeNode(NestPath);
! 
! 					memcpy(nest_path, path, sizeof(NestPath));
! 					joinpath = (JoinPath *) nest_path;
  				}
! 				break;
! 
! 			case T_MergeJoin:
  				{
! 					MergePath  *merge_path = makeNode(MergePath);
  
! 					memcpy(merge_path, path, sizeof(MergePath));
! 					joinpath = (JoinPath *) merge_path;
  				}
! 				break;
! 
! 			default:
! 
! 				/*
! 				 * Just skip anything else. We don't know if corresponding
! 				 * plan would build the output row from whole-row references
! 				 * of base relations and execute the EPQ checks.
! 				 */
! 				break;
! 		}
! 
! 		/* This path isn't good for us, check next. */
! 		if (!joinpath)
! 			continue;
! 
! 		/*
! 		 * If either inner or outer path is a ForeignPath corresponding to a
! 		 * pushed down join, replace it with the fdw_outerpath, so that we
! 		 * maintain path for EPQ checks built entirely of local join
! 		 * strategies.
! 		 */
! 		if (IsA(joinpath->outerjoinpath, ForeignPath))
! 		{
! 			ForeignPath *foreign_path;
! 
! 			foreign_path = (ForeignPath *) joinpath->outerjoinpath;
! 			if (foreign_path->path.parent->reloptkind == RELOPT_JOINREL)
! 				joinpath->outerjoinpath = foreign_path->fdw_outerpath;
! 		}
! 
! 		if (IsA(joinpath->innerjoinpath, ForeignPath))
! 		{
! 			ForeignPath *foreign_path;
! 
! 			foreign_path = (ForeignPath *) joinpath->innerjoinpath;
! 			if (foreign_path->path.parent->reloptkind == RELOPT_JOINREL)
! 				joinpath->innerjoinpath = foreign_path->fdw_outerpath;
! 		}
! 
! 		return (Path *) joinpath;
  	}
! 	return NULL;
  }
--- 695,871 ----
  	return oid;
  }
  
+ 
  /*
!  * Make a local join path for a foreign join
   *
!  * Note: currently, this function only supports unparameterized foreign joins.
   */
! Path *
! CreateLocalJoinPath(PlannerInfo *root,
! 					RelOptInfo *joinrel,
! 					RelOptInfo *outerrel,
! 					RelOptInfo *innerrel,
! 					JoinType jointype,
! 					JoinPathExtraData *extra)
  {
! 	Path	   *outer_path = outerrel->cheapest_total_path;
! 	Path	   *inner_path = innerrel->cheapest_total_path;
! 	Path	   *result;
! 	Relids		required_outer;
! 	JoinCostWorkspace workspace;
  
