The class org.hibernate.ogm.model.key.spi.IdSourceKey is currently designed to handle keys which are mapped on multiple columns, with multiple values.

This is actually never the case as both a native sequence and a JPA TableGenerated will use at most a "segment" name, which is the single column we need to know about.

Also, the value object is always of String type.

I also noticed that several dialects actually made some assumptions about this being only a singleton array, and in some cases the value was being converted to String, so this cleans up some otherwise dodgy looking code.