Fix another oversight in checking if a join with LATERAL refs is legal.

It was possible for the planner to decide to join a LATERAL subquery to
the outer side of an outer join before the outer join itself is completed.
Normally that's fine because of the associativity rules, but it doesn't
work if the subquery contains a lateral reference to the inner side of the
outer join.  In such a situation the outer join *must* be done first.
join_is_legal() missed this consideration and would allow the join to be
attempted, but the actual path-building code correctly decided that no
valid join path could be made, sometimes leading to planner errors such as
"failed to build any N-way joins".

Per report from Andreas Seltenreich.  Back-patch to 9.3 where LATERAL
support was added.

diff --git a/src/backend/optimizer/path/joinrels.c b/src/backend/optimizer/path/joinrels.c
index 125b33f48e7c66b331b54b54c2a9258a4bdf6042..7e2816aaec6974b55f941d6e6eb2dd85b65c41d4 100644 (file)
--- a/src/backend/optimizer/path/joinrels.c
+++ b/src/backend/optimizer/path/joinrels.c
@@ -334,6 +334,7 @@ join_is_legal(PlannerInfo *root, RelOptInfo *rel1, RelOptInfo *rel2,
    bool        must_be_leftjoin;
    bool        lateral_fwd;
    bool        lateral_rev;
+   Relids      join_lateral_rels;
    ListCell   *l;
 
    /*
@@ -569,6 +570,35 @@ join_is_legal(PlannerInfo *root, RelOptInfo *rel1, RelOptInfo *rel2,
        }
    }
 
+   /*
+    * LATERAL references could also cause problems later on if we accept this
+    * join: if the join's minimum parameterization includes any rels that
+    * would have to be on the inside of an outer join with this join rel,
+    * then it's never going to be possible to build the complete query using
+    * this join.  We should reject this join not only because it'll save
+    * work, but because if we don't, the clauseless-join heuristics might
+    * think that legality of this join means that some other join rel need
+    * not be formed, and that could lead to failure to find any plan at all.
+    * It seems best not to merge this check into the main loop above, because
+    * it is concerned with SJs that are not otherwise relevant to this join.
+    */
+   join_lateral_rels = min_join_parameterization(root, joinrelids);
+   if (join_lateral_rels)
+   {
+       foreach(l, root->join_info_list)
+       {
+           SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(l);
+
+           if (bms_overlap(sjinfo->min_righthand, join_lateral_rels) &&
+               bms_overlap(sjinfo->min_lefthand, joinrelids))
+               return false;   /* will not be able to join to min_righthand */
+           if (sjinfo->jointype == JOIN_FULL &&
+               bms_overlap(sjinfo->min_lefthand, join_lateral_rels) &&
+               bms_overlap(sjinfo->min_righthand, joinrelids))
+               return false;   /* will not be able to join to min_lefthand */
+       }
+   }
+
    /* Otherwise, it's a valid join */
    *sjinfo_p = match_sjinfo;
    *reversed_p = reversed;

diff --git a/src/backend/optimizer/util/relnode.c b/src/backend/optimizer/util/relnode.c
index 4d80cfbe0e1793e6f17540e1bbd731e0577e2899..3fc7e5cf67a2a3a92f8e13dbb5ac5279d9bdf5d1 100644 (file)
--- a/src/backend/optimizer/util/relnode.c
+++ b/src/backend/optimizer/util/relnode.c
@@ -465,6 +465,45 @@ build_join_rel(PlannerInfo *root,
    return joinrel;
 }
 
+/*
+ * min_join_parameterization
+ *
+ * Determine the minimum possible parameterization of a joinrel, that is, the
+ * set of other rels it contains LATERAL references to.
+ */
+Relids
+min_join_parameterization(PlannerInfo *root, Relids joinrelids)
+{
+   Relids      result;
+   ListCell   *lc;
+
+   /* Easy if there are no lateral references */
+   if (root->lateral_info_list == NIL)
+       return NULL;
+
+   /*
+    * Scan lateral_info_list to find all the lateral references occurring in
+    * or below this join.
+    */
+   result = NULL;
+   foreach(lc, root->lateral_info_list)
+   {
+       LateralJoinInfo *ljinfo = (LateralJoinInfo *) lfirst(lc);
+
+       if (bms_is_subset(ljinfo->lateral_rhs, joinrelids))
+           result = bms_add_members(result, ljinfo->lateral_lhs);
+   }
+
+   /* Remove any rels that are already included in the join */
+   result = bms_del_members(result, joinrelids);
+
+   /* Maintain invariant that result is exactly NULL if empty */
+   if (bms_is_empty(result))
+       result = NULL;
+
+   return result;
+}
+
 /*
  * build_joinrel_tlist
  *   Builds a join relation's target list from an input relation.