! 	/*
! 	 * Generate a nestloop path (or hashjoin/mergejoin path, if full join)
! 	 * from the cheapest-total paths for the outer and inner relations.
! 	 */
! 	switch (jointype)
  	{
! 		case JOIN_INNER:
! 		case JOIN_LEFT:
! 		case JOIN_SEMI:
! 		case JOIN_ANTI:
! 			/*
! 			 * If the cheapest-total outer path is parameterized by the inner
! 			 * rel, we can't generate a nestloop path.  (There's no use
! 			 * looking for alternative outer paths, since it should already be
! 			 * the least-parameterized available path.)
! 			 */
! 			if (PATH_PARAM_BY_REL(outer_path, innerrel))
! 				return NULL;
! 			/* If proposed path is still parameterized, we can't use it. */
! 			required_outer = calc_nestloop_required_outer(outer_path,
! 														  inner_path);
! 			if (required_outer)
! 			{
! 				bms_free(required_outer);
! 				return NULL;
! 			}
! 			/* Get initial estimates */
! 			initial_cost_nestloop(root, &workspace, jointype,
! 								  outer_path, inner_path,
! 								  extra->sjinfo, &extra->semifactors);
! 			/* Generate a nestloop path */
! 			result = (Path *) create_nestloop_path(root,
! 												   joinrel,
! 												   jointype,
! 												   &workspace,
! 												   extra->sjinfo,
! 												   &extra->semifactors,
! 												   outer_path,
! 												   inner_path,
! 												   extra->restrictlist,
! 												   NIL, NULL);
! 			break;
! 		case JOIN_FULL:
! 			/*
! 			 * (1) if either cheapest-total path is parameterized by the other
! 			 * rel, we can't generate a hashjoin/mergejoin path, and (2) if
! 			 * proposed path is still parameterized (ie, the required_outer
! 			 * set calculated by calc_non_nestloop_required_outer isn't NULL),
! 			 * we can't use it; which means that both the cheapest-total paths
! 			 * should be unparameterized.
! 			 */
! 			if (outer_path->param_info || inner_path->param_info)
! 				return NULL;
! 			/*
! 			 * Create a hashjoin path, if possible.  Else create a mergejoin
! 			 * path, if possible.
! 			 */
! 			if (extra->hashclauses)
! 			{
! 				/* Get initial estimates */
! 				initial_cost_hashjoin(root, &workspace, jointype,
! 									  extra->hashclauses,
! 									  outer_path, inner_path,
! 									  extra->sjinfo, &extra->semifactors);
! 				/* Generate a hashjoin path */
! 				result = (Path *) create_hashjoin_path(root,
! 													   joinrel,
! 													   jointype,
! 													   &workspace,
! 													   extra->sjinfo,
! 													   &extra->semifactors,
! 													   outer_path,
! 													   inner_path,
! 													   extra->restrictlist,
! 													   NULL,
! 													   extra->hashclauses);
! 			}
! 			else if (extra->mergejoin_allowed)
! 			{
! 				/*
! 				 * If special case: for "x FULL JOIN y ON true", there will be
! 				 * no join clauses at all; create a clauseless mergejoin path.
! 				 * Else create a mergejoin path by explicitly sorting both the
! 				 * outer and inner relations.
! 				 */
! 				if (!extra->mergeclause_list)
  				{
! 					/* Get initial estimates */
! 					initial_cost_mergejoin(root, &workspace, jointype, NIL,
! 										   outer_path, inner_path, NIL, NIL,
! 										   extra->sjinfo);
! 					/* Generate a mergejoin path */
! 					result = (Path *) create_mergejoin_path(root,
! 															joinrel,
! 															jointype,
! 															&workspace,
! 															extra->sjinfo,
! 															outer_path,
! 															inner_path,
! 															extra->restrictlist,
! 															NIL, NULL,
! 															NIL, NIL, NIL);
  				}
! 				else
  				{
! 					List	   *outerkeys = extra->outersortkeys;
! 					List	   *innerkeys = extra->innersortkeys;
  
! 					/*
! 					 * It's possible that the cheapest-total paths will
! 					 * already be sorted properly; if so, suppress an explicit
! 					 * sort.
! 					 */
! 					if (outerkeys &&
! 						pathkeys_contained_in(outerkeys, outer_path->pathkeys))
! 						outerkeys = NIL;
! 					if (innerkeys &&
! 						pathkeys_contained_in(innerkeys, inner_path->pathkeys))
! 						innerkeys = NIL;
! 					/* Get initial estimates */
! 					initial_cost_mergejoin(root, &workspace, jointype,
! 										   extra->mergeclauses,
! 										   outer_path, inner_path,
! 										   outerkeys, innerkeys,
! 										   extra->sjinfo);
! 					/* Generate a mergejoin path */
! 					result = (Path *) create_mergejoin_path(root,
! 															joinrel,
! 															jointype,
! 															&workspace,
! 															extra->sjinfo,
! 															outer_path,
! 															inner_path,
! 															extra->restrictlist,
! 															NIL,
! 															NULL,
! 															extra->mergeclauses,
! 															outerkeys,
! 															innerkeys);
  				}
! 			}
! 			else
! 				result = NULL;
! 			break;
! 		default:
! 			/* other values not expected here */
! 			elog(ERROR, "unrecognized join type: %d",
! 				 (int) jointype);
! 			result = NULL; /* keep compiler quiet */
! 			break;
  	}
! 
! 	return result;
  }
*** a/src/backend/optimizer/path/joinpath.c
--- b/src/backend/optimizer/path/joinpath.c
***************
*** 25,33 ****
  /* 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))
- 
  static void sort_inner_and_outer(PlannerInfo *root, RelOptInfo *joinrel,
  					 RelOptInfo *outerrel, RelOptInfo *innerrel,
  					 JoinType jointype, JoinPathExtraData *extra);
--- 25,30 ----
***************
*** 85,97 **** add_paths_to_joinrel(PlannerInfo *root,
  					 List *restrictlist)
  {
  	JoinPathExtraData extra;
- 	bool		mergejoin_allowed = true;
  	ListCell   *lc;
  
  	extra.restrictlist = restrictlist;
  	extra.mergeclause_list = NIL;
  	extra.sjinfo = sjinfo;
  	extra.param_source_rels = NULL;
  
