Spark 3.4: Adapt to hash function under clickhouse-core

ClickHouse · Jul 25, 2023 · 4e201d6 · 4e201d6
1 parent 85a025f
commit 4e201d6
Show file tree

Hide file tree

Showing 16 changed files with 60 additions and 493 deletions.
diff --git a/...rc/test/scala/org/apache/spark/sql/clickhouse/cluster/ClickHouseClusterHashUDFSuite.scala b/...rc/test/scala/org/apache/spark/sql/clickhouse/cluster/ClickHouseClusterHashUDFSuite.scala
@@ -15,8 +15,12 @@
 package org.apache.spark.sql.clickhouse.cluster
 
 import org.apache.spark.sql.clickhouse.TestUtils.om
-import xenon.clickhouse.func.{CompositeFunctionRegistry, DynamicFunctionRegistry, StaticFunctionRegistry}
-import xenon.clickhouse.func.clickhouse.ClickHouseXxHash64Shard
+import xenon.clickhouse.func.{
+  ClickHouseXxHash64Shard,
+  CompositeFunctionRegistry,
+  DynamicFunctionRegistry,
+  StaticFunctionRegistry
+}
 
 import java.lang.{Long => JLong}
 
@@ -30,15 +34,6 @@ class ClickHouseClusterHashUDFSuite extends SparkClickHouseClusterTest {
     new CompositeFunctionRegistry(Array(StaticFunctionRegistry, dynamicFunctionRegistry))
   }
 
-  def product[A](xs: Seq[Seq[A]]): Seq[Seq[A]] =
-    xs.toList match {
-      case Nil => Seq(Seq())
-      case head :: tail => for {
-          h <- head
-          t <- product(tail)
-        } yield h +: t
-    }
-
   def runTest(funcSparkName: String, funcCkName: String, stringVal: String): Unit = {
     val sparkResult = spark.sql(
       s"""SELECT

diff --git a/spark-3.4/clickhouse-spark/src/main/scala/xenon/clickhouse/ClickHouseCatalog.scala b/spark-3.4/clickhouse-spark/src/main/scala/xenon/clickhouse/ClickHouseCatalog.scala
@@ -26,8 +26,7 @@ import xenon.clickhouse.Constants._
 import xenon.clickhouse.client.NodeClient
 import xenon.clickhouse.exception.CHClientException
 import xenon.clickhouse.exception.ClickHouseErrCode._
-import xenon.clickhouse.func.clickhouse.ClickHouseXxHash64Shard
-import xenon.clickhouse.func.{FunctionRegistry, _}
+import xenon.clickhouse.func.{ClickHouseXxHash64Shard, FunctionRegistry, _}
 import xenon.clickhouse.spec._
 
 import java.time.ZoneId

diff --git a/...ickhouse/func/clickhouse/CityHash64.scala → ...la/xenon/clickhouse/func/CityHash64.scala b/...ickhouse/func/clickhouse/CityHash64.scala → ...la/xenon/clickhouse/func/CityHash64.scala
@@ -12,29 +12,15 @@
  * limitations under the License.
  */
 
-package xenon.clickhouse.func.clickhouse
+package xenon.clickhouse.func
 
-import io.netty.buffer.{ByteBuf, Unpooled}
-import org.apache.spark.unsafe.types.UTF8String
-import xenon.clickhouse.func.MultiArgsHash
-import xenon.clickhouse.func.clickhouse.cityhash.{CityHash_v1_0_2, UInt128}
+import xenon.clickhouse.hash
 
-object CityHash64 extends MultiArgsHash {
+object CityHash64 extends MultiStringArgsHash {
   // https://github.com/ClickHouse/ClickHouse/blob/v23.5.3.24-stable/src/Functions/FunctionsHashing.h#L694
 
   override protected def funcName: String = "clickhouse_cityHash64"
   override val ckFuncNames: Array[String] = Array("cityHash64")
 
-  def convertToByteBuf(array: Array[Byte]): ByteBuf = {
-    val byteBuf = Unpooled.buffer(array.length).writeBytes(array)
-    byteBuf
-  }
-
-  override def invokeBase(value: UTF8String): Long = {
-    // ignore UInt64 vs Int64
-    val data = value.getBytes
-    CityHash_v1_0_2.CityHash64(convertToByteBuf(data), 0, data.length)
-  }
-
-  override def combineHashes(v1: Long, v2: Long): Long = CityHash_v1_0_2.Hash128to64(new UInt128(v1, v2))
+  override def applyHash(input: Array[Any]): Long = hash.CityHash64(input)
 }
diff --git a/...non/clickhouse/func/clickhouse/Days.scala → ...in/scala/xenon/clickhouse/func/Days.scala b/...non/clickhouse/func/clickhouse/Days.scala → ...in/scala/xenon/clickhouse/func/Days.scala
@@ -12,11 +12,10 @@
  * limitations under the License.
  */
 
