Please clarify documentation on how to use Coral-Spark with no hive about coral HOT 2 OPEN

But I've requirement of converting pure (LogicalPlan -> String) to re-generate sql.

Could you please clarify what kind of logical plan representation this refers to? If you are referring to Spark logical plan string representation, we have a WIP module that converts Spark's logical plan to a RelNode representation. You can see some test cases here. This is an example where HiveQL is not the input format to Coral. Actually, in this example the metadata source is not necessarily the HMS either.

In general, the way the input logic is expressed can be decoupled from the metadata store/engine. In the current implementation there are two options for sources of metadata: the Hive Metastore, and a local file on the local file system. Other metadata sources can be added as well. All metadata sources, including HMS, can be used with HiveQL or other input languages.

from coral.

Sorry for replying so late.

I think you got my point correctly.
If my requirement is to just convert a spark's logical plan object into any kind of output sql (for example take presto / trino) then HMS or any kind of metastore is redundant.
Spark's logical plan can be analyzed and further schema can be resolved for upstreams.

The tests that you have mentioned still requires HMS (after 3 months).

So wanted to know, can we independently convert spark logical plans into supported output sqls using coral? Or is it like we will always require a metadata store/engine?

from coral.