[HUDI-3236] use fields'comments persisted in catalog to fill in schema (#4587)

hudi-spark-datasource/hudi-spark-common/src/main/scala/org/apache/spark/sql/catalyst/catalog/HoodieCatalogTable.scala

@@ -17,6 +17,7 @@
 
 package org.apache.spark.sql.catalyst.catalog
 
+import org.apache.hudi.AvroConversionUtils
 import org.apache.hudi.HoodieWriterUtils._
 import org.apache.hudi.common.config.DFSPropertiesConfiguration
 import org.apache.hudi.common.model.HoodieTableType
@@ -30,7 +31,7 @@ import org.apache.spark.internal.Logging
 import org.apache.spark.sql.{AnalysisException, SparkSession}
 import org.apache.spark.sql.avro.SchemaConverters
 import org.apache.spark.sql.catalyst.TableIdentifier
-import org.apache.spark.sql.hudi.{HoodieOptionConfig, HoodieSqlCommonUtils}
+import org.apache.spark.sql.hudi.HoodieOptionConfig
 import org.apache.spark.sql.hudi.HoodieSqlCommonUtils._
 import org.apache.spark.sql.types.{StructField, StructType}
 
@@ -62,7 +63,7 @@ class HoodieCatalogTable(val spark: SparkSession, val table: CatalogTable) exten
    * hoodie table's location.
    * if create managed hoodie table, use `catalog.defaultTablePath`.
    */
-  val tableLocation: String = HoodieSqlCommonUtils.getTableLocation(table, spark)
+  val tableLocation: String = getTableLocation(table, spark)
 
   /**
    * A flag to whether the hoodie table exists.
@@ -114,17 +115,27 @@ class HoodieCatalogTable(val spark: SparkSession, val table: CatalogTable) exten
 
   /**
    * The schema of table.
-   * Make StructField nullable.
+   * Make StructField nullable and fill the comments in.
    */
   lazy val tableSchema: StructType = {
+    val resolver = spark.sessionState.conf.resolver
     val originSchema = getTableSqlSchema(metaClient, includeMetadataFields = true).getOrElse(table.schema)
-    StructType(originSchema.map(_.copy(nullable = true)))
+    val fields = originSchema.fields.map { f =>
+      val nullableField: StructField = f.copy(nullable = true)
+      val catalogField = findColumnByName(table.schema, nullableField.name, resolver)
+      if (catalogField.isDefined) {
+        catalogField.get.getComment().map(nullableField.withComment).getOrElse(nullableField)
+      } else {
+        nullableField
+      }
+    }
+    StructType(fields)
   }
 
   /**
    * The schema without hoodie meta fields
    */
-  lazy val tableSchemaWithoutMetaFields: StructType = HoodieSqlCommonUtils.removeMetaFields(tableSchema)
+  lazy val tableSchemaWithoutMetaFields: StructType = removeMetaFields(tableSchema)
 
   /**
    * The schema of data fields
@@ -136,7 +147,7 @@ class HoodieCatalogTable(val spark: SparkSession, val table: CatalogTable) exten
   /**
    * The schema of data fields not including hoodie meta fields
    */
-  lazy val dataSchemaWithoutMetaFields: StructType = HoodieSqlCommonUtils.removeMetaFields(dataSchema)
+  lazy val dataSchemaWithoutMetaFields: StructType = removeMetaFields(dataSchema)
 
   /**
    * The schema of partition fields
@@ -146,7 +157,7 @@ class HoodieCatalogTable(val spark: SparkSession, val table: CatalogTable) exten
   /**
    * All the partition paths
    */
-  def getAllPartitionPaths: Seq[String] = HoodieSqlCommonUtils.getAllPartitionPaths(spark, table)
+  def getPartitionPaths: Seq[String] = getAllPartitionPaths(spark, table)
 
