minor improvements from comments

Signed-off-by: ekexium <[email protected]>

Author: ekexium

internal/unionstore/arena/arena.go

@@ -226,6 +226,7 @@ func (cp *MemDBCheckpoint) IsSamePosition(other *MemDBCheckpoint) bool {
 	return cp.blocks == other.blocks && cp.offsetInBlock == other.offsetInBlock
 }
 
+// LessThan compares two checkpoints.
 func (cp *MemDBCheckpoint) LessThan(cp2 *MemDBCheckpoint) bool {
 	if cp == nil || cp2 == nil {
 		logutil.BgLogger().Panic("unexpected nil checkpoint", zap.Any("cp", cp), zap.Any("cp2", cp2))

internal/unionstore/art/art.go

@@ -54,8 +54,8 @@ type ART struct {
 	missCount         atomic.Uint64
 
 	// The counter of every write operation, used to invalidate iterators that were created before the write operation.
-	SeqNo int
-	// increased by 1 when an operation that may affect the content returned by "snapshot iter" (i.e. stage[0]) happens.
+	WriteSeqNo int
+	// Increased by 1 when an operation that may affect the content returned by "snapshot iter" (i.e. stage[0]) happens.
 	// It's used to invalidate snapshot iterators.
 	// invariant: no concurrent access to it
 	SnapshotSeqNo int
@@ -122,7 +122,7 @@ func (t *ART) Set(key artKey, value []byte, ops ...kv.FlagsOp) error {
 		}
 	}
 
-	t.SeqNo++
+	t.WriteSeqNo++
 	if len(t.stages) == 0 {
 		t.dirty = true
 	}
@@ -487,7 +487,7 @@ func (t *ART) RevertToCheckpoint(cp *arena.MemDBCheckpoint) {
 	t.allocator.vlogAllocator.RevertToCheckpoint(t, cp)
 	t.allocator.vlogAllocator.Truncate(cp)
 	t.allocator.vlogAllocator.OnMemChange()
-	t.SeqNo++
+	t.WriteSeqNo++
 	if len(t.stages) == 0 || t.stages[0].LessThan(cp) {
 		t.SnapshotSeqNo++
 	}
@@ -510,7 +510,7 @@ func (t *ART) Release(h int) {
 	if h != len(t.stages) {
 		panic("cannot release staging buffer")
 	}
-	t.SeqNo++
+	t.WriteSeqNo++
 	if h == 1 {
 		t.SnapshotSeqNo++
 		tail := t.checkpoint()
@@ -533,7 +533,7 @@ func (t *ART) Cleanup(h int) {
 		panic(fmt.Sprintf("cannot cleanup staging buffer, h=%v, len(tree.stages)=%v", h, len(t.stages)))
 	}
 
-	t.SeqNo++
+	t.WriteSeqNo++
 	if h == 1 {
 		t.SnapshotSeqNo++
 	}
@@ -562,7 +562,7 @@ func (t *ART) Reset() {
 	t.allocator.vlogAllocator.Reset()
 	t.lastTraversedNode.Store(arena.NullU64Addr)
 	t.SnapshotSeqNo++
-	t.SeqNo++
+	t.WriteSeqNo++
 }
 
 // DiscardValues releases the memory used by all values.

internal/unionstore/art/art_iterator.go

@@ -59,7 +59,7 @@ func (t *ART) iter(lowerBound, upperBound []byte, reverse, includeFlags bool) (*
 		// this avoids the initial value of currAddr equals to endAddr.
 		currAddr: arena.BadAddr,
 		endAddr:  arena.NullAddr,
-		seqNo:    t.SeqNo,
+		seqNo:    t.WriteSeqNo,
 	}
 	it.init(lowerBound, upperBound)
 	if !it.valid {
@@ -88,11 +88,11 @@ type Iterator struct {
 }
 
 func (it *Iterator) checkSeqNo() {
-	if it.seqNo != it.tree.SeqNo && !it.ignoreSeqNo {
+	if it.seqNo != it.tree.WriteSeqNo && !it.ignoreSeqNo {
 		logutil.BgLogger().Panic(
 			"seqNo mismatch",
 			zap.Int("it seqNo", it.seqNo),
-			zap.Int("art seqNo", it.tree.SeqNo),
+			zap.Int("art seqNo", it.tree.WriteSeqNo),
 			zap.Stack("stack"),
 		)
 	}

internal/unionstore/art/art_snapshot.go

@@ -21,7 +21,9 @@ import (
 	"github.com/tikv/client-go/v2/internal/unionstore/arena"
 )
 
-func (t *ART) getSnapshot() arena.MemDBCheckpoint {
+// GetSnapshot returns the "snapshot" for snapshotGetter or snapshotIterator, which is usually the snapshot
+// of stage[0]
+func (t *ART) GetSnapshot() arena.MemDBCheckpoint {
 	if len(t.stages) > 0 {
 		return t.stages[0]
 	}
@@ -32,7 +34,7 @@ func (t *ART) getSnapshot() arena.MemDBCheckpoint {
 func (t *ART) SnapshotGetter() *SnapGetter {
 	return &SnapGetter{
 		tree: t,
-		cp:   t.getSnapshot(),
+		cp:   t.GetSnapshot(),
 	}
 }
 
@@ -52,7 +54,7 @@ func (t *ART) newSnapshotIterator(start, end []byte, desc bool) *SnapIter {
 	inner.ignoreSeqNo = true
 	it := &SnapIter{
 		Iterator: inner,
-		cp:       t.getSnapshot(),
+		cp:       t.GetSnapshot(),
 	}
 	it.tree.allocator.snapshotInc()
 	for !it.setValue() && it.Valid() {

internal/unionstore/memdb_art.go

@@ -17,6 +17,7 @@ package unionstore
 import (
 	"context"
 	"fmt"
+	"go.uber.org/zap"
 	"sync"
 
 	"github.com/tikv/client-go/v2/internal/logutil"
@@ -210,6 +211,9 @@ type snapshotBatchedIter struct {
 	pos       int
 	batchSize int
 	nextKey   []byte
+
+	// only used to check if the snapshot ever changes between batches. It is not supposed to change.
+	snapshot MemDBCheckpoint
 }
 
 func (db *artDBWithContext) BatchedSnapshotIter(lower, upper []byte, reverse bool) Iterator {
@@ -225,14 +229,29 @@ func (db *artDBWithContext) BatchedSnapshotIter(lower, upper []byte, reverse boo
 		batchSize:             32,
 	}
 
-	iter.fillBatch()
+	iter.snapshot = db.GetSnapshot()
+	err := iter.fillBatch()
+	if err != nil {
+		logutil.BgLogger().Error("failed to fill batch for snapshotBatchedIter", zap.Error(err))
+	}
 	return iter
 }
 
 func (it *snapshotBatchedIter) fillBatch() error {
 	if it.snapshotTruncateSeqNo != it.db.SnapshotSeqNo {
-		return errors.New(fmt.Sprintf("invalid iter: truncation happened, iter's=%d, db's=%d",
-			it.snapshotTruncateSeqNo, it.db.SnapshotSeqNo))
+		return errors.Errorf(
+			"invalid iter: truncation happened, iter's=%d, db's=%d",
+			it.snapshotTruncateSeqNo,
+			it.db.SnapshotSeqNo,
+		)
+	}
+
+	if it.db.GetSnapshot() != it.snapshot {
+		return errors.Errorf(
+			"snapshot changed between batches, expected=%v, actual=%v",
+			it.snapshot,
+			it.db.GetSnapshot(),
+		)
 	}
 
 	it.db.RLock()