[SPARK-53656][SQL] Refactor MemoryStream to use SparkSession instead of SQLContext

Author: ganeshas-db

sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/runtime/memory.scala

@@ -24,7 +24,7 @@ import javax.annotation.concurrent.GuardedBy
 import scala.collection.mutable.ListBuffer
 
 import org.apache.spark.internal.Logging
-import org.apache.spark.sql.{Encoder, SQLContext}
+import org.apache.spark.sql.{Encoder, SparkSession}
 import org.apache.spark.sql.catalyst.InternalRow
 import org.apache.spark.sql.catalyst.encoders.{encoderFor, ExpressionEncoder}
 import org.apache.spark.sql.catalyst.expressions.UnsafeRow
@@ -47,28 +47,28 @@ object MemoryStream {
   protected val currentBlockId = new AtomicInteger(0)
   protected val memoryStreamId = new AtomicInteger(0)
 
-  def apply[A : Encoder](implicit sqlContext: SQLContext): MemoryStream[A] =
-    new MemoryStream[A](memoryStreamId.getAndIncrement(), sqlContext)
+  def apply[A : Encoder](implicit sparkSession: SparkSession): MemoryStream[A] =
+    new MemoryStream[A](memoryStreamId.getAndIncrement(), sparkSession)
 
-  def apply[A : Encoder](numPartitions: Int)(implicit sqlContext: SQLContext): MemoryStream[A] =
-    new MemoryStream[A](memoryStreamId.getAndIncrement(), sqlContext, Some(numPartitions))
+  def apply[A : Encoder](numPartitions: Int)(implicit sparkSession: SparkSession): MemoryStream[A] =
+    new MemoryStream[A](memoryStreamId.getAndIncrement(), sparkSession, Some(numPartitions))
 }
 
 /**
  * A base class for memory stream implementations. Supports adding data and resetting.
  */
-abstract class MemoryStreamBase[A : Encoder](sqlContext: SQLContext) extends SparkDataStream {
+abstract class MemoryStreamBase[A : Encoder](sparkSession: SparkSession) extends SparkDataStream {
   val encoder = encoderFor[A]
   protected val attributes = toAttributes(encoder.schema)
 
   protected lazy val toRow: ExpressionEncoder.Serializer[A] = encoder.createSerializer()
 
   def toDS(): Dataset[A] = {
-    Dataset[A](sqlContext.sparkSession, logicalPlan)
+    Dataset[A](sparkSession, logicalPlan)
   }
 
   def toDF(): DataFrame = {
-    Dataset.ofRows(sqlContext.sparkSession, logicalPlan)
+    Dataset.ofRows(sparkSession, logicalPlan)
   }
 
   def addData(data: A*): OffsetV2 = {
@@ -156,9 +156,9 @@ class MemoryStreamScanBuilder(stream: MemoryStreamBase[_]) extends ScanBuilder w
  */
 case class MemoryStream[A : Encoder](
     id: Int,
-    sqlContext: SQLContext,
+    sparkSession: SparkSession,
     numPartitions: Option[Int] = None)
-  extends MemoryStreamBase[A](sqlContext)
+  extends MemoryStreamBase[A](sparkSession)
   with MicroBatchStream
   with SupportsTriggerAvailableNow
   with Logging {

sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/sources/ContinuousMemoryStream.scala

@@ -27,7 +27,7 @@ import org.json4s.jackson.Serialization
 
 import org.apache.spark.{SparkEnv, TaskContext}
 import org.apache.spark.rpc.RpcEndpointRef
-import org.apache.spark.sql.{Encoder, SQLContext}
+import org.apache.spark.sql.{Encoder, SparkSession}
 import org.apache.spark.sql.catalyst.InternalRow
 import org.apache.spark.sql.catalyst.expressions.UnsafeRow
 import org.apache.spark.sql.connector.read.InputPartition
@@ -44,8 +44,9 @@ import org.apache.spark.util.RpcUtils
  *    ContinuousMemoryStreamInputPartitionReader instances to poll. It returns the record at
  *    the specified offset within the list, or null if that offset doesn't yet have a record.
  */
-class ContinuousMemoryStream[A : Encoder](id: Int, sqlContext: SQLContext, numPartitions: Int = 2)
-  extends MemoryStreamBase[A](sqlContext) with ContinuousStream {
+class ContinuousMemoryStream[A : Encoder](id: Int,
+                                          sparkSession: SparkSession, numPartitions: Int = 2)
+  extends MemoryStreamBase[A](sparkSession) with ContinuousStream {
 
   private implicit val formats: Formats = Serialization.formats(NoTypeHints)
 
