[SPARK-27402][SQL][test-hadoop3.2][test-maven] Fix hadoop-3.2 test issue(except the hive-thriftserver module)

dev/sparktestsupport/modules.py

@@ -15,9 +15,17 @@
 # limitations under the License.
 #
 
+from __future__ import print_function
 from functools import total_ordering
 import itertools
 import re
+import os
+
+if os.environ.get("AMPLAB_JENKINS"):
+    hadoop_version = os.environ.get("AMPLAB_JENKINS_BUILD_PROFILE", "hadoop2.7")
+else:
+    hadoop_version = os.environ.get("HADOOP_PROFILE", "hadoop2.7")
+print("[info] Choosing supported modules with Hadoop profile", hadoop_version)
 
 all_modules = []
 
@@ -72,7 +80,11 @@ def __init__(self, name, dependencies, source_file_regexes, build_profile_flags=
         self.dependent_modules = set()
         for dep in dependencies:
             dep.dependent_modules.add(self)
-        all_modules.append(self)
+        # TODO: Skip hive-thriftserver module for hadoop-3.2. remove this once hadoop-3.2 support it
+        if name == "hive-thriftserver" and hadoop_version == "hadoop3.2":
+            print("[info] Skip unsupported module:", name)
+        else:
+            all_modules.append(self)
 
     def contains_file(self, filename):
         return any(re.match(p, filename) for p in self.source_file_prefixes)

sql/core/src/test/scala/org/apache/spark/sql/execution/command/DDLSuite.scala

@@ -1371,7 +1371,7 @@ abstract class DDLSuite extends QueryTest with SQLTestUtils {
       // if (isUsingHiveMetastore) {
       //  assert(storageFormat.properties.get("path") === expected)
       // }
-      assert(storageFormat.locationUri === Some(expected))
+      assert(storageFormat.locationUri.map(_.getPath) === Some(expected.getPath))
     }
     // set table location
     sql("ALTER TABLE dbx.tab1 SET LOCATION '/path/to/your/lovely/heart'")

sql/core/src/test/scala/org/apache/spark/sql/execution/datasources/orc/OrcSourceSuite.scala

@@ -118,7 +118,7 @@ abstract class OrcSuite extends OrcTest with BeforeAndAfterAll {
     }
   }
 
-  protected def testSelectiveDictionaryEncoding(isSelective: Boolean) {
+  protected def testSelectiveDictionaryEncoding(isSelective: Boolean, isHive23: Boolean = false) {
     val tableName = "orcTable"
 
     withTempDir { dir =>
@@ -171,7 +171,7 @@ abstract class OrcSuite extends OrcTest with BeforeAndAfterAll {
           // Hive 0.11 and RLE v2 is introduced in Hive 0.12 ORC with more improvements.
           // For more details, see https://orc.apache.org/specification/
           assert(stripe.getColumns(1).getKind === DICTIONARY_V2)
-          if (isSelective) {
+          if (isSelective || isHive23) {
             assert(stripe.getColumns(2).getKind === DIRECT_V2)
           } else {
             assert(stripe.getColumns(2).getKind === DICTIONARY_V2)

sql/hive/src/main/scala/org/apache/spark/sql/hive/client/HiveClientImpl.scala

@@ -52,6 +52,7 @@ import org.apache.spark.sql.catalyst.parser.{CatalystSqlParser, ParseException}
 import org.apache.spark.sql.execution.QueryExecutionException
 import org.apache.spark.sql.execution.command.DDLUtils
 import org.apache.spark.sql.hive.HiveExternalCatalog.{DATASOURCE_SCHEMA, DATASOURCE_SCHEMA_NUMPARTS, DATASOURCE_SCHEMA_PART_PREFIX}
+import org.apache.spark.sql.hive.HiveUtils
 import org.apache.spark.sql.hive.client.HiveClientImpl._
 import org.apache.spark.sql.types._
 import org.apache.spark.util.{CircularBuffer, Utils}
@@ -191,7 +192,29 @@ private[hive] class HiveClientImpl(
   }
 
