Initial commit for multi-part lists

dgraph/cmd/debug/run.go

@@ -112,6 +112,11 @@ func uidToVal(itr *badger.Iterator, prefix string) map[uint64]int {
 		if pk.IsSchema() {
 			continue
 		}
+		if pk.StartUid > 0 {
+			// This key is part of a multi-part posting list. Skip it and only read
+			// the main key, which is the entry point to read the whole list.
+			continue
+		}
 
 		pl, err := posting.ReadPostingList(item.KeyCopy(nil), itr)
 		if err != nil {
@@ -501,6 +506,9 @@ func printKeys(db *badger.DB) {
 		if pk.Uid > 0 {
 			fmt.Fprintf(&buf, " uid: %d ", pk.Uid)
 		}
+		if pk.StartUid > 0 {
+			fmt.Fprintf(&buf, " startUid: %d ", pk.StartUid)
+		}
 		fmt.Fprintf(&buf, " key: %s", hex.EncodeToString(item.Key()))
 		if opt.itemMeta {
 			fmt.Fprintf(&buf, " item: [%d, b%04b]", item.EstimatedSize(), item.UserMeta())

posting/index.go

@@ -527,15 +527,18 @@ func (r *rebuild) Run(ctx context.Context) error {
 		if le == 0 {
 			continue
 		}
-		kv, err := pl.MarshalToKv()
+		kvs, err := pl.Rollup()
 		if err != nil {
 			return err
 		}
-		// We choose to write the PL at r.startTs, so it won't be read by txns,
-		// which occurred before this schema mutation. Typically, we use
-		// kv.Version as the timestamp.
-		if err = writer.SetAt(kv.Key, kv.Value, kv.UserMeta[0], r.startTs); err != nil {
-			return err
+
+		for _, kv := range kvs {
+			// We choose to write the PL at r.startTs, so it won't be read by txns,
+			// which occurred before this schema mutation. Typically, we use
+			// kv.Version as the timestamp.
+			if err = writer.SetAt(kv.Key, kv.Value, kv.UserMeta[0], r.startTs); err != nil {
+				return err
+			}
 		}
 		// This locking is just to catch any future issues.  We shouldn't need
 		// to release this lock, because each posting list must only be accessed