feat(csharp/src/Drivers/Databricks): Design of SEA support for Databricks C# driver#3576
Conversation
| Thrift --> TStmt[DatabricksStatement<br/>Thrift] | ||
| Thrift --> TCF[CloudFetch<br/>Thrift-based] | ||
|
|
||
| REST --> RConn[StatementExecutionConnection] |
There was a problem hiding this comment.
Why do we need separation at this layer? Can't we reuse existing Statement and Connection classes, and reuse the session as well, and just change the underlying client integration client?
There was a problem hiding this comment.
both statement and connection will have different implementations between thrift and SEA, such as how to execute query and how to get metadata.
|
|
||
| REST --> RConn[StatementExecutionConnection] | ||
| REST --> RClient[StatementExecutionClient] | ||
| REST --> RStmt[StatementExecutionStatement] |
There was a problem hiding this comment.
This approach looks to me as prone to lots of redundant code, which can lead to maintenance cost for bug fixing.
|
we can reduce the redundant code by having a common parent class, this can be done during PR. |
CurtHagenlocher
left a comment
There was a problem hiding this comment.
I'm going to check this in as it looks fine to me and if there's any disagreement then iteration can continue against the checked-in design.
This PR introduces a comprehensive design document for adding Databricks Statement Execution API support to the
Databricks C# ADBC driver as an alternative to the current Thrift-based protocol. The design focuses on code
reuse, backward compatibility, and simplification.