Remove HeapBitmapScan's skip_fetch optimization

The optimization does not take the removal of TIDs by a concurrent vacuum into
account. The concurrent vacuum can remove dead TIDs and make pages ALL_VISIBLE
while those dead TIDs are referenced in the bitmap. This can lead to a
skip_fetch scan returning too many tuples.

It likely would be possible to implement this optimization safely, but we
don't have the necessary infrastructure in place. Nor is it clear that it's
worth building that infrastructure, given how limited the skip_fetch
optimization is.

In the backbranches we just disable the optimization by always passing
need_tuples=true to table_beginscan_bm(). We can't perform API/ABI changes in
the backbranches and we want to make the change as minimal as possible.

Author: Matthias van de Meent <boekewurm+postgres@gmail.com>
Reported-By: Konstantin Knizhnik <knizhnik@garret.ru>
Discussion: https://postgr.es/m/CAEze2Wg3gXXZTr6_rwC+s4-o2ZVFB5F985uUSgJTsECx6AmGcQ@mail.gmail.com
Backpatch-through: 13

2 files changed, 10 insertions, 97 deletions

diff --git a/src/backend/access/heap/heapam.c b/src/backend/access/heap/heapam.c
index cedaa195cb6..5b3fe4a1d3b 100644
--- a/src/backend/access/heap/heapam.c
+++ b/src/backend/access/heap/heapam.c
@@ -314,31 +314,6 @@ bitmapheap_stream_read_next(ReadStream *pgsr, void *private_data,
 			tbmres->blockno >= hscan->rs_nblocks)
 			continue;
 
-		/*
-		 * We can skip fetching the heap page if we don't need any fields from
-		 * the heap, the bitmap entries don't need rechecking, and all tuples
-		 * on the page are visible to our transaction.
-		 */
-		if (!(sscan->rs_flags & SO_NEED_TUPLES) &&
-			!tbmres->recheck &&
-			VM_ALL_VISIBLE(sscan->rs_rd, tbmres->blockno, &bscan->rs_vmbuffer))
-		{
-			OffsetNumber offsets[TBM_MAX_TUPLES_PER_PAGE];
-			int			noffsets;
-
-			/* can't be lossy in the skip_fetch case */
-			Assert(!tbmres->lossy);
-			Assert(bscan->rs_empty_tuples_pending >= 0);
-
-			/*
-			 * We throw away the offsets, but this is the easiest way to get a
-			 * count of tuples.
-			 */
-			noffsets = tbm_extract_page_tuple(tbmres, offsets, TBM_MAX_TUPLES_PER_PAGE);
-			bscan->rs_empty_tuples_pending += noffsets;
-			continue;
-		}
-
 		return tbmres->blockno;
 	}
 
@@ -1124,8 +1099,10 @@ heap_beginscan(Relation relation, Snapshot snapshot,
 	{
 		BitmapHeapScanDesc bscan = palloc(sizeof(BitmapHeapScanDescData));
 
-		bscan->rs_vmbuffer = InvalidBuffer;
-		bscan->rs_empty_tuples_pending = 0;
+		/*
+		 * Bitmap Heap scans do not have any fields that a normal Heap Scan
+		 * does not have, so no special initializations required here.
+		 */
 		scan = (HeapScanDesc) bscan;
 	}
 	else
@@ -1280,23 +1257,10 @@ heap_rescan(TableScanDesc sscan, ScanKey key, bool set_params,
 		scan->rs_cbuf = InvalidBuffer;
 	}
 
-	if (scan->rs_base.rs_flags & SO_TYPE_BITMAPSCAN)
-	{
-		BitmapHeapScanDesc bscan = (BitmapHeapScanDesc) scan;
-
-		/*
-		 * Reset empty_tuples_pending, a field only used by bitmap heap scan,
-		 * to avoid incorrectly emitting NULL-filled tuples from a previous
-		 * scan on rescan.
-		 */
-		bscan->rs_empty_tuples_pending = 0;
-
-		if (BufferIsValid(bscan->rs_vmbuffer))
-		{
-			ReleaseBuffer(bscan->rs_vmbuffer);
-			bscan->rs_vmbuffer = InvalidBuffer;
-		}
-	}
+	/*
+	 * SO_TYPE_BITMAPSCAN would be cleaned up here, but it does not hold any
+	 * additional data vs a normal HeapScan
+	 */
 
 	/*
 	 * The read stream is reset on rescan. This must be done before
@@ -1325,15 +1289,6 @@ heap_endscan(TableScanDesc sscan)
 	if (BufferIsValid(scan->rs_cbuf))
 		ReleaseBuffer(scan->rs_cbuf);
 
-	if (scan->rs_base.rs_flags & SO_TYPE_BITMAPSCAN)
-	{
-		BitmapHeapScanDesc bscan = (BitmapHeapScanDesc) sscan;
-
-		bscan->rs_empty_tuples_pending = 0;
-		if (BufferIsValid(bscan->rs_vmbuffer))
-			ReleaseBuffer(bscan->rs_vmbuffer);
-	}
-
 	/*
 	 * Must free the read stream before freeing the BufferAccessStrategy.
 	 */
diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c
index 24d3765aa20..ac082fefa77 100644
--- a/src/backend/access/heap/heapam_handler.c
+++ b/src/backend/access/heap/heapam_handler.c
@@ -2138,32 +2138,6 @@ heapam_scan_bitmap_next_tuple(TableScanDesc scan,
 	while (hscan->rs_cindex >= hscan->rs_ntuples)
 	{
 		/*
-		 * Emit empty tuples before advancing to the next block
-		 */
-		if (bscan->rs_empty_tuples_pending > 0)
-		{
-			/*
-			 * If we don't have to fetch the tuple, just return nulls.
-			 */
-			ExecStoreAllNullTuple(slot);
-			bscan->rs_empty_tuples_pending--;
-
-			/*
-			 * We do not recheck all NULL tuples. Because the streaming read
-			 * API only yields TBMIterateResults for blocks actually fetched
-			 * from the heap, we must unset `recheck` ourselves here to ensure
-			 * correct results.
-			 *
-			 * Our read stream callback accrues a count of empty tuples to
-			 * emit and then emits them after emitting tuples from the next
-			 * fetched block. If no blocks need fetching, we'll emit the
-			 * accrued count at the end of the scan.
-			 */
-			*recheck = false;
-			return true;
-		}
-
-		/*
 		 * Returns false if the bitmap is exhausted and there are no further
 		 * blocks we need to scan.
 		 */
@@ -2516,24 +2490,8 @@ BitmapHeapScanNextBlock(TableScanDesc scan,
 
 	if (BufferIsInvalid(hscan->rs_cbuf))
 	{
-		if (BufferIsValid(bscan->rs_vmbuffer))
-		{
-			ReleaseBuffer(bscan->rs_vmbuffer);
-			bscan->rs_vmbuffer = InvalidBuffer;
-		}
-
-		/*
-		 * The bitmap is exhausted. Now emit any remaining empty tuples. The
-		 * read stream API only returns TBMIterateResults for blocks actually
-		 * fetched from the heap. Our callback will accrue a count of empty
-		 * tuples to emit for all blocks we skipped fetching. So, if we skip
-		 * fetching heap blocks at the end of the relation (or no heap blocks
-		 * are fetched) we need to ensure we emit empty tuples before ending
-		 * the scan. We don't recheck empty tuples so ensure `recheck` is
-		 * unset.
-		 */
-		*recheck = false;
-		return bscan->rs_empty_tuples_pending > 0;
+		/* the bitmap is exhausted */
+		return false;
 	}
 
 	Assert(per_buffer_data);