  	/*
  	 * Find potential mergejoin clauses.  We can skip this if we are not
--- 82,98 ----
  					 List *restrictlist)
  {
  	JoinPathExtraData extra;
  	ListCell   *lc;
  
  	extra.restrictlist = restrictlist;
  	extra.mergeclause_list = NIL;
+ 	extra.mergejoin_allowed = true;
  	extra.sjinfo = sjinfo;
  	extra.param_source_rels = NULL;
+ 	extra.hashclauses = NIL;
+ 	extra.mergeclauses = NIL;
+ 	extra.outersortkeys = NIL;
+ 	extra.innersortkeys = NIL;
  
  	/*
  	 * Find potential mergejoin clauses.  We can skip this if we are not
***************
*** 106,112 **** add_paths_to_joinrel(PlannerInfo *root,
  														  innerrel,
  														  restrictlist,
  														  jointype,
! 														  &mergejoin_allowed);
  
  	/*
  	 * If it's SEMI or ANTI join, compute correction factors for cost
--- 107,113 ----
  														  innerrel,
  														  restrictlist,
  														  jointype,
! 													&extra.mergejoin_allowed);
  
  	/*
  	 * If it's SEMI or ANTI join, compute correction factors for cost
***************
*** 169,175 **** add_paths_to_joinrel(PlannerInfo *root,
  	 * 1. Consider mergejoin paths where both relations must be explicitly
  	 * sorted.  Skip this if we can't mergejoin.
  	 */
! 	if (mergejoin_allowed)
  		sort_inner_and_outer(root, joinrel, outerrel, innerrel,
  							 jointype, &extra);
  
--- 170,176 ----
  	 * 1. Consider mergejoin paths where both relations must be explicitly
  	 * sorted.  Skip this if we can't mergejoin.
  	 */
! 	if (extra.mergejoin_allowed)
  		sort_inner_and_outer(root, joinrel, outerrel, innerrel,
  							 jointype, &extra);
  
***************
*** 180,186 **** add_paths_to_joinrel(PlannerInfo *root,
  	 * (That's okay because we know that nestloop can't handle right/full
  	 * joins at all, so it wouldn't work in the prohibited cases either.)
  	 */
! 	if (mergejoin_allowed)
  		match_unsorted_outer(root, joinrel, outerrel, innerrel,
  							 jointype, &extra);
  
--- 181,187 ----
  	 * (That's okay because we know that nestloop can't handle right/full
  	 * joins at all, so it wouldn't work in the prohibited cases either.)
  	 */
! 	if (extra.mergejoin_allowed)
  		match_unsorted_outer(root, joinrel, outerrel, innerrel,
  							 jointype, &extra);
  
***************
*** 197,203 **** add_paths_to_joinrel(PlannerInfo *root,
  	 * those made by match_unsorted_outer when add_paths_to_joinrel() is
  	 * invoked with the two rels given in the other order.
  	 */
! 	if (mergejoin_allowed)
  		match_unsorted_inner(root, joinrel, outerrel, innerrel,
  							 jointype, &extra);
  #endif
--- 198,204 ----
  	 * those made by match_unsorted_outer when add_paths_to_joinrel() is
  	 * invoked with the two rels given in the other order.
  	 */
! 	if (extra.mergejoin_allowed)
  		match_unsorted_inner(root, joinrel, outerrel, innerrel,
  							 jointype, &extra);
  #endif
***************
*** 631,636 **** clause_sides_match_join(RestrictInfo *rinfo, RelOptInfo *outerrel,
--- 632,644 ----
   * 'innerrel' is the inner join relation
   * 'jointype' is the type of join to do
   * 'extra' contains additional input values
+  *
+  * This function returns the following info into fields of the
+  * JoinPathExtraData struct for possible use by the FDW:
+  *
+  *	mergeclauses	RestrictInfos to use as merge clauses in a mergejoin
+  *	outersortkeys	Sort pathkeys for the outer relation of the mergejoin
+  *	innersortkeys	Sort pathkeys for the inner relation of the mergejoin
   */
  static void
  sort_inner_and_outer(PlannerInfo *root,
***************
*** 773,778 **** sort_inner_and_outer(PlannerInfo *root,
--- 781,794 ----
  						   innerkeys,
  						   jointype,
  						   extra);
+ 
+ 		/* Save first mergejoin data for possible use by the FDW */
+ 		if (outerkeys == all_pathkeys)
+ 		{
+ 			extra->mergeclauses = cur_mergeclauses;
+ 			extra->outersortkeys = outerkeys;
+ 			extra->innersortkeys = innerkeys;
+ 		}
  	}
  }
  
