Github user rdblue commented on a diff in the pull request:
https://github.com/apache/spark/pull/21122#discussion_r186152832
--- Diff:
sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveDDLSuite.scala
---
@@ -1354,7 +1354,8 @@ class HiveDDLSuite
val indexName = tabName + "_index"
withTable(tabName) {
// Spark SQL does not support creating index. Thus, we have to use
Hive client.
- val client =
spark.sharedState.externalCatalog.asInstanceOf[HiveExternalCatalog].client
+ val client =
+
spark.sharedState.externalCatalog.unwrapped.asInstanceOf[HiveExternalCatalog].client
--- End diff --
There are several places that access the `client`, like
`HiveSessionStateBuilder` and `SaveAsHiveFile`. My point is that assuming the
catalog is a `HiveExternalCatalog`, casting, and accessing a field of the
catalog to get the client isn't a good way to pass the client to where it is
needed. Why doesn't this code pass both `client` and `catalog` to these classes
to avoid this problem?
I don't think it matters that Hive may be removed in the future. Unless
removing it is going to happen in a week or two, I'm skeptical that it is going
to be soon enough that we should ignore problems with the implementation.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