-package xenon.clickhouse.func.clickhouse
+package xenon.clickhouse.func
 
 import org.apache.spark.sql.connector.catalog.functions.{BoundFunction, ScalarFunction, UnboundFunction}
 import org.apache.spark.sql.types._
-import xenon.clickhouse.func.ClickhouseEquivFunction
 
 import java.time.LocalDate
 import java.time.format.DateTimeFormatter

diff --git a/spark-3.4/clickhouse-spark/src/main/scala/xenon/clickhouse/func/FunctionRegistry.scala b/spark-3.4/clickhouse-spark/src/main/scala/xenon/clickhouse/func/FunctionRegistry.scala
@@ -15,7 +15,6 @@
 package xenon.clickhouse.func
 
 import org.apache.spark.sql.connector.catalog.functions.UnboundFunction
-import xenon.clickhouse.func.clickhouse._
 
 import scala.collection.mutable
 

diff --git a/...on/clickhouse/func/clickhouse/Hours.scala → ...n/scala/xenon/clickhouse/func/Hours.scala b/...on/clickhouse/func/clickhouse/Hours.scala → ...n/scala/xenon/clickhouse/func/Hours.scala
@@ -12,13 +12,12 @@
  * limitations under the License.
  */
 
-package xenon.clickhouse.func.clickhouse
+package xenon.clickhouse.func
 
 import org.apache.spark.sql.connector.catalog.functions.{BoundFunction, ScalarFunction, UnboundFunction}
 import org.apache.spark.sql.types._
-import xenon.clickhouse.func.ClickhouseEquivFunction
 
-import java.sql.{Date, Timestamp}
+import java.sql.Timestamp
 import java.text.SimpleDateFormat
 
 object Hours extends UnboundFunction with ScalarFunction[Int] with ClickhouseEquivFunction {

diff --git a/...enon/clickhouse/func/clickhouse/Mod.scala → ...ain/scala/xenon/clickhouse/func/Mod.scala b/...enon/clickhouse/func/clickhouse/Mod.scala → ...ain/scala/xenon/clickhouse/func/Mod.scala
@@ -12,11 +12,10 @@
  * limitations under the License.
  */
 
-package xenon.clickhouse.func.clickhouse
+package xenon.clickhouse.func
 
 import org.apache.spark.sql.connector.catalog.functions.{BoundFunction, ScalarFunction, UnboundFunction}
 import org.apache.spark.sql.types._
-import xenon.clickhouse.func.ClickhouseEquivFunction
 
 object Mod extends UnboundFunction with ScalarFunction[Long] with ClickhouseEquivFunction {
 

diff --git a/...n/clickhouse/func/clickhouse/Months.scala → .../scala/xenon/clickhouse/func/Months.scala b/...n/clickhouse/func/clickhouse/Months.scala → .../scala/xenon/clickhouse/func/Months.scala
@@ -12,11 +12,10 @@
  * limitations under the License.
  */
 
-package xenon.clickhouse.func.clickhouse
+package xenon.clickhouse.func
 
 import org.apache.spark.sql.connector.catalog.functions.{BoundFunction, ScalarFunction, UnboundFunction}
 import org.apache.spark.sql.types._
-import xenon.clickhouse.func.ClickhouseEquivFunction
 
 import java.time.LocalDate
 import java.time.format.DateTimeFormatter

diff --git a/...xenon/clickhouse/func/MultiArgsHash.scala → ...clickhouse/func/MultiStringArgsHash.scala b/...xenon/clickhouse/func/MultiArgsHash.scala → ...clickhouse/func/MultiStringArgsHash.scala
@@ -19,32 +19,41 @@ import org.apache.spark.sql.connector.catalog.functions.{BoundFunction, ScalarFu
 import org.apache.spark.sql.types._
 import org.apache.spark.unsafe.types.UTF8String
 
