Fix rescan logic in nodeCtescan.

The previous coding essentially assumed that nodes would be rescanned in
the same order they were initialized in; or at least that the "leader" of
a group of CTEscans would be rescanned before any others were required to
execute.  Unfortunately, that isn't even a little bit true.  It's possible
to devise queries in which the leader isn't rescanned until other CTEscans
on the same CTE have run to completion, or even in which the leader never
gets a rescan call at all.

The fix makes the leader specially responsible only for initial creation
and final destruction of the tuplestore; rescan resets are now a
symmetrically shared responsibility.  This means that we might reset the
tuplestore multiple times when restarting a plan subtree containing
multiple CTEscans; but resetting an already-empty tuplestore is cheap
enough that that doesn't seem like a problem.

Per report from Adam Mackler; the new regression test cases are based on
his example query.

Back-patch to 8.4 where CTE scans were introduced.

diff --git a/src/backend/executor/nodeCtescan.c b/src/backend/executor/nodeCtescan.c
index 7ab15ac63e1020dd00b3ef27950615dc108b795b..7e10b8a530deb2850938c254e80eae9fe09d4cd9 100644 (file)
--- a/src/backend/executor/nodeCtescan.c
+++ b/src/backend/executor/nodeCtescan.c
@@ -205,7 +205,7 @@ ExecInitCteScan(CteScan *node, EState *estate, int eflags)
         * The Param slot associated with the CTE query is used to hold a pointer
         * to the CteState of the first CteScan node that initializes for this
         * CTE.  This node will be the one that holds the shared state for all the
-        * CTEs.
+        * CTEs, particularly the shared tuplestore.
         */
        prmdata = &(estate->es_param_exec_vals[node->cteParam]);
        Assert(prmdata->execPlan == NULL);
@@ -294,7 +294,10 @@ ExecEndCteScan(CteScanState *node)
         * If I am the leader, free the tuplestore.
         */
        if (node->leader == node)
+       {
                tuplestore_end(node->cte_table);
+               node->cte_table = NULL;
+       }
 }
 
 /* ----------------------------------------------------------------
@@ -312,26 +315,26 @@ ExecReScanCteScan(CteScanState *node)
 
        ExecScanReScan(&node->ss);
 
-       if (node->leader == node)
+       /*
+        * Clear the tuplestore if a new scan of the underlying CTE is required.
+        * This implicitly resets all the tuplestore's read pointers.  Note that
+        * multiple CTE nodes might redundantly clear the tuplestore; that's OK,
+        * and not unduly expensive.  We'll stop taking this path as soon as
+        * somebody has attempted to read something from the underlying CTE
+        * (thereby causing its chgParam to be cleared).
+        */
+       if (node->leader->cteplanstate->chgParam != NULL)
        {
-               /*
-                * The leader is responsible for clearing the tuplestore if a new scan
-                * of the underlying CTE is required.
-                */
-               if (node->cteplanstate->chgParam != NULL)
-               {
-                       tuplestore_clear(tuplestorestate);
-                       node->eof_cte = false;
-               }
-               else
-               {
-                       tuplestore_select_read_pointer(tuplestorestate, node->readptr);
-                       tuplestore_rescan(tuplestorestate);
-               }
+               tuplestore_clear(tuplestorestate);
+               node->leader->eof_cte = false;
        }
        else
        {
-               /* Not leader, so just rewind my own pointer */
+               /*
+                * Else, just rewind my own pointer.  Either the underlying CTE
+                * doesn't need a rescan (and we can re-read what's in the tuplestore
+                * now), or somebody else already took care of it.
+                */
                tuplestore_select_read_pointer(tuplestorestate, node->readptr);
                tuplestore_rescan(tuplestorestate);
        }

