wzx140 commented on code in PR #6745:
URL: https://github.com/apache/hudi/pull/6745#discussion_r982461731
##########
hudi-client/hudi-spark-client/src/main/scala/org/apache/spark/sql/hudi/HoodieSparkRecordSerializer.scala:
##########
@@ -38,25 +40,20 @@ import scala.collection.mutable
* schema, as to reduce network IO.
* Actions like parsing or compressing schemas are computationally expensive
so the serializer
* caches all previously seen values as to reduce the amount of work needed to
do.
- * @param schemas a map where the keys are unique IDs for spark schemas and
the values are the
- * string representation of the Avro schema, used to decrease
the amount of data
- * that needs to be serialized.
*/
-class SparkStructTypeSerializer(schemas: Map[Long, StructType]) extends
KSerializer[HoodieSparkRecord] {
+class HoodieSparkRecordSerializer extends KSerializer[HoodieSparkRecord] {
/** Used to reduce the amount of effort to compress the schema */
private val compressCache = new mutable.HashMap[StructType, Array[Byte]]()
private val decompressCache = new mutable.HashMap[ByteBuffer, StructType]()
- /** Fingerprinting is very expensive so this alleviates most of the work */
- private val fingerprintCache = new mutable.HashMap[StructType, Long]()
- private val schemaCache = new mutable.HashMap[Long, StructType]()
-
// GenericAvroSerializer can't take a SparkConf in the constructor b/c then
it would become
// a member of KryoSerializer, which would make KryoSerializer not
Serializable. We make
// the codec lazy here just b/c in some unit tests, we use a KryoSerializer
w/out having
// the SparkEnv set (note those tests would fail if they tried to serialize
avro data).
private lazy val codec = CompressionCodec.createCodec(SparkEnv.get.conf)
+ private var objSerializerMap = new ConcurrentHashMap[Kryo,
FieldSerializer[HoodieSparkRecord]]
Review Comment:
"We should only serialize the UnsafeRow that HoodieSparkRecord holds (see my
comment above)".
We can not do this. Because there are other fields(HoodieKey,
HoodieRecordLocation, HoodieOperation, etc.) in HoodieSparkRecord
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