-abstract class MultiArgsHash extends UnboundFunction with ClickhouseEquivFunction {
+abstract class MultiStringArgsHash extends UnboundFunction with ClickhouseEquivFunction {
+
+  def applyHash(input: Array[Any]): Long
+
+  protected def funcName: String
+
+  override val ckFuncNames: Array[String]
+
+  override def description: String = s"$name: (value: string, ...) => hash_value: long"
+
   private def isExceptedType(dt: DataType): Boolean =
     dt.isInstanceOf[StringType]
 
   final override def name: String = funcName
+
   final override def bind(inputType: StructType): BoundFunction = {
     val inputDataTypes = inputType.fields.map(_.dataType)
-    if (inputDataTypes.forall(isExceptedType)) new ScalarFunction[Long] {
-      override def inputTypes(): Array[DataType] = inputDataTypes
-      override def name: String = funcName
-      override def canonicalName: String = s"clickhouse.$name"
-      override def resultType: DataType = LongType
-      override def toString: String = name
-      override def produceResult(input: InternalRow): Long = {
-        val inputStrings: Seq[UTF8String] =
-          input.toSeq(Seq.fill(input.numFields)(StringType)).asInstanceOf[Seq[UTF8String]]
-        inputStrings.map(invokeBase).reduce(combineHashes)
+    if (inputDataTypes.forall(isExceptedType)) {
+      // need to new a ScalarFunction instance for each bind,
+      // because we do not know the number of arguments in advance
+      new ScalarFunction[Long] {
+        override def inputTypes(): Array[DataType] = inputDataTypes
+        override def name: String = funcName
+        override def canonicalName: String = s"clickhouse.$name"
+        override def resultType: DataType = LongType
+        override def toString: String = name
+        override def produceResult(input: InternalRow): Long = {
+          val inputStrings: Array[Any] =
+            input.toSeq(Seq.fill(input.numFields)(StringType)).asInstanceOf[Seq[UTF8String]].toArray
+              .map(_.getBytes)
+          applyHash(inputStrings)
+        }
       }
-    }
-    else throw new UnsupportedOperationException(s"Expect multiple STRING argument. $description")
+    } else throw new UnsupportedOperationException(s"Expect multiple STRING argument. $description")
 
   }
 
-  protected def funcName: String
-  override val ckFuncNames: Array[String]
-  override def description: String = s"$name: (value: string, ...) => hash_value: long"
-  def invokeBase(value: UTF8String): Long
-  def combineHashes(v1: Long, v2: Long): Long
 }
diff --git a/...ckhouse/func/clickhouse/MurmurHash2.scala → ...a/xenon/clickhouse/func/MurmurHash2.scala b/...ckhouse/func/clickhouse/MurmurHash2.scala → ...a/xenon/clickhouse/func/MurmurHash2.scala
@@ -12,40 +12,25 @@
  * limitations under the License.
  */
 
-package xenon.clickhouse.func.clickhouse
+package xenon.clickhouse.func
 
-import org.apache.commons.codec.digest.{MurmurHash2, MurmurHash3}
-import org.apache.spark.sql.connector.catalog.functions.{BoundFunction, ScalarFunction, UnboundFunction}
-import org.apache.spark.sql.types._
-import org.apache.spark.unsafe.types.UTF8String
-import xenon.clickhouse.func.{ClickhouseEquivFunction, MultiArgsHash, Util}
+import xenon.clickhouse.hash
+import xenon.clickhouse.hash.HashUtils
 
-object MurmurHash2_64 extends MultiArgsHash {
+object MurmurHash2_64 extends MultiStringArgsHash {
   // https://github.com/ClickHouse/ClickHouse/blob/v23.5.3.24-stable/src/Functions/FunctionsHashing.h#L460
 
   override protected def funcName: String = "clickhouse_murmurHash2_64"
   override val ckFuncNames: Array[String] = Array("murmurHash2_64")
 
-  override def invokeBase(value: UTF8String): Long = {
-    // ignore UInt64 vs Int64
-    val data = value.getBytes
-    MurmurHash2.hash64(data, data.length, 0)
-  }
-
-  override def combineHashes(v1: Long, v2: Long): Long = Util.intHash64Impl(v1) ^ v2
+  override def applyHash(input: Array[Any]): Long = hash.Murmurhash2_64(input)
 }
 
-object MurmurHash2_32 extends MultiArgsHash {
+object MurmurHash2_32 extends MultiStringArgsHash {
   // https://github.com/ClickHouse/ClickHouse/blob/v23.5.3.24-stable/src/Functions/FunctionsHashing.h#L519
 
   override protected def funcName: String = "clickhouse_murmurHash2_32"
   override val ckFuncNames: Array[String] = Array("murmurHash2_32")
 
