address comments.

Author: gatorsmile

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

@@ -83,12 +83,23 @@ class SQLBuilder(logicalPlan: LogicalPlan, sqlContext: SQLContext) extends Loggi
 
     // TABLESAMPLE is part of tableSource clause in the parser,
     // and thus we must handle it with subquery.
-    case Sample(lb, ub, withReplacement, _, child @ Subquery(alias, grandChild))
-      if !withReplacement && lb <= (ub + RandomSampler.roundingEpsilon) =>
+    case p @ Sample(lb, ub, withReplacement, _, _)
+        if !withReplacement && lb <= (ub + RandomSampler.roundingEpsilon) =>
       val fraction = math.min(100, math.max(0, (ub - lb) * 100))
-      val aliasName = if (grandChild.isInstanceOf[Subquery]) alias else ""
-      val plan = if (grandChild.isInstanceOf[Subquery]) grandChild else child
-      s"${toSQL(plan)} TABLESAMPLE($fraction PERCENT) $aliasName"
+      p.child match {
+        case m: MetastoreRelation =>
+          val aliasName = m.alias.getOrElse("")
+          build(
+            s"`${m.databaseName}`.`${m.tableName}`",
+            "TABLESAMPLE(" + fraction + " PERCENT)",
+            aliasName)
+        case s: Subquery =>
+          val aliasName = if (s.child.isInstanceOf[Subquery]) s.alias else ""
+          val plan = if (s.child.isInstanceOf[Subquery]) s.child else s
+          build(toSQL(plan), "TABLESAMPLE(" + fraction + " PERCENT)", aliasName)
+        case _ =>
+          build(toSQL(p.child), "TABLESAMPLE(" + fraction + " PERCENT)")
+      }
 
     case p: Filter =>
       val whereOrHaving = p.child match {

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

@@ -26,24 +26,32 @@ class LogicalPlanToSQLSuite extends SQLBuilderTest with SQLTestUtils {
   import testImplicits._
 
   protected override def beforeAll(): Unit = {
+    sql("DROP TABLE IF EXISTS parquet_t0")
+    sql("DROP TABLE IF EXISTS parquet_t1")
+    sql("DROP TABLE IF EXISTS parquet_t2")
     sql("DROP TABLE IF EXISTS t0")
-    sql("DROP TABLE IF EXISTS t1")
-    sql("DROP TABLE IF EXISTS t2")
-    sqlContext.range(10).write.saveAsTable("t0")
+
+    sqlContext.range(10).write.saveAsTable("parquet_t0")
+    sql("CREATE TABLE t0 AS SELECT * FROM parquet_t0")
 
     sqlContext
       .range(10)
       .select('id as 'key, concat(lit("val_"), 'id) as 'value)
       .write
-      .saveAsTable("t1")
+      .saveAsTable("parquet_t1")
 
-    sqlContext.range(10).select('id as 'a, 'id as 'b, 'id as 'c, 'id as 'd).write.saveAsTable("t2")
+    sqlContext
+      .range(10)
+      .select('id as 'a, 'id as 'b, 'id as 'c, 'id as 'd)
+      .write
+      .saveAsTable("parquet_t2")
   }
 
   override protected def afterAll(): Unit = {
+    sql("DROP TABLE IF EXISTS parquet_t0")
+    sql("DROP TABLE IF EXISTS parquet_t1")
+    sql("DROP TABLE IF EXISTS parquet_t2")
     sql("DROP TABLE IF EXISTS t0")
-    sql("DROP TABLE IF EXISTS t1")
-    sql("DROP TABLE IF EXISTS t2")
   }
 
   private def checkHiveQl(hiveQl: String): Unit = {
@@ -82,96 +90,123 @@ class LogicalPlanToSQLSuite extends SQLBuilderTest with SQLTestUtils {
   }
 
   test("in") {
-    checkHiveQl("SELECT id FROM t0 WHERE id IN (1, 2, 3)")
+    checkHiveQl("SELECT id FROM parquet_t0 WHERE id IN (1, 2, 3)")
   }
 
