Github user budde commented on a diff in the pull request:
https://github.com/apache/spark/pull/16944#discussion_r102554381
--- Diff:
sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveMetastoreCatalog.scala ---
@@ -161,22 +164,70 @@ private[hive] class
HiveMetastoreCatalog(sparkSession: SparkSession) extends Log
bucketSpec,
Some(partitionSchema))
+ val catalogTable = metastoreRelation.catalogTable
val logicalRelation = cached.getOrElse {
val sizeInBytes =
metastoreRelation.stats(sparkSession.sessionState.conf).sizeInBytes.toLong
val fileIndex = {
- val index = new CatalogFileIndex(
- sparkSession, metastoreRelation.catalogTable, sizeInBytes)
+ val index = new CatalogFileIndex(sparkSession, catalogTable,
sizeInBytes)
if (lazyPruningEnabled) {
index
} else {
index.filterPartitions(Nil) // materialize all the
partitions in memory
}
}
val partitionSchemaColumnNames =
partitionSchema.map(_.name.toLowerCase).toSet
- val dataSchema =
- StructType(metastoreSchema
- .filterNot(field =>
partitionSchemaColumnNames.contains(field.name.toLowerCase)))
+ val filteredMetastoreSchema = StructType(metastoreSchema
+ .filterNot(field =>
partitionSchemaColumnNames.contains(field.name.toLowerCase)))
+
+ // Infer a case-sensitive schema when the metastore doesn't
return one, if configured.
+ val inferenceMode =
sparkSession.sessionState.conf.caseSensitiveInferenceMode
+ val inferredSchema = if (!catalogTable.schemaPreservesCase &&
+ inferenceMode != HiveCaseSensitiveInferenceMode.NEVER_INFER)
{
+ logInfo(s"Inferring case-sensitive schema for table
$tableIdentifier")
+ val fileStatuses = fileIndex.listFiles(Nil).flatMap(_.files)
+ val inferred = defaultSource.inferSchema(sparkSession,
options, fileStatuses)
+ if (fileType.equals("parquet")) {
+
inferred.map(ParquetFileFormat.mergeMetastoreParquetSchema(metastoreSchema, _))
+ } else {
+ inferred
+ }
+ } else {
+ None
+ }
+
+ // If configured, save the inferred case-sensitive schema to the
table properties and
+ // fetch the updated CatalogTable record for use in the
LogicalRelation.
+ val updatedCatalogTable = if (!catalogTable.schemaPreservesCase
&&
+ inferenceMode ==
HiveCaseSensitiveInferenceMode.INFER_AND_SAVE) {
+ inferredSchema.flatMap { schema =>
+ logInfo(s"Saving case-sensitive schema for table
$tableIdentifier to table " +
+ "properties")
+ val updatedTable = catalogTable.copy(schema = schema)
+ try {
+ val catalog = sparkSession.sharedState.externalCatalog
+ catalog.alterTable(updatedTable)
+
Option(catalog.getTable(updatedTable.identifier.database.get,
+ updatedTable.identifier.table))
+ } catch {
+ case NonFatal(ex) =>
+ logError(s"Error saving case-sensitive schema for table
$tableIdentifier: $ex")
+ None
+ }
+ }
+ } else {
+ None
+ }
+
+ val dataSchema = if (!catalogTable.schemaPreservesCase) {
+ inferredSchema.getOrElse {
+ logWarning(s"Unable to infer schema for table
$tableIdentifier from file format " +
+ s"$defaultSource; using metastore schema.")
+ filteredMetastoreSchema
+ }
+ } else {
+ filteredMetastoreSchema
+ }
--- End diff --
I started by trying to keep this as consistent as possible with the
previous schema inference code but I agree that this function is a bit
unwieldy. I'll refactor some of this code into helper functions.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