   /**
    * Check if table is a partitioned table
@@ -171,10 +182,12 @@ class HoodieCatalogTable(val spark: SparkSession, val table: CatalogTable) exten
         .setTableCreateSchema(SchemaConverters.toAvroType(finalSchema).toString())
         .initTable(hadoopConf, tableLocation)
     } else {
+      val (recordName, namespace) = AvroConversionUtils.getAvroRecordNameAndNamespace(table.identifier.table)
+      val schema = SchemaConverters.toAvroType(finalSchema, false, recordName, namespace)
       HoodieTableMetaClient.withPropertyBuilder()
         .fromProperties(properties)
         .setTableName(table.identifier.table)
-        .setTableCreateSchema(SchemaConverters.toAvroType(finalSchema).toString())
+        .setTableCreateSchema(schema.toString())
         .setPartitionFields(table.partitionColumnNames.mkString(","))
         .initTable(hadoopConf, tableLocation)
     }
@@ -239,7 +252,7 @@ class HoodieCatalogTable(val spark: SparkSession, val table: CatalogTable) exten
       originTableConfig: Map[String, String] = Map.empty): Map[String, String] = {
     val extraConfig = mutable.Map.empty[String, String]
     if (isTableExists) {
-      val allPartitionPaths = getAllPartitionPaths
+      val allPartitionPaths = getPartitionPaths
       if (originTableConfig.contains(HoodieTableConfig.HIVE_STYLE_PARTITIONING_ENABLE.key)) {
         extraConfig(HoodieTableConfig.HIVE_STYLE_PARTITIONING_ENABLE.key) =
           originTableConfig(HoodieTableConfig.HIVE_STYLE_PARTITIONING_ENABLE.key)

hudi-spark-datasource/hudi-spark-common/src/main/scala/org/apache/spark/sql/hudi/HoodieSqlCommonUtils.scala

@@ -19,6 +19,7 @@ package org.apache.spark.sql.hudi
 
 import org.apache.hadoop.conf.Configuration
 import org.apache.hadoop.fs.Path
+
 import org.apache.hudi.client.common.HoodieSparkEngineContext
 import org.apache.hudi.common.config.{DFSPropertiesConfiguration, HoodieMetadataConfig}
 import org.apache.hudi.common.fs.FSUtils
@@ -26,9 +27,10 @@ import org.apache.hudi.common.model.HoodieRecord
 import org.apache.hudi.common.table.timeline.{HoodieActiveTimeline, HoodieInstantTimeGenerator}
 import org.apache.hudi.common.table.{HoodieTableMetaClient, TableSchemaResolver}
 import org.apache.hudi.{AvroConversionUtils, SparkAdapterSupport}
+
 import org.apache.spark.api.java.JavaSparkContext
 import org.apache.spark.sql.catalyst.TableIdentifier
-import org.apache.spark.sql.catalyst.analysis.UnresolvedRelation
+import org.apache.spark.sql.catalyst.analysis.{Resolver, UnresolvedRelation}
 import org.apache.spark.sql.catalyst.catalog.{CatalogTable, CatalogTableType}
 import org.apache.spark.sql.catalyst.expressions.{And, Attribute, Expression}
 import org.apache.spark.sql.catalyst.plans.logical.{LogicalPlan, SubqueryAlias}
@@ -40,6 +42,7 @@ import org.apache.spark.sql.{Column, DataFrame, SparkSession}
 import java.net.URI
 import java.text.SimpleDateFormat
 import java.util.{Locale, Properties}
+
 import scala.collection.JavaConverters._
 import scala.collection.immutable.Map
 
@@ -301,4 +304,12 @@ object HoodieSqlCommonUtils extends SparkAdapterSupport {
       true
     }
   }
+
+  // Find the origin column from schema by column name, throw an AnalysisException if the column
+  // reference is invalid.
+  def findColumnByName(schema: StructType, name: String, resolver: Resolver):Option[StructField] = {
+    schema.fields.collectFirst {
+      case field if resolver(field.name, name) => field
+    }
+  }
 }

hudi-spark-datasource/hudi-spark-common/src/main/scala/org/apache/spark/sql/hudi/command/AlterHoodieTableChangeColumnCommand.scala