   test("aggregate function in having clause") {
-    checkHiveQl("SELECT COUNT(value) FROM t1 GROUP BY key HAVING MAX(key) > 0")
+    checkHiveQl("SELECT COUNT(value) FROM parquet_t1 GROUP BY key HAVING MAX(key) > 0")
   }
 
   test("aggregate function in order by clause") {
-    checkHiveQl("SELECT COUNT(value) FROM t1 GROUP BY key ORDER BY MAX(key)")
+    checkHiveQl("SELECT COUNT(value) FROM parquet_t1 GROUP BY key ORDER BY MAX(key)")
   }
 
   // When there are multiple aggregate functions in ORDER BY clause, all of them are extracted into
   // Aggregate operator and aliased to the same name "aggOrder".  This is OK for normal query
   // execution since these aliases have different expression ID.  But this introduces name collision
   // when converting resolved plans back to SQL query strings as expression IDs are stripped.
   test("aggregate function in order by clause with multiple order keys") {
-    checkHiveQl("SELECT COUNT(value) FROM t1 GROUP BY key ORDER BY key, MAX(key)")
+    checkHiveQl("SELECT COUNT(value) FROM parquet_t1 GROUP BY key ORDER BY key, MAX(key)")
   }
 
   test("type widening in union") {
-    checkHiveQl("SELECT id FROM t0 UNION ALL SELECT CAST(id AS INT) AS id FROM t0")
+    checkHiveQl("SELECT id FROM parquet_t0 UNION ALL SELECT CAST(id AS INT) AS id FROM parquet_t0")
   }
 
   test("self join") {
-    checkHiveQl("SELECT x.key FROM t1 x JOIN t1 y ON x.key = y.key")
+    checkHiveQl("SELECT x.key FROM parquet_t1 x JOIN parquet_t1 y ON x.key = y.key")
   }
 
   test("self join with group by") {
-    checkHiveQl("SELECT x.key, COUNT(*) FROM t1 x JOIN t1 y ON x.key = y.key group by x.key")
+    checkHiveQl(
+      "SELECT x.key, COUNT(*) FROM parquet_t1 x JOIN parquet_t1 y ON x.key = y.key group by x.key")
   }
 
   test("three-child union") {
-    checkHiveQl("SELECT id FROM t0 UNION ALL SELECT id FROM t0 UNION ALL SELECT id FROM t0")
+    checkHiveQl(
+      """
+        |SELECT id FROM parquet_t0
+        |UNION ALL SELECT id FROM parquet_t0
+        |UNION ALL SELECT id FROM parquet_t0
+      """.stripMargin)
   }
 
   test("case") {
-    checkHiveQl("SELECT CASE WHEN id % 2 > 0 THEN 0 WHEN id % 2 = 0 THEN 1 END FROM t0")
+    checkHiveQl("SELECT CASE WHEN id % 2 > 0 THEN 0 WHEN id % 2 = 0 THEN 1 END FROM parquet_t0")
   }
 
   test("case with else") {
-    checkHiveQl("SELECT CASE WHEN id % 2 > 0 THEN 0 ELSE 1 END FROM t0")
+    checkHiveQl("SELECT CASE WHEN id % 2 > 0 THEN 0 ELSE 1 END FROM parquet_t0")
   }
 
   test("case with key") {
-    checkHiveQl("SELECT CASE id WHEN 0 THEN 'foo' WHEN 1 THEN 'bar' END FROM t0")
+    checkHiveQl("SELECT CASE id WHEN 0 THEN 'foo' WHEN 1 THEN 'bar' END FROM parquet_t0")
   }
 
   test("case with key and else") {
-    checkHiveQl("SELECT CASE id WHEN 0 THEN 'foo' WHEN 1 THEN 'bar' ELSE 'baz' END FROM t0")
+    checkHiveQl("SELECT CASE id WHEN 0 THEN 'foo' WHEN 1 THEN 'bar' ELSE 'baz' END FROM parquet_t0")
   }
 
