In a non-hashed Agg node, reset the "aggcontext" at group boundaries, instead

of individually pfree'ing pass-by-reference transition values.  This should
be at least as fast as the prior coding, and it has the major advantage of
clearing out any working data an aggregate function may have stored in or
underneath the aggcontext.  This avoids memory leakage when an aggregate
such as array_agg() is used in GROUP BY mode.  Per report from Chris Spotts.

Back-patch to 8.4.  In principle the problem could arise in prior versions,
but since they didn't have array_agg the issue seems not critical.

diff --git a/src/backend/executor/nodeAgg.c b/src/backend/executor/nodeAgg.c
index cf0c242245257a1c98c80d8c06a0f6fec8d64782..d7cccc534f15618681ac99eb5eb62de4cfd8248a 100644 (file)
--- a/src/backend/executor/nodeAgg.c
+++ b/src/backend/executor/nodeAgg.c
@@ -55,6 +55,8 @@
  *       in either case its value need not be preserved.  See int8inc() for an
  *       example.      Notice that advance_transition_function() is coded to avoid a
  *       data copy step when the previous transition value pointer is returned.
+ *       Also, some transition functions make use of the aggcontext to store
+ *       working state.
  *
  *
  * Portions Copyright (c) 1996-2009, PostgreSQL Global Development Group
@@ -272,18 +274,6 @@ initialize_aggregates(AggState *aggstate,
                                                                          work_mem, false);
                }
 
-               /*
-                * If we are reinitializing after a group boundary, we have to free
-                * any prior transValue to avoid memory leakage.  We must check not
-                * only the isnull flag but whether the pointer is NULL; since
-                * pergroupstate is initialized with palloc0, the initial condition
-                * has isnull = 0 and null pointer.
-                */
-               if (!peraggstate->transtypeByVal &&
-                       !pergroupstate->transValueIsNull &&
-                       DatumGetPointer(pergroupstate->transValue) != NULL)
-                       pfree(DatumGetPointer(pergroupstate->transValue));
-
                /*
                 * (Re)set transValue to the initial value.
                 *
@@ -911,10 +901,15 @@ agg_retrieve_direct(AggState *aggstate)
                }
 
                /*
-                * Clear the per-output-tuple context for each group
+                * Clear the per-output-tuple context for each group, as well as
+                * aggcontext (which contains any pass-by-ref transvalues of the
+                * old group).  We also clear any child contexts of the aggcontext;
+                * some aggregate functions store working state in such contexts.
                 */
                ResetExprContext(econtext);
 
+               MemoryContextResetAndDeleteChildren(aggstate->aggcontext);
+
                /*
                 * Initialize working state for a new input tuple group
                 */
@@ -1234,7 +1229,8 @@ ExecInitAgg(Agg *node, EState *estate, int eflags)
         * structures and transition values.  NOTE: the details of what is stored
         * in aggcontext and what is stored in the regular per-query memory
         * context are driven by a simple decision: we want to reset the
-        * aggcontext in ExecReScanAgg to recover no-longer-wanted space.
+        * aggcontext at group boundaries (if not hashing) and in ExecReScanAgg
+        * to recover no-longer-wanted space.
         */
        aggstate->aggcontext =
                AllocSetContextCreate(CurrentMemoryContext,

diff --git a/src/backend/utils/adt/array_userfuncs.c b/src/backend/utils/adt/array_userfuncs.c
index aa219f585bfb9d959195f3f839e4f3b1bb407533..c7b383d150a37201a72d1061f7446d1e3912fb3e 100644 (file)
--- a/src/backend/utils/adt/array_userfuncs.c
+++ b/src/backend/utils/adt/array_userfuncs.c
@@ -539,7 +539,9 @@ array_agg_finalfn(PG_FUNCTION_ARGS)
 
        /*
         * Make the result.  We cannot release the ArrayBuildState because
-        * sometimes aggregate final functions are re-executed.
+        * sometimes aggregate final functions are re-executed.  Rather, it
+        * is nodeAgg.c's responsibility to reset the aggcontext when it's
+        * safe to do so.
         */
        result = makeMdArrayResult(state, 1, dims, lbs,
                                                           CurrentMemoryContext,