Github user ericl commented on a diff in the pull request:
https://github.com/apache/spark/pull/16944#discussion_r102843314
--- Diff:
sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveMetastoreCatalog.scala ---
@@ -161,22 +164,45 @@ 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 inferredSchema = inferSchema(
+ catalogTable,
+ metastoreSchema,
+ options,
+ defaultSource,
+ fileType,
+ fileIndex)
+
+ // 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 = updateCatalogTable(catalogTable,
inferredSchema)
+
+ val dataSchema = inferenceMode match {
+ case (INFER_AND_SAVE | INFER_ONLY) if
(!catalogTable.schemaPreservesCase) =>
+ inferredSchema.getOrElse {
+ logWarning(s"Unable to infer schema for table
$tableIdentifier from file format " +
+ s"$defaultSource (inference mode: $inferenceMode); using
metastore schema.")
+ filteredMetastoreSchema
+ }
+ case _ =>
+ filteredMetastoreSchema
+ }
--- End diff --
I'm still finding this a bit awkward to read because the inferenceMode
config flag is referenced in three separate places. How about centralizing the
logic a little, moving the config checks to the body of this function?
```
val shouldInferSchema = !catalogTable.schemaPreservesCase && inferenceMode
!= NEVER_INFER
val dataSchema = if (shouldInferSchema) {
val inferredSchema = inferSchema(...)
if (inferenceMode == INFER_AND_SAVE) {
try {
updateCatalogTableSchema(...)
} catch {
...
}
}
inferredSchema
} else {
filteredMetastoreSchema
}
```
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
