Github user liancheng commented on a diff in the pull request:
https://github.com/apache/spark/pull/3640#discussion_r21512962
--- Diff: sql/hive/src/main/scala/org/apache/spark/sql/hive/hiveUdfs.scala
---
@@ -54,47 +54,95 @@ private[hive] abstract class HiveFunctionRegistry
val functionClassName = functionInfo.getFunctionClass.getName
if (classOf[UDF].isAssignableFrom(functionInfo.getFunctionClass)) {
- HiveSimpleUdf(functionClassName, children)
+ HiveSimpleUdf(new HiveFunctionCache(functionClassName), children)
} else if
(classOf[GenericUDF].isAssignableFrom(functionInfo.getFunctionClass)) {
- HiveGenericUdf(functionClassName, children)
+ HiveGenericUdf(new HiveFunctionCache(functionClassName), children)
} else if (
classOf[AbstractGenericUDAFResolver].isAssignableFrom(functionInfo.getFunctionClass))
{
- HiveGenericUdaf(functionClassName, children)
+ HiveGenericUdaf(new HiveFunctionCache(functionClassName), children)
} else if
(classOf[UDAF].isAssignableFrom(functionInfo.getFunctionClass)) {
- HiveUdaf(functionClassName, children)
+ HiveUdaf(new HiveFunctionCache(functionClassName), children)
} else if
(classOf[GenericUDTF].isAssignableFrom(functionInfo.getFunctionClass)) {
- HiveGenericUdtf(functionClassName, Nil, children)
+ HiveGenericUdtf(new HiveFunctionCache(functionClassName), Nil,
children)
} else {
sys.error(s"No handler for udf ${functionInfo.getFunctionClass}")
}
}
}
-private[hive] trait HiveFunctionFactory {
- val functionClassName: String
-
- def createFunction[UDFType]() =
-
getContextOrSparkClassLoader.loadClass(functionClassName).newInstance.asInstanceOf[UDFType]
-}
-
-private[hive] abstract class HiveUdf extends Expression with Logging with
HiveFunctionFactory {
- self: Product =>
+/**
+ * This class provides the UDF creation and also the UDF instance
serialization and
+ * de-serialization cross process boundary.
+ *
+ * We use class instead of trait, seems property variables of trait cannot
be serialized when
+ * bundled with Case Class; in the other hand, we need to intercept the
UDF instance ser/de.
+ * the "Has-a" probably better than "Is-a".
+ * @param functionClassName UDF class name
+ */
+class HiveFunctionCache(var functionClassName: String) extends
java.io.Externalizable {
+ // for Serialization
+ def this() = this(null)
+
+ private var instance: Any = null
+
+ def writeExternal(out: java.io.ObjectOutput) {
+ // output the function name
+ out.writeUTF(functionClassName)
+
+ // Write a flag if instance is null or not
+ out.writeBoolean(instance != null)
+ if (instance != null) {
+ // Some of the UDF are serializable, but some others are not
+ // Hive Utilities can handle both cases
+ val baos = new java.io.ByteArrayOutputStream()
+ HiveShim.serializePlan(instance, baos)
+ val functionInBytes = baos.toByteArray
+
+ // output the function bytes
+ out.writeInt(functionInBytes.length)
+ out.write(functionInBytes, 0, functionInBytes.length)
+ }
+ }
- type UDFType
- type EvaluatedType = Any
+ def readExternal(in: java.io.ObjectInput) {
+ // read the function name
+ functionClassName = in.readUTF()
- def nullable = true
+ if (in.readBoolean()) {
+ // if the instance is not null
+ // read the function in bytes
+ val functionInBytesLength = in.readInt()
+ val functionInBytes = new Array[Byte](functionInBytesLength)
+ in.read(functionInBytes, 0, functionInBytesLength)
- lazy val function = createFunction[UDFType]()
+ // deserialize the function object via Hive Utilities
+ instance = HiveShim.deserializePlan(new
java.io.ByteArrayInputStream(functionInBytes),
+ getContextOrSparkClassLoader.loadClass(functionClassName))
+ }
+ }
- override def toString =
s"$nodeName#$functionClassName(${children.mkString(",")})"
+ def createFunction[UDFType](alwaysCreateNewInstance: Boolean = false) = {
+ if (alwaysCreateNewInstance) {
+
getContextOrSparkClassLoader.loadClass(functionClassName).newInstance.asInstanceOf[UDFType]
+ } else {
+ if (instance == null) {
+ instance =
getContextOrSparkClassLoader.loadClass(functionClassName).newInstance
+ }
+ instance.asInstanceOf[UDFType]
+ }
--- End diff --
Actually, how about removing the `alwaysCreateNewInstance` argument (which
is confusing), and define a new `HiveSimpleUdfWrapper` that overrides
`createFunction`, and always return a new instance?
---
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