***************
*** 1275,1280 **** consider_parallel_nestloop(PlannerInfo *root,
--- 1291,1301 ----
   * 'innerrel' is the inner join relation
   * 'jointype' is the type of join to do
   * 'extra' contains additional input values
+  *
+  * This function returns the following info into a field of the
+  * JoinPathExtraData struct for possible use by the FDW:
+  *
+  *	hashclauses		RestrictInfos to use as hash clauses in a hashjoin
   */
  static void
  hash_inner_and_outer(PlannerInfo *root,
***************
*** 1320,1325 **** hash_inner_and_outer(PlannerInfo *root,
--- 1341,1349 ----
  		hashclauses = lappend(hashclauses, restrictinfo);
  	}
  
+ 	/* Save hashclauses for possible use by the FDW */
+ 	extra->hashclauses = hashclauses;
+ 
  	/* If we found any usable hashclauses, make paths */
  	if (hashclauses)
  	{
*** a/src/include/foreign/fdwapi.h
--- b/src/include/foreign/fdwapi.h
***************
*** 235,240 **** extern FdwRoutine *GetFdwRoutineByRelId(Oid relid);
  extern FdwRoutine *GetFdwRoutineForRelation(Relation relation, bool makecopy);
  extern bool IsImportableForeignTable(const char *tablename,
  						 ImportForeignSchemaStmt *stmt);
! extern Path *GetExistingLocalJoinPath(RelOptInfo *joinrel);
  
  #endif   /* FDWAPI_H */
--- 235,242 ----
  extern FdwRoutine *GetFdwRoutineForRelation(Relation relation, bool makecopy);
  extern bool IsImportableForeignTable(const char *tablename,
  						 ImportForeignSchemaStmt *stmt);
! extern Path *CreateLocalJoinPath(PlannerInfo *root, RelOptInfo *joinrel,
! 					RelOptInfo *outerrel, RelOptInfo *innerrel,
! 					JoinType jointype, JoinPathExtraData *extra);
  
  #endif   /* FDWAPI_H */
*** a/src/include/nodes/relation.h
--- b/src/include/nodes/relation.h
***************
*** 904,909 **** typedef struct Path
--- 904,913 ----
  #define PATH_REQ_OUTER(path)  \
  	((path)->param_info ? (path)->param_info->ppi_req_outer : (Relids) NULL)
  
+ /* Macro for determining whether path is parameterized by rel */
+ #define PATH_PARAM_BY_REL(path, rel)  \
+ 	((path)->param_info && bms_overlap(PATH_REQ_OUTER(path), (rel)->relids))
+ 
  /*----------
   * IndexPath represents an index scan over a single index.
   *
***************
*** 2006,2022 **** typedef struct SemiAntiJoinFactors
--- 2010,2039 ----
   *		clauses that apply to this join
   * mergeclause_list is a list of RestrictInfo nodes for available
   *		mergejoin clauses in this join
+  * mergejoin_allowed is a flag to indicate whether mergejoins are allowed
   * sjinfo is extra info about special joins for selectivity estimation
   * semifactors is as shown above (only valid for SEMI or ANTI joins)
   * param_source_rels are OK targets for parameterization of result paths
+  *
+  * Remaining fields are set only for possible use by the FDW:
+  *
+  * hashclauses are the RestrictInfos to use as hash clauses in a hashjoin
+  * mergeclauses are the RestrictInfos to use as merge clauses in a mergejoin
+  * outersortkeys are the sort pathkeys for the outer side of the mergejoin
+  * innersortkeys are the sort pathkeys for the inner side of the mergejoin
   */
  typedef struct JoinPathExtraData
  {
  	List	   *restrictlist;
  	List	   *mergeclause_list;
+ 	bool		mergejoin_allowed;
  	SpecialJoinInfo *sjinfo;
  	SemiAntiJoinFactors semifactors;
  	Relids		param_source_rels;
+ 	List	   *hashclauses;
+ 	List	   *mergeclauses;
+ 	List	   *outersortkeys;
+ 	List	   *innersortkeys;
  } JoinPathExtraData;
  
  /*
-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

Reply via email to