[SPARK-34806][SQL] Add Observation helper for Dataset.observe

### What changes were proposed in this pull request?
This pull request introduces a helper class that simplifies usage of `Dataset.observe()` for batch datasets:

    val observation = Observation("name")
    val observed = ds.observe(observation, max($"id").as("max_id"))
    observed.count()
    val metrics = observation.get

### Why are the changes needed?
Currently, users are required to implement the `QueryExecutionListener` interface to retrieve the metrics, as well as apply some knowledge on threading and locking to pull the metrics over to the main thread. With the helper class, metrics can be retrieved from batch dataset processing with three lines of code (the action on the observed dataset does not count as a line of code here).

### Does this PR introduce _any_ user-facing change?
Yes, one new class and one `Dataset`` method.

### How was this patch tested?
Adds a unit test to `DataFrameSuite`, similar to `"get observable metrics by callback"` in `DataFrameCallbackSuite`.

Closes #33422 from EnricoMi/branch-observation.

Authored-by: Enrico Minack <github@enrico.minack.dev>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>

sql/core/src/main/scala/org/apache/spark/sql/Dataset.scala

@@ -19,6 +19,7 @@ package org.apache.spark.sql
 
 import java.io.{ByteArrayOutputStream, CharArrayWriter, DataOutputStream}
 
+import scala.annotation.varargs
 import scala.collection.JavaConverters._
 import scala.collection.mutable.{ArrayBuffer, HashSet}
 import scala.reflect.runtime.universe.TypeTag
@@ -1947,6 +1948,32 @@ class Dataset[T] private[sql](
     CollectMetrics(name, (expr +: exprs).map(_.named), logicalPlan)
   }
 
+  /**
+   * Observe (named) metrics through an `org.apache.spark.sql.Observation` instance.
+   * This is equivalent to calling `observe(String, Column, Column*)` but does not require
+   * adding `org.apache.spark.sql.util.QueryExecutionListener` to the spark session.
+   * This method does not support streaming datasets.
+   *
+   * A user can retrieve the metrics by accessing `org.apache.spark.sql.Observation.get`.
+   *
+   * {{{
+   *   // Observe row count (rows) and highest id (maxid) in the Dataset while writing it
+   *   val observation = Observation("my_metrics")
+   *   val observed_ds = ds.observe(observation, count(lit(1)).as("rows"), max($"id").as("maxid"))
+   *   observed_ds.write.parquet("ds.parquet")
+   *   val metrics = observation.get
+   * }}}
+   *
+   * @throws IllegalArgumentException If this is a streaming Dataset (this.isStreaming == true)
+   *
+   * @group typedrel
+   * @since 3.3.0
+   */
+  @varargs
+  def observe(observation: Observation, expr: Column, exprs: Column*): Dataset[T] = {
+    observation.on(this, expr, exprs: _*)
+  }
+
   /**
    * Returns a new Dataset by taking the first `n` rows. The difference between this function
    * and `head` is that `head` is an action and returns an array (by triggering query execution)

