support compaction filter

db_test.go

@@ -1155,6 +1155,54 @@ func TestMinReadTs(t *testing.T) {
 	})
 }
 
+func TestCompactionFilter(t *testing.T) {
+	dir, err := ioutil.TempDir("", "badger")
+	require.NoError(t, err)
+	defer os.RemoveAll(dir)
+	opts := getTestOptions(dir)
+	opts.ValueThreshold = 8 * 1024
+	opts.MaxTableSize = 32 * 1024
+	opts.NumMemtables = 2
+	opts.NumLevelZeroTables = 1
+	opts.NumLevelZeroTablesStall = 2
+	opts.CompactionFilter = func(key, val, userMeta []byte) (skip bool) {
+		// Only keep the keys with user meta.
+		if len(userMeta) == 0 {
+			return true
+		}
+		return false
+	}
+	db, err := Open(opts)
+	require.NoError(t, err)
+	val := make([]byte, 1024*4)
+	// Insert 100 entries to trigger some compaction.
+	for i := 0; i < 100; i++ {
+		db.Update(func(txn *Txn) error {
+			key := []byte(fmt.Sprintf("key%d", i))
+			if i%2 == 0 {
+				txn.Set(key, val)
+			} else {
+				txn.SetWithMetaSlice(key, val, []byte{0})
+			}
+			return nil
+		})
+	}
+	// The first 50 entries must have been compacted already.
+	db.View(func(txn *Txn) error {
+		for i := 0; i < 50; i++ {
+			key := []byte(fmt.Sprintf("key%d", i))
+			item, _ := txn.Get(key)
+			if i%2 == 0 {
+				require.Nil(t, item)
+			} else {
+				require.NotNil(t, item)
+				require.Len(t, item.UserMeta(), 1)
+			}
+		}
+		return nil
+	})
+}
+
 func ExampleOpen() {
 	dir, err := ioutil.TempDir("", "badger")
 	if err != nil {

levels.go

@@ -325,6 +325,8 @@ func (s *levelsController) compactBuildTables(level int, cd compactDef, limiter
 	// would affect the snapshot view guarantee provided by transactions.
 	minReadTs := s.kv.orc.readMark.MinReadTs()
 
+	filter := s.kv.opt.CompactionFilter
+
 	var numVersions int
 	var lastKey, skipKey []byte
 	var newTables []*table.Table
@@ -334,7 +336,7 @@ func (s *levelsController) compactBuildTables(level int, cd compactDef, limiter
 	if start != nil {
 		it.Seek(start)
 	}
-	for ; it.Valid() && (end == nil || y.CompareKeys(it.Key(), end) < 0); {
+	for it.Valid() && (end == nil || y.CompareKeys(it.Key(), end) < 0) {
 		timeStart := time.Now()
 		fileID := s.reserveFileID()
 		fd, err := y.CreateSyncedFile(table.NewFilename(fileID, s.kv.opt.Dir), false)
@@ -400,6 +402,13 @@ func (s *levelsController) compactBuildTables(level int, cd compactDef, limiter
 					}
 				}
 			}
+			if filter != nil {
+				skip := filter(it.Key(), vs.Value, vs.UserMeta)
+				if skip {
+					discardStats.collect(vs)
+					continue
+				}
+			}
 			numKeys++
 			builder.Add(it.Key(), it.Value())
 		}

options.go

@@ -109,6 +109,11 @@ type Options struct {
 	TableBuilderOptions options.TableBuilderOptions
 
 	ValueLogWriteOptions options.ValueLogWriterOptions
+
+	// The compaction process invokes this method for kv that is being compacted. A return value of false
+	// indicates that the kv should be preserved in the output of this compaction run and a return value
+	// of true indicates that this key-value should be removed from the output of the compaction.
+	CompactionFilter func(key, val, userMeta []byte) (skip bool)
 }
 
 // DefaultOptions sets a list of recommended options for good performance.