…lt === result and not show error trace Closes #50
…n xml. Also correcting typo of multiple attributes with the same name on the same node in the README sample configuration xml.
…o enable huge performance gains in using many short-lived child injectors At the same time, the list of injection points has been made weak so that types from child ApplicationDomains don't keep the ApplicationDomain from being GCed if nothing else keeps it alive. As XML configuration doesn't expect to interact with metadata configuration at all, passing XML to the Injector ctor disables the global cache. All tests were adapted to call Injector.purgeInjectionPointsCache to make sure there aren't any unexpected interactions between different tests.
…e an injector for gathering their configuration settings. The Injector parameter is left available but made optional in all injection points to guarantee backwards compatibility in terms of semantic versioning.
Additionally, added some more tests for InjectionConfig. Closes #39
…#applyInjection to make stricter code analysis tools happy
…tantiatable Closes #37
…ed note about trick to use metadata based injection point configuration to section about XML based configuration
…arent to not work
…hing to do with getInstance
… to some ApplicationDomain problems
… to the parent injector once a missing dependency is encountered, the dependency is satisfied by the parent injector but the child injector retains control and processes subsequent injections further down the object tree
domains and flash.utils.getDefinitionByName. The loaded module
…the API actually does. getInstance implies that I will recieve an instance no matter what. Instead I get an error if it isn't mapped. This is an unexpected result. Sure, getInstance is nice and short, but a little verbosity here makes the intent of the API much clearer. An alternative would be to use Injector#instantiate to return an injected instance if the mapping is not available.
…ort a proper true result for hasMapping on a child injector.