Github user rdblue commented on the issue:
https://github.com/apache/spark/pull/20387
> I'd suggest that we just propogate the paths parameter to options, and
data source implementations are free to interprete the path option to whatever
they want, e.g. table and database names.
What about code paths that expect table names? In our branch, we've added
support for converting Hive relations (which have a `TableIdentifier`, not a
path) and using `insertInto`. Table names are paths are the two main ways to
identify tables and I think both should be supported.
This is a new API, so it doesn't matter that `load` and `save` currently
use paths. We can easily update that support for tables. If we don't, then
there will be no common way to refer to tables: some implementations will use
`table`, some will pass `db` separately, and some might use `database`.
Standardizing this and adding support in Spark will produce more consistent
behavior across data sources.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