diff --git a/src/test/regress/expected/with.out b/src/test/regress/expected/with.out
index a491b2ca61f82bee3978a1b2bf262bc646fe91bd..a7491bf6b93038b8b23368516d0681a563fa9831 100644 (file)
--- a/src/test/regress/expected/with.out
+++ b/src/test/regress/expected/with.out
@@ -1209,6 +1209,81 @@ SELECT * FROM outermost;
 ERROR:  recursive reference to query "outermost" must not appear within a subquery
 LINE 2:   WITH innermost as (SELECT 2 FROM outermost) 
                                            ^
+--
+-- Test CTEs read in non-initialization orders
+--
+WITH RECURSIVE
+  tab(id_key,link) AS (VALUES (1,17), (2,17), (3,17), (4,17), (6,17), (5,17)),
+  iter (id_key, row_type, link) AS (
+      SELECT 0, 'base', 17
+    UNION ALL (
+      WITH remaining(id_key, row_type, link, min) AS (
+        SELECT tab.id_key, 'true'::text, iter.link, MIN(tab.id_key) OVER ()
+        FROM tab INNER JOIN iter USING (link)
+        WHERE tab.id_key > iter.id_key
+      ),
+      first_remaining AS (
+        SELECT id_key, row_type, link
+        FROM remaining
+        WHERE id_key=min
+      ),
+      effect AS (
+        SELECT tab.id_key, 'new'::text, tab.link
+        FROM first_remaining e INNER JOIN tab ON e.id_key=tab.id_key
+        WHERE e.row_type = 'false'
+      )
+      SELECT * FROM first_remaining
+      UNION ALL SELECT * FROM effect
+    )
+  )
+SELECT * FROM iter;
+ id_key | row_type | link 
+--------+----------+------
+      0 | base     |   17
+      1 | true     |   17
+      2 | true     |   17
+      3 | true     |   17
+      4 | true     |   17
+      5 | true     |   17
+      6 | true     |   17
+(7 rows)
+
+WITH RECURSIVE
+  tab(id_key,link) AS (VALUES (1,17), (2,17), (3,17), (4,17), (6,17), (5,17)),
+  iter (id_key, row_type, link) AS (
+      SELECT 0, 'base', 17
+    UNION (
+      WITH remaining(id_key, row_type, link, min) AS (
+        SELECT tab.id_key, 'true'::text, iter.link, MIN(tab.id_key) OVER ()
+        FROM tab INNER JOIN iter USING (link)
+        WHERE tab.id_key > iter.id_key
+      ),
+      first_remaining AS (
+        SELECT id_key, row_type, link
+        FROM remaining
+        WHERE id_key=min
+      ),
+      effect AS (
+        SELECT tab.id_key, 'new'::text, tab.link
+        FROM first_remaining e INNER JOIN tab ON e.id_key=tab.id_key
+        WHERE e.row_type = 'false'
+      )
+      SELECT * FROM first_remaining
+      UNION ALL SELECT * FROM effect
+    )
+  )
+SELECT * FROM iter;
+ id_key | row_type | link 
+--------+----------+------
+      0 | base     |   17
+      1 | true     |   17
+      2 | true     |   17
+      3 | true     |   17
+      4 | true     |   17
+      5 | true     |   17
+      6 | true     |   17
+(7 rows)
+
 --
 -- Data-modifying statements in WITH
 --

diff --git a/src/test/regress/sql/with.sql b/src/test/regress/sql/with.sql
index 22fdddc4eeafa2de36c1eaf57b20ec0d44d27df3..684c8f13db1a692c78ba4b87355a03377bec83c0 100644 (file)
--- a/src/test/regress/sql/with.sql
+++ b/src/test/regress/sql/with.sql
@@ -574,6 +574,62 @@ WITH RECURSIVE outermost(x) AS (
 )
 SELECT * FROM outermost;
 
+--
+-- Test CTEs read in non-initialization orders
+--
+
+WITH RECURSIVE
+  tab(id_key,link) AS (VALUES (1,17), (2,17), (3,17), (4,17), (6,17), (5,17)),
+  iter (id_key, row_type, link) AS (
+      SELECT 0, 'base', 17
+    UNION ALL (
+      WITH remaining(id_key, row_type, link, min) AS (
+        SELECT tab.id_key, 'true'::text, iter.link, MIN(tab.id_key) OVER ()
+        FROM tab INNER JOIN iter USING (link)
+        WHERE tab.id_key > iter.id_key
+      ),
+      first_remaining AS (
+        SELECT id_key, row_type, link
+        FROM remaining
+        WHERE id_key=min
+      ),
+      effect AS (
+        SELECT tab.id_key, 'new'::text, tab.link
+        FROM first_remaining e INNER JOIN tab ON e.id_key=tab.id_key
+        WHERE e.row_type = 'false'
+      )
+      SELECT * FROM first_remaining
+      UNION ALL SELECT * FROM effect
+    )
+  )
+SELECT * FROM iter;
+
+WITH RECURSIVE
+  tab(id_key,link) AS (VALUES (1,17), (2,17), (3,17), (4,17), (6,17), (5,17)),
+  iter (id_key, row_type, link) AS (
+      SELECT 0, 'base', 17
+    UNION (
+      WITH remaining(id_key, row_type, link, min) AS (
+        SELECT tab.id_key, 'true'::text, iter.link, MIN(tab.id_key) OVER ()
+        FROM tab INNER JOIN iter USING (link)
+        WHERE tab.id_key > iter.id_key
+      ),
+      first_remaining AS (
+        SELECT id_key, row_type, link
+        FROM remaining
+        WHERE id_key=min
+      ),
+      effect AS (
+        SELECT tab.id_key, 'new'::text, tab.link
+        FROM first_remaining e INNER JOIN tab ON e.id_key=tab.id_key
+        WHERE e.row_type = 'false'
+      )
+      SELECT * FROM first_remaining
+      UNION ALL SELECT * FROM effect
+    )
+  )
+SELECT * FROM iter;
+
 --
 -- Data-modifying statements in WITH
 --