@@ -112,11 +113,11 @@ class ContinuousMemoryStream[A : Encoder](id: Int, sqlContext: SQLContext, numPa
 object ContinuousMemoryStream {
   protected val memoryStreamId = new AtomicInteger(0)
 
-  def apply[A : Encoder](implicit sqlContext: SQLContext): ContinuousMemoryStream[A] =
-    new ContinuousMemoryStream[A](memoryStreamId.getAndIncrement(), sqlContext)
+  def apply[A : Encoder](implicit sparkSession: SparkSession): ContinuousMemoryStream[A] =
+    new ContinuousMemoryStream[A](memoryStreamId.getAndIncrement(), sparkSession)
 
-  def singlePartition[A : Encoder](implicit sqlContext: SQLContext): ContinuousMemoryStream[A] =
-    new ContinuousMemoryStream[A](memoryStreamId.getAndIncrement(), sqlContext, 1)
+  def singlePartition[A : Encoder](implicit sparkSession: SparkSession): ContinuousMemoryStream[A] =
+    new ContinuousMemoryStream[A](memoryStreamId.getAndIncrement(), sparkSession, 1)
 }
 
 /**

sql/core/src/test/scala/org/apache/spark/sql/DatasetSuite.scala

@@ -1012,7 +1012,7 @@ class DatasetSuite extends QueryTest
     assert(err.getMessage.contains("An Observation can be used with a Dataset only once"))
 
     // streaming datasets are not supported
-    val streamDf = new MemoryStream[Int](0, sqlContext).toDF()
+    val streamDf = new MemoryStream[Int](0, spark).toDF()
     val streamObservation = Observation("stream")
     val streamErr = intercept[IllegalArgumentException] {
       streamDf.observe(streamObservation, avg($"value").cast("int").as("avg_val"))

sql/core/src/test/scala/org/apache/spark/sql/execution/streaming/AsyncProgressTrackingMicroBatchExecutionSuite.scala

@@ -67,9 +67,9 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
 
   class MemoryStreamCapture[A: Encoder](
       id: Int,
-      sqlContext: SQLContext,
+      sparkSession: SparkSession,
       numPartitions: Option[Int] = None)
-    extends MemoryStream[A](id, sqlContext, numPartitions = numPartitions) {
+    extends MemoryStream[A](id, sparkSession: SparkSession, numPartitions = numPartitions) {
 
     val commits = new ListBuffer[streaming.Offset]()
     val commitThreads = new ListBuffer[Thread]()
@@ -136,7 +136,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
   test("async WAL commits recovery") {
     val checkpointLocation = Utils.createTempDir(namePrefix = "streaming.metadata").getCanonicalPath
 
-    val inputData = new MemoryStream[Int](id = 0, sqlContext = sqlContext)
+    val inputData = new MemoryStream[Int](id = 0, spark)
     val ds = inputData.toDF()
 
     var index = 0
@@ -204,7 +204,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
   }
 
   test("async WAL commits turn on and off") {
-    val inputData = new MemoryStream[Int](id = 0, sqlContext = sqlContext)
+    val inputData = new MemoryStream[Int](id = 0, spark)
     val ds = inputData.toDS()
 
     val checkpointLocation = Utils.createTempDir(namePrefix = "streaming.metadata").getCanonicalPath
@@ -308,7 +308,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
   }
 
   test("Fail with once trigger") {
-    val inputData = new MemoryStream[Int](id = 0, sqlContext = sqlContext)
+    val inputData = new MemoryStream[Int](id = 0, spark)
     val ds = inputData.toDF()
 
     val e = intercept[IllegalArgumentException] {
@@ -323,7 +323,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
 
   test("Fail with available now trigger") {
 
-    val inputData = new MemoryStream[Int](id = 0, sqlContext = sqlContext)
+    val inputData = new MemoryStream[Int](id = 0, spark)
     val ds = inputData.toDF()
 
     val e = intercept[IllegalArgumentException] {
@@ -339,7 +339,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
   test("switching between async wal commit enabled and trigger once") {
     val checkpointLocation = Utils.createTempDir(namePrefix = "streaming.metadata").getCanonicalPath
 