-  override def invokeBase(value: UTF8String): Long = {
-    val data = value.getBytes
-    val v = MurmurHash2.hash32(data, data.length, 0)
-    Util.toUInt32Range(v)
-  }
-
-  override def combineHashes(v1: Long, v2: Long): Long = Util.toUInt32Range(Util.int32Impl(v1) ^ v2)
+  override def applyHash(input: Array[Any]): Long = HashUtils.toUInt32(hash.Murmurhash2_32(input))
 }
diff --git a/...ckhouse/func/clickhouse/MurmurHash3.scala → ...a/xenon/clickhouse/func/MurmurHash3.scala b/...ckhouse/func/clickhouse/MurmurHash3.scala → ...a/xenon/clickhouse/func/MurmurHash3.scala
@@ -12,41 +12,25 @@
  * limitations under the License.
  */
 
-package xenon.clickhouse.func.clickhouse
+package xenon.clickhouse.func
 
-import org.apache.commons.codec.digest.MurmurHash3
-import org.apache.spark.sql.connector.catalog.functions.{BoundFunction, ScalarFunction, UnboundFunction}
-import org.apache.spark.sql.types._
-import org.apache.spark.unsafe.types.UTF8String
-import xenon.clickhouse.func.{ClickhouseEquivFunction, MultiArgsHash, Util}
+import xenon.clickhouse.hash
+import xenon.clickhouse.hash.HashUtils
 
-object MurmurHash3_64 extends MultiArgsHash {
+object MurmurHash3_64 extends MultiStringArgsHash {
   // https://github.com/ClickHouse/ClickHouse/blob/v23.5.3.24-stable/src/Functions/FunctionsHashing.h#L543
 
   override protected def funcName: String = "clickhouse_murmurHash3_64"
   override val ckFuncNames: Array[String] = Array("murmurHash3_64")
 
-  override def invokeBase(value: UTF8String): Long = {
-    // ignore UInt64 vs Int64
-    val data = value.getBytes
-    val hashes = MurmurHash3.hash128x64(data, 0, data.length, 0)
-    hashes(0) ^ hashes(1)
-  }
-
-  override def combineHashes(v1: Long, v2: Long): Long = Util.intHash64Impl(v1) ^ v2
+  override def applyHash(input: Array[Any]): Long = hash.Murmurhash3_64(input)
 }
 
-object MurmurHash3_32 extends MultiArgsHash {
+object MurmurHash3_32 extends MultiStringArgsHash {
   // https://github.com/ClickHouse/ClickHouse/blob/v23.5.3.24-stable/src/Functions/FunctionsHashing.h#L519
 
   override protected def funcName: String = "clickhouse_murmurHash3_32"
   override val ckFuncNames: Array[String] = Array("murmurHash3_32")
 
-  override def invokeBase(value: UTF8String): Long = {
-    val data = value.getBytes
-    val v = MurmurHash3.hash32x86(data, 0, data.length, 0)
-    Util.toUInt32Range(v)
-  }
-
-  override def combineHashes(v1: Long, v2: Long): Long = Util.toUInt32Range(Util.int32Impl(v1) ^ v2)
+  override def applyHash(input: Array[Any]): Long = HashUtils.toUInt32(hash.Murmurhash3_32(input))
 }
diff --git a/...clickhouse/func/clickhouse/XxHash64.scala → ...cala/xenon/clickhouse/func/XxHash64.scala b/...clickhouse/func/clickhouse/XxHash64.scala → ...cala/xenon/clickhouse/func/XxHash64.scala
@@ -12,13 +12,12 @@
  * limitations under the License.
  */
 
-package xenon.clickhouse.func.clickhouse
+package xenon.clickhouse.func
 
 import org.apache.spark.sql.catalyst.expressions.XxHash64Function
 import org.apache.spark.sql.connector.catalog.functions.{BoundFunction, ScalarFunction, UnboundFunction}
 import org.apache.spark.sql.types._
 import org.apache.spark.unsafe.types.UTF8String
-import xenon.clickhouse.func.ClickhouseEquivFunction
 import xenon.clickhouse.spec.{ClusterSpec, ShardUtils}
 
 /**

diff --git a/...on/clickhouse/func/clickhouse/Years.scala → ...n/scala/xenon/clickhouse/func/Years.scala b/...on/clickhouse/func/clickhouse/Years.scala → ...n/scala/xenon/clickhouse/func/Years.scala
@@ -12,11 +12,10 @@
  * limitations under the License.
  */
 
-package xenon.clickhouse.func.clickhouse
+package xenon.clickhouse.func
 
 import org.apache.spark.sql.connector.catalog.functions.{BoundFunction, ScalarFunction, UnboundFunction}
 import org.apache.spark.sql.types._
-import xenon.clickhouse.func.ClickhouseEquivFunction
 
 import java.time.LocalDate
 import java.time.format.DateTimeFormatter