Gang, Writing the Shiro library and now the Scheduler one I'm facing the very same issue : circular dependencies between libraries and extensions when writing unit tests.
In other words, when writing unit tests for a library I often ends using the indexing-rdf extension for in-memory indexing. This creates a circular dependency between libraries and extensions. This is not acceptable so for example Shiro library unit tests were moved to the qi4j-tests release unit / repository because of that. I feel it to be very unconvenient and would like to start a discussion on about to change this fact. I see the point in having a qi4j-test release unit separated from libraries & extensions for tests that needs external setup or performance unit tests that are tooled right in qi4j-tests. I'm only talking about unit tests of libraries doing indexing/queries. In my case Shiro and Scheduler unit tests both depends on in-memory indexing, would it be possible to have such a module in a place that do not create circular dependencies ? I understand indexing is much more complex that entitystore and that it's why there is an in-memory entitystore implementation in qi4j-core but no indexing. As of today we have two potential in-memory indexing implementations : rdf, sql (solr seems to be file based only). I'm not giving a solution to my issue because I don't know (yet?) what would be a good one. WDYT ? /Paul _______________________________________________ qi4j-dev mailing list [email protected] http://lists.ops4j.org/mailman/listinfo/qi4j-dev
