properly redistribute WindowAgg input, fix plan change in window.out

We can't push down the whole WindowAgg (at least now for now), so
we need to redistribute the input properly. Luckily thanks to the
parallel aggregates, we get plans like this:

                               QUERY PLAN
    -----------------------------------------------------------------
     WindowAgg
       ->  Finalize GroupAggregate
             Group Key: (tenk1.ten + tenk1.four)
             ->  Remote Subquery Scan on all (datanode_1,datanode_2)
                   ->  Sort
                         Sort Key: ((tenk1.ten + tenk1.four))
                         ->  Partial HashAggregate
                               Group Key: (tenk1.ten + tenk1.four)
                               ->  Seq Scan on tenk1
    (9 rows)

That however means mild breakage in window.out, due to the partial
aggregates etc.

diff --git a/src/backend/optimizer/plan/planner.c b/src/backend/optimizer/plan/planner.c
index 4ea83f0b0d3dec59efdee0de1313fe87f93072e8..fffcddd0ff0a1447fb3e7ce53bcfd8f55a354100 100644 (file)
--- a/src/backend/optimizer/plan/planner.c
+++ b/src/backend/optimizer/plan/planner.c
@@ -167,7 +167,7 @@ static bool groupingsets_distribution_match(PlannerInfo *root, Query *parse,
 static Path *adjust_path_distribution(PlannerInfo *root, Query *parse,
                                          Path *path);
 static bool can_push_down_grouping(PlannerInfo *root, Query *parse, Path *path);
-
+static bool can_push_down_window(PlannerInfo *root, Path *path);
 
 /*****************************************************************************
  *
@@ -4820,6 +4820,10 @@ create_one_window_path(PlannerInfo *root,
                        window_target = output_target;
                }
 
+               /* We can't really push down window functions for now. */
+               if (!can_push_down_window(root, path))
+                       path = create_remotesubplan_path(root, path, NULL);
+
                path = (Path *)
                        create_windowagg_path(root, window_rel, path, window_target,
                                                                  wflists->windowFuncs[wc->winref],
@@ -6340,3 +6344,13 @@ can_push_down_grouping(PlannerInfo *root, Query *parse, Path *path)
 
        return grouping_distribution_match(root, parse, path, parse->groupClause);
 }
+
+static bool
+can_push_down_window(PlannerInfo *root, Path *path)
+{
+       /*  */
+       if (! path->distribution)
+               return true;
+
+       return false;
+}

diff --git a/src/backend/optimizer/util/pathnode.c b/src/backend/optimizer/util/pathnode.c
index 7bd784078685b286c6f337e5c1d9addfc092ca7a..3c7553051db8271ec3c9e383593252d154cc9a15 100644 (file)
--- a/src/backend/optimizer/util/pathnode.c
+++ b/src/backend/optimizer/util/pathnode.c
@@ -4189,6 +4189,8 @@ create_windowagg_path(PlannerInfo *root,
        pathnode->winclause = winclause;
        pathnode->winpathkeys = winpathkeys;
 
+       pathnode->path.distribution = copyObject(subpath->distribution);
+
        /*
         * For costing purposes, assume that there are no redundant partitioning
         * or ordering columns; it's not worth the trouble to deal with that

diff --git a/src/test/regress/expected/window.out b/src/test/regress/expected/window.out
index a844fb357bd19b4ba5c8508da7d9415c4665e6c7..2d92a32e4fe541cd5147ecc0f5517066b505f958 100644 (file)
--- a/src/test/regress/expected/window.out
+++ b/src/test/regress/expected/window.out
@@ -598,13 +598,15 @@ select first_value(max(x)) over (), y
                            QUERY PLAN                            
 -----------------------------------------------------------------
  WindowAgg
-   ->  HashAggregate
+   ->  Finalize GroupAggregate
          Group Key: (tenk1.ten + tenk1.four)
          ->  Remote Subquery Scan on all (datanode_1,datanode_2)
-               ->  HashAggregate
-                     Group Key: (tenk1.ten + tenk1.four)
-                     ->  Seq Scan on tenk1
-(7 rows)
+               ->  Sort
+                     Sort Key: ((tenk1.ten + tenk1.four))
+                     ->  Partial HashAggregate
+                           Group Key: (tenk1.ten + tenk1.four)
+                           ->  Seq Scan on tenk1
+(9 rows)
 
 -- test non-default frame specifications
 SELECT four, ten,