[SPARK-20728][SQL] Make OrcFileFormat configurable between sql/hive and sql/core

sql/catalyst/src/main/scala/org/apache/spark/sql/internal/SQLConf.scala

@@ -363,6 +363,14 @@ object SQLConf {
     .checkValues(Set("none", "uncompressed", "snappy", "zlib", "lzo"))
     .createWithDefault("snappy")
 
+  val ORC_USE_NEW_VERSION = buildConf("spark.sql.orc.useNewVersion")
+    .doc("When true, use new OrcFileFormat in sql/core module instead of the one in sql/hive. " +
+      "Since new OrcFileFormat uses Apache ORC library instead of ORC library Hive 1.2.1, it is " +
+      "more stable and faster.")
+    .internal()
+    .booleanConf
+    .createWithDefault(true)
+
   val ORC_FILTER_PUSHDOWN_ENABLED = buildConf("spark.sql.orc.filterPushdown")
     .doc("When true, enable filter pushdown for ORC files.")
     .booleanConf

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

@@ -182,7 +182,7 @@ class DataFrameReader private[sql](sparkSession: SparkSession) extends Logging {
         "read files of Hive data source directly.")
     }
 
-    val cls = DataSource.lookupDataSource(source)
+    val cls = DataSource.lookupDataSource(sparkSession.sessionState.conf, source)
     if (classOf[DataSourceV2].isAssignableFrom(cls)) {
       val options = new DataSourceV2Options(extraOptions.asJava)
 

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

@@ -234,7 +234,7 @@ final class DataFrameWriter[T] private[sql](ds: Dataset[T]) {
 
     assertNotBucketed("save")
 
-    val cls = DataSource.lookupDataSource(source)
+    val cls = DataSource.lookupDataSource(df.sparkSession.sessionState.conf, source)
     if (classOf[DataSourceV2].isAssignableFrom(cls)) {
       cls.newInstance() match {
         case ds: WriteSupport =>

sql/core/src/main/scala/org/apache/spark/sql/execution/command/tables.scala

@@ -40,6 +40,7 @@ import org.apache.spark.sql.execution.datasources.{DataSource, PartitioningUtils
 import org.apache.spark.sql.execution.datasources.csv.CSVFileFormat
 import org.apache.spark.sql.execution.datasources.json.JsonFileFormat
 import org.apache.spark.sql.execution.datasources.parquet.ParquetFileFormat
+import org.apache.spark.sql.internal.SQLConf
 import org.apache.spark.sql.types._
 import org.apache.spark.sql.util.SchemaUtils
 import org.apache.spark.util.Utils
@@ -190,7 +191,7 @@ case class AlterTableAddColumnsCommand(
     colsToAdd: Seq[StructField]) extends RunnableCommand {
   override def run(sparkSession: SparkSession): Seq[Row] = {
     val catalog = sparkSession.sessionState.catalog
-    val catalogTable = verifyAlterTableAddColumn(catalog, table)
+    val catalogTable = verifyAlterTableAddColumn(sparkSession.sessionState.conf, catalog, table)
 
     try {
       sparkSession.catalog.uncacheTable(table.quotedString)
@@ -216,6 +217,7 @@ case class AlterTableAddColumnsCommand(
    * For datasource table, it currently only supports parquet, json, csv.
    */
   private def verifyAlterTableAddColumn(
+      conf: SQLConf,
       catalog: SessionCatalog,
       table: TableIdentifier): CatalogTable = {
     val catalogTable = catalog.getTempViewOrPermanentTableMetadata(table)
@@ -229,7 +231,7 @@ case class AlterTableAddColumnsCommand(
     }
 
     if (DDLUtils.isDatasourceTable(catalogTable)) {
-      DataSource.lookupDataSource(catalogTable.provider.get).newInstance() match {
+      DataSource.lookupDataSource(conf, catalogTable.provider.get).newInstance() match {
         // For datasource table, this command can only support the following File format.
         // TextFileFormat only default to one column "value"
         // Hive type is already considered as hive serde table, so the logic will not

sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/DataSource.scala

@@ -36,8 +36,10 @@ import org.apache.spark.sql.catalyst.util.CaseInsensitiveMap
 import org.apache.spark.sql.execution.datasources.csv.CSVFileFormat
 import org.apache.spark.sql.execution.datasources.jdbc.JdbcRelationProvider
 import org.apache.spark.sql.execution.datasources.json.JsonFileFormat
+import org.apache.spark.sql.execution.datasources.orc.OrcFileFormat
 import org.apache.spark.sql.execution.datasources.parquet.ParquetFileFormat
 import org.apache.spark.sql.execution.streaming._
+import org.apache.spark.sql.internal.SQLConf
 import org.apache.spark.sql.sources._
 import org.apache.spark.sql.streaming.OutputMode
 import org.apache.spark.sql.types.{CalendarIntervalType, StructType}
@@ -85,7 +87,8 @@ case class DataSource(
 
   case class SourceInfo(name: String, schema: StructType, partitionColumns: Seq[String])
 
-  lazy val providingClass: Class[_] = DataSource.lookupDataSource(className)
+  lazy val providingClass: Class[_] =
+    DataSource.lookupDataSource(sparkSession.sessionState.conf, className)
   lazy val sourceInfo: SourceInfo = sourceSchema()
   private val caseInsensitiveOptions = CaseInsensitiveMap(options)
   private val equality = sparkSession.sessionState.conf.resolver
@@ -568,8 +571,12 @@ object DataSource extends Logging {
     "org.apache.spark.Logging")
 
   /** Given a provider name, look up the data source class definition. */
-  def lookupDataSource(provider: String): Class[_] = {
-    val provider1 = backwardCompatibilityMap.getOrElse(provider, provider)
+  def lookupDataSource(conf: SQLConf, provider: String): Class[_] = {
+    var provider1 = backwardCompatibilityMap.getOrElse(provider, provider)
+    if (Seq("orc").contains(provider1.toLowerCase) && conf.getConf(SQLConf.ORC_USE_NEW_VERSION)) {
+      logInfo(s"$provider1 is replaced with ${classOf[OrcFileFormat].getCanonicalName}")
+      provider1 = classOf[OrcFileFormat].getCanonicalName
+    }
     val provider2 = s"$provider1.DefaultSource"
     val loader = Utils.getContextOrSparkClassLoader
     val serviceLoader = ServiceLoader.load(classOf[DataSourceRegister], loader)

sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/rules.scala

@@ -108,8 +108,9 @@ case class PreprocessTableCreation(sparkSession: SparkSession) extends Rule[Logi
       }
 
       // Check if the specified data source match the data source of the existing table.
-      val existingProvider = DataSource.lookupDataSource(existingTable.provider.get)
-      val specifiedProvider = DataSource.lookupDataSource(tableDesc.provider.get)
+      val conf = sparkSession.sessionState.conf
+      val existingProvider = DataSource.lookupDataSource(conf, existingTable.provider.get)
+      val specifiedProvider = DataSource.lookupDataSource(conf, tableDesc.provider.get)
       // TODO: Check that options from the resolved relation match the relation that we are
       // inserting into (i.e. using the same compression).
       if (existingProvider != specifiedProvider) {

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

@@ -28,6 +28,8 @@ import org.apache.spark.scheduler.{SparkListener, SparkListenerJobStart}
 import org.apache.spark.sql.catalyst.util.StringUtils
 import org.apache.spark.sql.execution.aggregate
 import org.apache.spark.sql.execution.aggregate.{HashAggregateExec, SortAggregateExec}
+import org.apache.spark.sql.execution.datasources.{HadoopFsRelation, LogicalRelation}
+import org.apache.spark.sql.execution.datasources.orc.OrcFileFormat
 import org.apache.spark.sql.execution.joins.{BroadcastHashJoinExec, CartesianProductExec, SortMergeJoinExec}
 import org.apache.spark.sql.functions._
 import org.apache.spark.sql.internal.SQLConf
@@ -2782,4 +2784,23 @@ class SQLQuerySuite extends QueryTest with SharedSQLContext {
       assert(spark.read.format(orc).load(path).collect().length == 2)
     }
   }
+
+  test("SPARK-20728 Make ORCFileFormat configurable between sql/hive and sql/core") {
+    withSQLConf(SQLConf.ORC_USE_NEW_VERSION.key -> "false") {
+      val e = intercept[AnalysisException] {
+        sql("CREATE TABLE spark_20728(a INT) USING ORC")
+      }
+      assert(e.message.contains("The ORC data source must be used with Hive support enabled"))
+    }
+
+    withSQLConf(SQLConf.ORC_USE_NEW_VERSION.key -> "true") {
+      withTable("spark_20728") {
+        sql("CREATE TABLE spark_20728(a INT) USING ORC")
+        val fileFormat = sql("SELECT * FROM spark_20728").queryExecution.analyzed.collectFirst {
+          case l: LogicalRelation => l.relation.asInstanceOf[HadoopFsRelation].fileFormat.getClass
+        }
+        assert(fileFormat == Some(classOf[OrcFileFormat]))
+      }
+    }
+  }
 }

