Latest push

[Mark Rotteveel]

On 22-10-2018 17:27, Douglas Surber wrote:
> We just pushed updates to ADBA and AoJ.
> 
> ADBA changes:
> 
> 	Removed Session.activate and Session.deactivate
> 	Revised Session.Lifecycle
> 	Revised Result.Column to be more consistent
> 
> AoJ changes:
> 
> 	Updates to conform to ADBA changes
> 	Implemented all Operations
> 	Implemented real tests
> 
> We have only run the tests with Oracle Database. If you can run on another database let us know what you find.

Some notes (they are bit random in order and importance, just putting 
things when I find them):

ADBA:
- The enums in Session.java have a lot of Javadoc syntax errors (eg 
using {@link ATTACHED} instead of {@link #ATTACHED}.

AoJ
- README.md has replaced java.sql.DriverManager.getConnection with 
java.sql.DriverManager.getSession (probably when renaming the ADBA 
connection to session).
- Session.java mentions java.sql.Session instead of java.sql.Connection
- Session.java method jdbcValidate has logging for Session.isValid and 
Session.isClosed, which should be Connection as it references the JDBC 
connection. Other methods have similar problems (eg 
Session.prepareStatement while it references Connection.prepareStatement
- The code seems to favour the new lambda based logging methods where 
the 'normal' logging should probably be preferred to avoid creation of 
lambdas.
- Both TransactionCompletion.java and TransactionEnd.java contain the 
same class (TransactionCompletion), and I'm not sure which is the right 
one (they are very similar).
- Javadoc on OperationGroup field `held` references 
releaseProhibitingMoreOperations and submitHoldingForMoreOperations 
which no longer exist; also the Javadoc syntax for @see is not valid 
(should be @see #submit())
- DataSource.registerSession seems to be unused
- MultiOperation class javadoc has reference to method 
rowProcessorOperation that does not exist, should that be 
rowPublisherOperation instead?
- I notice that with some regularity, raw types, unchecked casts, etc 
are used which throws away type safety
- LocalOperation.execute() calls Thread.interrupted() but ignores it, 
should it be calling Thread.interrupt() instead or is this intentionally 
done to clear the interrupt state?
- Operation ADBATYPE_TO_JDBCTYPE: use an EnumMap
- I see a lot of casts from CompletionStage to CompletableFuture (eg in 
MultiOperation) without checks, this suggests that some usages of 
CompletionStage should be explicitly replaced with CompletableFuture.
- A lot of fields are protected, while their access should be restricted 
to private or package private. In some cases this yields logically 
incorrect code (eg ParameterizedOperation is public, but as 
ParameterValue is package private, subclass in other packages would not 
be able to access the protected field setParameters)
- RowBaseOperation.enquoteIdentifier is never used
- RowPublisherOperation.completeQuery casts a CompletionStage<? extends 
T> to T which can't be correct
- Session.sessionPropertyValue seems to be unused
- module-info.java is missing `requires java.logging;`

I haven't migrated the tests yet, maybe tomorrow or otherwise later this 
week.
-- 
Mark Rotteveel