[GitHub] felixcheung commented on a change in pull request #23746: [SPARK-26761][SQL][R] Vectorized R gapply() implementation

[GitBox <gi...@apache.org>]

felixcheung commented on a change in pull request #23746: [SPARK-26761][SQL][R] Vectorized R gapply() implementation
URL: https://github.com/apache/spark/pull/23746#discussion_r255848613
 
 

 ##########
 File path: sql/core/src/main/scala/org/apache/spark/sql/execution/r/ArrowRRunner.scala
 ##########
 @@ -0,0 +1,205 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements.  See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License.  You may obtain a copy of the License at
+ *
+ *    http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.spark.sql.execution.r
+
+import java.io._
+import java.nio.channels.Channels
+
+import scala.collection.JavaConverters._
+
+import org.apache.arrow.vector.VectorSchemaRoot
+import org.apache.arrow.vector.ipc.{ArrowStreamReader, ArrowStreamWriter}
+import org.apache.arrow.vector.util.ByteArrayReadableSeekableByteChannel
+
+import org.apache.spark.{SparkException, TaskContext}
+import org.apache.spark.api.r._
+import org.apache.spark.api.r.SpecialLengths
+import org.apache.spark.broadcast.Broadcast
+import org.apache.spark.sql.catalyst.InternalRow
+import org.apache.spark.sql.execution.arrow.{ArrowUtils, ArrowWriter}
+import org.apache.spark.sql.types.StructType
+import org.apache.spark.sql.vectorized.{ArrowColumnVector, ColumnarBatch, ColumnVector}
+import org.apache.spark.util.Utils
+
+
+/**
+ * Similar to `ArrowPythonRunner`, but exchange data with R worker via Arrow stream.
+ */
+class ArrowRRunner(
+    func: Array[Byte],
+    packageNames: Array[Byte],
+    broadcastVars: Array[Broadcast[Object]],
+    schema: StructType,
+    timeZoneId: String)
+  extends RRunner[ColumnarBatch](
+    func,
+    "arrow",
+    "arrow",
+    packageNames,
+    broadcastVars,
+    numPartitions = -1,
+    isDataFrame = true,
+    schema.fieldNames,
+    RRunnerModes.DATAFRAME_GAPPLY) {
+
+  protected override def writeData(
+      dataOut: DataOutputStream,
+      printOut: PrintStream,
+      iter: Iterator[_]): Unit = if (iter.hasNext) {
+    val inputIterator = iter.asInstanceOf[Iterator[(Array[Byte], Iterator[InternalRow])]]
+    val arrowSchema = ArrowUtils.toArrowSchema(schema, timeZoneId)
+    val allocator = ArrowUtils.rootAllocator.newChildAllocator(
+      "stdout writer for R", 0, Long.MaxValue)
+    val root = VectorSchemaRoot.create(arrowSchema, allocator)
+    val out = new ByteArrayOutputStream()
+    val keys = collection.mutable.ArrayBuffer.empty[Array[Byte]]
+
+    Utils.tryWithSafeFinally {
+      val arrowWriter = ArrowWriter.create(root)
+      val writer = new ArrowStreamWriter(root, null, Channels.newChannel(out))
+      writer.start()
+
+      while (inputIterator.hasNext) {
+        val (key, nextBatch) = inputIterator.next()
+        keys.append(key)
+
+        while (nextBatch.hasNext) {
+          arrowWriter.write(nextBatch.next())
+        }
+
+        arrowWriter.finish()
+        writer.writeBatch()
+        arrowWriter.reset()
+      }
+      writer.end()
+    } {
+      // Don't close root and allocator in TaskCompletionListener to prevent
+      // a race condition. See `ArrowPythonRunner`.
+      root.close()
+      allocator.close()
+    }
+
+    // Currently, there looks no way to read batch by batch by socket connection in R side,
+    // See ARROW-4512. Therefore, it writes the whole Arrow streaming-formatted binary at
+    // once for now.
+    val data = out.toByteArray
+    dataOut.writeInt(data.length)
+    dataOut.write(data)
+
+    keys.foreach(dataOut.write)
+  }
+
+  protected override def newReaderIterator(
+      dataStream: DataInputStream, errThread: BufferedStreamThread): Iterator[ColumnarBatch] = {
+    new Iterator[ColumnarBatch] {
+      private val allocator = ArrowUtils.rootAllocator.newChildAllocator(
+        "stdin reader for R", 0, Long.MaxValue)
+
+      private var reader: ArrowStreamReader = _
+      private var root: VectorSchemaRoot = _
+      private var vectors: Array[ColumnVector] = _
+
+      TaskContext.get().addTaskCompletionListener[Unit] { _ =>
+        if (reader != null) {
+          reader.close(false)
+        }
+        allocator.close()
+      }
+
+      private var batchLoaded = true
+      private var nextObj: ColumnarBatch = _
+      private var eos = false
+
+      override def hasNext: Boolean = nextObj != null || {
+        if (!eos) {
+          nextObj = read()
+          hasNext
+        } else {
+          false
+        }
+      }
+
+      override def next(): ColumnarBatch = {
+        if (hasNext) {
+          val obj = nextObj
+          nextObj = null.asInstanceOf[ColumnarBatch]
+          obj
 
 Review comment:
   this and hasNext is a bit weird, but I assume there's a reason for it..

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org