   /** Returns the configuration for the current session. */
-  def conf: HiveConf = state.getConf
+  def conf: HiveConf = if (!HiveUtils.isHive23) {
+    state.getConf
+  } else {
+    val hiveConf = state.getConf
+    // Hive changed the default of datanucleus.schema.autoCreateAll from true to false
+    // and hive.metastore.schema.verification from false to true since Hive 2.0.
+    // For details, see the JIRA HIVE-6113, HIVE-12463 and HIVE-1841.
+    // isEmbeddedMetaStore should not be true in the production environment.
+    // We hard-code hive.metastore.schema.verification and datanucleus.schema.autoCreateAll to allow
+    // bin/spark-shell, bin/spark-sql and sbin/start-thriftserver.sh to automatically create the
+    // Derby Metastore when running Spark in the non-production environment.
+    val isEmbeddedMetaStore = {
+      val msUri = hiveConf.getVar(ConfVars.METASTOREURIS)
+      val msConnUrl = hiveConf.getVar(ConfVars.METASTORECONNECTURLKEY)
+      (msUri == null || msUri.trim().isEmpty) &&
+        (msConnUrl != null && msConnUrl.startsWith("jdbc:derby"))
+    }
+    if (isEmbeddedMetaStore) {
+      hiveConf.setBoolean("hive.metastore.schema.verification", false)
+      hiveConf.setBoolean("datanucleus.schema.autoCreateAll", true)
+    }
+    hiveConf
+  }
 
   private val userName = conf.getUser
 

sql/hive/src/main/scala/org/apache/spark/sql/hive/client/IsolatedClientLoader.scala

@@ -28,6 +28,7 @@ import org.apache.commons.io.{FileUtils, IOUtils}
 import org.apache.commons.lang3.{JavaVersion, SystemUtils}
 import org.apache.hadoop.conf.Configuration
 import org.apache.hadoop.hive.conf.HiveConf.ConfVars
+import org.apache.hadoop.hive.shims.ShimLoader
 
 import org.apache.spark.SparkConf
 import org.apache.spark.deploy.SparkSubmitUtils
@@ -196,6 +197,7 @@ private[hive] class IsolatedClientLoader(
   protected def isBarrierClass(name: String): Boolean =
     name.startsWith(classOf[HiveClientImpl].getName) ||
     name.startsWith(classOf[Shim].getName) ||
+    name.startsWith(classOf[ShimLoader].getName) ||
     barrierPrefixes.exists(name.startsWith)
 
   protected def classToPath(name: String): String =

sql/hive/src/main/scala/org/apache/spark/sql/hive/test/TestHive.scala

@@ -63,6 +63,9 @@ object TestHive
         // SPARK-8910
         .set(UI_ENABLED, false)
         .set(config.UNSAFE_EXCEPTION_ON_MEMORY_LEAK, true)
+        // Hive changed the default of hive.metastore.disallow.incompatible.col.type.changes
+        // from false to true. For details, see the JIRA HIVE-12320 and HIVE-17764.
+        .set("spark.hadoop.hive.metastore.disallow.incompatible.col.type.changes", "false")
         // Disable ConvertToLocalRelation for better test coverage. Test cases built on
         // LocalRelation will exercise the optimization rules better by disabling it as
         // this rule may potentially block testing of other optimization rules such as
@@ -116,8 +119,10 @@ class TestHiveContext(
     @transient override val sparkSession: TestHiveSparkSession)
   extends SQLContext(sparkSession) {
 
-  val HIVE_CONTRIB_JAR: String = "hive-contrib-0.13.1.jar"
-  val HIVE_HCATALOG_CORE_JAR: String = "hive-hcatalog-core-0.13.1.jar"
+  val HIVE_CONTRIB_JAR: String =
+    if (HiveUtils.isHive23) "hive-contrib-2.3.4.jar" else "hive-contrib-0.13.1.jar"
+  val HIVE_HCATALOG_CORE_JAR: String =
+    if (HiveUtils.isHive23) "hive-hcatalog-core-2.3.4.jar" else "hive-hcatalog-core-0.13.1.jar"
 
   /**
    * If loadTestTables is false, no test tables are loaded. Note that this flag can only be true

sql/hive/src/test/scala/org/apache/spark/sql/hive/ClasspathDependenciesSuite.scala