@@ -26,6 +26,7 @@ import org.apache.hudi.exception.HoodieException
 import org.apache.spark.sql.{AnalysisException, Row, SparkSession}
 import org.apache.spark.sql.catalyst.TableIdentifier
 import org.apache.spark.sql.catalyst.catalog.HoodieCatalogTable
+import org.apache.spark.sql.hudi.HoodieSqlCommonUtils._
 import org.apache.spark.sql.types.{StructField, StructType}
 
 import scala.util.control.NonFatal
@@ -40,16 +41,19 @@ case class AlterHoodieTableChangeColumnCommand(
   extends HoodieLeafRunnableCommand {
 
   override def run(sparkSession: SparkSession): Seq[Row] = {
+    val resolver = sparkSession.sessionState.conf.resolver
     val hoodieCatalogTable = HoodieCatalogTable(sparkSession, tableIdentifier)
 
-    val resolver = sparkSession.sessionState.conf.resolver
-    if (!resolver(columnName, newColumn.name)) {
-      throw new AnalysisException(s"Can not support change column name for hudi table currently.")
-    }
+    // Find the origin column from dataSchema by column name.
+    val originColumn = findColumnByName(hoodieCatalogTable.dataSchema, columnName, resolver).getOrElse(
+      throw new AnalysisException(s"Can't find column `$columnName` given table data columns " +
+        s"${hoodieCatalogTable.dataSchema.fieldNames.mkString("[`", "`, `", "`]")}")
+    )
+
     // Get the new schema
     val newTableSchema = StructType(
       hoodieCatalogTable.tableSchema.fields.map { field =>
-      if (resolver(field.name, columnName)) {
+      if (field.name == originColumn.name) {
         newColumn
       } else {
         field
@@ -57,7 +61,7 @@ case class AlterHoodieTableChangeColumnCommand(
     })
     val newDataSchema = StructType(
       hoodieCatalogTable.dataSchema.fields.map { field =>
-        if (resolver(field.name, columnName)) {
+        if (field.name == columnName) {
           newColumn
         } else {
           field

hudi-spark-datasource/hudi-spark-common/src/main/scala/org/apache/spark/sql/hudi/command/AlterHoodieTableDropPartitionCommand.scala

@@ -148,7 +148,7 @@ case class AlterHoodieTableDropPartitionCommand(
       hoodieCatalogTable: HoodieCatalogTable,
       normalizedSpecs: Seq[Map[String, String]]): String = {
     val table = hoodieCatalogTable.table
-    val allPartitionPaths = hoodieCatalogTable.getAllPartitionPaths
+    val allPartitionPaths = hoodieCatalogTable.getPartitionPaths
     val enableHiveStylePartitioning = isHiveStyledPartitioning(allPartitionPaths, table)
     val enableEncodeUrl = isUrlEncodeEnabled(allPartitionPaths, table)
     val partitionsToDrop = normalizedSpecs.map { spec =>

hudi-spark-datasource/hudi-spark-common/src/main/scala/org/apache/spark/sql/hudi/command/ShowHoodieTablePartitionsCommand.scala

@@ -48,10 +48,10 @@ case class ShowHoodieTablePartitionsCommand(
 
     if (partitionColumnNamesOpt.isPresent && partitionColumnNamesOpt.get.nonEmpty && schemaOpt.nonEmpty) {
       if (specOpt.isEmpty) {
-        hoodieCatalogTable.getAllPartitionPaths.map(Row(_))
+        hoodieCatalogTable.getPartitionPaths.map(Row(_))
       } else {
         val spec = specOpt.get
-        hoodieCatalogTable.getAllPartitionPaths.filter { partitionPath =>
+        hoodieCatalogTable.getPartitionPaths.filter { partitionPath =>
           val part = PartitioningUtils.parsePathFragment(partitionPath)
           spec.forall { case (col, value) =>
             PartitionPathEncodeUtils.escapePartitionValue(value) == part.getOrElse(col, null)