This branch includes all changes for a 1.3.0 release of Oracle R2DBC.
I have verified this branch by running our test suite on both JDK 11 and JDK 23, and connecting to both Oracle Database 19(ADB) and 23. I also ran the OracleR2dbcRepositoryIntegrationTests from the Spring Data R2DBC Project.
The Oracle JDBC dependency is updated to version 23.6, and the Project Reactor dependency to version 3.6.11.
README Changes
Added basic release notes for 1.3.0
Revised JDBC connection properties section. I especially wanted to point out the alternative ways to configure properties in cases where an Option is not defined.
Revised the VECTOR section, providing more information about mappings between Java arrays and VECTOR dimension types, including BINARY. Also removed section about a defect in 23.5.
Implementation Changes
The only significant change in our implementation is the addition of an addCloseTask method in OracleConnectionImpl. This was added to workaround a defect (bug #37160069) in the 23.6 release of Oracle JDBC. The defect causes closeAsyncOracle to fail when a Blob, Clob, or a SELECT ... FOR UPDATE ResultSet are still open. To avoid this scenario, we can use addCloseTask to ensure that LOBs and ResultSets are freed/closed before calling closeAsyncOracle.
This change brought a slight refactoring to the constructors of many classes. These classes will now be constructed with a reference to the OracleConnectionImpl that created them. This allows the classes to access the addCloseTask method. My hope is that any functionality we add in the future can also be accessed through a method of OracleConnectionImpl.
Another change in our implementation is that Blob.discard() and Clob.discard() will no longer be called when a Blob or Clob is set as a bind value. It is not clear to me why this call was present in the OracleStatementImpl code to begin with. I am fairly sure it is not correct to do this though. If Oracle R2DBC did not create the Blob/Clob, then it should not be discarding it; The user who created the Blob/Clob should do instead.
Test Code Changes
Fixed compiler warnings
Correct the JDBC URL when PROTOCOL=tcps
Added checks for disabling pipelining
Added tests for Blob/Clob binds
Changed warning tests for DB 19 (forcing a view to reference a non-existent table is an error, not a warning)
This branch includes all changes for a 1.3.0 release of Oracle R2DBC.
I have verified this branch by running our test suite on both JDK 11 and JDK 23, and connecting to both Oracle Database 19(ADB) and 23. I also ran the OracleR2dbcRepositoryIntegrationTests from the Spring Data R2DBC Project.
The Oracle JDBC dependency is updated to version 23.6, and the Project Reactor dependency to version 3.6.11.
README Changes
Optionis not defined.Implementation Changes The only significant change in our implementation is the addition of an
addCloseTaskmethod inOracleConnectionImpl. This was added to workaround a defect (bug #37160069) in the 23.6 release of Oracle JDBC. The defect causescloseAsyncOracleto fail when a Blob, Clob, or aSELECT ... FOR UPDATEResultSet are still open. To avoid this scenario, we can useaddCloseTaskto ensure that LOBs and ResultSets are freed/closed before callingcloseAsyncOracle. This change brought a slight refactoring to the constructors of many classes. These classes will now be constructed with a reference to theOracleConnectionImplthat created them. This allows the classes to access theaddCloseTaskmethod. My hope is that any functionality we add in the future can also be accessed through a method ofOracleConnectionImpl.Another change in our implementation is that
Blob.discard()andClob.discard()will no longer be called when aBloborClobis set as a bind value. It is not clear to me why this call was present in theOracleStatementImplcode to begin with. I am fairly sure it is not correct to do this though. If Oracle R2DBC did not create the Blob/Clob, then it should not be discarding it; The user who created the Blob/Clob should do instead.Test Code Changes