Github user cloud-fan commented on the issue: https://github.com/apache/spark/pull/20387 > This is a new API... Are you saying you wanna add a new method in `DataFreameReader` that is different than `load`? In Scala, parameter name is part of the method signature, so for `def load(path: String)`, we can't change its semantic, the parameter is a path. It's fine if a data source impelementation teach its users that path will interpreted as database/tables by it, but this should not be a contract in Spark. I do agree that Spark should set a standard for specifying database and table, as it's very common. We can even argue that path is not a general concept for data sources, but we still provide special APIs for path. My proposal: How about we add a new methods `table` in `DataFrameReader`? The usage would look like: `spark.read.format("iceberg").table("db.table").load()`, what do you think? We should not specify `database`, as if we may have catalog federation and table name may have 3 parts `catalog.db.table`. Let's keep it general and let the data source to interprete it.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org