It took a few years of pondering and a few hours to write the actual code. So
now a use can do something like this:
String[] configs = new String[] { "cayenne-x.xml", "cayenne-y.xml",
"cayenne-z.xml"};
ServerRuntime runtime = new ServerRuntime(configs);
The merge rules, as documented in Javadocs for
DefaultDataChannelDescriptorMerger are:
The general rule of merge is that the order of descriptors on the merge list
matters. If there are two conflicting metadata objects belonging to two
descriptors, an object from the last descriptor takes precedence over the
object from the first one. This way it is easy to override pieces of metadata.
This is also similar to how DI modules are merged in Cayenne. So this is how
the merge works:
* Merged descriptor name is the same as the name of the last descriptor on the
merge list.
* Merged descriptor properties are the same as the properties of the last
descriptor on the merge list. I.e. properties are not merged to avoid invalid
combinations and unexpected runtime behavior.
* If there are two or more DataMaps with the same name, only one DataMap is
placed in the merged descriptor, the rest are discarded. DataMap with highest
index in the descriptor array is chosen per precedence rule above.
* If there are two or more DataNodes with the same name, only one DataNodes is
placed in the merged descriptor, the rest are discarded. DataNodes with highest
index in the descriptor array is chosen per precedence rule above.
Of course none of that would have been easy without DI container and
simplifying the project structure to just one domain per project. Anyways, time
for 3.1M3, and IMO the Beta is already on the horizon.
Andrus