   test("select distinct without aggregate functions") {
-    checkHiveQl("SELECT DISTINCT id FROM t0")
+    checkHiveQl("SELECT DISTINCT id FROM parquet_t0")
   }
 
   test("cluster by") {
-    checkHiveQl("SELECT id FROM t0 CLUSTER BY id")
+    checkHiveQl("SELECT id FROM parquet_t0 CLUSTER BY id")
   }
 
   test("distribute by") {
-    checkHiveQl("SELECT id FROM t0 DISTRIBUTE BY id")
+    checkHiveQl("SELECT id FROM parquet_t0 DISTRIBUTE BY id")
   }
 
   test("distribute by with sort by") {
-    checkHiveQl("SELECT id FROM t0 DISTRIBUTE BY id SORT BY id")
+    checkHiveQl("SELECT id FROM parquet_t0 DISTRIBUTE BY id SORT BY id")
   }
 
   test("distinct aggregation") {
-    checkHiveQl("SELECT COUNT(DISTINCT id) FROM t0")
+    checkHiveQl("SELECT COUNT(DISTINCT id) FROM parquet_t0")
   }
 
   test("TABLESAMPLE") {
-    checkHiveQl("SELECT * FROM t0 TABLESAMPLE(100 PERCENT) s")
+    // Project [id#2L]
+    // +- Sample 0.0, 1.0, false, ...
+    //    +- Subquery s
+    //       +- Subquery parquet_t0
+    //          +- Relation[id#2L] ParquetRelation
+    checkHiveQl("SELECT s.id FROM parquet_t0 TABLESAMPLE(100 PERCENT) s")
+
+    // Project [id#2L]
+    // +- Sample 0.0, 1.0, false, ...
+    //    +- Subquery parquet_t0
+    //       +- Relation[id#2L] ParquetRelation
+    checkHiveQl("SELECT * FROM parquet_t0 TABLESAMPLE(100 PERCENT)")
+
+    // Project [id#21L]
+    // +- Sample 0.0, 1.0, false, ...
+    //    +- MetastoreRelation default, t0, Some(s)
+    checkHiveQl("SELECT s.id FROM t0 TABLESAMPLE(100 PERCENT) s")
+
+    // Project [id#24L]
+    // +- Sample 0.0, 1.0, false, ...
+    //    +- MetastoreRelation default, t0, None
     checkHiveQl("SELECT * FROM t0 TABLESAMPLE(100 PERCENT)")
+
     // When a sampling fraction is not 100%, the returned results are random.
     // Thus, added an always-false filter here to check if the generated plan can be successfully
     // executed.
-    checkHiveQl("SELECT s.id FROM t0 TABLESAMPLE(0.1 PERCENT) s WHERE 1=0")
-    checkHiveQl("SELECT * FROM t0 TABLESAMPLE(0.1 PERCENT) WHERE 1=0")
+    checkHiveQl("SELECT s.id FROM parquet_t0 TABLESAMPLE(0.1 PERCENT) s WHERE 1=0")
+    checkHiveQl("SELECT * FROM parquet_t0 TABLESAMPLE(0.1 PERCENT) WHERE 1=0")
   }
 
   // TODO Enable this
   // Query plans transformed by DistinctAggregationRewriter are not recognized yet
   ignore("multi-distinct columns") {
-    checkHiveQl("SELECT a, COUNT(DISTINCT b), COUNT(DISTINCT c), SUM(d) FROM t2 GROUP BY a")
+    checkHiveQl("SELECT a, COUNT(DISTINCT b), COUNT(DISTINCT c), SUM(d) FROM parquet_t2 GROUP BY a")
   }
 
   test("persisted data source relations") {
     Seq("orc", "json", "parquet").foreach { format =>
-      val tableName = s"${format}_t0"
+      val tableName = s"${format}_parquet_t0"
       withTable(tableName) {
         sqlContext.range(10).write.format(format).saveAsTable(tableName)
         checkHiveQl(s"SELECT id FROM $tableName")