sql/core/src/main/scala/org/apache/spark/sql/Observation.scala

@@ -0,0 +1,156 @@
+/*
+ * 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
+
+import java.util.UUID
+
+import org.apache.spark.sql.execution.QueryExecution
+import org.apache.spark.sql.util.QueryExecutionListener
+
+
+/**
+ * Helper class to simplify usage of `Dataset.observe(String, Column, Column*)`:
+ *
+ * {{{
+ *   // Observe row count (rows) and highest id (maxid) in the Dataset while writing it
+ *   val observation = Observation("my metrics")
+ *   val observed_ds = ds.observe(observation, count(lit(1)).as("rows"), max($"id").as("maxid"))
+ *   observed_ds.write.parquet("ds.parquet")
+ *   val metrics = observation.get
+ * }}}
+ *
+ * This collects the metrics while the first action is executed on the observed dataset. Subsequent
+ * actions do not modify the metrics returned by [[get]]. Retrieval of the metric via [[get]]
+ * blocks until the first action has finished and metrics become available.
+ *
+ * This class does not support streaming datasets.
+ *
+ * @param name name of the metric
+ * @since 3.3.0
+ */
+class Observation(name: String) {
+
+  /**
+   * Create an Observation instance without providing a name. This generates a random name.
+   */
+  def this() = this(UUID.randomUUID().toString)
+
+  private val listener: ObservationListener = ObservationListener(this)
+
+  @volatile private var sparkSession: Option[SparkSession] = None
+
+  @volatile private var row: Option[Row] = None
+
+  /**
+   * Attach this observation to the given [[Dataset]] to observe aggregation expressions.
+   *
+   * @param ds dataset
+   * @param expr first aggregation expression
+   * @param exprs more aggregation expressions
+   * @tparam T dataset type
+   * @return observed dataset
+   * @throws IllegalArgumentException If this is a streaming Dataset (ds.isStreaming == true)
+   */
+  private[spark] def on[T](ds: Dataset[T], expr: Column, exprs: Column*): Dataset[T] = {
+    if (ds.isStreaming) {
+      throw new IllegalArgumentException("Observation does not support streaming Datasets")
+    }
+    register(ds.sparkSession)
+    ds.observe(name, expr, exprs: _*)
+  }
+
+  /**
+   * Get the observed metrics. This waits for the observed dataset to finish its first action.
+   * Only the result of the first action is available. Subsequent actions do not modify the result.
+   *
+   * @return the observed metrics as a [[Row]]
+   * @throws InterruptedException interrupted while waiting
+   */
+  @throws[InterruptedException]
+  def get: Row = {
+    synchronized {
+      // we need to loop as wait might return without us calling notify
+      // https://en.wikipedia.org/w/index.php?title=Spurious_wakeup&oldid=992601610
+      while (this.row.isEmpty) {
+        wait()
+      }
+    }
+
+    this.row.get
+  }
+
+  private def register(sparkSession: SparkSession): Unit = {
+    // makes this class thread-safe:
+    // only the first thread entering this block can set sparkSession
+    // all other threads will see the exception, as it is only allowed to do this once
+    synchronized {
+      if (this.sparkSession.isDefined) {
+        throw new IllegalArgumentException("An Observation can be used with a Dataset only once")
+      }
+      this.sparkSession = Some(sparkSession)
+    }
+
+    sparkSession.listenerManager.register(this.listener)
+  }
+
+  private def unregister(): Unit = {
+    this.sparkSession.foreach(_.listenerManager.unregister(this.listener))
+  }
+
+  private[spark] def onFinish(qe: QueryExecution): Unit = {
+    synchronized {
+      if (this.row.isEmpty) {
+        this.row = qe.observedMetrics.get(name)
+        if (this.row.isDefined) {
+          notifyAll()
+          unregister()
+        }
+      }
+    }
+  }
+
+}
+
+private[sql] case class ObservationListener(observation: Observation)
+  extends QueryExecutionListener {
+
+  override def onSuccess(funcName: String, qe: QueryExecution, durationNs: Long): Unit =
+    observation.onFinish(qe)
+
+  override def onFailure(funcName: String, qe: QueryExecution, exception: Exception): Unit =
+    observation.onFinish(qe)
+
+}
+
+/**
+ * (Scala-specific) Create instances of Observation via Scala `apply`.
+ * @since 3.3.0
+ */
+object Observation {
+
+  /**
+   * Observation constructor for creating an anonymous observation.
+   */
+  def apply(): Observation = new Observation()
+
+  /**
+   * Observation constructor for creating a named observation.
+   */
+  def apply(name: String): Observation = new Observation(name)
+
+}