@@ -58,11 +58,19 @@ class ClasspathDependenciesSuite extends SparkFunSuite {
   }
 
   test("shaded Protobuf") {
-    assertLoads("org.apache.hive.com.google.protobuf.ServiceException")
+    if (HiveUtils.isHive23) {
+      assertLoads("com.google.protobuf.ServiceException")
+    } else {
+      assertLoads("org.apache.hive.com.google.protobuf.ServiceException")
+    }
   }
 
   test("shaded Kryo") {
-    assertLoads("org.apache.hive.com.esotericsoftware.kryo.Kryo")
+    if (HiveUtils.isHive23) {
+      assertLoads("com.esotericsoftware.kryo.Kryo")
+    } else {
+      assertLoads("org.apache.hive.com.esotericsoftware.kryo.Kryo")
+    }
   }
 
   test("hive-common") {
@@ -81,7 +89,12 @@ class ClasspathDependenciesSuite extends SparkFunSuite {
   }
 
   test("parquet-hadoop-bundle") {
-    assertLoads("parquet.hadoop.ParquetOutputFormat")
-    assertLoads("parquet.hadoop.ParquetInputFormat")
+    if (HiveUtils.isHive23) {
+      assertLoads("org.apache.parquet.hadoop.ParquetOutputFormat")
+      assertLoads("org.apache.parquet.hadoop.ParquetInputFormat")
+    } else {
+      assertLoads("parquet.hadoop.ParquetOutputFormat")
+      assertLoads("parquet.hadoop.ParquetInputFormat")
+    }
   }
 }

sql/hive/src/test/scala/org/apache/spark/sql/hive/HiveExternalCatalogVersionsSuite.scala

@@ -180,6 +180,8 @@ class HiveExternalCatalogVersionsSuite extends SparkSubmitTestUtils {
         "--master", "local[2]",
         "--conf", "spark.ui.enabled=false",
         "--conf", "spark.master.rest.enabled=false",
+        "--conf", "spark.sql.hive.metastore.version=1.2.1",
+        "--conf", "spark.sql.hive.metastore.jars=maven",
         "--conf", s"spark.sql.warehouse.dir=${wareHousePath.getCanonicalPath}",
         "--conf", s"spark.sql.test.version.index=$index",
         "--driver-java-options", s"-Dderby.system.home=${wareHousePath.getCanonicalPath}",
@@ -197,6 +199,8 @@ class HiveExternalCatalogVersionsSuite extends SparkSubmitTestUtils {
       "--master", "local[2]",
       "--conf", "spark.ui.enabled=false",
       "--conf", "spark.master.rest.enabled=false",
+      "--conf", "spark.sql.hive.metastore.version=1.2.1",
+      "--conf", "spark.sql.hive.metastore.jars=maven",
       "--conf", s"spark.sql.warehouse.dir=${wareHousePath.getCanonicalPath}",
       "--driver-java-options", s"-Dderby.system.home=${wareHousePath.getCanonicalPath}",
       unusedJar.toString)

sql/hive/src/test/scala/org/apache/spark/sql/hive/HiveMetastoreCatalogSuite.scala

@@ -206,7 +206,13 @@ class DataSourceWithHiveMetastoreCatalogSuite
         assert(columns.map(_.dataType) === Seq(DecimalType(10, 3), StringType))
 
         checkAnswer(table("t"), testDF)
-        assert(sparkSession.metadataHive.runSqlHive("SELECT * FROM t") === Seq("1.1\t1", "2.1\t2"))
+        if (HiveUtils.isHive23) {
+          assert(sparkSession.metadataHive.runSqlHive("SELECT * FROM t") ===
+            Seq("1.100\t1", "2.100\t2"))
+        } else {
+          assert(sparkSession.metadataHive.runSqlHive("SELECT * FROM t") ===
+            Seq("1.1\t1", "2.1\t2"))
+        }
       }
     }
 
@@ -238,8 +244,13 @@ class DataSourceWithHiveMetastoreCatalogSuite
           assert(columns.map(_.dataType) === Seq(DecimalType(10, 3), StringType))
 
           checkAnswer(table("t"), testDF)
-          assert(sparkSession.metadataHive.runSqlHive("SELECT * FROM t") ===
-            Seq("1.1\t1", "2.1\t2"))
+          if (HiveUtils.isHive23) {
+            assert(sparkSession.metadataHive.runSqlHive("SELECT * FROM t") ===
+              Seq("1.100\t1", "2.100\t2"))
+          } else {
+            assert(sparkSession.metadataHive.runSqlHive("SELECT * FROM t") ===
+              Seq("1.1\t1", "2.1\t2"))
+          }
         }
       }
     }

sql/hive/src/test/scala/org/apache/spark/sql/hive/HiveShimSuite.scala

@@ -35,10 +35,18 @@ class HiveShimSuite extends SparkFunSuite {
 
     // test when READ_COLUMN_NAMES_CONF_STR is empty
     HiveShim.appendReadColumns(conf, ids, names)
-    assert(names.asJava === ColumnProjectionUtils.getReadColumnNames(conf))
+    if (HiveUtils.isHive23) {
+      assert(names === ColumnProjectionUtils.getReadColumnNames(conf))
+    } else {
+      assert(names.asJava === ColumnProjectionUtils.getReadColumnNames(conf))
+    }
 
