Github user srowen commented on the issue:
https://github.com/apache/spark/pull/22466
There is ... see
https://cwiki.apache.org/confluence/display/Hive/LanguageManual+DDL#LanguageManualDDL-ManagedandExternalTables
I think Spark conflates the two. It's rare (?) but possible to specify a
custom location of a managed table, but, typically occurs for `EXTERNAL`
tables. So maybe this is OK.
```
private def createTable(tableIdent: TableIdentifier): Unit = {
val storage =
DataSource.buildStorageFormatFromOptions(extraOptions.toMap)
val tableType = if (storage.locationUri.isDefined) {
CatalogTableType.EXTERNAL
} else {
CatalogTableType.MANAGED
}
```
And in `SqlParser`:
```
// If location is defined, we'll assume this is an external table.
// Otherwise, we may accidentally delete existing data.
val tableType = if (external || location.isDefined) {
CatalogTableType.EXTERNAL
} else {
CatalogTableType.MANAGED
}
```
So if `LOCATION` implies `EXTERNAL` in Spark, then I get this. `EXTERNAL`
tables shouldn't be deleted.
I agree that the Hive impl doesn't seem to take this into account, on the
code paths that call `dropDatabase`. CC @andrewor14 in case he is available to
comment on the original implementaiton. WDYT @cloud-fan
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
