Improve performance of index-only scans with many index columns.

StoreIndexTuple was a loop over index_getattr, which is O(N^2)
if the index columns are variable-width, and the performance
impact is already quite visible at ten columns.  The obvious
move is to replace that with a call to index_deform_tuple ...
but that's *also* a loop over index_getattr.  Improve it to
be essentially a clone of heap_deform_tuple.

(There are a few other places that loop over all index columns
with index_getattr, and perhaps should be changed likewise,
but most of them don't seem performance-critical.  Anyway, the
rest would mostly only be interested in the index key columns,
which there aren't likely to be so many of.  Wide index tuples
are a new thing with INCLUDE.)

Konstantin Knizhnik

Discussion: https://postgr.es/m/e06b2d27-04fc-5c0e-bb8c-ecd72aa24959@postgrespro.ru

1 files changed, 7 insertions, 13 deletions

diff --git a/src/backend/executor/nodeIndexonlyscan.c b/src/backend/executor/nodeIndexonlyscan.c
index 72c04b528f5..26758e77039 100644
--- a/src/backend/executor/nodeIndexonlyscan.c
+++ b/src/backend/executor/nodeIndexonlyscan.c
@@ -269,23 +269,17 @@ IndexOnlyNext(IndexOnlyScanState *node)
 static void
 StoreIndexTuple(TupleTableSlot *slot, IndexTuple itup, TupleDesc itupdesc)
 {
-	int			nindexatts = itupdesc->natts;
-	Datum	   *values = slot->tts_values;
-	bool	   *isnull = slot->tts_isnull;
-	int			i;
-
 	/*
-	 * Note: we must use the tupdesc supplied by the AM in index_getattr, not
-	 * the slot's tupdesc, in case the latter has different datatypes (this
-	 * happens for btree name_ops in particular).  They'd better have the same
-	 * number of columns though, as well as being datatype-compatible which is
-	 * something we can't so easily check.
+	 * Note: we must use the tupdesc supplied by the AM in index_deform_tuple,
+	 * not the slot's tupdesc, in case the latter has different datatypes
+	 * (this happens for btree name_ops in particular).  They'd better have
+	 * the same number of columns though, as well as being datatype-compatible
+	 * which is something we can't so easily check.
 	 */
-	Assert(slot->tts_tupleDescriptor->natts == nindexatts);
+	Assert(slot->tts_tupleDescriptor->natts == itupdesc->natts);
 
 	ExecClearTuple(slot);
-	for (i = 0; i < nindexatts; i++)
-		values[i] = index_getattr(itup, i + 1, itupdesc, &isnull[i]);
+	index_deform_tuple(itup, itupdesc, slot->tts_values, slot->tts_isnull);
 	ExecStoreVirtualTuple(slot);
 }