[SPARK-29498][SQL] CatalogTable to HiveTable should not change the table's ownership

[wangyum]

What changes were proposed in this pull request?

CatalogTable to HiveTable will change the table's ownership. How to reproduce:

import org.apache.spark.sql.catalyst.TableIdentifier
import org.apache.spark.sql.catalyst.catalog.{CatalogStorageFormat, CatalogTable, CatalogTableType}
import org.apache.spark.sql.types.{LongType, StructType}

val identifier = TableIdentifier("spark_29498", None)
val owner = "SPARK-29498"
val newTable = CatalogTable(
  identifier,
  tableType = CatalogTableType.EXTERNAL,
  storage = CatalogStorageFormat(
    locationUri = None,
    inputFormat = None,
    outputFormat = None,
    serde = None,
    compressed = false,
    properties = Map.empty),
  owner = owner,
  schema = new StructType().add("i", LongType, false),
  provider = Some("hive"))

spark.sessionState.catalog.createTable(newTable, false)
// The owner is not SPARK-29498
println(spark.sessionState.catalog.getTableMetadata(identifier).owner)

This PR makes it set the HiveTable's owner to CatalogTable's owner if it's owner is not empty when converting CatalogTable to HiveTable.

Why are the changes needed?

We should not change the ownership of the table when converting CatalogTable to HiveTable.

Does this PR introduce any user-facing change?

No

How was this patch tested?

unit test

[SparkQA]

Test build #112271 has finished for PR 26160 at commit 86d8325.

• This patch fails Spark unit tests.
• This patch merges cleanly.
• This patch adds no public classes.

[SparkQA]

Test build #112274 has finished for PR 26160 at commit 507fced.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[wangyum]

cc @yaooqinn @cloud-fan

[yaooqinn]

IIUC, the ownership of a table only be created/changed(#26068) when we create or alter a table.
CreateTable - the owner <=> the creator <=> the spark runner
AlterTable - the original owner or the spark runner if none or empty
So I guess both cases are handled with or without this change.

SPARK-29507, We now just lack the ability to modify the owner of a table.

[wangyum]

I think the owner of the HiveTable is always the same as the owner of the CatalogTable if the owner of the CatalogTable is not empty.

This is how we convert HiveTable to CatalogTable:

spark/sql/hive/src/main/scala/org/apache/spark/sql/hive/client/HiveClientImpl.scala

Line 519 in 86d8325

owner = Option(h.getOwner).getOrElse(""),

[yaooqinn]

Anyway, this change can keep the consistency of ownership between catalog table and hive table, so LGTM

[dongjoon-hyun]

cc @gatorsmile

[cloud-fan]

LGTM, merging to master!

[dongjoon-hyun]

Thank you, @cloud-fan , @wangyum .
Since this is a bug reported for the old releases, can we have this in our LTS branch, branch-2.4, too?

[HyukjinKwon]

Thanks @wangyum. Looks fine to me too