-    val inputData = new MemoryStream[Int](id = 0, sqlContext = sqlContext)
+    val inputData = new MemoryStream[Int](id = 0, spark)
     val ds = inputData.toDF()
 
     var index = 0
@@ -500,7 +500,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
   test("switching between async wal commit enabled and available now") {
     val checkpointLocation = Utils.createTempDir(namePrefix = "streaming.metadata").getCanonicalPath
 
-    val inputData = new MemoryStream[Int](id = 0, sqlContext = sqlContext)
+    val inputData = new MemoryStream[Int](id = 0, spark)
     val ds = inputData.toDF()
 
     var index = 0
@@ -669,7 +669,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
   }
 
   def testAsyncWriteErrorsAlreadyExists(path: String): Unit = {
-    val inputData = new MemoryStream[Int](id = 0, sqlContext = sqlContext)
+    val inputData = new MemoryStream[Int](id = 0, spark)
     val ds = inputData.toDS()
     val checkpointLocation = Utils.createTempDir(namePrefix = "streaming.metadata").getCanonicalPath
 
@@ -720,7 +720,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
   }
 
   def testAsyncWriteErrorsPermissionsIssue(path: String): Unit = {
-    val inputData = new MemoryStream[Int](id = 0, sqlContext = sqlContext)
+    val inputData = new MemoryStream[Int](id = 0, spark)
     val ds = inputData.toDS()
     val checkpointLocation = Utils.createTempDir(namePrefix = "streaming.metadata").getCanonicalPath
     val commitDir = new File(checkpointLocation + path)
@@ -778,7 +778,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
 
     val checkpointLocation = Utils.createTempDir(namePrefix = "streaming.metadata").getCanonicalPath
 
-    val inputData = new MemoryStreamCapture[Int](id = 0, sqlContext = sqlContext)
+    val inputData = new MemoryStreamCapture[Int](id = 0, spark)
 
     val ds = inputData.toDF()
 
@@ -852,7 +852,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
   }
 
   test("interval commits and recovery") {
-    val inputData = new MemoryStreamCapture[Int](id = 0, sqlContext = sqlContext)
+    val inputData = new MemoryStreamCapture[Int](id = 0, spark)
     val ds = inputData.toDS()
 
     val checkpointLocation = Utils.createTempDir(namePrefix = "streaming.metadata").getCanonicalPath
@@ -934,7 +934,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
   }
 
   test("recovery when first offset is not zero and not commit log entries") {
-    val inputData = new MemoryStreamCapture[Int](id = 0, sqlContext = sqlContext)
+    val inputData = new MemoryStreamCapture[Int](id = 0, spark)
     val ds = inputData.toDS()
 
     val checkpointLocation = Utils.createTempDir(namePrefix = "streaming.metadata").getCanonicalPath
@@ -961,7 +961,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
     /**
      * start new stream
      */
-    val inputData2 = new MemoryStreamCapture[Int](id = 0, sqlContext = sqlContext)
+    val inputData2 = new MemoryStreamCapture[Int](id = 0, spark)
     val ds2 = inputData2.toDS()
     testStream(ds2, extraOptions = Map(
       ASYNC_PROGRESS_TRACKING_ENABLED -> "true",
@@ -995,7 +995,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
   }
 
   test("recovery non-contiguous log") {
-    val inputData = new MemoryStreamCapture[Int](id = 0, sqlContext = sqlContext)
+    val inputData = new MemoryStreamCapture[Int](id = 0, spark)
     val ds = inputData.toDS()
 
     val checkpointLocation = Utils.createTempDir(namePrefix = "streaming.metadata").getCanonicalPath
@@ -1088,7 +1088,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
   }
 
   test("Fail on pipelines using unsupported sinks") {
-    val inputData = new MemoryStream[Int](id = 0, sqlContext = sqlContext)
+    val inputData = new MemoryStream[Int](id = 0, spark)
     val ds = inputData.toDF()
 