sql/core/src/test/scala/org/apache/spark/sql/sources/DDLSourceLoadSuite.scala

@@ -18,6 +18,7 @@
 package org.apache.spark.sql.sources
 
 import org.apache.spark.sql.{AnalysisException, SQLContext}
+import org.apache.spark.sql.internal.SQLConf
 import org.apache.spark.sql.test.SharedSQLContext
 import org.apache.spark.sql.types._
 
@@ -54,11 +55,17 @@ class DDLSourceLoadSuite extends DataSourceTest with SharedSQLContext {
       .load().schema == StructType(Seq(StructField("stringType", StringType, nullable = false))))
   }
 
-  test("should fail to load ORC without Hive Support") {
-    val e = intercept[AnalysisException] {
-      spark.read.format("orc").load()
+  test("should fail to load ORC only if spark.sql.orc.enabled=false and without Hive Support") {
+    Seq(
+      (true, "Unable to infer schema for ORC. It must be specified manually"),
+      (false, "The ORC data source must be used with Hive support")).foreach { case (value, m) =>
+      withSQLConf(SQLConf.ORC_USE_NEW_VERSION.key -> s"$value") {
+        val e = intercept[AnalysisException] {
+          spark.read.format("orc").load()
+        }
+        assert(e.message.contains(m))
+      }
     }
-    assert(e.message.contains("The ORC data source must be used with Hive support enabled"))
   }
 }
 

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

@@ -155,7 +155,6 @@ class DataFrameReaderWriterSuite extends QueryTest with SharedSQLContext with Be
     }
   }
 
-
   test("resolve default source") {
     spark.read
       .format("org.apache.spark.sql.test")
@@ -478,42 +477,56 @@ class DataFrameReaderWriterSuite extends QueryTest with SharedSQLContext with Be
       spark.read.schema(userSchema).parquet(Seq(dir, dir): _*), expData ++ expData, userSchema)
   }
 
-  /**
-   * This only tests whether API compiles, but does not run it as orc()
-   * cannot be run without Hive classes.
-   */
-  ignore("orc - API") {
-    // Reader, with user specified schema
-    // Refer to csv-specific test suites for behavior without user specified schema
-    spark.read.schema(userSchema).orc()
-    spark.read.schema(userSchema).orc(dir)
-    spark.read.schema(userSchema).orc(dir, dir, dir)
-    spark.read.schema(userSchema).orc(Seq(dir, dir): _*)
-    Option(dir).map(spark.read.schema(userSchema).orc)
+  test("orc - API and behavior regarding schema") {
+    withSQLConf(SQLConf.ORC_USE_NEW_VERSION.key -> "true") {
+      // Writer
+      spark.createDataset(data).toDF("str").write.mode(SaveMode.Overwrite).orc(dir)
+      val df = spark.read.orc(dir)
+      checkAnswer(df, spark.createDataset(data).toDF())
+      val schema = df.schema
 
-    // Writer
-    spark.range(10).write.orc(dir)
+      // Reader, without user specified schema
+      intercept[AnalysisException] {
+        testRead(spark.read.orc(), Seq.empty, schema)
+      }
+      testRead(spark.read.orc(dir), data, schema)
+      testRead(spark.read.orc(dir, dir), data ++ data, schema)
+      testRead(spark.read.orc(Seq(dir, dir): _*), data ++ data, schema)
+      // Test explicit calls to single arg method - SPARK-16009
+      testRead(Option(dir).map(spark.read.orc).get, data, schema)
+
+      // Reader, with user specified schema, data should be nulls as schema in file different
+      // from user schema
+      val expData = Seq[String](null, null, null)
+      testRead(spark.read.schema(userSchema).orc(), Seq.empty, userSchema)
+      testRead(spark.read.schema(userSchema).orc(dir), expData, userSchema)
+      testRead(spark.read.schema(userSchema).orc(dir, dir), expData ++ expData, userSchema)
+      testRead(
+        spark.read.schema(userSchema).orc(Seq(dir, dir): _*), expData ++ expData, userSchema)
+    }
   }
 
