Cache by-reference missing values in a long lived context

Attribute missing values might be needed past the lifetime of the tuple
descriptors from which they are extracted. To avoid possibly using
pointers for by-reference values which might thus be left dangling, we
cache a datumCopy'd version of the datum in the TopMemoryContext. Since
we first search for the value this only needs to be done once per
session for any such value.

Original complaint from Tom Lane, idea for mitigation by Andrew Dunstan,
tweaked by Tom Lane.

Backpatch to version 11 where missing values were introduced.

Discussion: https://postgr.es/m/1306569.1687978174@sss.pgh.pa.us

diff --git a/src/backend/access/common/heaptuple.c b/src/backend/access/common/heaptuple.c
index 1d03b7dd136e516fcca787246c6de913e7c411fe..1cf87386344210f2795bafcfbee29567363a4dd3 100644 (file)
--- a/src/backend/access/common/heaptuple.c
+++ b/src/backend/access/common/heaptuple.c
@@ -61,7 +61,11 @@
 #include "access/tupdesc_details.h"
 #include "access/tuptoaster.h"
 #include "executor/tuptable.h"
+#include "utils/datum.h"
 #include "utils/expandeddatum.h"
+#include "utils/hashutils.h"
+#include "utils/hsearch.h"
+#include "utils/memutils.h"
 
 
 /* Does att's datatype allow packing into the 1-byte-header varlena format? */
@@ -71,6 +75,58 @@
 #define VARLENA_ATT_IS_PACKABLE(att) \
        ((att)->attstorage != 'p')
 
+/*
+ * Setup for cacheing pass-by-ref missing attributes in a way that survives
+ * tupleDesc destruction.
+ */
+
+typedef struct
+{
+       int                     len;
+       Datum           value;
+} missing_cache_key;
+
+static HTAB *missing_cache = NULL;
+
+static uint32
+missing_hash(const void *key, Size keysize)
+{
+       const missing_cache_key *entry = (missing_cache_key *) key;
+
+       return DatumGetUInt32(
+               hash_any((const unsigned char *) entry->value, entry->len));
+}
+
+static int
+missing_match(const void *key1, const void *key2, Size keysize)
+{
+       const missing_cache_key *entry1 = (missing_cache_key *) key1;
+       const missing_cache_key *entry2 = (missing_cache_key *) key2;
+
+       if (entry1->len != entry2->len)
+               return entry1->len > entry2->len ? 1 : -1;
+
+       return memcmp(DatumGetPointer(entry1->value),
+                                 DatumGetPointer(entry2->value),
+                                 entry1->len);
+}
+
+static void
+init_missing_cache()
+{
+       HASHCTL         hash_ctl;
+
+       hash_ctl.keysize = sizeof(missing_cache_key);
+       hash_ctl.entrysize = sizeof(missing_cache_key);
+       hash_ctl.hcxt = TopMemoryContext;
+       hash_ctl.hash = missing_hash;
+       hash_ctl.match = missing_match;
+       missing_cache =
+               hash_create("Missing Values Cache",
+                                       32,
+                                       &hash_ctl,
+                                       HASH_ELEM | HASH_CONTEXT | HASH_FUNCTION | HASH_COMPARE);
+}
 
 /* ----------------------------------------------------------------
  *                                             misc support routines
@@ -102,8 +158,41 @@ getmissingattr(TupleDesc tupleDesc,
 
                if (attrmiss->am_present)
                {
+                       missing_cache_key key;
+                       missing_cache_key *entry;
+                       bool            found;
+                       MemoryContext oldctx;
+
                        *isnull = false;
-                       return attrmiss->am_value;
+
+                       /* no  need to cache by-value attributes */
+                       if (att->attbyval)
+                               return attrmiss->am_value;
+
+                       /* set up cache if required */
+                       if (missing_cache == NULL)
+                               init_missing_cache();
+
+                       /* check if there's a cache entry */
+                       Assert(att->attlen > 0 || att->attlen == -1);
+                       if (att->attlen > 0)
+                               key.len = att->attlen;
+                       else
+                               key.len = VARSIZE_ANY(attrmiss->am_value);
+                       key.value = attrmiss->am_value;
+
+                       entry = hash_search(missing_cache, &key, HASH_ENTER, &found);
+
+                       if (!found)
+                       {
+                               /* cache miss, so we need a non-transient copy of the datum */
+                               oldctx = MemoryContextSwitchTo(TopMemoryContext);
+                               entry->value =
+                                       datumCopy(attrmiss->am_value, false, att->attlen);
+                               MemoryContextSwitchTo(oldctx);
+                       }
+
+                       return entry->value;
                }
        }
 

diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index c463596ea7ba331509e7eca80458aa6d379b354f..e11a650b1c3b76cdb2c8905ec132a34dae52928d 100644 (file)
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2910,6 +2910,7 @@ mblen_converter
 mbverifier
 md5_ctxt
 metastring
+missing_cache_key
 mix_data_t
 mixedStruct
 mode_t