     val e = intercept[IllegalArgumentException] {
@@ -1109,7 +1109,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
 
     withSQLConf(SQLConf.MIN_BATCHES_TO_RETAIN.key -> "2", SQLConf.ASYNC_LOG_PURGE.key -> "false") {
       withTempDir { checkpointLocation =>
-        val inputData = new MemoryStreamCapture[Int](id = 0, sqlContext = sqlContext)
+        val inputData = new MemoryStreamCapture[Int](id = 0, spark)
         val ds = inputData.toDS()
 
         val clock = new StreamManualClock
@@ -1243,7 +1243,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
   test("with async log purging") {
     withSQLConf(SQLConf.MIN_BATCHES_TO_RETAIN.key -> "2", SQLConf.ASYNC_LOG_PURGE.key -> "true") {
       withTempDir { checkpointLocation =>
-        val inputData = new MemoryStreamCapture[Int](id = 0, sqlContext = sqlContext)
+        val inputData = new MemoryStreamCapture[Int](id = 0, spark)
         val ds = inputData.toDS()
 
         val clock = new StreamManualClock
@@ -1381,7 +1381,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
   }
 
   test("test multiple gaps in offset and commit logs") {
-    val inputData = new MemoryStreamCapture[Int](id = 0, sqlContext = sqlContext)
+    val inputData = new MemoryStreamCapture[Int](id = 0, spark)
     val ds = inputData.toDS()
 
     val checkpointLocation = Utils.createTempDir(namePrefix = "streaming.metadata").getCanonicalPath
@@ -1427,7 +1427,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
     /**
      * start new stream
      */
-    val inputData2 = new MemoryStreamCapture[Int](id = 0, sqlContext = sqlContext)
+    val inputData2 = new MemoryStreamCapture[Int](id = 0, spark)
     val ds2 = inputData2.toDS()
     testStream(ds2, extraOptions = Map(
       ASYNC_PROGRESS_TRACKING_ENABLED -> "true",
@@ -1460,7 +1460,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
   }
 
   test("recovery when gaps exist in offset and commit log") {
-    val inputData = new MemoryStreamCapture[Int](id = 0, sqlContext = sqlContext)
+    val inputData = new MemoryStreamCapture[Int](id = 0, spark)
     val ds = inputData.toDS()
 
     val checkpointLocation = Utils.createTempDir(namePrefix = "streaming.metadata").getCanonicalPath
@@ -1494,7 +1494,7 @@ class AsyncProgressTrackingMicroBatchExecutionSuite
     /**
      * start new stream
      */
-    val inputData2 = new MemoryStreamCapture[Int](id = 0, sqlContext = sqlContext)
+    val inputData2 = new MemoryStreamCapture[Int](id = 0, spark)
     val ds2 = inputData2.toDS()
     testStream(ds2, extraOptions = Map(
       ASYNC_PROGRESS_TRACKING_ENABLED -> "true",

sql/core/src/test/scala/org/apache/spark/sql/execution/streaming/MicroBatchExecutionSuite.scala

@@ -54,7 +54,7 @@ class MicroBatchExecutionSuite extends StreamTest with BeforeAndAfter with Match
   test("async log purging") {
     withSQLConf(SQLConf.MIN_BATCHES_TO_RETAIN.key -> "2", SQLConf.ASYNC_LOG_PURGE.key -> "true") {
       withTempDir { checkpointLocation =>
-        val inputData = new MemoryStream[Int](id = 0, sqlContext = sqlContext)
+        val inputData = new MemoryStream[Int](id = 0, spark)
         val ds = inputData.toDS()
         testStream(ds)(
           StartStream(checkpointLocation = checkpointLocation.getCanonicalPath),
@@ -99,7 +99,7 @@ class MicroBatchExecutionSuite extends StreamTest with BeforeAndAfter with Match
   test("error notifier test") {
     withSQLConf(SQLConf.MIN_BATCHES_TO_RETAIN.key -> "2", SQLConf.ASYNC_LOG_PURGE.key -> "true") {
       withTempDir { checkpointLocation =>
-        val inputData = new MemoryStream[Int](id = 0, sqlContext = sqlContext)
+        val inputData = new MemoryStream[Int](id = 0, spark)
         val ds = inputData.toDS()
         val e = intercept[StreamingQueryException] {
 

sql/core/src/test/scala/org/apache/spark/sql/execution/streaming/state/StateStoreCoordinatorSuite.scala