diff --git a/src/include/optimizer/pathnode.h b/src/include/optimizer/pathnode.h
index 4fa990fb2c531f2e6d7fac8f874fc40b9b5e0b0b..47c74ee6d2db2c4f2fc02e48bc65300b7d888585 100644 (file)
--- a/src/include/optimizer/pathnode.h
+++ b/src/include/optimizer/pathnode.h
@@ -142,6 +142,7 @@ extern RelOptInfo *build_join_rel(PlannerInfo *root,
               RelOptInfo *inner_rel,
               SpecialJoinInfo *sjinfo,
               List **restrictlist_ptr);
+extern Relids min_join_parameterization(PlannerInfo *root, Relids joinrelids);
 extern AppendRelInfo *find_childrel_appendrelinfo(PlannerInfo *root,
                            RelOptInfo *rel);
 extern RelOptInfo *find_childrel_top_parent(PlannerInfo *root, RelOptInfo *rel);

diff --git a/src/test/regress/expected/join.out b/src/test/regress/expected/join.out
index 82699c3a5a193b25ecc16d762363ce72833d28d4..6105a688f41d4f282b63e168df32d7225b1cff27 100644 (file)
--- a/src/test/regress/expected/join.out
+++ b/src/test/regress/expected/join.out
@@ -3465,6 +3465,47 @@ select * from
  doh! | 123 | 456 | hi de ho neighbor |   
 (2 rows)
 
+--
+-- test for appropriate join order in the presence of lateral references
+--
+explain (verbose, costs off)
+select * from
+  text_tbl t1
+  left join int8_tbl i8
+  on i8.q2 = 123,
+  lateral (select i8.q1, t2.f1 from text_tbl t2 limit 1) as ss
+where t1.f1 = ss.f1;
+                    QUERY PLAN                    
+--------------------------------------------------
+ Nested Loop
+   Output: t1.f1, i8.q1, i8.q2, (i8.q1), t2.f1
+   Join Filter: (t1.f1 = t2.f1)
+   ->  Nested Loop Left Join
+         Output: t1.f1, i8.q1, i8.q2
+         ->  Seq Scan on public.text_tbl t1
+               Output: t1.f1
+         ->  Materialize
+               Output: i8.q1, i8.q2
+               ->  Seq Scan on public.int8_tbl i8
+                     Output: i8.q1, i8.q2
+                     Filter: (i8.q2 = 123)
+   ->  Limit
+         Output: (i8.q1), t2.f1
+         ->  Seq Scan on public.text_tbl t2
+               Output: i8.q1, t2.f1
+(16 rows)
+
+select * from
+  text_tbl t1
+  left join int8_tbl i8
+  on i8.q2 = 123,
+  lateral (select i8.q1, t2.f1 from text_tbl t2 limit 1) as ss
+where t1.f1 = ss.f1;
+  f1  |        q1        | q2  |        q1        |  f1  
+------+------------------+-----+------------------+------
+ doh! | 4567890123456789 | 123 | 4567890123456789 | doh!
+(1 row)
+
 --
 -- test ability to push constants through outer join clauses
 --

diff --git a/src/test/regress/sql/join.sql b/src/test/regress/sql/join.sql
index d04bad720507262c55a594491b4782598e9c43e3..85245e7f254afd8c7611fb84d14a3f378b959646 100644 (file)
--- a/src/test/regress/sql/join.sql
+++ b/src/test/regress/sql/join.sql
@@ -1083,6 +1083,25 @@ select * from
   left join int4_tbl i4
   on i8.q1 = i4.f1;
 
+--
+-- test for appropriate join order in the presence of lateral references
+--
+
+explain (verbose, costs off)
+select * from
+  text_tbl t1
+  left join int8_tbl i8
+  on i8.q2 = 123,
+  lateral (select i8.q1, t2.f1 from text_tbl t2 limit 1) as ss
+where t1.f1 = ss.f1;
+
+select * from
+  text_tbl t1
+  left join int8_tbl i8
+  on i8.q2 = 123,
+  lateral (select i8.q1, t2.f1 from text_tbl t2 limit 1) as ss
+where t1.f1 = ss.f1;
+
 --
 -- test ability to push constants through outer join clauses
 --