Could anyone explain how EclipseLink detects the first access to field to make lazy fetching? It seems there is some magic and I can't find a way how it is done.
If I understand it right for @OneToMany relation EclipseLink creates intance of org.eclipse.persistence.indirection.IndirectList which controls the access. But I can't understand what EclipseLink does for @ManyToOne relation.
In the ManyToOne or OneToOne case EclipseLink does code enhancement to detect the access. That means instead of the real object you have a proxy in place.
There are docs:
https://www.eclipse.org/eclipselink/documentation/2.7/concepts/mappingintro002.htm#CHDJAHDC
Also interesting to read is the chapter about weaving. https://www.eclipse.org/eclipselink/documentation/2.7/concepts/app_dev005.htm#CCHJEDFH