     // test when READ_COLUMN_NAMES_CONF_STR is non-empty
     HiveShim.appendReadColumns(conf, moreIds, moreNames)
-    assert((names ++ moreNames).asJava === ColumnProjectionUtils.getReadColumnNames(conf))
+    if (HiveUtils.isHive23) {
+      assert((names ++ moreNames) === ColumnProjectionUtils.getReadColumnNames(conf))
+    } else {
+      assert((names ++ moreNames).asJava === ColumnProjectionUtils.getReadColumnNames(conf))
+    }
   }
 }

sql/hive/src/test/scala/org/apache/spark/sql/hive/StatisticsSuite.scala

@@ -100,8 +100,14 @@ class StatisticsSuite extends StatisticsCollectionTestBase with TestHiveSingleto
             .asInstanceOf[HiveTableRelation]
 
           val properties = relation.tableMeta.ignoredProperties
-          assert(properties("totalSize").toLong <= 0, "external table totalSize must be <= 0")
-          assert(properties("rawDataSize").toLong <= 0, "external table rawDataSize must be <= 0")
+          if (HiveUtils.isHive23) {
+            // Since HIVE-6727, Hive fixes table-level stats for external tables are incorrect.
+            assert(properties("totalSize").toLong == 6)
+            assert(properties.get("rawDataSize").isEmpty)
+          } else {
+            assert(properties("totalSize").toLong <= 0, "external table totalSize must be <= 0")
+            assert(properties("rawDataSize").toLong <= 0, "external table rawDataSize must be <= 0")
+          }
 
           val sizeInBytes = relation.stats.sizeInBytes
           assert(sizeInBytes === BigInt(file1.length() + file2.length()))
@@ -865,17 +871,25 @@ class StatisticsSuite extends StatisticsCollectionTestBase with TestHiveSingleto
         val totalSize = extractStatsPropValues(describeResult, "totalSize")
         assert(totalSize.isDefined && totalSize.get > 0, "totalSize is lost")
 
-        // ALTER TABLE SET/UNSET TBLPROPERTIES invalidates some Hive specific statistics, but not
-        // Spark specific statistics. This is triggered by the Hive alterTable API.
         val numRows = extractStatsPropValues(describeResult, "numRows")
-        assert(numRows.isDefined && numRows.get == -1, "numRows is lost")
-        val rawDataSize = extractStatsPropValues(describeResult, "rawDataSize")
-        assert(rawDataSize.isDefined && rawDataSize.get == -1, "rawDataSize is lost")
-
-        if (analyzedBySpark) {
+        if (HiveUtils.isHive23) {
+          // Since HIVE-15653(Hive 2.3.0), Hive fixs some ALTER TABLE commands drop table stats.
+          assert(numRows.isDefined && numRows.get == 500)
+          val rawDataSize = extractStatsPropValues(describeResult, "rawDataSize")
+          assert(rawDataSize.isDefined && rawDataSize.get == 5312)
           checkTableStats(tabName, hasSizeInBytes = true, expectedRowCounts = Some(500))
         } else {
-          checkTableStats(tabName, hasSizeInBytes = true, expectedRowCounts = None)
+          // ALTER TABLE SET/UNSET TBLPROPERTIES invalidates some Hive specific statistics, but not
+          // Spark specific statistics. This is triggered by the Hive alterTable API.
+          assert(numRows.isDefined && numRows.get == -1, "numRows is lost")
+          val rawDataSize = extractStatsPropValues(describeResult, "rawDataSize")
+          assert(rawDataSize.isDefined && rawDataSize.get == -1, "rawDataSize is lost")
+
+          if (analyzedBySpark) {
+            checkTableStats(tabName, hasSizeInBytes = true, expectedRowCounts = Some(500))
+          } else {
+            checkTableStats(tabName, hasSizeInBytes = true, expectedRowCounts = None)
+          }
         }
       }
     }