[hibernate-dev] SchemaCreatorImpl always creating a poolable sequence

[Mark Rotteveel]

On 28-7-2016 12:22, Vlad Mihalcea wrote:
> Hi Mark,
>
> I've just run the PooledHiLoSequenceIdentifierTest test after modifying
> the TABLE to SEQUENCE to see how it works and this is what I found:
>
> 1. SchemaCreatorImpl calls String[] getCreateSequenceStrings(String
> sequenceName, int initialValue, int incrementSize)
> 2. The String getCreateSequenceString(String sequenceName, int
> initialValue, int incrementSize) is then called which does this:
>
> if ( supportsPooledSequences() ) {
>    return getCreateSequenceString( sequenceName ) + " start with " + initialValue + " increment by " + incrementSize;
> }
> throw new MappingException( getClass().getName() + " does not support pooled sequences" );

Yes, and the above exception is my problem, however this default 
implementation suggests that this method should only be called if pooled 
sequences are supported (quod non), but in the current code it is called 
always in org.hibernate.tool.schema.internal.SchemaCreatorImpl.

> Now, this method is overridden in PostgreSQL and HSQLDB dialects.

Yes, but both support pooled sequences.

> I think we should create the sequence generator as indicated in this
> Firebird documentation:
>
> http://firebirdsql.org/manual/generatorguide-sqlsyntax.html

The current dialect already does that, but it isn't working because of 
above code. Providing the following override does seem to work, but this 
seems at odds with the exception thrown in the default implementation.

@Override
protected String getCreateSequenceString(String sequenceName, int 
initialValue, int incrementSize) throws MappingException {
	// Ignore initialValue and incrementSize
	return getCreateSequenceString( sequenceName );
}

For Firebird 2.5, 
http://www.firebirdsql.org/file/documentation/reference_manuals/fblangref25-en/html/fblangref25-ddl-sequence.html 
is more up-to-date.

I'm currently creating a new Firebird 2.5 specific dialect, and then I'm 
going to add a Firebird 3.0 dialect.

I might clean up the old InterbaseDialect and FirebirdDialect, but I 
don't want to touch those too much (especially since Firebird and 
Interbase should now be considered two distinct databases).

> Shouldn't we use GEN_ID for the pooled optimizers?

Yes GEN_ID could be used for that, but as far as I can tell the current 
support for pooled sequences seems to require specifying the increment 
size in the sequence DDL, which is not supported by Firebird.

Is there a specific extension point to override this, is it the 
getDefaultMultiTableBulkIdStrategy()?

Mark
-- 
Mark Rotteveel