[SPARK-23518][SQL] Avoid metastore access when the users only want to read and write data frames

## What changes were proposed in this pull request?

#18944 added one patch, which allowed a spark session to be created when the hive metastore server is down. However, it did not allow running any commands with the spark session. This brings troubles to the user who only wants to read / write data frames without metastore setup.

## How was this patch tested?

Added some unit tests to read and write data frames based on the original HiveMetastoreLazyInitializationSuite.

Please review http://spark.apache.org/contributing.html before opening a pull request.

Author: Feng Liu <[email protected]>

Closes #20681 from liufengdb/completely-lazy.

Author: Feng Liu

R/pkg/tests/fulltests/test_sparkSQL.R

@@ -67,6 +67,8 @@ sparkSession <- if (windows_with_hadoop()) {
     sparkR.session(master = sparkRTestMaster, enableHiveSupport = FALSE)
   }
 sc <- callJStatic("org.apache.spark.sql.api.r.SQLUtils", "getJavaSparkContext", sparkSession)
+# materialize the catalog implementation
+listTables()
 
 mockLines <- c("{\"name\":\"Michael\"}",
                "{\"name\":\"Andy\", \"age\":30}",

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/catalog/SessionCatalog.scala

@@ -54,8 +54,8 @@ object SessionCatalog {
  * This class must be thread-safe.
  */
 class SessionCatalog(
-    val externalCatalog: ExternalCatalog,
-    globalTempViewManager: GlobalTempViewManager,
+    externalCatalogBuilder: () => ExternalCatalog,
+    globalTempViewManagerBuilder: () => GlobalTempViewManager,
     functionRegistry: FunctionRegistry,
     conf: SQLConf,
     hadoopConf: Configuration,
@@ -70,8 +70,8 @@ class SessionCatalog(
       functionRegistry: FunctionRegistry,
       conf: SQLConf) {
     this(
-      externalCatalog,
-      new GlobalTempViewManager("global_temp"),
+      () => externalCatalog,
+      () => new GlobalTempViewManager("global_temp"),
       functionRegistry,
       conf,
       new Configuration(),
@@ -87,6 +87,9 @@ class SessionCatalog(
       new SQLConf().copy(SQLConf.CASE_SENSITIVE -> true))
   }
 
+  lazy val externalCatalog = externalCatalogBuilder()
+  lazy val globalTempViewManager = globalTempViewManagerBuilder()
+
   /** List of temporary views, mapping from table name to their logical plan. */
   @GuardedBy("this")
   protected val tempViews = new mutable.HashMap[String, LogicalPlan]

sql/core/src/main/scala/org/apache/spark/sql/internal/BaseSessionStateBuilder.scala

@@ -130,8 +130,8 @@ abstract class BaseSessionStateBuilder(
    */
   protected lazy val catalog: SessionCatalog = {
     val catalog = new SessionCatalog(
-      session.sharedState.externalCatalog,
-      session.sharedState.globalTempViewManager,
+      () => session.sharedState.externalCatalog,
+      () => session.sharedState.globalTempViewManager,
       functionRegistry,
       conf,
       SessionState.newHadoopConf(session.sparkContext.hadoopConfiguration, conf),

sql/hive-thriftserver/src/test/scala/org/apache/spark/sql/hive/HiveMetastoreLazyInitializationSuite.scala

@@ -38,6 +38,20 @@ class HiveMetastoreLazyInitializationSuite extends SparkFunSuite {
       // We should be able to run Spark jobs without Hive client.
       assert(spark.sparkContext.range(0, 1).count() === 1)
 
+      // We should be able to use Spark SQL if no table references.
+      assert(spark.sql("select 1 + 1").count() === 1)
+      assert(spark.range(0, 1).count() === 1)
+
+      // We should be able to use fs
+      val path = Utils.createTempDir()
+      path.delete()
+      try {
+        spark.range(0, 1).write.parquet(path.getAbsolutePath)
+        assert(spark.read.parquet(path.getAbsolutePath).count() === 1)
+      } finally {
+        Utils.deleteRecursively(path)
+      }
+
       // Make sure that we are not using the local derby metastore.
       val exceptionString = Utils.exceptionString(intercept[AnalysisException] {
         spark.sql("show tables")

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

@@ -39,17 +39,17 @@ import org.apache.spark.sql.types.{DecimalType, DoubleType}
 
 
 private[sql] class HiveSessionCatalog(
-    externalCatalog: HiveExternalCatalog,
-    globalTempViewManager: GlobalTempViewManager,
+    externalCatalogBuilder: () => HiveExternalCatalog,
+    globalTempViewManagerBuilder: () => GlobalTempViewManager,
     val metastoreCatalog: HiveMetastoreCatalog,
     functionRegistry: FunctionRegistry,
     conf: SQLConf,
     hadoopConf: Configuration,
     parser: ParserInterface,
     functionResourceLoader: FunctionResourceLoader)
   extends SessionCatalog(
-      externalCatalog,
-      globalTempViewManager,
+      externalCatalogBuilder,
+      globalTempViewManagerBuilder,
       functionRegistry,
       conf,
       hadoopConf,

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

@@ -42,17 +42,16 @@ class HiveSessionStateBuilder(session: SparkSession, parentState: Option[Session
    * Create a Hive aware resource loader.
    */
   override protected lazy val resourceLoader: HiveSessionResourceLoader = {
-    val client: HiveClient = externalCatalog.client
-    new HiveSessionResourceLoader(session, client)
+    new HiveSessionResourceLoader(session, () => externalCatalog.client)
   }
 
   /**
    * Create a [[HiveSessionCatalog]].
    */
   override protected lazy val catalog: HiveSessionCatalog = {
     val catalog = new HiveSessionCatalog(
-      externalCatalog,
-      session.sharedState.globalTempViewManager,
+      () => externalCatalog,
+      () => session.sharedState.globalTempViewManager,
       new HiveMetastoreCatalog(session),
       functionRegistry,
       conf,
@@ -105,8 +104,9 @@ class HiveSessionStateBuilder(session: SparkSession, parentState: Option[Session
 
 class HiveSessionResourceLoader(
     session: SparkSession,
-    client: HiveClient)
+    clientBuilder: () => HiveClient)
   extends SessionResourceLoader(session) {
+  private lazy val client = clientBuilder()
   override def addJar(path: String): Unit = {
     client.addJar(path)
     super.addJar(path)