Use CP_SMALL_TLIST for hash aggregate

Commit 1f39bce021 added disk-based hash aggregation, which may spill
incoming tuples to disk. It however did not request projection to make
the tuples as narrow as possible, which may mean having to spill much
more data than necessary (increasing I/O, pushing other stuff from page
cache, etc.).

This adds CP_SMALL_TLIST in places that may use hash aggregation - we do
that only for AGG_HASHED. It's unnecessary for AGG_SORTED, because that
either uses explicit Sort (which already does projection) or pre-sorted
input (which does not need spilling to disk).

Author: Tomas Vondra
Reviewed-by: Jeff Davis
Discussion: https://postgr.es/m/20200519151202.u2p2gpiawoaznsv2%40development

diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out
index 90db550b92140ad9216b1f662bd288e699975574..82fc1290ef269042a69dc16fe72828c55e2599e0 100644 (file)
--- a/contrib/postgres_fdw/expected/postgres_fdw.out
+++ b/contrib/postgres_fdw/expected/postgres_fdw.out
@@ -2715,7 +2715,7 @@ select sum(c1) from ft1 group by c2 having avg(c1 * (random() <= 1)::int) > 100
          Group Key: ft1.c2
          Filter: (avg((ft1.c1 * ((random() <= '1'::double precision))::integer)) > '100'::numeric)
          ->  Foreign Scan on public.ft1
-               Output: c1, c2
+               Output: c2, c1
                Remote SQL: SELECT "C 1", c2 FROM "S 1"."T 1"
 (10 rows)
 
@@ -2964,7 +2964,7 @@ select sum(c1) filter (where (c1 / c1) * random() <= 1) from ft1 group by c2 ord
          Output: sum(c1) FILTER (WHERE ((((c1 / c1))::double precision * random()) <= '1'::double precision)), c2
          Group Key: ft1.c2
          ->  Foreign Scan on public.ft1
-               Output: c1, c2
+               Output: c2, c1
                Remote SQL: SELECT "C 1", c2 FROM "S 1"."T 1"
 (9 rows)
 

diff --git a/src/backend/optimizer/plan/createplan.c b/src/backend/optimizer/plan/createplan.c
index 9941dfe65e4696c3fea3b88476de94f65e3b4d4f..744eed187d410e9cc0a6c5220a045a6888c973b5 100644 (file)
--- a/src/backend/optimizer/plan/createplan.c
+++ b/src/backend/optimizer/plan/createplan.c
@@ -2113,12 +2113,22 @@ create_agg_plan(PlannerInfo *root, AggPath *best_path)
    Plan       *subplan;
    List       *tlist;
    List       *quals;
+   int         flags;
 
    /*
     * Agg can project, so no need to be terribly picky about child tlist, but
-    * we do need grouping columns to be available
+    * we do need grouping columns to be available. We are a bit more careful
+    * with hash aggregate, where we explicitly request small tlist to minimize
+    * I/O needed for spilling (we can't be sure spilling won't be necessary,
+    * so we just do it every time).
     */
-   subplan = create_plan_recurse(root, best_path->subpath, CP_LABEL_TLIST);
+   flags = CP_LABEL_TLIST;
+
+   /* ensure small tlist for hash aggregate */
+   if (best_path->aggstrategy == AGG_HASHED)
+       flags |= CP_SMALL_TLIST;
+
+   subplan = create_plan_recurse(root, best_path->subpath, flags);
 
    tlist = build_path_tlist(root, &best_path->path);
 
@@ -2200,6 +2210,7 @@ create_groupingsets_plan(PlannerInfo *root, GroupingSetsPath *best_path)
    int         maxref;
    List       *chain;
    ListCell   *lc;
+   int         flags;
 
    /* Shouldn't get here without grouping sets */
    Assert(root->parse->groupingSets);
@@ -2207,9 +2218,18 @@ create_groupingsets_plan(PlannerInfo *root, GroupingSetsPath *best_path)
 
    /*
     * Agg can project, so no need to be terribly picky about child tlist, but
-    * we do need grouping columns to be available
+    * we do need grouping columns to be available. We are a bit more careful
+    * with hash aggregate, where we explicitly request small tlist to minimize
+    * I/O needed for spilling (we can't be sure spilling won't be necessary,
+    * so we just do it every time).
     */
-   subplan = create_plan_recurse(root, best_path->subpath, CP_LABEL_TLIST);
+   flags = CP_LABEL_TLIST;
+
+   /* ensure small tlist for hash aggregate */
+   if (best_path->aggstrategy == AGG_HASHED)
+       flags |= CP_SMALL_TLIST;
+
+   subplan = create_plan_recurse(root, best_path->subpath, flags);
 
    /*
     * Compute the mapping from tleSortGroupRef to column index in the child's