Support DDL

[meltsufin]

In Cloud Spanner, DDL statements are processed through a different API than SQL and DML.
See: spanner_database_admin.proto

[ChengyuanZhao]

Is the goal to use the same Statement type to run DDL ?

I'm not sure the r2dbc spec had that in mind: https://github.com/r2dbc/r2dbc-spi/blob/master/r2dbc-spec/src/main/asciidoc/statements.adoc

Furthermore, their tests specifically use an external thing to run the DDL : https://github.com/r2dbc/r2dbc-spi/blob/master/r2dbc-spi-test/src/main/java/io/r2dbc/spi/test/Example.java

I don't think they expected to use DDL inside R2DBC. Besides, while DML and SQL may happen arbitrary number of times in normal usage I think DML (creating deleting tables) is inherently an offline task right?

[elefeint]

Hmm, perhaps that's right. In SPI tests, the DDL operations are done outside of the reactive logic by getting hold of a separate JDBC template.

@mp911de Mark -- could you confirm whether R2DBC was meant to support DDL?

[mp911de]

R2DBC's Connection.createStatement(…) should be able to run any kind of queries, similar to how JDBC's Statement.execute() may run any kind of statements. Typically, DDL does not produce any rows nor result counts, therefore map(…) and getRowsUpdated() terminate empty for that kind of statements.

We're using JdbcTemplate in the SPi tests because it does not require the synchronization dance of reactive APIs.

[elefeint]

Good to know, thank you!

Cloud Spanner is a bit unusual in that it does not provide a single API to run any generic query through, so we'll have to do query detection in the driver. I've filed #92 to track doing so for DML, and this issue will track the DDL side.

[mp911de]

I'd suggest sticking with what your JDBC driver does in most cases. R2DBC is in a lot of ways similar to what JDBC does.

[ChengyuanZhao]

closing, this was done in #106