sql/core/src/test/java/test/org/apache/spark/sql/JavaDataFrameSuite.java

@@ -34,6 +34,7 @@ import org.junit.*;
 import org.apache.spark.api.java.JavaRDD;
 import org.apache.spark.api.java.JavaSparkContext;
 import org.apache.spark.sql.Dataset;
+import org.apache.spark.sql.Observation;
 import org.apache.spark.sql.Row;
 import org.apache.spark.sql.RowFactory;
 import org.apache.spark.sql.expressions.UserDefinedFunction;
@@ -523,4 +524,50 @@ public class JavaDataFrameSuite {
       .map(row -> row.get(0).toString() + row.getString(1)).toArray(String[]::new);
     Assert.assertArrayEquals(expected, result);
   }
+
+  /**
+   * Tests the Java API of Observation and Dataset.observe(Observation, Column, Column*).
+   */
+  @Test
+  public void testObservation() {
+    Observation namedObservation = new Observation("named");
+    Observation unnamedObservation = new Observation();
+
+    Dataset<Long> df = spark
+      .range(100)
+      .observe(
+        namedObservation,
+        min(col("id")).as("min_val"),
+        max(col("id")).as("max_val"),
+        sum(col("id")).as("sum_val"),
+        count(when(pmod(col("id"), lit(2)).$eq$eq$eq(0), 1)).as("num_even")
+      )
+      .observe(
+        unnamedObservation,
+        avg(col("id")).cast("int").as("avg_val")
+      );
+
+    df.collect();
+    List<?> namedMetrics = null;
+    List<?> unnamedMetrics = null;
+
+    try {
+      namedMetrics = JavaConverters.seqAsJavaList(namedObservation.get().toSeq());
+      unnamedMetrics = JavaConverters.seqAsJavaList(unnamedObservation.get().toSeq());
+    } catch (InterruptedException e) {
+      Assert.fail();
+    }
+    Assert.assertEquals(Arrays.asList(0L, 99L, 4950L, 50L), namedMetrics);
+    Assert.assertEquals(Arrays.asList(49), unnamedMetrics);
+
+    // we can get the result multiple times
+    try {
+      namedMetrics = JavaConverters.seqAsJavaList(namedObservation.get().toSeq());
+      unnamedMetrics = JavaConverters.seqAsJavaList(unnamedObservation.get().toSeq());
+    } catch (InterruptedException e) {
+      Assert.fail();
+    }
+    Assert.assertEquals(Arrays.asList(0L, 99L, 4950L, 50L), namedMetrics);
+    Assert.assertEquals(Arrays.asList(49), unnamedMetrics);
+  }
 }

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

@@ -42,6 +42,7 @@ import org.apache.spark.sql.execution.{FilterExec, QueryExecution, WholeStageCod
 import org.apache.spark.sql.execution.adaptive.AdaptiveSparkPlanHelper
 import org.apache.spark.sql.execution.aggregate.HashAggregateExec
 import org.apache.spark.sql.execution.exchange.{BroadcastExchangeExec, ReusedExchangeExec, ShuffleExchangeExec}
+import org.apache.spark.sql.execution.streaming.MemoryStream
 import org.apache.spark.sql.expressions.{Aggregator, Window}
 import org.apache.spark.sql.functions._
 import org.apache.spark.sql.internal.SQLConf
@@ -2382,6 +2383,49 @@ class DataFrameSuite extends QueryTest
     }
   }
 
+  test("SPARK-34806: observation on datasets") {
+    val namedObservation = Observation("named")
+    val unnamedObservation = Observation()
+
+    val df = spark
+      .range(100)
+      .observe(
+        namedObservation,
+        min($"id").as("min_val"),
+        max($"id").as("max_val"),
+        sum($"id").as("sum_val"),
+        count(when($"id" % 2 === 0, 1)).as("num_even")
+      )
+      .observe(
+        unnamedObservation,
+        avg($"id").cast("int").as("avg_val")
+      )
+
+    def checkMetrics(namedMetric: Row, unnamedMetric: Row): Unit = {
+      assert(namedMetric === Row(0L, 99L, 4950L, 50L))
+      assert(unnamedMetric === Row(49))
+    }
+
+    df.collect()
+    // we can get the result multiple times
+    checkMetrics(namedObservation.get, unnamedObservation.get)
+    checkMetrics(namedObservation.get, unnamedObservation.get)
+
+    // an observation can be used only once
+    val err = intercept[IllegalArgumentException] {
+      spark.range(100).observe(namedObservation, sum($"id").as("sum_val"))
+    }
+    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 streamObservation = Observation("stream")
+    val streamErr = intercept[IllegalArgumentException] {
+      streamDf.observe(streamObservation, avg($"value").cast("int").as("avg_val"))
+    }
+    assert(streamErr.getMessage.contains("Observation does not support streaming Datasets"))
+  }
+
   test("SPARK-25159: json schema inference should only trigger one job") {
     withTempPath { path =>
       // This test is to prove that the `JsonInferSchema` does not use `RDD#toLocalIterator` which