@@ -123,11 +123,10 @@ class StateStoreCoordinatorSuite extends SparkFunSuite with SharedSparkContext {
   test("query stop deactivates related store providers") {
     var coordRef: StateStoreCoordinatorRef = null
     try {
-      val spark = SparkSession.builder().sparkContext(sc).getOrCreate()
+      implicit val spark: SparkSession = SparkSession.builder().sparkContext(sc).getOrCreate()
       SparkSession.setActiveSession(spark)
       import spark.implicits._
       coordRef = spark.streams.stateStoreCoordinator
-      implicit val sqlContext = spark.sqlContext
       spark.conf.set(SQLConf.SHUFFLE_PARTITIONS.key, "1")
 
       // Start a query and run a batch to load state stores
@@ -254,7 +253,7 @@ class StateStoreCoordinatorSuite extends SparkFunSuite with SharedSparkContext {
       ) {
         case (coordRef, spark) =>
           import spark.implicits._
-          implicit val sqlContext = spark.sqlContext
+          implicit val sparkSession: SparkSession = spark
           val inputData = MemoryStream[Int]
           val query = setUpStatefulQuery(inputData, "query")
           // Add, commit, and wait multiple times to force snapshot versions and time difference
@@ -290,7 +289,7 @@ class StateStoreCoordinatorSuite extends SparkFunSuite with SharedSparkContext {
       ) {
         case (coordRef, spark) =>
           import spark.implicits._
-          implicit val sqlContext = spark.sqlContext
+          implicit val sparkSession: SparkSession = spark
           // Start a join query and run some data to force snapshot uploads
           val input1 = MemoryStream[Int]
           val input2 = MemoryStream[Int]
@@ -333,7 +332,7 @@ class StateStoreCoordinatorSuite extends SparkFunSuite with SharedSparkContext {
     ) {
       case (coordRef, spark) =>
         import spark.implicits._
-        implicit val sqlContext = spark.sqlContext
+        implicit val sparkSession: SparkSession = spark
         // Start and run two queries together with some data to force snapshot uploads
         val input1 = MemoryStream[Int]
         val input2 = MemoryStream[Int]
@@ -400,7 +399,7 @@ class StateStoreCoordinatorSuite extends SparkFunSuite with SharedSparkContext {
     ) {
       case (coordRef, spark) =>
         import spark.implicits._
-        implicit val sqlContext = spark.sqlContext
+        implicit val sparkSession: SparkSession = spark
         // Start a query and run some data to force snapshot uploads
         val inputData = MemoryStream[Int]
         val query = setUpStatefulQuery(inputData, "query")
@@ -444,7 +443,7 @@ class StateStoreCoordinatorSuite extends SparkFunSuite with SharedSparkContext {
     ) {
       case (coordRef, spark) =>
         import spark.implicits._
-        implicit val sqlContext = spark.sqlContext
+        implicit val sparkSession: SparkSession = spark
         // Start a query and run some data to force snapshot uploads
         val inputData = MemoryStream[Int]
         val query = setUpStatefulQuery(inputData, "query")

sql/core/src/test/scala/org/apache/spark/sql/execution/streaming/state/StateStoreSuite.scala

@@ -1206,9 +1206,8 @@ class StateStoreSuite extends StateStoreSuiteBase[HDFSBackedStateStoreProvider]
   test("SPARK-21145: Restarted queries create new provider instances") {
     try {
       val checkpointLocation = Utils.createTempDir().getAbsoluteFile
-      val spark = SparkSession.builder().master("local[2]").getOrCreate()
+      implicit val spark: SparkSession = SparkSession.builder().master("local[2]").getOrCreate()
       SparkSession.setActiveSession(spark)
-      implicit val sqlContext = spark.sqlContext
       spark.conf.set(SQLConf.SHUFFLE_PARTITIONS.key, "1")
       import spark.implicits._
       val inputData = MemoryStream[Int]