Github user davies commented on a diff in the pull request:
https://github.com/apache/spark/pull/8971#discussion_r41415962
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnType.scala ---
@@ -399,24 +409,158 @@ private[sql] object BINARY extends
ByteArrayColumnType(16) {
override def getField(row: InternalRow, ordinal: Int): Array[Byte] = {
row.getBinary(ordinal)
}
+
+ def serialize(value: Array[Byte]): Array[Byte] = value
+ def deserialize(bytes: Array[Byte]): Array[Byte] = bytes
}
-// Used to process generic objects (all types other than those listed
above). Objects should be
-// serialized first before appending to the column `ByteBuffer`, and is
also extracted as serialized
-// byte array.
-private[sql] case class GENERIC(dataType: DataType) extends
ByteArrayColumnType(16) {
- override def setField(row: MutableRow, ordinal: Int, value:
Array[Byte]): Unit = {
- row.update(ordinal, SparkSqlSerializer.deserialize[Any](value))
+private[sql] case class DECIMAL(precision: Int, scale: Int)
+ extends ByteArrayColumnType[Decimal](20) {
+
+ override val dataType: DataType = DecimalType(precision, scale)
+
+ override def getField(row: InternalRow, ordinal: Int): Decimal = {
+ row.getDecimal(ordinal, precision, scale)
}
- override def getField(row: InternalRow, ordinal: Int): Array[Byte] = {
- SparkSqlSerializer.serialize(row.get(ordinal, dataType))
+ override def setField(row: MutableRow, ordinal: Int, value: Decimal):
Unit = {
+ row.setDecimal(ordinal, value, precision)
+ }
+
+ override def serialize(value: Decimal): Array[Byte] = {
+ value.toJavaBigDecimal.unscaledValue().toByteArray
+ }
+
+ override def deserialize(bytes: Array[Byte]): Decimal = {
+ val javaDecimal = new BigDecimal(new BigInteger(bytes), scale)
+ Decimal.apply(javaDecimal, precision, scale)
+ }
+}
+
+private[sql] case class STRUCT(dataType: DataType)
+ extends ByteArrayColumnType[InternalRow](20) {
+
+ private val projection: UnsafeProjection =
+ UnsafeProjection.create(dataType.asInstanceOf[StructType])
+ private val numOfFields: Int =
dataType.asInstanceOf[StructType].fields.size
+
+ override def setField(row: MutableRow, ordinal: Int, value:
InternalRow): Unit = {
+ row.update(ordinal, value)
+ }
+
+ override def getField(row: InternalRow, ordinal: Int): InternalRow = {
+ row.getStruct(ordinal, numOfFields)
+ }
+
+ override def serialize(value: InternalRow): Array[Byte] = {
+ val unsafeRow = if (value.isInstanceOf[UnsafeRow]) {
+ value.asInstanceOf[UnsafeRow]
+ } else {
+ projection(value)
+ }
+ unsafeRow.getBytes
}
+
+ override def deserialize(bytes: Array[Byte]): InternalRow = {
+ val unsafeRow = new UnsafeRow
+ unsafeRow.pointTo(bytes, numOfFields, bytes.length)
+ unsafeRow
+ }
+
+ override def clone(v: InternalRow): InternalRow = v.copy()
+}
+
+private[sql] case class ARRAY(dataType: DataType)
+ extends ByteArrayColumnType[ArrayData](16) {
+
+ private lazy val projection = UnsafeProjection.create(Array(dataType))
+ private val mutableRow = new GenericMutableRow(new Array[Any](1))
+
+ override def setField(row: MutableRow, ordinal: Int, value: ArrayData):
Unit = {
+ row.update(ordinal, value)
+ }
+
+ override def getField(row: InternalRow, ordinal: Int): ArrayData = {
+ row.getArray(ordinal)
+ }
+
+ override def serialize(value: ArrayData): Array[Byte] = {
+ val unsafeArray = if (value.isInstanceOf[UnsafeArrayData]) {
+ value.asInstanceOf[UnsafeArrayData]
+ } else {
+ mutableRow(0) = value
+ projection(mutableRow).getArray(0)
--- End diff --
We could, but only saving two lines, it may not worth it.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