   test("column nullability and comment - write and then read") {
-    Seq("json", "parquet", "csv").foreach { format =>
-      val schema = StructType(
-        StructField("cl1", IntegerType, nullable = false).withComment("test") ::
-          StructField("cl2", IntegerType, nullable = true) ::
-          StructField("cl3", IntegerType, nullable = true) :: Nil)
-      val row = Row(3, null, 4)
-      val df = spark.createDataFrame(sparkContext.parallelize(row :: Nil), schema)
-
-      val tableName = "tab"
-      withTable(tableName) {
-        df.write.format(format).mode("overwrite").saveAsTable(tableName)
-        // Verify the DDL command result: DESCRIBE TABLE
-        checkAnswer(
-          sql(s"desc $tableName").select("col_name", "comment").where($"comment" === "test"),
-          Row("cl1", "test") :: Nil)
-        // Verify the schema
-        val expectedFields = schema.fields.map(f => f.copy(nullable = true))
-        assert(spark.table(tableName).schema == schema.copy(fields = expectedFields))
+    withSQLConf(SQLConf.ORC_USE_NEW_VERSION.key -> "true") {
+      Seq("json", "orc", "parquet", "csv").foreach { format =>
+        val schema = StructType(
+          StructField("cl1", IntegerType, nullable = false).withComment("test") ::
+            StructField("cl2", IntegerType, nullable = true) ::
+            StructField("cl3", IntegerType, nullable = true) :: Nil)
+        val row = Row(3, null, 4)
+        val df = spark.createDataFrame(sparkContext.parallelize(row :: Nil), schema)
+
+        val tableName = "tab"
+        withTable(tableName) {
+          df.write.format(format).mode("overwrite").saveAsTable(tableName)
+          // Verify the DDL command result: DESCRIBE TABLE
+          checkAnswer(
+            sql(s"desc $tableName").select("col_name", "comment").where($"comment" === "test"),
+            Row("cl1", "test") :: Nil)
+          // Verify the schema
+          val expectedFields = schema.fields.map(f => f.copy(nullable = true))
+          assert(spark.table(tableName).schema == schema.copy(fields = expectedFields))
+        }
       }
     }
   }

sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveStrategies.scala

@@ -195,8 +195,18 @@ case class RelationConversions(
         .convertToLogicalRelation(relation, options, classOf[ParquetFileFormat], "parquet")
     } else {
       val options = relation.tableMeta.storage.properties
-      sessionCatalog.metastoreCatalog
-        .convertToLogicalRelation(relation, options, classOf[OrcFileFormat], "orc")
+      if (conf.getConf(SQLConf.ORC_USE_NEW_VERSION)) {
+        sessionCatalog.metastoreCatalog.convertToLogicalRelation(
+          relation,
+          options,
+          classOf[org.apache.spark.sql.execution.datasources.orc.OrcFileFormat],
+          "orc")
+      } else {
+        sessionCatalog.metastoreCatalog.convertToLogicalRelation(
+          relation,
+          options,
+          classOf[org.apache.spark.sql.hive.orc.OrcFileFormat], "orc")
+      }
     }
   }
 

sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/SQLQuerySuite.scala

@@ -2153,4 +2153,21 @@ class SQLQuerySuite extends QueryTest with SQLTestUtils with TestHiveSingleton {
       }
     }
   }
+
+  test("SPARK-20728 Make ORCFileFormat configurable between sql/hive and sql/core") {
+    Seq(
+      (true, classOf[org.apache.spark.sql.execution.datasources.orc.OrcFileFormat]),
+      (false, classOf[org.apache.spark.sql.hive.orc.OrcFileFormat])).foreach { case (v, format) =>
+
+      withSQLConf(SQLConf.ORC_USE_NEW_VERSION.key -> s"$v") {
+        withTable("spark_20728") {
+          sql("CREATE TABLE spark_20728(a INT) USING ORC")
+          val fileFormat = sql("SELECT * FROM spark_20728").queryExecution.analyzed.collectFirst {
+            case l: LogicalRelation => l.relation.asInstanceOf[HadoopFsRelation].fileFormat.getClass
+          }
+          assert(fileFormat == Some(format))
+        }
+      }
+    }
+  }
 }