Thanks Andrus, so the approach I would take now is NOT to have cayenne in any project-model jar.
Those deployment units should only have *.map.xml file and then all applications needing one or more of the data objects from those deployment units have a custom cayenne.xml. This would be awkward for the modeller if at all possible? The other scenario is of course for each top project to have copies of *.map.xml files of those deployment units holding data objects. This violates DRY principle which brings a maintenance responsibility for syncing mapping files across all projects. I can also imagine having one giant project-model with all the data objects defined in it for all the different databases in a company (across all company projects), but that doesn't seem very modular. Anyone with any good practice advice? Cheers, Borut 2010/2/13 Andrus Adamchik <[email protected]> > > On Feb 12, 2010, at 1:57 PM, Borut Bolčina wrote: > > >> What is the easiest way of doing this? I was reading >> http://cwiki.apache.org/CAY/runtime-metadata-merging.html, but am I doing >> something wrong? >> > > This page is a design document. Cayenne doesn't support merging multiple > projects out of the box, but hopefully will in 3.1. For now you'd have to > implement your own merging algorithm, that shouldn't be too hard for a > specific case. > > On a side note, we need a distinct style for the wiki scratch pages, so > that they are not mistaken for the official documentation. > > Andrus > >
